Moving Target Detection And Tracking Based On Background Subtraction And Mean Shift

Moving target detection and tracking is the core issue of computer vision, which combines computer image processing, video, image processing, pattern recognition, artificial intelligence and many other fields. Using this method can work efficiently and improve the performance of monitoring system. It is a very active research direction in the field of computer vision. It has wide application in intelligent transportation, TV monitoring, human-computer interaction, video compression coding and other fields, which have good research value. Because of the reality application of the complexity environments, research based on robust detection and tracking algorithm has important theoretical and practical significance. For this reason, we will do the research based on the following areas.Firstly of all, for a large quantity of calculation caused by the traditional Gaussian mixture distribution model, pixel classification modeling algorithm is presented. A single Gaussian distribution modeling is used in a single-mode state using and the multi-modal pixel Gaussian distribution model is used in mixed modeling. Classification based on different issues is presented in order to improve the system in real time.Secondly, for target feature extraction vulnerable issue on the impact of the background pixel in the Mean Shift algorithm, background similarity algorithm is presented to reduce the weight of the background pixels as well as optimize the extraction of target feature. Meanwhile, the edge detection algorithms extract the target measure is introduced to adapt the Mean Shift algorithm of changing in target scale issues, and the target template update strategy is presented.Then, because the Mean Shift algorithm can not meet the goal of rapid and large area of velocity shelter problem, the Kalman algorithm is introduced to predict the target location as well as improve robust on tracking of the Mean Shift algorithm. And the template matching algorithm is used to re-lock the lost tracked target.Finally, a test platform is built. The proposed algorithm based on sampling of the large number videos is simulated. The test results show that the robustness of detection and tracking has been improved, which verify the effectiveness of the algorithm.