A REAL-TIME ABANDONED OBJECTS DETECTION ALGORITHM BASED ON MULTI-TARGET TRACKING

Most existing abandoned objects detection algorithms embody poor flexibility and scalability. To solve these problems and achieve the purpose of real-time performance, an algorithm for detecting discarded objects based on target tracking is proposed. The algorithm consisted of three steps: detection, tracking and correlation analysis. The position of interesting targets was extracted via the detector. In order to realize the aim of abandoned objects detection, a series of operations were completed, including tracking, correlation analysis and static state analysis. The method was verified on AVSS AB2007 dataset, with accuracy reaching 95.87%. The experimental results show that the proposed algorithm can meet the real-time requirements, and has a strong robustness and a good flexibility.