An Algorithm Researchof Moving Target Detection And Tracking Under Complex Background

Recently, intelligent video surveillance stands out in security, of which inthe field the research of moving target detection and tracking has being a focus，andthe performance of them directly affects that of the whole system. So we make acomprehensive study of target detection and tracking in this paper.The research is based on the daily video monitoring scenario, where existingdynamic-complex background, such as tree swaying and illumination change. It ismainly divided into two aspects of moving target detection and target tracking..First, select the codebook algorithm, which does well under complex backgroundand runs in real-time. List the algorithm’s issues in the range of brightness and colordistance threshold, which resulting in the inaccuracy of detection result, reasonrespectively from theory to experiment and put forward the improved strategy to solvethem.Second, at the aspect of target tracking, adopt the simple and efficient real-timecompressive tracking algorithm. Make the target-detection result as the input sourceof tracking, to solve the problem that tracking target needed to be selected manually,by introducing an algorithm of target detection. At the same time, restrict the updateof classifier based on the detection result of codebook, when the two or more targetsoccurring occlusion, and then use the previous classifier to overcome the target losswhen the targets start to separating,.In conclusion, we improve the codebook algorithm and combine it with thecompressive tracking algorithm. Use the detection algorithm to locate the effectivetarget in the field of vision, which used in the compressive tracking algorithm.Experiment results show that in daily video monitoring scenario, the combinedalgorithm can meet the requirement of real-time and accuracy.