Dissertations / Theses on the topic 'Rule-Based Moving Object Tracking'

Detecting and tracking moving objects are important topics in computer vision research. Classical methods perform well in applications of steady cameras. However, these techniques are not suitable for the applications of moving cameras because the unconstrained nature of realistic environments and sudden camera movement makes cues to object positions rather fickle. A major difficulty is that every pixel moves and new background keeps showing up when a handheld or car-mounted camera moves. In this dissertation, a novel estimation method of camera motion parameters will be discussed first. Based on the estimated camera motion parameters, two detection algorithms are developed using Bayes' rule and belief propagation. Next, an MCMC-based feature-guided particle filtering method is presented to track detected moving objects. In addition, two detection algorithms without using camera motion parameters will be further discussed. These two approaches require no pre-defined class or model to be trained in advance. The experiment results will demonstrate robust detecting and tracking performance in object sizes and positions.

Video surveillance has long been in use to monitor security sensitive areas such as banks, department stores, crowded public places and borders. The rise in computer speed, availability of cheap large-capacity storage devices and high speed network infrastructure enabled the way for cheaper, multi sensor video surveillance systems. In this thesis, the problem of tracking multiple targets with multiple cameras has been discussed. Cameras have been located so that they have overlapping fields of vision. A dynamic background-modeling algorithm is described for segmenting moving objects from the background, which is capable of adapting to dynamic scene changes and periodic motion, such as illumination change and swaying of trees. After segmentation of foreground scene, the objects to be tracked have been acquired by morphological operations and connected component analysis. For the purpose of tracking the moving objects, an active contour model (snakes) is one of the approaches, in addition to a Kalman tracker. As the main tracking algorithm, a rule based tracker has been developed first for a single camera, and then extended to multiple cameras. Results of used and proposed methods are given in detail.

As the number of surveillance cameras and mobile platforms with cameras increases, automated detection and tracking of objects on these systems gain importance. There are various tracking methods designed for stationary or moving cameras. For stationary cameras, correspondence based tracking methods along with background subtraction have various advantages such as enabling detection of object entry and exit in a scene. They also provide robust tracking when the camera is static. However, they fail when the camera is moving. Conversely, histogram based methods such as mean shift enables object tracking on moving camera cases. Though, with mean shift object&rsquo
s entry and exit cannot be detected automatically which means a new object&rsquo
s manual initialization is required. In this thesis, we propose a dual-mode object tracking system which combines the benefits of correspondence based tracking and mean shift tracking. For each frame, a reliability measure based on background update rate is calculated. Interquartile Range is used for finding outliers on this measure and camera movement is detected. If the camera is stationary, correspondence based tracking is used and when camera is moving, the system switches to the mean shift tracking mode until the reliability of correspondence based tracking is sufficient according to the reliability measure. The results demonstrate that, in stationary camera mode, new objects can be detected automatically by correspondence based tracking along with background subtraction. When the camera starts to move, generation of false objects by correspondence based tracking is prevented by switching to mean shift tracking mode and handing over the correct bounding boxes with a seamless operation which enables continuous tracking.

碩士
國立中山大學
機械與機電工程學系研究所
96
As a result of everlasting developments in multimedia technologies, all kinds of objects tracking theory using machine vision or image process methods have been proposed. Most of the methods are based on shape of the object. For this reason, the profile of the tracked object must be known in advance. In many situations, we expect to track the object whose shape is unknown but speed or direction is explicit. For instance, speed or moving direction of the object is known. This thesis presents a spatio-temporal tracking technique, which extracts image information depending on speed of the moving object regardless of its shape. Furthermore, combination of the proposed method in spatio-temporal domain and the optical flow scheme makes the whole tracking system even more robust.

碩士
國立交通大學
資訊科學與工程研究所
95
Predication-based techniques are able to reduce the energy consumption in object tracking sensor networks. Prior works exploit mining object moving patterns for prediction-based object tracking sensor network and developed a hierarchical architecture to efficiently track objects. Note that sensors are inherently storage-constrained. Clearly, mining and storing individual object moving patterns unavoidably need a considerable amount of storage spaces in sensor nodes, which is not of practical. Thus, in this paper, we propose a group-based object tracking sensor network (abbreviated as GBOT) which explores the feature of group mobility of objects for storage-constrained object tracking sensor networks. Specifically, we first formulate a dissimilarity function among object moving patterns, where object moving patterns are viewed as emission trees. In light of the dissimilarity function, the dissimilarity relationships among objects are derived. Given dissimilarity relationships among objects, we further propose two clustering schemes to discover group mobility patterns of objects. Furthermore, for each group, we judiciously select one representative emission tree and utilize this emission tree for prediction. In addition, a maintenance algorithm is derived to preserve the prediction accuracy when moving behaviors of objects vary. Experimental results show that GBOT not only effectively reduces storage cost but also has a good prediction accuracy in storage-constrained sensor networks.

碩士
雲林科技大學
光學電子工程研究所
96
Intelligent surveillance system is a very popular research subject. In the majority of research, the researchers merely segment simple static background. However in the scenario, there are some changes in background both indoor and outdoor, such as change of illumination, variation of shadow, waving tree, undulate of water and flutter of flag, etc. In this thesis, I propose a method of building background model which can complete the adaptive update determined by the environmental changes. This system is able to fast build static background and update real time. Besides, it could build dynamic background with low space memory. Therefore, it could detect moving object correctly in chaotic background. We take Histogram based matching technique to promote tracking performance and accuracy in the object tracking part. Finally, we use TI Davinci development-platform as the proof. It is able to reach the rate of 7 ~ 8 frames per second by using simple ARM.

碩士
國立彰化師範大學
電子工程學系
97
This thesis proposed an object detection and tracking framework based on an entropy-based background modeling technique. Our research group previously proposed a video object detection method, called ASLEI (Adaptive State-Labeling Entropy Image), which has the advantages of high detection quality, robustness, and low computational complexity. However, as other object detection methods based on temporal differencing, ASLET fails to detect a still object, which keeps still in the video for several consecutive frames. This thesis proposed a new background modeling technique derived from ASLEI. The proposed entropy-based background modeling can achieve a better background registration rate than other background modeling techniques using pixel-based temporal differencing. Also, the still objects in the video can be easily detected by employing the proposed background model. We use the ASLEI algorithm to find the moving object mask (MOM) and use the background model to find the static object mask (SOM). Both MOM and SOM are combined to extract the object blobs from the video sequences. The extracted object blobs are classified as either moving blobs or static blobs according the percentage of blob pixels which belongs to MOM. Then, those static blobs are further classified as static foreground blobs or static background blobs by measuring the foreground similarity for every static blob. Furthermore, we perform a feature analysis to classify the static background blobs as either removed objects or ghosts, and classify the static foreground blobs as either abandoned objects or halted humans. In order to achieve reliable object tracking, we proposed the intersection matching of minimum displacement and the maximum overlapping test, which can help to construct a more reliable match table for object tracking. The experimental results have shown that the proposed framework can perform object detection, recognition, and tracking correctly even in complex scenes. Thus, the proposed object detection and tracking framework is very suitable for building intelligent video surveillance systems.

碩士
國立臺灣科技大學
電機工程系
98
In the last decade, object tracking systems have been widely applied in many different fields due to the rapid development of computer vision techniques and faster computing ability, such as Surveillance System, Health-Care System. In this field, many approaches require establishing background in preprocessing step. This limits tracking algorithm only be executed under a fixed camera. However, many applications are taking place in a moving camera. Accordingly, we propose a new algorithm to track rigid or non-rigid object by a moving camera. The proposed tracking algorithm use rotation-invariant texture feature and color feature to increase the tracking correctness. The target is jointly modeled by color and texture information. We adjust the weight of each feature, so it is less sensitive to different circumstances such as partial occlusions. When fully occluded, we extend search region and double the particle number to avoid missing target if the occlusion disappear. The experimental results reveal that our tracking method can efficiently and successfully track rigid or non-rigid object under appearance and illumination changes. Also, fewer samples are used to achieve better result than the traditional particle filter method.

碩士
國立臺北科技大學
電子工程系研究所
105
In recent years, automation becomes more and more important in the world. Autonomous car is an example in our life around. It uses the 3D LiDAR to receive the surrounding information. 3D LiDAR (3D Light Detection and Ranging) is the scanner that can catch the point cloud map real time to record depth information around. We purpose a moving object tracking system with 3D LiDAR and color camera. The sensor of system is VLP-16 3D LiDAR that shoots 16 channels infrared light and calculates the distance between scanner and obstacle using TOF (Time of Flight) technology. Because it is not enough to perform object tracking using point cloud map, this work adopts color image that is taken by RGB camera (Intel Creative Senz3D Camera) and combines point cloud map to achieve moving object tracking. Moreover, we designed the system hardware architecture and implemented a chip via the Cell-based IC Design Flow.

碩士
淡江大學
機械與機電工程學系碩士班
99
In this thesis, the visual simultaneous localization, mapping and moving object tracking (SLAMMOT) is developed using the extended Kalman filter (EKF). The research is divided into two parts: first, one monocular vision is utilized as the only sensor for the SLAMMOT system. The camera calibration is replaced by an image correction model to simplify the linearization derivation of the measurement equation in state estimation. Second, the algorithm of moving object detection (MOD) is developed based on the constraint condition of static image features on the epipolar line. We also develop the procedures of data association and map management for the SLAM task with moving objects. Finally, the EKF SLAMMOT with the proposed algorithm is implemented on a monocular vision system. The integrated system has successfully tested the basic capabilities of SLAM, including loop-closure, ground truth, long-distance navigation, and moving object detection and tracking for the system in the indoor environment.

碩士
國立臺灣大學
電機工程學研究所
102
The objective of this thesis is to develop simultaneous localization and mapping (SLAM) with capability of tracking moving object in indoor environments. SLAM can help build environment map, while detection and tracking of moving object separate the environment into static and dynamic parts. The map can help detect the moving object, on the other hand, the moving object tracking can help separate the stationary and moving objects, thus we can separate them in the map. By augmenting the moving objects state and related constraints into the robot and objects graph, the general graph-based framework for SLAM issues can be extended to jointly optimize the SLAM and moving object tracking result. By incorporating the moving object prediction and moving object Retro-BestGuess, the later measurement of moving object can help the estimation of the previous state and vice versa. Consequently, the trajectory of robot together with the trajectories of moving objects is optimized. Furthermore, the SLAM with moving object tracking issues in the cluttered indoor environment are analyzed, the moving object may have different size and characteristics difficult to modelling, and the data association is difficult. The multi-frame moving object detection is applied to detect the moving object without the need of prior knowledge, by which even the slightly movement can be detected. The multi-sensor fusion methodologies can help increase the data association accuracy. The experimental results shown that our algorithm is feasible in cluttered indoor environment, graph-based SLAM incorporating moving objects can decrease the pose estimation uncertainty compare to the one not incorporating them.

碩士
逢甲大學
航太與系統工程學系
107
The automated factories are very common nowadays, machine vision has become an indispensable technology for improve the practicality and adaptability of robots. Other than the more expensive off-the-shelf 3D camera, this paper use LabVIEW to develop the vision-based robotic system, which is composed of the 2D camera, robotic arm and conveyor, for tracking, picking and placing a moving object tasks. Functions such as: the camera adjustment and setting, the camera calibration and compensation for the object coordinate offset, the eye to hand calibration by using the coordinate transformation with pixel coordinate and robot coordinate that can make robotic arm moving to the accurate object position, the conveyor control and calibration with vision processing, the pattern matching with the calculation of the final picking coordinate of the object, and controlling robotic arm to pick and place the object. …etc. will be included in this system. The thesis will illustrate all functions and verifications of the system tests. The practicality and adaptability of this system will be confirmed after the revision and system integration based on the experiment results are achieved. Besides, the matured system will also compensate the robotic arm and conveyor velocity by vision-based algorithm and benefit the system with the advantages of automation and adaptation.

碩士
國立高雄應用科技大學
電機工程系博碩士班
102
Object detection and tracking plays an important role in video surveillance systems. Because security service personnel may not be able to continuously focus on monitors, it is necessary to apply intelligent video surveillance systems to assist security service personnel in monitoring. However, current traditional surveillance systems mainly use for fixed cameras. Because traditional cameras are fixed, they certainly result blind spot in monitoring. Therefore, this thesis proposes an intelligent surveillance system, which can detect and track the suspicious moving object by using active cameras in the environmental monitoring. In moving object detection, this research utilizes similarity measures between the camera's current image and the background image, and applies the threshold that is adjusted automatically with the surveillance space to execute thresholding with the image of similarity measure. Then, we use morphology operation to filter the noise of binary image, which is generated from thresholding. In moving object tracking, we calculate the statistical features of the binary image in predefined image regions to determine which region the moving object locates in, and we control the stepper motor to track the moving object. Finally, we execute the edge detection with the binary image to find the mean location of the moving object, and display a window at this mean location. In this thesis, we utilize Verilog HDL (Verilog Hardware Description Language) to implement this system in the FPGA (Field Programmable Gate Array) development boards, and we apply some simple and efficient methods to achieve the purpose, which can track the moving object rapidly with 60 FPS (Frames Per Second) and reduce FPGA resource consumption. In addition, we combine this system with infrared control, and we can use infrared remote controller to rotate active cameras to monitor the position which we want to observe. Finally, we utilize the cooperation between two tracking systems to expand the range of monitoring.

碩士
淡江大學
機械與機電工程學系碩士班
99
This thesis presents a visual simultaneous localization, mapping and moving object tracking (SLAMMOT) based on extended Kalman filter (EKF). First, we use the geometric constraints of static landmarks in three-dimensional space to design the algorithms of data association and map management. Since these algorithms are independent of the EKF estimator, the SLAMMOT system can recover from the problem of robot kidnapped automatically. Second, we use the same geometric constraints to develop the algorithm for moving object detection. The developed algorithms are integrated with the EKF estimator to carry out the experiments of SLAMMOT tasks in indoor environments.

碩士
國立臺灣大學
資訊工程學研究所
100
Scene understanding is one of the most important foundations for a mobile robot to operate in human-habited environments. As the real environments are typically dynamic, moving object tracking becomes an inescapable problem. While the tracking algorithm becomes more and more elaborate, however, its performance in real world still can not be guaranteed. The major reason is that so far we do not have enough real data with ground-truth to evaluate and analysis the state of the art tracking algorithms . In this thesis, we explore the laser-based moving object tracking problem and propose an annotation system that allows the user to annotate the ground-truth of segmentation and data association with 2D laser measurements. As the annotating task is difficult and tedious, the system is designed to achieve higher accuracy and reduce the task loading in the annotation process. To prove the usefulness of our system, real data sequences are collected and annotated by multiple users in our experiments. The results shows that the annotation performance varies but the system keeps helpful across different users. In particular, the V-measure reaches to 0.995 bits and the false positive rate and the false negative rate are reduced to 0.341% and 1.239%. At last, the ground-truth data is also generated by validating the annotated data carefully and repeatedly.

Thesis submitted to the Department of Electrical and Computer Engineering of the Faculty of Science and Technology of the University of Coimbra in partial fulfillment of the requirements for the Degree of Doctor of Philosophy
Nesta tese é proposto um novo sistema multissensorial de detecção de obstáculos e objetos usando um LIDAR-3D, uma câmara monocular a cores e um sistema de posicionamento baseado em sensores inerciais e GPS, com aplicação a sistemas de condução autónoma. Em primeiro lugar, propõe-se a criação de um sistema de deteção de obstáculos, que incorpora dados 4D (3D espacial + tempo) e é composto por dois módulos principais: (i) uma estimativa do perfil do chão através de uma aproximação planar por partes e (ii) um modelo baseado numa grelha de voxels para a deteção de obstáculos estáticos e dinâmicos recorrendo à informação do próprio movimento do veículo. As funcionalidade do systemo foram posteriormente aumentado para permitir a Deteção e Seguimento de Objetos Móveis (DATMO) permitindo a percepção ao nível do objeto em cenas dinâmicas. De seguida procede-se à fusão dos dados obtidos pelo LIDAR-3D com os dados obtidos por uma câmara para melhorar o desempenho da função de seguimento do sistema DATMO. Em segundo lugar, é proposto um sistema de deteção de objetos baseado nos paradigmas de geração e verificação de hipóteses, usando dados obtidos pelo LIDAR-3D, recorrendo à utilização de redes neurais convolucionais (ConvNets). A geração de hipóteses é realizada aplicando um agrupamento de dados ao nível da nuvem de pontos. Na fase de verificação de hipóteses, é gerado um mapa de profundidade a partir dos dados do LIDAR-3D, sendo que esse mapa é inserido numa ConvNet para a deteção de objetos. Finalmente, é proposta uma detecção multimodal de objetos usando uma rede neuronal híbrida, composta por Deep ConvNets e uma rede neural do tipo Multi-Layer Perceptron (MLP). As modalidades sensoriais consideradas são: mapas de profundidade, mapas de reflectância geradas a partir do LIDAR-3D e imagens a cores. São definidos três detetores de objetos que individualmente, em cada modalidade, recorrendo a uma ConvNet detetam as bounding boxes do objeto. As deteções em cada uma das modalidades são depois consideradas em conjunto e fundidas por uma estratégia de fusão baseada em MLP. O propósito desta fusão é reduzir a taxa de erro na deteção de cada modalidade, o que leva a uma deteção mais precisa. Foram realizadas avaliações quantitativas e qualitativas dos métodos propostos, utilizando conjuntos de dados obtidos a partir dos datasets "Avaliação de Detecção de Objetos" e "Avaliação de Rastreamento de Objetos" do KITTI Vision Benchmark Suite. Os resultados obtidos demonstram a aplicabilidade e a eficiência da abordagem proposta para a deteção de obstáculos e objetos em cenários urbanos.
In this thesis, we propose on-board multisensor obstacle and object detection systems using a 3D-LIDAR, a monocular color camera and a GPS-aided Inertial Navigation System (INS) positioning data, with application in self-driving road vehicles. Firstly, an obstacle detection system is proposed that incorporates 4D data (3D spatial data and time), and composed by two main modules: (i) a ground surface estimation using piecewise planes, and (ii) a voxel grid model for static and moving obstacles detection using ego-motion information. An extension of the proposed obstacle detection system to a Detection And Tracking Moving Object (DATMO) system is proposed to achieve an object-level perception of dynamic scenes, followed by the fusion of 3D-LIDAR with camera data to improve the tracking function of the DATMO system. The obstacle detection we propose is to effectively model dynamic driving environment. The proposed DATMO method is able to deal with the localization error of the position sensing system when computing the motion. The proposed fusion tracking module integrates multiple sensors to improve object tracking. Secondly, an object detection system based on the hypothesis generation and verification paradigms is proposed using 3D-LIDAR data and Convolutional Neural Networks (ConvNets). Hypothesis generation is performed by applying clustering on point cloud data. In the hypothesis verification phase, a depth map is generated using 3D-LIDAR data, and the depth map values are inputted to a ConvNet for object detection. Finally, a multimodal object detection is proposed using a hybrid neural network, composed by deep ConvNets and a Multi-Layer Perceptron (MLP) neural network. Three modalities, depth and reflectance maps (both generated from 3D-LIDAR data) and a color image, are used as inputs. Three deep ConvNet-based object detectors run individually on each modality to detect the object bounding boxes. Detections on each one of the modalities are jointly learned and fused by an MLP-based late-fusion strategy. The purpose of the multimodal detection fusion is to reduce the misdetection rate from each modality, which leads to a more accurate detection. Quantitative and qualitative evaluations were performed using ‘Object Detection Evaluation’ dataset and ‘Object Tracking Evaluation’ based derived datasets from the KITTI Vision Benchmark Suite. Reported results demonstrate the applicability and efficiency of the proposed obstacle and object detection approaches in urban scenarios.

碩士
元智大學
機械工程學系
94
Visual tracking means a camera can track a moving target while the spatial relationship between the camera and the target is changing. Visual tracking has found many applications in security surveillance, robot assembly, autonomous vehicles navigations and moving objects tracking. It is easy to track moving objects in a fixed camera because of the static background. But it is a challenging and hard task for tracking moving objects in a moving camera. In this paper, we describe how image sequence taken by a moving camera may be processed to detect and track moving objects against a moving background in real-time. Motion is found by tracking image features, segmentation is based on feature’s velocity, and to filter background features using Gaussian normal distribution. Finally, we use weighting approach for filtering the error of optical flow computation and enhancing moving objects tracking.

碩士
龍華科技大學
工程技術研究所
95
This study is mainly to construct a real-time multiple objects image tracking system based on moving edges detection technique. The moving edges detection technique is utilized as the main detection rule, the moving target shifting method and background compensation method are also applied to solve the shortcoming of moving edges detection. This tracking system can find out the moving object correctly under a complicated environment, in addition, the template matching method is applied to search for multiple moving objects simultaneously, therefore, this system is suitable for more extensive situation.　The Back-Propagation Neural Network technology is utilized to compensate the gray level in template, and to solve the influence of unevenness of the luminance in image. Finally, the detected object position information is used for the control system to track or monitor the object movement. This whole real-time multiple objects image tracking system can be divided by the software and hardware. In software, a Visual Basic program is developed as the operation interfaces, and the Halcon image processing software library is applied as the image processing developing tool. In hardware, a color CCD camera with image acquisition card are used as the image source, and X-Y table with servo motors controlled by a motion control card are used as the movement tracking platform.

碩士
國立交通大學
電子工程系所
94
An intelligent video surveillance system usually performs the tasks of background modeling, motion detection, and tracking. In this thesis, a level set function is used to record the moving objects for these three operations. The background model is first constructed before the background subtraction is performed for the motion detection. Then, a mobile camera keeps tracking the moving objects with a region tracking model. The original region tracking model is modified to alleviate the interference of cluttered environment. The relation between two level surfaces of successive two frames is taken into consideration. The probability model built from the statistic property of an image is also included. Finally, an integrated surveillance system is proposed. Different units in the surveillance system may choose appropriate contour models to solve their problems.

碩士
中原大學
資訊工程研究所
84
This thesis suggests a method to track moving objects in image sequences which are taken by a stationary camera. In this algorithm, the number of moving objects and their moving direction are unlimited, and their moving paths can be crossed. But the objects must move in a plain, and they can''t rotate themselves. Beside, the objects have to own rigid body and inertia characteristic. The first basic idea of this algorithm is to detect the changing parts of the images. These changing parts are parts of moving objects. We call these changing parts "Moving blob". After determined where are the moving blobs, we need use shape analysis to describe them. Then we set up some constraints of these features to match blobs in two consecutive image frames. These relationship is called "Motion vector". This kind of relationship represent the motion track in a small period of time. As mentioned above, these motion vectors consist of many features, so they are mapped into a high dimension feature space. In this feature space, motion vectors with similar motion behavior and appearance features will be together to form a cluster. So we use cluster analysis to separate them. Obviously, each cluster represent a moving blob. With the same idea, we group moving blobs with similar motion behavior into moving objects. After moving objects are found, we extract the motion track coordinates of the objects. In the end, we use pseudo-inverse method to map image coordinates into real world coordinates. In experiments, we use three different kinds of image sequences to test our algorithm. The algorithm finds out the motion path of objects. But the algorithm has some limitations. First, the algorithm can''t analyze rotational motion. Second, if the light is changing dramatically, the algorithm won''t achieve its goal. The color and texture of moving objects and background may also influence the result greatly. These are what we need to study further to make the algorithm more robust.

碩士
元智大學
資訊工程學系
98
To provide longer wireless sensor networks lifetime in moving objects tracking applications. We proposed a Message-Pruning Tree based scheme which capable auto adapt locality. Since a proper locality adapting can save more energy in Message-Pruning Tree based scheme, The most approaches for Message-Pruning Tree demand a fixed locality which often raise the communication cost during forward query and update messages. By heuristic search proper locality, the simulation results show that our proposed scheme provides outperformance than other conventional fixed locality in terms of query and update cost.

碩士
國立臺灣科技大學
資訊工程系
93
In the last decade, due to the popularization of video products and the rapid development of computer vision techniques, the detection and tracking methods for dynamic images have been widely applied in many kinds of fields, such as video surveillance, intelligent transportation, and parking area management systems. They can replace a lot of bored and time-wasting work, and avoid mannal mistakes caused by fatigue of human. On the effectiveness for a given period of time, these visual detection and tracking systems possess the ability of reporting sudden situations in real time, so that the whole time costs of such systems can be greatly reduced. In this thesis, the detection phase of our developed system consists of four parts: background generation, foreground detection, shadow elimination, and background maintenance. In the background generation part, the median method is used for constructing background images from the past N frames. In the foreground detection part, an extraction function is applied to indirectly perform differencing to obtain foreground images. In the shadow elimination part, a deterministic nonmodel-based method is adopted to remove shadows. As to the background maintenance part, a history map which records the number of times of the changes of corresponding pixels is employed to maintain background images. In the tracking phase of the system, this thesis exploits a particle filter to track moving objects. The color distribution of a moving object is chosen as its features represented by a color probability histogram. In order to raise the accuracy of tracking, the background information serves as the increase candidate weight of a moving object. The experimental results reveal that in general situations our system can achieve real-time processing and can obtain robust detection and tracking results for multiple moving objects.