The Algorithm Of Exception Object Detection In The Transport Hub

Exception object detection is an important part of intelligent monitoring. With terrorist attacks against the transport hub which has a large flow of people continue to occur, more attention has been paid on exception object detection in these scenes. However there are some problems such as false alarm and block widespread in the existing algorithms. And these algorithms also can not distinguish between people and objects. How to judge whether the people near the abnormal target is its owner is also a difficult problem for them to be solved.To solve these problems, we designed a set of exception object detection based on PC. The program uses dual background and evidence accumulating to solve false alarm problem and occlusion. HOG features and SVM classifier is used to distinguish between the target. Color,shape and other characteristics are used to determine whether the people near the target is its original owner, then to determine whether the target is abnormal left.Experimental results show that this program can solve the false alarm problem and occlusion effectively. The phenomena that the people who keep still at the scene were detected as abnormal objects and that the owner never left its luggage but the luggage is detected as abnormal object are efficient avoided. The system is stable, fast processing speed and can meet the practical application of real-time requirements. This system is conductive to large scale applications.