Research On Moving Objects Detection And Tracking Based On Background Subtraction With Illumination Robustness

Moving object detection and tracking based on video is an emerging area of research focus in computer vision and pattern recognition. It can be used in intelligent video surveillance and driver assistance system. The intelligent video surveillance system use camera work to replace the human eyes. It has the capability of collecting real and meaningful information and making appropriate decisions in order to meet needs of public safety. The key technology in intelligent video surveillance is moving target detection and tracking technology. There are a lot of factor which have impact on the result. One of the main factors is illumination. The research of video-based moving target detection and tracking technology with illumination robustness for the superiority of intelligent video surveillance system is very significant.Many famous domestic and foreign research institutions and researchers have begun to research in-depth and developed intelligent video surveillance system, such as the U.S. Carnegie Melfon University, Chinese Academy of Sciences Institute of Automation; have realized the intelligent video surveillance platform. In order to improve system reliability and intelligence, we must further improve its robustness, accuracy, and universal applicability, which involve a lot of problems.In this paper, moving object detection algorithm is in-depth and meticulous researched. Firstly, the commonly used methods:optical flow method, background subtraction, inter-frame difference method were discussed, then focus on the basis of the background difference method to study New background modeling and background updates method. According to the different indoor and outdoor scenes, illumination robustness algorithms are studied. Solutions are given to adapt to complex conditions, including the outdoor in day and night, indoor lighting mutation scenes, then shadow is suppressed to avoid the interference of the shadow of the detection target. Mutation for indoor lighting scenes, the existing background subtraction and improved algorithm is difficult to quickly adapt to light mutation, so that large areas of background pixels will be mistakenly detected as foreground pixels. The moving target is as a precondition, a fast detection algorithm is proposed to adapt the lighting mutation based the idea of block and classification Through the experiment, the method is faster, more accurate detection rate, can quickly adapt to light mutation and update the background image in a timely manner, while meeting the requirements of real-time and accuracy. The detection methods base LBP and its modification descriptors are discussed in day and night scene in the paper and the best LBP descriptor is chosen. In the day scene, a new descriptor is proposed, named MCS-LBP. In order to increase rate, a new object area detection method based on MCS-LBP with kalman filter or frame difference is proposed. Insufficient for night illumination, low contrast image at night, the image enhancement algorithm is analyzed in the paper and as pre-processing algorithm for target detection algorithm. The CLBP descriptor is also used in the object detection at night. Then two methods are compared and analyzed through the experiment. At last, a completed solution about moving target detection and tracking based on video is proposed.