Research On Moving Object Detection And Tracking In Video Images

Moving object detection and tracking is one of the important research direction s in computer vision,it is the foundation and key technology in intelligent transportation,human-computer interaction,visual navigation and intelligent monitoring application,the technology mainly involves digital image processing,computer vision,pattern recognition and artificial intelligence etc,therefore,it has a strong theoretical research and engineering application value.This thesis focus moving object detection and tracking algorithm in video images,the main contents of this article are:1.In moving object detection: this article analyzes hybrid Gauss and non parameter method for moving object detection,and then proposes a temporal entropy based kernel density estimation for background modeling.This method firstly constructs an entropy map according to the complexity of the scene,and divides the map into stable regions and dynamic regions based on the maximum entropy threshold principle.Moreover,in order to balance the efficiency and performance,different number of samples is adopted for respective regions.The experimental results show that this method can effectively separate the target,which has strong stability,robustness and faster update rate for the background modeling.2.In object tracking: Firstly,this paper firstly analyzes the compressive tracking method and then in order to deal with the robustness of object tracking because of single feature,proposes an improved compressive tracking method based on adaptive feature fusion.The experimental results verify the performance of this advanced method.Moreover,an optimal feature online selection mechanism is introduced to compressive tracking method,which can select the feature that can distinguish the positive and negative samples from the feature pool effectively.The method can effectively reduce the drift probability that caused by weak feature,and the update mechanism can effectively deal with the issue of selective partial occlusion.The experimental results show that the algorithm can improve the tracking accuracy and stability.3.Based on the above research results of the proposed tracking a lgorithm,a set of fatigue status detection system is designed.Based on the tracking algorithm,this system can locate the face position in real-time,and detect the eye in the range of facial position,then the software can judge the status of fatigue according to the eye closing degree.The final test results show that the system can recognize the fatigue and has good accuracy and robustness to a certain view variation.