Application And Research Of The Abandoned Object Detection Technology In The Video Surveillance

Intelligent video surveillance system can automatically monitor the scene, when found irregularities in the scene immediately trigger the alarm, thus greatly reduce the workload of staff, and improve the accuracy of detection, so intelligent video surveillance system compared with traditional video surveillance system is more suitable for security work. Currently placed abandoned objects is one of the main means of terrorist attacks, especially in followed places, abandoned object detection can’t be ignored. For densely populated public places and some of the higher level of security departments, a real time, 24 hours abandoned object detection becomes especially important.The main study in this thesis is detecting abandoned object in surveillance video. The main task of this algorithm is to detect abandoned objects and their owners. So far the abandoned event detection is mainly based on target detection and target tracking. Our algorithm is based on target detection.In this thesis,we made two improvements on the original algorithm. On the one hand, in order to reduce the complexity of the calculation, forground extraction method of the the original algorithm changed to hybrid differencing method; On the other hand, partially static method we added to the original algorithm.Specifically,in this thesis abandoned object detection algorithm based on stop object detection. The algorithm is divided into background maintenance, abandoned object detection, owner detection three stages. In background maintenance stage, hybrid differencing method is adopted to establish the background model and get the foreground image, this method not only can obtain the moving foreground objects but also can obtain the stationary foreground objects; In abandoned object detection stage, using the three dimensional modeling method to establish moving foreground and static foreground; This algorithm for "ghosting", and partly stationary objects and occlusion have a higher robustness. Owner detection is using the video backtracking method to find out a video data, the video from the abandoned object just entering the region to the end of completely enter the area,first of all calculate the color histogram for candidate owners of all the images, then calculate average value. The candidate owner regarded as the real owner, if its histogram is the most close to the average.In addition, describes in detail the experimental environment and the overall structure of the system. Set the main parameters of the algorithms, Functional testing and quality evaluation results will be described for several publicly available sequences. The experimental results show that this algorithm can better accomplish abandoned object detection and owner detection, and has good robustness and accuracy. Finally, with the abandoned object detection algorithm designed an abandoned object detection system.