To see the other types of publications on this topic, follow the link: OCR.
Dissertations / Theses on the topic 'OCR'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 dissertations / theses for your research on the topic 'OCR.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
1
McDonald, Mercedes Terre. "OCR: A STATISTICAL MODEL OF MULTI-ENGINE OCR SYSTEMS." Master's thesis, University of Central Florida, 2004. http://digital.library.ucf.edu/cdm/ref/collection/ETD/id/4459.
Full textAbstract:
This thesis is a benchmark performed on three commercial Optical Character Recognition (OCR) engines. The purpose of this benchmark is to characterize the performance of the OCR engines with emphasis on the correlation of errors between each engine. The benchmarks are performed for the evaluation of the effect of a multi-OCR system employing a voting scheme to increase overall recognition accuracy. This is desirable since currently OCR systems are still unable to recognize characters with 100% accuracy. The existing error rates of OCR engines pose a major problem for applications where a single error can possibly effect significant outcomes, such as in legal applications. The results obtained from this benchmark are the primary determining factor in the decision of implementing a voting scheme. The experiment performed displayed a very high accuracy rate for each of these commercial OCR engines. The average accuracy rate found for each engine was near 99.5% based on a less than 6,000 word document. While these error rates are very low, the goal is 100% accuracy in legal applications. Based on the work in this thesis, it has been determined that a simple voting scheme will help to improve the accuracy rate.M.S.
Department of Electrical and Computer Engineering
Engineering and Computer Science
Electrical and Computer Engineering
2
Peluch, Tibor. "OCR cíleně znehodnocených textů." Master's thesis, Vysoké učení technické v Brně. Fakulta elektrotechniky a komunikačních technologií, 2009. http://www.nusl.cz/ntk/nusl-218134.
Full textAbstract:
The thesis deals with programming of application in operating system Windows. Main features of application system Microsoft Foundation Class are resumed in brief here. In following part there is idea about implementing an application with graphic user interface that makes, using schema, work with data, possible. The third part deals with implementation of blocks into dynamic linked libraries and there is outlined a possibility to use data of this programme as an external module and a possibility of realtime data processing e.g. picture and sound. The verification of a good functionality of this application is in the last part. The application is really tested in diagnosing of devaluated texts for protecting web forms www.centrum.cz. There were designed blocks making picture read possible just from internet, preprocessing, segmentation, feature extraction, evaluationg in neural network and blocks that make possible to read and save processed data into the disc.
3
Belgiovine, Mauro. "Advanced industrial OCR using Autoencoders." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2017. http://amslaurea.unibo.it/13807/.
Full textAbstract:
Il contenuto di questa tesi di laurea descrive il lavoro svolto durante un tirocinio di sei mesi presso Datalogic ADC. L'obiettivo del lavoro è stato quello di utilizzare uno specifico tipo di rete neurale, chiamata Autoencoder, per scopi legati al riconoscimento o alla convalida di caratteri in un sistema OCR industriale. In primo luogo è stato creato un classificatore di immagini di caratteri basato su Denoising Autoencoder; successivamente, è stato studiato un metodo per utilizzare l'Autoencoder come un classificatore di secondo livello, per meglio distinguere le false attivazioni da quelle corrette in condizioni di incertezza di un classificatore generico. Entrambe le architetture sono state valutate su dataset reali di clienti di Datalogic e i risultati sperimentali ottenuti sono presentati in questa tesi.
4
Noghe, Petr. "Vyhodnocení testových formulářů pomocí OCR." Master's thesis, Vysoké učení technické v Brně. Fakulta elektrotechniky a komunikačních technologií, 2013. http://www.nusl.cz/ntk/nusl-219986.
Full textAbstract:
This thesis deals with the evaluation forms using optical character recognition. Image processing and methods used for OCR is described in the first part of thesis. In the practical part is created database of sample characters. The chosen method is based on correlation between patterns and recognized characters. The program is designed in a graphical environment MATLAB. Finally, several forms are evaluated and success rate of the proposed program is detected.
5
Sahiti, Ylli. "OCR algoritmers noggrannhet och snabbhet vid identifieringen av text på olika typer av bakgrund : En jämförelse mellan OCR - algoritmerna Tesseract och Google ML-Kit." Thesis, Jönköping University, JTH, Avdelningen för datateknik och informatik, 2021. http://urn.kb.se/resolve?urn=urn:nbn:se:hj:diva-53789.
Full textAbstract:
SyfteOCR, optical character recognition, algoritmer kan implementeras på olika sätt, de påverkar även resultatet både beroende på vilken implementation som används och vilket dataset som det används på. Därför är det viktigt att testa de olika OCR algoritmerna på just det dataset som är tänkt att användas för att få ett förutsägbart resultat. Metod60 bilder är tagna på innehållsförteckningar tryckta på svenska livsmedelsprodukter med tre olika bakgrundsytor, aluminium, konvexa ytor och mjukplast. Två OCR algoritmer, ML Kit och Tesseract, har jämförts med avseende på precision och hastighet i syfte att hitta svårigheter för de respektive algoritmerna. ResultatBåda undersökta OCR algoritmerna hade störst svårighet att identifiera tecken i bilder med reflektion. ML Kit hade färre problem med lågupplösta bilder, Tesseract hade anmärkningsvärt större problem med lågupplösta bilder. För ändamålet som studien avser så är ML Kit det bästa valet. BegränsningarEnbart två OCR algoritmer jämförs i studien. Livsmedelsprodukter med tre olika bakgrundsytor har studerats, aluminium, konvex yta och mjukplast. Endast precision samt hastighet har jämförts. Precisionen har jämförts i antalet rätt ord samt hur nära en prediktering är från att vara helt rätt (Levenshtein algoritm).
6
Dordevic, Larisa, and Ahlén Hanna Richter. "Motivation till och Upplevelsen av Hinderbanelopp : Obstacle Course Race (OCR)." Thesis, Högskolan i Halmstad, Akademin för hälsa och välfärd, 2017. http://urn.kb.se/resolve?urn=urn:nbn:se:hh:diva-33625.
Full textAbstract:
Föreliggande studie syftade till att använda självbestämmandeteorin för att öka kunskapen om motiv till OCR-deltagande i Sverige genom att studera a) deltagarnas upplevelse av självbestämmande utifrån motivationsregleringarna, b) om de grundläggande psykologiska behoven tillfredsställs i deltagandet och c) vilka egenskaper (t.ex. utformning av loppen, svårighetsgrad på hinder, omgivning) i själva OCR-eventet som kan relateras till dessa faktorer. OCR räknas som en av de snabbast växande sporterna i världen och verkar utmana samt attrahera en stor mångfald. En kvalitativ forskningsmetod användes där åtta deltagare (3 kvinnor och 5 män) från sydvästra Sverige intervjuades för att få en djupgående inblick i motivationen samt upplevelsen av eventen. Samtliga deltagare hade deltagit i minst ett OCRevent i Sverige innan intervjuerna. Intervjuerna samt analyserna var baserade på de tre grundläggande psykologiska behoven och motivationsregleringarna utifrån Självbestämmandeteorin (SDT). Studiens huvudsakliga fynd indikerade att deltagande i OCR bidrog till både inre och yttre motivation och att samtliga behov utifrån SDT tillfredsställdes, vilket genererade i att en hög nivå av självbestämmande upplevdes bland deltagarna. Vidare visade resultaten att det fanns motivationsfaktorer, exempelvis den sociala omgivningen, glädjen, utmaningen och utvecklingen, som hade större påverkan än andra faktorer. Det ansågs i den föreliggande studien att deltagarna upplevde en hög nivå av självbestämmande och att OCR kan bidra till ett ökat psykologiskt välbefinnande. Framtida forskning bör studera ett större och mer slumpmässigt urval för att öka generaliserbarheten men även ta del av de negativa hälsoeffekterna OCR (extremsport) skulle kunna bidra till.
7
Lund, Mikael. "Hur ser framtiden ut för OCR?" Thesis, Malmö högskola, Fakulteten för teknik och samhälle (TS), 2007. http://urn.kb.se/resolve?urn=urn:nbn:se:mau:diva-20561.
Full textAbstract:
Examensarbetet handlar om OCR (Optical Character Recognition). OCR-tekniken går utpå att konvertera inskannade bilder från maskinskriven eller handskriven text (siffror, bokstäver och symboler) till datorformat.Syftet med detta examensarbete är att utforska OCRs framtid och vilka användningsområden som finns idag för tekniken. Det intressanta är att se hur OCR klarar sig när mer och mer material är digitala.Genomförandet till detta examensarbete har gjorts med information från böcker, Internet,mejl och genom att tittat närmare på ett företag inom den grafiska branschen som använder sig av OCR, nämligen Aftonbladet. Jag har även testat ett OCR-program, ABBYYsFineReader 8 och gjort tester med några testteman, exempelvis matematiktest och olikatester på artiklar från några tidningar.Mina slutsatser är att OCR har en framtid men tekniken har en del förbättringsmöjligheter,exempelvis tolkning av handskrivna texter. OCR kan finnas kvar även när mer och mermaterial blir digitala om det integreras i befintliga tekniker, som i ett spam-filter för att tolka texten i bilden. Den nuvarande OCR-tekniken fungerar bra om materialet ärmaskinskrivet och i bra skick men den måste bli bättre på att tolka handskrivna texter för att kunna användas vid arkiveringsbehov av sådana texter.My examination subject is about OCR (Optical Character Recognition). The idea of OCRtechnology is to convert scanned images of machine-printed or handwritten text (numerals, letters and symbols) into a computer-processable format.The purpose of my examination subject is to explore the future of OCR and why to use it today. It’s interesting to see if OCR survives when more and more material is digital.The implementations to the examination subject have been made from books, Internet, e-mail and I have discovered how a company in the graphic industry are using OCR, namely Aftonbladet.I have also tested an OCR-program, ABBYYs FineReader 8, and done some testing with some testthemes, for example mathematics test and different tests on articles from a few magazines.My conclusions are that OCR has a future but the technology needs some improvements, forexample interpreting handwritten texts. OCR can exist, even when more and more material is digital, if its integrated with existing technologies, for example with a spam-filter to interpret the text within in the picture. The current OCR-technology works fine with machine-printed material, and when the document quality is good. However it needs to be on handwritten text to be used forarchiving needs.
8
Serafini, Sara. "Machine Learning applied to OCR tasks." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2019.
Find full textAbstract:
The content of this thesis describes the work done during a six-month
internship at Datalogic, in its research laboratories in Pasadena (CA). The aim of my research was to implement and evaluate a classifier as part of an industrial OCR system for learning purposes and to see how well it could work in comparison to current best Datalogic products, since it might be simpler/faster, it might be a good alternative for implementing on an embedded system (where current Datalogic products may not be able to run fast enough).
9
Nilsson, Elin. "Test av OCR-verktyg för Linux." Thesis, Linnaeus University, School of Computer Science, Physics and Mathematics, 2010. http://urn.kb.se/resolve?urn=urn:nbn:se:lnu:diva-5906.
Full textAbstract:
Denna rapport handlar om att ta fram ett OCR-verktyg för digitalisering av pappersdokument. Krav på detta verktyg är att bland annat det ska vara kompatibelt med Linux, det ska kunna ta kommandon via kommandoprompt och dessutom ska det kunna hantera skandinaviska tecken.
Tolv OCR-verktyg granskades, sedan valdes tre verktyg ut; Ocrad, Tesseract och OCR Shop XTR. För att testa dessa scannades två dokument in och digitaliserades i varje verktyg.
Resultatet av testerna är att Tesseract är de verktyget som är mest precist och Ocrad är det verktyget som är snabbast. OCR Shop XTR visar på sämst resultat både i tidtagning och i antal korrekta ord.
This report is about finding OCR software for digitizing paper documents. Requirements were to include those which were compatible with Linux, being able to run commands via the command line and also being able to handle the Scandinavian characters.
Twelve OCR softwares were reviewed, and three softwares were chosen; Ocrad, Tesseract and OCR Shop XTR. To test these, two document were scanned and digitized in each tool.
The results of the tests are that Tesseract is the tool which is the most precise and Ocrad is the tool which is the fastest. OCR Shop XTR shows the worst results both in timing and number of correct words.
10
Buchal, Petr. "Využití neanotovaných dat pro trénování OCR." Master's thesis, Vysoké učení technické v Brně. Fakulta informačních technologií, 2021. http://www.nusl.cz/ntk/nusl-445580.
Full textAbstract:
The creation of a high-quality optical character recognition system (OCR) requires a large amount of labeled data. Obtaining, or in other words creating, such a quantity of labeled data is a costly process. This thesis focuses on several methods which efficiently use unlabeled data for the training of an OCR neural network. The proposed methods fall into the category of self-training algorithms. The general approach of all proposed methods can be summarized as follows. Firstly, the seed model is trained on a limited amount of labeled data. Then, the seed model in combination with the language model is used for producing pseudo-labels for unlabeled data. Machine-labeled data are then combined with the training data used for the creation of the seed model and they are used again for the creation of the target model. The successfulness of individual methods is measured on the handwritten ICFHR 2014 Bentham dataset. Experiments were conducted on two datasets which represented different degrees of labeled data availability. The best model trained on the smaller dataset achieved 3.70 CER [%], which is a relative improvement of 42 % in comparison with the seed model, and the best model trained on the bigger dataset achieved 1.90 CER [%], which is a relative improvement of 26 % in comparison with the seed model. This thesis shows that the proposed methods can be efficiently used to improve the OCR error rate by means of unlabeled data.
11
Grönlund, Jakob, and Angelina Johansson. "Defect Detection and OCR on Steel." Thesis, Linköpings universitet, Datorseende, 2019. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-157508.
Full textAbstract:
In large scale productions of metal sheets, it is important to maintain an effective way to continuously inspect the products passing through the production line. The inspection mainly consists of detection of defects and tracking of ID numbers. This thesis investigates the possibilities to create an automatic inspection system by evaluating different machine learning algorithms for defect detection and optical character recognition (OCR) on metal sheet data. Digit recognition and defect detection are solved separately, where the former compares the object detection algorithm Faster R-CNN and the classical machine learning algorithm NCGF, and the latter is based on unsupervised learning using a convolutional autoencoder (CAE). The advantage of the feature extraction method is that it only needs a couple of samples to be able to classify new digits, which is desirable in this case due to the lack of training data. Faster R-CNN, on the other hand, needs much more training data to solve the same problem. NCGF does however fail to classify noisy images and images of metal sheets containing an alloy, while Faster R-CNN seems to be a more promising solution with a final mean average precision of 98.59%. The CAE approach for defect detection showed promising result. The algorithm learned how to only reconstruct images without defects, resulting in reconstruction errors whenever a defect appears. The errors are initially classified using a basic thresholding approach, resulting in a 98.9% accuracy. However, this classifier requires supervised learning, which is why the clustering algorithm Gaussian mixture model (GMM) is investigated as well. The result shows that it should be possible to use GMM, but that it requires a lot of GPU resources to use it in an end-to-end solution with a CAE.
12
Nogén, David, and Jennifer Jonsson. "Matbudgetapplikation." Thesis, KTH, Data- och elektroteknik, 2013. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-123690.
Full textAbstract:
Flera nya tjänster som Mina utgifter och Smartbudget vittnar om ett tilltagande intresse bland konsumenter att planera sin ekonomi. Matvaror utgör en stor del av det genomsnittliga hushållets budget och är därmed en kostnadspost som kan göra stor skillnad i hushållets ekonomi.Detta examensarbete ska undersöka möjligheten att jämföra matvarors pris på olika affärer med hjälp av en Android-applikation och genom att fotografera texten på kvitton. Texten kommer sen processas och sorteras för att få ut nödvändig data som sen kan sparas undan i en databas. Färdiga Algoritmer och OCR-motorer har utvärderats och implementerats i applikationen direkt genom så kallade C-Bibliotek. Dessa gör det möjligt att utan större problem vidareutveckla applikationen för iOS eller Windows Phone.Projektet och Android-applikationen visar på möjligheterna att använda färdiga C-bibliotek samt telefoners kamera för att enkelt sålla ut och spara undan den informationen som är relevant för konsumentens del.Multiple new services such as “Mina utgifter” and “Smartbudget” show that there is an increased interest among consumers to plan their economy. Groceries represent a large part of the average households budget and is thereby an important thing that can make a large difference in every households economy.This thesis will examine the possibilities to compare food prices with help of an Android application and by taking pictures of the text on receipts. The text will then be processed and sorted to get the necessary data which later can be saved into a database. Premade algorithms and OCR-engines have been evaluated and implemented directly into the Android application by using so called C-Libraries. This makes it possible without major efforts to further develop the application for IOS or Windows Phone.This project and the Android application show the possibilities to use premade libraries and the phones camera to extract and save the necessary information that is relevant for consumers.
13
Poli, Flavio. "Robust string text detection for industrial OCR." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2017. http://amslaurea.unibo.it/12885/.
Full textAbstract:
Tesi che propone un algoritmo per il ritrovamento di linee di testo per OCR industriali. Tramite un aproccio ad albero e sfruttando la conoscenza sulla stringa da cercare, vengono esplorate più soluzioni fino a trovare quella più promettente. Fornisce in uscita anche una stima su quanto l'algoritmo è confidente sul risultato.
14
Corsi, Giacomo. "Fast Neural Network Technique for Industrial OCR." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2018. http://amslaurea.unibo.it/15258/.
Full textAbstract:
The content of my thesis describes the work done during my internship at Datalogic in Pasadena. This project improves the performance of the Optical Character Recognition (OCR) solution with use of Deep Learning (DL) techniques. It enhances the character detection process that had been previously developed and relies on template matching done on the Histogram of Gradients (HOG) features. This approach had been already validated with good performance, but detects only those characters which do not vary in the dataset.
First, this document gives a introduction to OCR and DL topics, then describes the pipeline of the Datalogic OCR product.
After that, it is explained the technique that was usedto raise the accuracy of the previous solution. It consists in applying DL to improve the robustness and keep good detection rate even though the character variations (scale and rotation) are considerable.
The first phase was focused on speeding up the process and so the function used for gauging the matching with the templates, the Zero-mean Normalized Cross-Correlation, was replaced while a modified version, called Squared Normalization has been introduced.
Secondly, the original system was cast as a Convolutional Neural Network (CNN) by turning the HOG templates into convolutional kernels. It was necessary to rethink its training process as it was noticed that, using standard target values, there was no gain. A novel way of computing the targets, named Graceful Improvement, has been developed.
Then, the analysis on the results of this new solution showed that, even ifit detects characters that present variations with original templates, the false positive rate around the image was also higher. To decrease this negative side effect, a fast ROI (Region Of Interest) filter acting on the detections has been realized.
Finally, during the above development steps, performances in terms of accuracy and time have been evaluated on some real Datalogic's customer datasets.
15
Strohmaier, Christian M. "Methoden der lexikalischen Nachkorrektur OCR-erfasster Dokumente." Diss., lmu, 2005. http://nbn-resolving.de/urn:nbn:de:bvb:19-36743.
Full text
16
Fridolfsson, Olle. "Machine Learning : for Barcode Detection and OCR." Thesis, Linköpings universitet, Datorseende, 2015. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-119425.
Full textAbstract:
Machine learning can be utilized in many different ways in the field of automatic manufacturing and logistics. In this thesis supervised machine learning have been utilized to train a classifiers for detection and recognition of objects in images. The techniques AdaBoost and Random forest have been examined, both are based on decision trees. The thesis has considered two applications: barcode detection and optical character recognition (OCR). Supervised machine learning methods are highly appropriate in both applications since both barcodes and printed characters generally are rather distinguishable. The first part of this thesis examines the use of machine learning for barcode detection in images, both traditional 1D-barcodes and the more recent Maxi-codes, which is a type of two-dimensional barcode. In this part the focus has been to train classifiers with the technique AdaBoost. The Maxi-code detection is mainly done with Local binary pattern features. For detection of 1D-codes, features are calculated from the structure tensor. The classifiers have been evaluated with around 200 real test images, containing barcodes, and shows promising results. The second part of the thesis involves optical character recognition. The focus in this part has been to train a Random forest classifier by using the technique point pair features. The performance has also been compared with the more proven and widely used Haar-features. Although, the result shows that Haar-features are superior in terms of accuracy. Nevertheless the conclusion is that point pairs can be utilized as features for Random forest in OCR.
17
Kapusta, Ján. "OCR modul pro rozpoznání písmen a číslic." Master's thesis, Vysoké učení technické v Brně. Fakulta elektrotechniky a komunikačních technologií, 2010. http://www.nusl.cz/ntk/nusl-218623.
Full textAbstract:
This paper describes basic methods used for optical character recognition. It explains all procedures of recognition from adjustment of picture, processing, feature extracting to matching algorithms. It compares methods and algorithms for character recognition obtained graphically distorted or else modified image so-called „captcha“, used in present. Further it compares method based on invariant moments and neural network as final classifier and method based on correlation between normals and recognized characters.
18
Schaedel, Karin, and Tommy Söderberg. "Automatisk överföring av analog data från pappersenkäter till digital databas på Karolinska Universitetssjukhuset Huddinge." Thesis, KTH, Medicinteknik och hälsosystem, 2020. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-277627.
Full textAbstract:
På Karolinska Universitetssjukhuset Huddinge har en stor mängd knäprotesoperationsenkäter staplats på hög under två år. Svaren från dessa ska konverteras till digitalt format så dessa kan lagras i databasen REDCap för att kunna utföra kvalitetskontroll och prospektiv uppföljning i flera år. För att spara arbetstid efterfrågades ett program som kunde läsa in enkätsvaren automatiskt. I detta projekt skapades ett program i MATLAB med målsättning att klara av att läsa enkätmarkeringar samt minst 70 % av personnumren. Dessa personnummer skulle sedan läggas in i ett Excel-ark och övrig svarsdata i ett separat Excel-ark på grund av sekretesslagar. Resultatet blev att programmet inte klarade av att läsa av personnummer och annan handskriven text men klarade att läsa av markerade flervalsfrågor till 90 % säkerhet i just de enkäterna som programmet var designat för. Programmet kan i nuläget användas för smidigare inläsning tillsammans med korrekturläsning. Det rekommenderas dock att fortsätta vidareutveckla programmet innan användning sker.At Karolinska University Hospital Huddinge, many knee replacement surveys have been piled high for two years. The answers from these must be converted to digital format so that they can be stored in the REDCap database to be able to perform quality control and prospective follow-up for several years. To save working hours, a program that could read the questionnaires automatically was requested. In this project, a program was created in MATLAB with the goal of being able to read questionnaire markings and at least 70% of the social security numbers. These social security numbers were to be written on an Excel sheet and other answer data on a separate Excel sheet due to confidentiality laws. The result was that the program could not handle reading of social security numbers and other handwritten text but managed to read marked multiple-choice questions to 90% certainty in the surveys for which the program was designed. The program can currently be used for easier reading along with proofreading from staff. However, it is recommended to continue to develop the program before using it.
19
Spedicati, Marco. "Automatic generation of annotated datasets for industrial OCR." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2019. http://amslaurea.unibo.it/17385/.
Full textAbstract:
Machine learning algorithms need a lot of data, both for training and for
testing. However, not always appropriate data are in fact available. This
document presents the work that has been carried out at Datalogic USA’s
laboratories in Eugene, Oregon, USA, to create data for industrial Optical
Character Recognition (OCR) applications. It describes the automatic sys-
tem that has been built. The images are created by printing and capturing
strings of a variable layout, and they are ground truthed in a later stage, in
an automatic way. Two datasets are generated, of which one is employed to
asses a network’s performance.
20
Lamberti, Lorenzo. "A deep learning solution for industrial OCR applications." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2019. http://amslaurea.unibo.it/19777/.
Full textAbstract:
This thesis describes a project developed throughout a six months internship in the Machine Vision Laboratory of Datalogic based in Pasadena, California.
The project aims to develop a deep learning system as a possible solution for industrial optical character recognition applications. In particular, the focus falls on a specific algorithm called You Only Look Once (YOLO), which is a general-purpose object detector based on convolutional neural networks that currently offers state-of-the-art performances in terms of trade-off between speed and accuracy. This algorithm is indeed well known for reaching impressive processing speeds, but its intrinsic structure makes it struggle in detecting small objects clustered together, which unfortunately matches our scenario: we are trying to read alphanumerical codes by detecting each single character and then reconstructing the final string. The final goal of this thesis is to overcome this drawback and push the accuracy performances of a general object detector convolutional neural network to its limits, in order to meet the demanding requirements of industrial OCR applications. To accomplish this, first YOLO's unique detecting approach was mastered in its original framework called Darknet, written in C and CUDA, then all the code was translated into Python programming language for a better flexibility, which also allowed the deployment of a custom architecture. Four different datasets with increasing complexity were used as case-studies and the final performances reached were surprising: the accuracy varies between 99.75\% and 99.97\% with a processing time of 15 ms for images $1000\times1000$ big, largely outperforming in speed the current deep learning solution deployed by Datalogic. On the downsides, the training phase usually requires a very large amount of data and time and YOLO also showed some memorization behaviours if not enough variability is given at training time.
21
Larsson, Andreas, and Tony Segerås. "Automated invoice handling with machine learning and OCR." Thesis, KTH, Data- och elektroteknik, 2016. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-188202.
Full textAbstract:
Companies often process invoices manually, therefore automation could reduce manual labor. The aim of this thesis is to evaluate which OCR-engine, Tesseract or OCRopus, performs best at interpreting invoices. This thesis also evaluates if it is possible to use machine learning to automatically process invoices based on previously stored data. By interpreting invoices with the OCR-engines, it results in the output text having few spelling errors. However, the invoice structure is lost, making it impossible to interpret the corresponding fields. If Naïve Bayes is chosen as the algorithm for machine learning, the prototype can correctly classify recurring invoice lines after a set of data has been processed. The conclusion is, neither of the two OCR-engines can interpret the invoices to plain text making it understandable. Machine learning with Naïve Bayes works on invoices if there is enough previously processed data. The findings in this thesis concludes that machine learning and OCR can be utilized to automatize manual labor.Företag behandlar oftast fakturor manuellt och en automatisering skulle kunna minska fysiskt arbete. Målet med examensarbetet var att undersöka vilken av OCR-läsarna, Tesseract och OCRopus som fungerar bäst på att tolka en inskannad faktura. Även undersöka om det är möjligt med maskininlärning att automatiskt behandla fakturor utifrån tidigare sparad data. Genom att tolka text med hjälp av OCR-läsarna visade resultaten att den producerade texten blev språkligt korrekt, men att strukturen i fakturan inte behölls vilket gjorde det svårt att tolka vilka fält som hör ihop. Naïve Bayes valdes som algoritm till maskininlärningen och resultatet blev en prototyp som korrekt kunde klassificera återkommande fakturarader, efter att en mängd träningsdata var behandlad. Slutsatsen är att ingen av OCR-läsarna kunde tolka fakturor så att resultatet kunde användas vidare, och att maskininlärning med Naïve Bayes fungerar på fakturor om tillräckligt med tidigare behandlad data finns. Utfallet av examensarbetet är att maskininlärning och OCR kan användas för att automatisera fysiskt arbete.
22
Higham, Richard G. "A biophysical analysis of the Ocr protein gel." Thesis, University of Edinburgh, 2007. http://hdl.handle.net/1842/2569.
Full textAbstract:
Ocr is unusual among proteins in its ability to form a transparent gel at high ammonium sulphate concentrations. This transition was investigated using a combination of spectroscopic, microscopic and rheological techniques. It occurs sharply at a concentration of 3.2M ammonium sulphate and is not observed with other types of salt. Rheological measurements showed that rather than precipitating under such conditions, ocr forms a weak viscoelastic gel. Far UV circular dichroism spectra reveal that ocr does not denature in the gel phase, while near UV CD spectra suggest the formation of long, helical structures. Well resolved fibrils were observed using atomic force microscopy. They were over 1µm in length and varied between 2.6nm to 10.4nm in height, corresponding to the thickness and length of the ocr dimer. Ocr is a highly charged protein (-56e at pH 8) and is shaped like a banana. We argue that it is stabilized in specifically aggregated structures at large salt concentrations by these physical properties. Electrostatic repulsions between proteins are screened by salts, allowing proteins to approach close enough to aggregate. The charge on ocr is high enough to resist such precipitation. However, at 3.2M ammonium sulphate we suggest that the salt molecules bridge neighbouring ocr dimers via hydrogen bonds, connecting amino acid carboxyl groups with the ammonium groups of the salt. The banana-shaped dimers stack on top of each other, forming long helical fibrils that intertwine into a semi flexible network.
23
Nederhof, Mark-Jan. "OCR of hand-written transcriptions of hieroglyphic text." Universitätsbibliothek Leipzig, 2016. http://nbn-resolving.de/urn:nbn:de:bsz:15-qucosa-201704.
Full textAbstract:
Encoding hieroglyphic texts is time-consuming. If a text already exists as hand-written transcription, there is an alternative, namely OCR. Off-the-shelf OCR systems seem difficult to adapt to the peculiarities of Ancient Egyptian. Presented is a proof-of-concept tool that was designed to digitize texts of Urkunden IV in the hand-writing of Kurt Sethe. It automatically recognizes signs and produces a normalized encoding, suitable for storage in a database, or for printing on a screen or on paper, requiring little manual correction.
The encoding of hieroglyphic text is RES (Revised Encoding Scheme) rather than (common dialects of) MdC (Manuel de Codage). Earlier papers argued against MdC and in favour of RES for corpus development. Arguments in favour of RES include longevity of the encoding, as its semantics are font-independent. The present study provides evidence that RES is also much preferable to MdC in the context of OCR. With a well-understood parsing technique, relative positioning of scanned signs can be straightforwardly mapped to suitable primitives of the encoding.
24
Capra, Daniele. "Applicazione di sistemi ocr in contatori di consumo domestico." Bachelor's thesis, Alma Mater Studiorum - Università di Bologna, 2015. http://amslaurea.unibo.it/8788/.
Full textAbstract:
La tesi parla dei software OCR, della loro storia e di come si sono evoluti nel tempo, come sono strutturati e come funziona la logica alla base del loro funzionamento, inoltre studia come questi software vengano utilizzati nella lettura dei contatori di consumo domestici.
25
Holt, Adam 1971. "Scan your life : integrating OCR into your personal haystack!" Thesis, Massachusetts Institute of Technology, 2000. http://hdl.handle.net/1721.1/8562.
Full textAbstract:
Thesis (M.Eng. and S.B.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2000.Includes bibliographical references (p. 93-105).
I built a self-serve OCR station where anybody can scan in documents at high-speed - a public yet private ATM that accepts document deposits of a wider assortment than just checks. Depending on whether you scan a business card, an article or your entire filing cabinet, CPU-intensive recognition continues after you leave the station, and you are emailed options for secure web pickup. Users of MIT's Haystack personal repositories can even do "1-click" merging of offline literary artifacts into their online lives. The paperless pipe dream may never happen, but cheap digital optics and a mundane 40-year old technology (OCR) are converging to change the game. The mindless convenience of my $6000 kiosk suggests OCR will become a regulated munition* in the coming intellectual property and privacy wars. As OCR proliferates into cheap PDA's, neither publisher nor individual may ever again rely on humanity's oldest form of copy protection: paper. (*) The Digital Millennium Copyright Act (1998) bans technology that circumvents copyright locks.
by Adam Holt.
M.Eng.and S.B.
26
Candeias, Mariline Teixeira. "Estudo da Variação de Kₒ com OCR em Areias." Master's thesis, Faculdade de Ciências e Tecnologia, 2012. http://hdl.handle.net/10362/8537.
Full textAbstract:
Dissertação para obtenção do Grau de Mestre em
Engenharia Civil (Perfil de Estruturas e Geotecnia)O presente documento tem como objetivo o estudo da variação do coeficiente de impulso em repouso com o grau de sobreconsolidação, num solo arenoso. Este parâmetro é quantificado através da realização de ensaios triaxiais de consolidação Kₒ, em provetes reconstituídos de areia Toyoura. Para a realização dos ensaios utilizou-se equipamento laboratorial controlado computacionalmente através do GDSLAB, disponível no laboratório de Mecânica dos Solos do Departamento de Engenharia Civil da Nova (Universidade Nova de Lisboa). Foram verificadas as condições de validade, do módulo de operação do ensaio Kₒ, nomeadamente, a condição de deformação radial nula e o excesso de pressão intersticial nulo. Dos ensaios laboratoriais de validação, também se estudou a taxa de carregamento que se adequa ao solo normalmente consolidado e sobreconsolidado. Finalmente, procedeu-se ao ensaio final de consolidação Kₒ em que se realizou três ciclos de carga-descarga, por forma a tornar mais evidente a influência de OCR no valor de Kₒ. Deste ensaio, conclui-se que para todos os ciclos de carga-descarga realizados o valor máximo de Kₒ é sensivelmente 1,5. Os resultados experimentais são comparados com os resultados das fórmulas empíricas sugeridas por outros autores. Da qual se concluiu que a fórmula de Mayne e Kulhawy (1982) é a que se melhor aproxima dos resultados do presente estudo para um valor de OCR inferior ou igual a 5. Para valores de OCR superiores a 5, os valores de Kₒ seguem trajetórias lineares diferentes, para as quais foram propostas fórmulas empíricas.
27
Stephanou, Augoustinos S. "Biophysical study of the DNA charge mimicry displayed by the T7 Ocr protein." Thesis, University of Edinburgh, 2010. http://hdl.handle.net/1842/4348.
Full textAbstract:
The homodimeric Ocr protein of bacteriophage T7 is a molecular mimic of a bent double-stranded DNA molecule ~24 bp in length. As such, Ocr is a highly effective competitive inhibitor of the bacterial Type I restriction modification (R/M) system. Thus, Ocr facilitates phage infection of the bacterial cell to proceed unhindered by the action of the R/M defense system. The main aim of this work was to understand the basis of the DNA mimicry displayed by Ocr. The surface of the protein is replete with acidic residues, most or all of which mimic the phosphate backbone of DNA. Aspartate and glutamate residues on the surface of Ocr were either mutated or chemically modified in order to investigate their contribution to the tight binding between Ocr and the EcoKI Type I R/M enzyme. Single or double mutations of Ocr had no discernable effect on binding to EcoKI or its methyltransferase component (M.EcoKI). Chemical modification was then used to specifically modify the carboxyl moieties of Ocr, thereby neutralizing the negative charges on the protein surface. Ocr samples modified to varying degrees were analysed to establish the extent of derivatisation prior to extensive biophysical characterisation to assess the impact of these changes in terms of binding to the EcoKI R/M system. The results of this analysis revealed that the electrostatic mimicry of Ocr increases the binding affinity for its target enzyme by at least ~800-fold. In addition, based on the known 3-D structure of the protein, a set of multiple mutations were introduced into Ocr aimed at eliminating patches of negative charge from the protein surface. Specifically, between 5 and 17 acidic residues were targeted for mutation (Asp and Glu to Asn and Gln, respectively). Analysis of the in vivo activity of the mutant Ocr along with biophysical characterisation of the purified proteins was then performed. Results from these studies identified regions of the Ocr protein that were critical in forming a tight association with the EcoKI R/M system. Furthermore by comparing the relative contribution of different groups of acidic residues to the free energy of binding, the actual mechanism by which Ocr mimics the charge distribution of DNA has been delineated.
28
Larsson, Anders. "Framtagning av prototyp för att läsa och dokumentera kundspecifikationer." Thesis, Högskolan Dalarna, Informationsteknologi, 2006. http://urn.kb.se/resolve?urn=urn:nbn:se:du-2117.
Full textAbstract:
To increase the quality in their products ABB is working towards a cleared order concept. That means that all customer specified options are to be known before they start with order calculations and construction. As it is today they use paper checklists to achieve this. One order may have several reactors, and for a reactor there can be several different alternatives. For each alternative a new checklist must be filled out.As of today all reading of the customer specification and checklist fill in with is done by hand by different persons, and sometimes the same data is read more than one time. All data is also manually inserted into the calculation tools.To decrease the risk that data is left out or gets distorted they want to have a tool to aid with the reading of the specification and the documentation of that work. Already read data can be copied over to another alternative so that it not must be read one more time. The read data are to be stored in a database so that it easily can be per automation inserted into the different design tools.
29
Albertini, Federica. "Development and evaluation of an OCR system for industrial applications." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2016.
Find full textAbstract:
Questa tesi descrive il lavoro portato a termine durante un tirocinio di sei mesi svolto in Datalogic ADC, presso la sede di Pasadena, California. Durante questa esperienza sono stata coinvolta nella concretizzazione di un vero e proprio prodotto industriale: un software di riconoscimento ottico di caratteri (OCR) basato sugli istogrammi di gradienti orientati (HOG).
L'oggetto del lavoro di tesi comprende l'implementazione di un algoritmo di OCR in grado di leggere una o più righe di testo all'interno di una stessa regione di interesse garantendo velocità e accuratezza richieste in ambito industriale.
E’ possibile suddividere la realizzazione in 3 momenti: l'analisi del prototipo precedentemente realizzato, l'implementazione del nuovo prodotto e la fase finale di test per confrontarlo con i migliori algoritmi allo stato dell'arte.
Un'attenta fase di studio ha portato alla rielaborazione e ottimizzazione del precedente modello, al fine di porre le basi per la concretizzazione del nuovo tool. In seguito, integrando funzioni apposite al soddisfacimento dei requisiti, una prima versione del
prodotto è stata realizzata con particolare attenzione a modularità e correttezza del codice.
Il progetto è stato poi ulteriormente esteso al funzionamento con rappresentazione fixed-point per renderlo integrabile su smart camera: questa modifica, resa necessaria dalla mancanza di floating-point unit sui sistemi embedded, ha richiesto l'implementazione di un' apposita libreria di gestione.
A seguire, per testare robustezza e affidabilità dell'algoritmo, sono stati effettuati test su dataset forniti da reali clienti. Una finale fase di benchmark, confrontando il lavoro svolto con un noto tool commerciale allo stato dell'arte, è stata portata a termine al fine
di prevedere un possibile impatto sul mercato.
I risultati degli esperimenti condotti hanno portato ad interessanti risultati che potranno essere utilizzati per futuri sviluppi e miglioramenti del software OCR.
30
Benjamin, Didier. "Régularisation appliquée au traitement d'images : sélection d'architectures connexionnistes en OCR." Paris 13, 1997. http://www.theses.fr/1997PA132033.
Full textAbstract:
Cette thèse est consacrée à l'étude des techniques de régularisation appliquées au connexionnisme pour le traitement de données assimilables à des images. Pour ce type de traitement, on désire des systèmes adaptatifs aptes à tolérer certains changements de caractéritiques dans les données (taille, positionnement, rotation). Les réseaux connexionnistes sont d'excellents candidats du fait de leur capacité d'adaptation (mécanisme d'apprentissage). Ceci leur permet parmi l'ensemble des solutions à un problème donné d'en choisir une convenant aux données utilisées. Malheureusement cette solution n'est pas toujours celle que l'on aurait souhaitée, il devient donc nécessaire d'orienter cette adaptation. On cherche alors à inclure explicitement une information permettant par exemple la prise en compte des tolérances précitées. Si la régularisation représente une méthode possible d'inclusion explicite, la spécification d'une caractéristique et son exploitation sous forme de terme régularisant reste difficile. C'est sur ces deux derniers points que porte notre travail. La première partie de la thèse décrit les modèles connexionnistes sur lesquels nous nous proposons de travailler. Nous rappelons ensuite les outils de base permettant l'optimisation des poids d'un réseau. Le problème de la généralisation d'un réseau (aptitude à répondre correctement à des données non apprises) appelle plusieurs remarques sur sa quantification. Nous présentons quelques méthodes pour mesurer la performance d'un modèle. Nous rappelons pourquoi et comment mesurer la capacité en généralisation. Ceci nous amène à introduire des outils basés sur la définition de distributions que l'on trouve présentes naturellement dans l'apprentissage Bayésien. L'introduction d'une connaissance à priori sur les paramètres d'un réseau introduite par l'apprentissage Bayésien peut être vue comme une méthode de régularisation. Nous rappelons ce qu'est régulariser, pourquoi des résultats montrant l'équivalence de cette méthode à d'autres bien connues, la rend intéressante. Nous rappelons aussi que les techniques de régularisation posent un problème évident d'estimation des paramètres qu'elles introduisent. Ce problème induit des calculs lourds.
31
Edvartsen, Hannes. "OCR of dot peen markings : with deep learning and image analysis." Thesis, Luleå tekniska universitet, Institutionen för system- och rymdteknik, 2018. http://urn.kb.se/resolve?urn=urn:nbn:se:ltu:diva-71013.
Full textAbstract:
A way to follow products through the chain of production is important in the process industry and it is often solved by marking them with serial numbers. In some cases permanent markings such as dot peen marking is required. To ensure profitability in the industry and reduce errors, these markings must be read automatically. Automatic reading of dot peen markings using a camera can be hard since there is low contrast between the background and the numbers, the background can be uneven and different illuminations can affect the visibility. In this work, two different systems are implemented and evaluated to assess the possibility of developing a robust system. One system uses image analysis to segment the numbers before classifying them. The other system uses the recent advances in deep learning for object detection. Both implementations are shown to work in near real-time on a cpu. The deep learning object detection approach was able to classify all numbers correct in a image 60% of the time, while the other approach only succeeded in 20% of the time.
32
Cracknell, Christopher Robert William. "A software toolkit for handprinted form readers." Thesis, University of Essex, 1999. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.285796.
Full text
33
Vega-Cortes, Liselle. "Evaluation of Analysis Methods used for the Assessment of I-walls Stability." Thesis, Virginia Tech, 2007. http://hdl.handle.net/10919/31047.
Full textAbstract:
On Monday, 29 August 2005, Hurricane Katrina struck the U.S. gulf coast. The storm caused damage to 169 miles of the 284 miles that compose the Hurricane Protection System (HPS) of the area. The system suffered 46 breaches due to water levels overtopping and another four caused by instability due to soil foundation failure. The Interagency Performance Evaluation Task Force (IPET) conducted a study to analyze what happened on the I-wall breach of the various New Orleans flood control structures and looked for solutions to improve the design of these floodwalls.
The purpose of the investigation, describe in this document, is to evaluate different methods to improve the analysis model created by IPET, select the best possible analysis techniques, and apply them to a current cross-section that did not fail during Hurrican Katrina. The use of Finite Element (FE) analysis to obtain the vertical total stress distribution in the vicinity of the I-wall and to calculate pore pressures proved to be an effective enhancement. The influence of overconsolidation on the shear strength distribution of the foundation soils was examined as well.Master of Science
34
Onak, Onder Nazim. "Comparison Of Ocr Algorithms Using Fourier And Wavelet Based Feature Extraction." Master's thesis, METU, 2011. http://etd.lib.metu.edu.tr/upload/12612928/index.pdf.
Full textAbstract:
A lot of research have been carried in the field of optical character recognition. Selection of
a feature extraction scheme is probably the most important factor in achieving high recognition
performance. Fourier and wavelet transforms are among the popular feature extraction
techniques allowing rotation invariant recognition. The performance of a particular feature
extraction technique depends on the used dataset and the classifier. Dierent feature types
may need dierent types of classifiers. In this thesis Fourier and wavelet based features are
compared in terms of classification accuracy. The influence of noise with dierent intensities
is also analyzed. Character recognition system is implemented with Matlab. Isolated gray
scale character image first transformed into one dimensional function. Then, set of features
are extracted. The feature set are fed to a classifier. Two types of classifier were used, Nearest
Neighbor and Linear Discriminant Function. The performance of each feature extraction and
classification methods were tested on various rotated and scaled character images.
35
Atanasici, C. "Characterisation of OCR, the product of gene 0.3 from bacteriophage T7." Thesis, University of Edinburgh, 2001. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.641120.
Full textAbstract:
The OCR protein of bacteriophage T7 is a small, dimeric protein which inhibits the cleavage of the phage DNA by type I restriction enzymes present in the infected host cell. I have studied the structure and the stability of OCR and analysed its binding to the EcoKI type I restriction-modification enzyme. OCR has an unique Tryptophan residue in position 94 which is solvent exposed and important for protein stability but not activity. The protein has a melting temperature of 72.19 °C and a molar extinction coefficient of 32095 M-1 cm-1. Asparagine 4 residue from one monomer is in close proximity to Asparagine 4 in the other monomer. Serine 68 residues are also at the monomer-monomer interface. Six surface exposed amino acids were substituted with cysteine and labelled with cysteine-specific fluorophores. The interaction between labelled OCR(Cys) proteins and EcoKI methylase revealed a huge surface area buried at the interface of the two proteins. OCR binds tightly to the R and S subunits of EcoKI and weakly to the M subunit of EcoKI. One OCR dimer binds to EcoKI methylase and two dimers to EcoKI nuclease. Both OCR-EcoKI methylase and OCR-EcoKI nuclease complexes have a Kd of about 10-11 M.
36
Atanasiu, Constandache. "Characterization of Ocr, the product of gene 0.3 from bacteriophage T7." Thesis, University of Edinburgh, 2000. http://hdl.handle.net/1842/11642.
Full text
37
Kanwar, Nisha. "Mapping charge to function relationships of the DNA mimic protein Ocr." Thesis, University of Edinburgh, 2014. http://hdl.handle.net/1842/9374.
Full textAbstract:
This thesis investigates the functional consequences of neutralising the negative charges on the bacteriophage T7 antirestriction protein ocr. The ocr molecule is a small highly negatively charged, protein homodimer that mimics a short DNA duplex upon binding to the Type I Restriction Modification (RM) system. Thus, ocr facilitates phage infection by binding to and inactivating the host RM system. The aim of this study was to analyse the effect of reducing the negative charge on the ocr molecule by mutating the acidic residues of the protein. The ocr molecule (117 residues) is replete with Asp and Glu residues; each monomer of the homodimer contains 34 acidic residues. Our strategy was to begin with a synthetic gene in which all the acidic residues of ocr had been neutralised. This so called ‘positive ocr’ (or pocr) was used as a template to gradually reintroduce codons for acidic residues by adapting the ISOR strategy proposed by D.S.Tawfik. After each round of mutagenesis an average of 4-6 acidic residues were incorporated into pocr. In this fashion a series of mutant libraries in which acidic residues were progressively introduced into pocr was generated. A high-throughput in vivo selection assay was developed and validated by assessing the antirestriction behaviour of a number of mutants of the DNA mimic proteins wtOcr and Orf18 ArdA. Further to this, selective screening of the libraries allowed us to select clones that displayed antirestriction activity. These mutants were purified and in vitro characterisation confirmed these mutants as displaying the minimum number of acidic residues deemed critical for the activity of ocr. This in vitro process effectively simulated the evolution of the charge mimicry of ocr. Moreover, we were able to tune the high-throughput assay to different selection criteria in order to elucidate various levels of functionality and unexpected changes in phenotype. This approach enables us to map the “in vitro” evolution of ocr to identify acidic residues that are required for protein expression, solubility and function proceeding to a fully functional antirestriction protein.
38
Erlandsson, Zacharias. "Suitability of OCR Engines in Information Extraction Systems : a Comparative Evaluation." Thesis, KTH, Skolan för elektroteknik och datavetenskap (EECS), 2019. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-255021.
Full textAbstract:
Previous research has compared the performance of OCR (optical character recognition) engines strictly for character recognition purposes. However, comparisons of OCR engines and their suitability as an intermediate tool for information extraction systems has not previously been examined thoroughly. This thesis compares the two popular OCR engines Tesseract OCR and Google Cloud Vision for use in an information extraction system for automatic extraction of data from a financial PDF document. It also highlights findings regarding the most important features of an OCR engine for use in an information extraction system, when it comes to structure of output as well as accuracy of recognitions. The results show a statistically signifant increase in accuracy for the Tesseract implementation compared to the Google Cloud Vision one, despite previous research showing that Google Cloud Vision outperforms Tesseract in terms of accuracy. This was accredited to Tesseract producing more predictable output in terms of structure, as well as the nature of the document which allowed for smaller OCR processing mistakes to be corrected during the extraction stage. The extraction system makes use of the aforementioned OCR correctional procedures as well as an ad-hoc type system based on the nature of the document and its fields in order to further increase the accuracy of the holistic system. Results for each of the extraction modes for each OCR engine are presented in terms of average accuracy across the test suite consisting of 115 documents.Tidigare forskning har gjorts som jämför prestandan av OCR-motorer (optical character recognition) uteslutande för dess teckenläsande egenskaper. Jämförelser för OCR-motorer som verktyg för system för informationsextraktion har däremot inte gjorts tidigare. Det här examensarbetet jämför de två populära OCR-motorerna Tesseract OCR och Google Cloud Vision för användning i ett system som används för automatisk extraktion av data från ett finansiellt PDFdokument. Arbetet belyser även observationer angående vilka de viktigaste egenskaperna hos en OCR-motor är för användning i ett system för informationsextraktion. Resultaten visade en statistisk signifikant ökning i exakthet för implementationen med Tesseract jämfört med Google Cloud Vision, trots tidigare forskning som visar att Google Cloud Vision kan utföra teckenläsning mer exakt. Detta ackrediteras till det faktum att Tesseract producerar mer konsekvent utdata när det kommer till struktur, och att vissa felaktiga teckeninläsningar kan korrigeras av extraktionssystemet. Extraktionssystemet använder sig av ovan nämnd OCR-rättande metodik samt ett ad-hoc typsystem baserat på dokumentets innehåll för att öka exaktheten för det holistiska systemet. Dessa metoder kan även isoleras till enskilda extraktionslägen. Resultat för varje extraktionsläge presenteras genom genomsnittlig exakthet över testsviten som bestod av 115 dokument.
39
Nguyen, Thi Tuyet Hai. "Facilitating access to historical documents by improving digitisation results." Thesis, La Rochelle, 2020. http://www.theses.fr/2020LAROS004.
Full textAbstract:
Les documents papiers sont à la base de nos connaissances et renferment une myriade d'informations dont certaines sont très précieuses pour notre société. Dans un but de préservation et afin de les rendre plus accessibles, de nombreux projets de numérisation visent à convertir ce type de documents en textes numérisés, notamment en utilisant des logiciels de reconnaissance optique de caractères (OCR). Toutefois, certains problèmes inhérents aux techniques actuelles d’OCR rendent difficiles la recherche ou l’accès aux informations présentes dans ces collections numérisées, tant pour les utilisateurs que pour les processus automatiques, et limitent ainsi l’impact de ces efforts de numérisation. L’une des limitations de la numérisation repose sur le processus même puisque les documents numérisés ne sont pas immédiatement représentés sous leur forme logique (partie, chapitre, section, etc.), mais de façon physique. Ainsi, une œuvre sera numérisée page par page, ce qui ne correspond généralement qu’à une organisation physique et pas à l'intention rédactionnelle des auteurs. La structure logique des documents doit ainsi être extraite afin de permettre aux utilisateurs de naviguer dans les collections ou même de trouver des informations au sein d’un ouvrage. Un second verrou du processus de numérisation, qui en est également le plus important, correspond aux performances des moteurs d’OCR. En effet, celles-ci sont substantiellement réduites pour les documents patrimoniaux qui ont généralement subis des dégradations. Les erreurs d’OCR que cela induit ont un impact non négligeable sur la performance des outils de recherches et sur les systèmes de traitement du langage naturel puisqu’il faut par exemple apparier des besoins bien écrits à des textes mal reconnus. Cette thèse a pour objectif de faciliter l’accès aux documents historiques numérisés en étudiant les problèmes précédemment mentionnés. En vue de faciliter l’accès aux documents historiques, plusieurs approches sont proposées, visant à reconstruire les structures logiques des ouvrages et à améliorer la qualité des textes numérisés par OCR. En ce qui concerne l’extraction de la structure logique, nous avons développé des approches de fusion combinant des méthodes préexistantes afin d’extraire la table des matières d’ouvrages numérisés. Nos expériences ont démontré que cette approche surpasse l’état de l’art. La contribution majeure de cette thèse fournit, quant à elle, des méthodes pour la détection et la correction des erreurs d’OCR. Les caractéristiques communes et divergentes entre les erreurs d’OCR et celles des utilisateurs sont clarifiées pour mieux concevoir les traitements post-OCR. Normalement, un système de post-traitement détecte et rectifie les erreurs résiduelles. Toutefois, il peut être préférable de gérer ces erreurs séparément grâce à des applications qui permettent de filtrer, d’étiqueter, ou de traiter sélectivement de telles données. Dans cette étude, nous examinons différentes approches post-OCR basées sur la modélisation des erreurs typiques observées, et sur des modèles de réseaux de neurones. Les résultats montrent que les performances de nos méthodes sont comparables à plusieurs méthodes de référence sur des jeux de données en anglais utilisés lors des deux premières éditions de la compétition sur la correction des textes post-OCR organisée durant les conférence ICDAR en 2017 et 2019Born-analog documents contain enormous knowledge which is valuable to our society. For the purpose of preservation and easy accessibility, several digitisation projects have converted these documents into digital texts by using optical character recognition (OCR) software. Some existing problems of current OCR techniques prevent users and further processes from accessing, searching, or retrieving information on these digitised collections, and so limit the benefits of these above projects. A notable limitation is the fact that OCRed books are often split into pages with paragraphs, lines, and words. Certain meaningful structures such as chapters, sections, etc., are not available. Thus, it is not convenient for users to navigate or search information inside books. Another constraint is that the accuracy of modern OCR engines on historical documents substantially decreases. Erroneous OCR output considerably impacts on the performance of search engines and natural language processing systems. This thesis facilitates access to historical digitised documents by addressing such problems. In order to facilitate access to historical documents, several approaches are proposed within this thesis, aiming to reconstruct the logical book structures and to improve the quality of digitised text. The first contribution is to rebuild the logical book structures. An ensemble method is introduced to extract tables of contents of digitised books. Experimental results show that our approach outperforms the state-of-the-art for both evaluation metrics. The major contribution of this thesis is to provide methodologies to reduce OCR errors. Common and different features between OCR errors and human misspellings are clarified for better designing post-OCR processing. Normally, a post-processing system detects and corrects remaining errors. However, it is reasonable to treat them separately in some applications which allow to filter out, flag, or selectively reprocess such data. In this thesis, we examine different post-OCR approaches, ones based on error model and language model, and others that involve neural network models. Results reveal that the performance of our proposals is comparable to several strong baselines on English datasets of the first two rounds of the competition on post-OCR text correction organised in the International Conference on Document Analysis and Recognition in 2017 and 2019
40
Albertazzi, Riccardo. "A study on the application of generative adversarial networks to industrial OCR." Master's thesis, Alma Mater Studiorum - Università di Bologna, 2018.
Find full textAbstract:
High performance and nearly perfect accuracy are the standards required by OCR algorithms for industrial applications. In the last years research on Deep Learning has proven that Convolutional Neural Networks (CNNs) are a very powerful and robust tool for image analysis and classification; when applied to OCR tasks, CNNs are able to perform much better than previously adopted techniques and reach easily 99% accuracy.
However, Deep Learning models' effectiveness relies on the quality of the data used to train them; this can become a problem since OCR tools can run for months without interruption, and during this period unpredictable variations (printer errors, background modifications, light conditions) could affect the accuracy of the trained system. We cannot expect that the final user who trains the tool will take thousands of training pictures under different conditions until all imaginable variations have been captured; we then have to be able to generate these variations programmatically.
Generative Adversarial Networks (GANs) are a recent breakthrough in machine learning; these networks are able to learn the distribution of the input data and therefore generate realistic samples belonging to that distribution. This thesis' objective is learning how GANs work in detail and perform experiments on generative models that allow to create unseen variations of OCR training characters, thus allowing the whole OCR system to be more robust to future character variations.
41
Chen, Qing. "Evaluation of OCR algorithms for images with different spatial resolutions and noises." Thesis, University of Ottawa (Canada), 2004. http://hdl.handle.net/10393/26601.
Full textAbstract:
Various shaped-based image invariants are popular algorithms used in optical character recognitions (OCR), 3D object recognitions, and pattern recognitions. The shape-based image invariants can be divided into two different categories: boundary based image invariants such as Fourier descriptors and chain code; region-based image invariants including various moment-based invariants such as Hu's seven moment invariants and Zernike moments.
This thesis introduced and evaluated different shape-based image invariants from the perspective of their invariance property to image transformations including scaling, translation, rotation and different image spatial resolutions. The influence caused by salt and pepper noises of different intensities is also analyzed. The image reconstruction ability of Zernike moments is implemented and discussed as well. An OCR engine is implemented with MATLAB scripts to perform the image feature extraction and image recognition. The OCR engine is also used in the overall performance evaluation of Fourier descriptors and Hu's seven moment invariants.
42
RUBIO, VILLALBA IGNACIO. "Analysis of the OCR System Application in Intermodal Terminals : Malmö Intermodal Terminal." Thesis, KTH, Transportplanering, 2020. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-278856.
Full textAbstract:
The analysis carried out in this thesis is made from two different points of view, the qualitative and the quantitative, by using the case study of Malmö intermodal terminal. The first analysis is focused on how the intermodal terminals works and which elements of it interact and how, in order to achieve the purpose of the terminal, and how the Intelligent Video Gate is able to affect in any way to this functioning, mainly in a positive way that allows the better functioning of the terminal.From the quantitative point of view what is carried out is a timing and economic analysis of the Malmö Intermodal Terminal, which is based on the information obtained from the qualitative analysis and from the data provided by the terminal operators that allow to make different simulations to compare the effect of the Intelligent Video Gate implementation in this specific terminal, and that could be extended to similar intermodal terminals located in regions with similar labour conditions and that as the European Union have a huge standardized freight system.Finally, what is stated with the provided data, despite not allowing to make the most complex and representative simulation, is that the aim of the Intelligent Video Gate is reached successfully with a great improvement of the efficiency what allows to ensure with quite certainty that the system implementation is recommended in this kind of terminals.
43
Lundqvist, Melvin, and Agnes Forsberg. "A comparison of OCR methods on natural images in different image domains." Thesis, KTH, Skolan för elektroteknik och datavetenskap (EECS), 2020. http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-280286.
Full textAbstract:
Optical character recognition (OCR) is a blanket term for methods that convert printed or handwritten text into machine-encoded text. As the digital world keeps growing the amount of digital images with text increases, and the need for OCR methods that can handle more than plain text documents as well. There are OCR engines that can convert images of clean documents with an over 99% recognition rate. OCR for natural images is getting more and more attention, but because natural images can be far more diverse than plain text documents it also leads to complications. To combat these issues it needs to be clear in what areas the OCR methods of today struggle. This thesis aims to answer this by testing three popular, readily available, OCR methods on a dataset comprised only of natural images containing text. The results show that one of the methods, GOCR, can not handle natural images as its test results were very far from correct. For the other two methods, ABBYY FineReader and Tesseract, the results were better but also show that there still is a long way to go, especially when it comes to images with special font. However when the images are less complicated some of our methods performed above our expectations.Optical character recognition (OCR) är en samlingsterm för metoder som konverterar tryckt eller handskriven text till maskinkod. När den digitala världen växer så växer även antalet digitala bilder med text, och även behovet för OCR metoder som kan hantera mer än vanliga textdokument. Det finns idag OCR motorer som kan konvertera bilder av rena dokument till maskinkod med över 99% korrekthet. OCR för fotografier får mer och mer uppmärksamhet, men eftersom fotografier har mycket större mångfaldhet än rena textdokument leder detta också till problem. För att hantera detta krävs klarhet inom vilka områden som dagens OCR-metoder har problem. Denna uppsats ämnar svara på denna fråga genom att undersöka och testa tre populära, enkelt tillgängliga OCR metoder på ett dataset som endast innehåller fotografier av naturliga miljöer med text. Resultaten visade att en av metoderna, GOCR, inte kan hantera fotografier. GOCRs testresultat var långt från det korrekta. För de andra metoderna, ABBYY FineReader och Tesseract, var resultaten bättre men visade att det fortfarande finns mycket arbete att göra inom området, särskilt när det kommer till bilder med speciella typsnitt. När det däremot kommer till bilder som är mindre komplicerade blev vi förvånade över hur bra resultatet var för några av metoderna.
44
Mishra, Vishal Vijayshankar. "Sequence-to-Sequence Learning using Deep Learning for Optical Character Recognition (OCR)." University of Toledo / OhioLINK, 2017. http://rave.ohiolink.edu/etdc/view?acc_num=toledo1513273051760905.
Full text
45
Feijó, José Victor Feijó de Araujo. "Análise e Classificação de imagens para aplicação de OCR em cupons fiscais." Florianópolis, SC, 2017. https://repositorio.ufsc.br/xmlui/handle/123456789/182212.
Full textAbstract:
TCC(graduação) - Universidade Federal de Santa Catarina. Centro Tecnológico. Ciências da Computação.Submitted by José Victor Feijo de Araujo null (victor.feijo@ufsc.br) on 2017-12-12T00:28:08Z No. of bitstreams: 1 TCC_JOSE_VICTOR_FEIJÓ.pdf: 18256303 bytes, checksum: 6f566a4daec3603fa7cc31bf1d8da5c8 (MD5)
Approved for entry into archive by Renato Cislaghi (renato.cislaghi@ufsc.br) on 2017-12-13T21:10:36Z (GMT) No. of bitstreams: 1 TCC_JOSE_VICTOR_FEIJÓ.pdf: 18256303 bytes, checksum: 6f566a4daec3603fa7cc31bf1d8da5c8 (MD5)
Made available in DSpace on 2017-12-13T21:10:36Z (GMT). No. of bitstreams: 1 TCC_JOSE_VICTOR_FEIJÓ.pdf: 18256303 bytes, checksum: 6f566a4daec3603fa7cc31bf1d8da5c8 (MD5)
A proposta sugerida por este trabalho foi de analisar o impacto de um modelo de classificação, seguido de técnicas de PDI e OCR para extração de texto em cupons fiscais, classificando-os em subgrupos. Técnicas selecionadas de PDI foram aplicadas para cada grupo com suas devidas características, por fim extraindo texto dessas imagens através de um algoritmo de OCR. Foi realizado um estudo sobre os algoritmos clássicos de classificação na área de aprendizado de máquinas, com foco nos algoritmos de “clusterização” e sua correlação com a classificação de imagens em um modelo de aprendizado não supervisionado. Também foi feita uma análise sobre as características das imagens de cupons fiscais e das possíveis técnicas de PDI que podem ser aplicadas. Em relação ao OCR, também foi realizado um estudo para verificar possíveis soluções na extração de texto e entender seu comportamento, possibilitando desta maneira implementar a arquitetura proposta. Sendo assim, foram desenvolvidos métodos para classificar as imagens em clusters utilizando algoritmos de “clusterização”. Também foram propostas três técnicas de PDI, a primeira aplicando uma série de realces, a segunda uma binarização adaptativa e a terceira técnica utilizando a compressão de dados JPEG. Essas imagens foram enviadas para o serviço de OCR do Google Vision, onde foi possível extrair o texto das imagens em formato de blocos. Os resultados do modelo desenvolvido foram avaliados comparando a taxa de acerto do OCR com os valores de texto reais presentes nos cupons fiscais, onde foi possível analisar a precisão de cada técnica proposta e da arquitetura como um todo. Foram obtidos resultados positivos utilizando o modelo desenvolvido, melhorando a extração do valor total da compra em aproximadamente 6%. Além disso, os resultados da compressão JPEG melhoraram também a extração de outros dados do cupom fiscal, como por exemplo o CNPJ e a data da compra.
46
Maurer, Yves. "Improving the quality of the text, a pilot project to assess and correct the OCR in a multilingual environment." Sächsische Landesbibliothek - Staats- und Universitätsbibliothek Dresden, 2017. https://slub.qucosa.de/id/qucosa%3A16445.
Full textAbstract:
The user expectation from a digitized collection is that a full text search can be performed and that it will retrieve all the relevant results. The reality is, however, that the errors introduced during Optical Character Recognition (OCR) degrade the results significantly and users do not get what they expect. The National Library of Luxembourg started its digitization program in 2000 and in 2005 started performing OCR on the scanned images. The OCR was always performed by the scanning suppliers, so over the years quite a lot of different OCR programs in different versions have been used. The manual parts of the digitization chain (handling, scanning, zoning, …) are difficult, costly and mostly incompressible, so the library thought that the supplier should focus on a high quality level for these parts. OCR is an automated process and so the library believed that the text recognized by the OCR could be improved automatically since OCR software improves over the years. This is why the library has never asked the supplier for a minimum recognition rate.
The author is proposing to test this assumption by first evaluating the base quality of the text extracted by the original supplier, followed by running a contemporary OCR program and finally comparing its quality to the first extraction. The corpus used is the collection of digitized newspapers from Luxembourg, published from the 18th century to the 20th century. A complicating element is that the corpus consists of three main languages, German, French and Luxembourgish, which are often present on a single newspaper page together. A preliminary step is hence added to detect the language used in a block of text so that the correct dictionaries and OCR engines can be used.
47
Skoglund, Jesper, and Lukas Vikström. "Automating the process of dividing a map image into sections : Using Tesseract OCR and pixel traversing." Thesis, Linköpings universitet, Institutionen för datavetenskap, 2018. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-148319.
Full textAbstract:
This paper presents an algorithm with the purpose of automatically dividing a simple floor plan into sections. Sections include names, size and location on the image, all of which will be automatically extracted by the algorithm as a step of converting a simple image into an interactive map. The labels for each section utilizes tesseract-OCR wrapper tesseractJS to extract text and label location. In regards to section borders pixel traversing is employed coupled with CIE76 for color comparison which results in the discovery of size and location of the section. Performance of the algorithm was measured on three different maps using metrics such as correctness, quality, completeness, jaccard index and name accuracy. The metrics showed the potential of such an algorithm in terms of automating the task of sectioning an image. With results ranging between lowest percentage of 48% and highest of 100% on three different maps looking at correctness, quality, completeness, average jaccard index and average name accuracy per map.
48
Raza, Ghulam. "Algorithms for the recognition of poor quality documents." Thesis, Nottingham Trent University, 1998. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.241828.
Full text
49
Rodrigues, Antonio Jose Nunes Navarro. "A robust off-line hand written character recognition system using dynamic features." Thesis, University of Newcastle Upon Tyne, 1996. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.295503.
Full text
50
Senior, Andrew William. "Off-line cursive handwriting recognition using recurrent neural networks." Thesis, University of Cambridge, 1994. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.338024.
Full text