Multi-Object Detection And Tracking Based On Background Subtraction

With the rapid development of the "global village", intelligent video surveillance will usher in the age of big data. It is the great challenge of intelligent video surveillance must face how to extract useful information in the large amount of data generated every day.Multi-object detection and tracking is the core technology in an intelligent video surveillance system. It can analysis the information in video automatically without human intervention and achieve the scene dynamic object detection and tracking by using computer vision, pattern matching and artificial intelligence technology. On this basis, it can analysis the behavior of object and timely feedback on the abnormal behavior and emergency.In recent years, a large number of scholars engaged in the exploration and research in the field, and made a lot of new algorithms and models. But there are still many issues unresolved. In this paper, we proposed a multi-object detection and tracking algorithm based on background subtraction. It does not need any a priori knowledge. By combining color and edge information and the use of adaptive threshold method, the proposed background subtraction algorithm is robust to illumination change and the sudden movement of objects in the background. We extract KLT features and region features in foreground. Then we compute the similarity between two feature and cluster features to regions. According to the information between regions, such as distance, foreground, speed, we optimize the configuration of region by using simulated annealing algorithm. Finally, we can get the motion information of objects. The experiment results show our method is applicability and validity.