Research On Target Detection And Tracking Technology In Video Surveillance For Safe City

Throughout the present situation of domestic and foreign security of the city, as the safety accidents from the metro, aviation, the home area and road transportation and other public places, continue to occur, so the public security problem has been paid more and more attention by people. In recent years, local government departments in domestic are also increasing the human, material and financial resources devoted to public safety, are trying them best to build a "harmonious society, peaceful city" slogan. Among them, in order to maintain social stability and harmonious development of society, the law department begins to focus on strengthening the construction of informatization, relies on information means to supervise security and stability of public places. Therefore, this paper presents the research on object detection and tracking technology in video surveillance for safe city. In this research, we systematically discuss the target detection, target feature representation and extraction technology and the framework of particle filter algorithm. For these three aspects and meeting the actual needs, we propose the target detection that is the fusion method between background difference method and frame difference method, and target feature representation and extraction technology using the HOG feature and the particle filter technique using particle image to complete the predict of target position and size.Through the background introduction of the current city security and monitoring technology, this paper raises the importance of constructing video monitoring technology for safe city strengthening police with science and technology. And we emphatically introduce the significance and the value for modern living with the development of monitoring technology. In addition, for the research on this topic, we have also conducted comprehensive investigation of research status at home and abroad, and analyzed the research direction and development trend of video monitoring technology. A good target feature representation and description method will plays a very significant role for the accuracy improvement of target tracking algorithm. Therefore, we mainly introduce the important and traditional target feature representation, and give the concrete elaboration. We detailed analyze characteristics of the color feature, texture feature, Haar feature and SIFT feature, and from their respective feature representation methods and feature extraction, we illustrates the characteristics of traditional classical image representation method. This article summarizes the meaning of dynamic target detection and its important status, and for research on popular dynamic target detection algorithm including frame difference method and the background difference method, we systematically and comprehensively understand the work process of each algorithm, and the main research direction of the dynamic target detection, which also provides strong theoretical basis for the research of our proposed dynamic target detection algorithm. Research on multi-target detection technology in video surveillance is based on the technology of multi-target tracking. Here, we propose the fusion method between frame difference method and background difference method to realize the multiple targets detection in the monitored area. In this process, we introduce the target detection and segmentation technology of moving object from the background image.Finally, for the study of particle filter algorithm, this paper comprehensively propose the research on target detection and tracking technology in video surveillance for safe city. Here, we study the detection of the target feature representation and extraction, because it is only given the good object feature representation, we can realize accurate object tracking. So, here we use the current popular image feature extraction technology, namely the HOG. Because the HOG feature descriptor is a typical local feature descriptor, which was firstly proposed in image matching field, the actual effect is very significant and it can solve the problem of image rotation and scale invariance. Therefore, we will use the HOG feature representation to target tracking. Finally, we detailed propose multi-target tracking algorithm based on particle filter. Through the research on combining the target detection, HOG feature representation and particle filter algorithm, we realize the effective monitoring and tracking of moving targets in video surveillance on Crossing test data.