Research On Moving Object Tracking Method In Video Surveillance

In order to alleviate the human resources investment in the video surveillance system,the moving object tracking technology is introduced into the video surveillance field to achieve the purpose of behavior analysis and the detection of the abnormal behavior.In recent years,in order to solve various problems in object tracking,many methods have been proposed.However,in the current research of object tracking problem,the problem of object deformation and occlusion is still the main problem affecting the accuracy of object tracking,especially in the context of multi-object tracking,the problem of mutual occlusion and adhesion between targets.It also has a great impact on the research of object tracking problems.In this paper,a patch-based model based on Extended Kalman Filter(EKF)is proposed to model the object,and the SIFT feature points are extracted from the target region.At the same time,maintaining the relative positional relationship among the patches;The extended Kalman filter algorithm is used to predict the position of the moving object at the next moment according to the probabilistic model,so as to reduce the matching range of the object,thereby improving the matching efficiency and accuracy.The model is fed back and corrected by matching the successful object position,and the values of the object model are updated.Through the relative positional relationship between the patches,the occlusion problem of the object can be effectively solved,and the robustness of the object model can be maintained.At the same time,the patches that are too far apart can be eliminated,and the mismatch rate can be reduced.Aiming at the multi-object tracking problem whose background changes,a background subtraction strategy is proposed.The background is corrected by the relative motion relationship,and the matching background area is subtracted in the image frame which is to be matched.At the same time the remaining area is matched as the target area with the moving object,thereby it can reduce the influence of the background on the object and improve the accuracy of the object match.MHT algorithm is used to track multiple objects,and the structure of patches is used to solve the problem of mutual occlusion and adhesion among the objects.After the matching is completed,the object and background models are updated online to ensure the validity of the object and background models.Through tracking experiments on single object and multi-objects whose background changes,it can be seen that the algorithm of this paper can effectively improve the accuracy of object tracking,and it can solve the object tracking problem when the object is occluded very well.