Abandoned Object Detection Based On Video Image Sequences

With the fast development of highways and tunnels, traffic accidents are increasing. The abandoned object incident is a frequent occurrence of traffic accidents, and the traffic accidents and the potential security risks caused by it has become an urgent problem. How to timely and accurately detect the occurrence of such incidents as soon as possible to exclude security risks and protect the safety and smoothness of highways and tunnels has become a hot issue.In this paper, the dual background detecting method is applied to the abandoned object incident detection on the basis of the analysis of related detection algorithm. An improvement is put forward and the misjudgment caused by the dual background detecting method is eliminated combined with the classification results of the Support Vector Machine (SVM) classifier, in order to adapt to the traffic environment and detect the occurrence of such incidents accurately. Abandoned object incident detection can be divided into three parts, which are moving target detecting, moving target tracking and abandoned object detecting. First of all the background difference method is used to ensure the real-time performance based on the analysis of commonly used moving target detection algorithms. Moving targets is detected through the steps including image preprocessing, the background extracting and updating, background difference, shadow eliminating and so on. The fusion algorithm for the splitting contours is proposed. Experiment shows that detection accuracy has been greatly improved after the fusion of splitting contours. Secondly, two commonly used moving target tracking algorithms which are Kalman filtering and region centroid tracking method are implemented. At the same time, the calculating method of traffic parameters like instantaneous speed and traffic flow is given. The moving target records are built and updated to achieve the goal of real-time tracking. Finally, the dual background detecting method is applied to the abandoned object incident detection based on the front work. In order to adapt to the environment traffic, SVM classifier is used for the classification of the moving targets. Abandoned object detection is based on the moving targets which are classified as non-car targets. Misjudgment caused by the dual background detecting method is eliminated. Experiment shows that the improved dual background detection algorithm which is combined with the classification results of SVM and applied to abandoned object incident detection, can well adapt to the traffic environment and achieve a good detection results.