Detection And Tracking Of Moving Targets In Complex Background

In life,video surveillance system can be seen everywhere.More and more moving objects are detected and tracked based on video surveillance.The background of the video image is a real environment,and the background is complex,such as noise,illumination change,background disturbance and shadow of moving target.But the detection algorithm is not very ideal,and it still needs to be improved.Therefore,based on video surveillance images,this paper studies the detection and tracking of moving objects in complex background.First,this paper propose an anisotropic diffusion based image denoising method based on PDE(Partial Differential Equation),which is based on the P-M model.This method constructs a new diffusion function based on the model,and uses the structure tensor as the independent variable of the function.The improved method can remove noise while preserving the details of image edges.Secondly,we study the algorithm of VIBE(Visual Background Extractor)detection,analyze the existing problems,and propose a new method to establish and initialize the VIBE background based on GMM(Gaussian mixture model).Meanwhile,combined with the target extraction method of SOBS(self-Organizing through artificial neural networks),improve the target extraction of VIBE algorithm,and a target detection algorithm based on heuristic rule is put forward.Experiments show that the improved method effectively improves the ghost phenomenon existing in the VIBE algorithm,and the target detection effect is better.Finally,based on the improved VIBE algorithm,we remove the moving object’s shadow and track the moving target.The method of multi feature fusion is used to detect the shadow.The minimum circumscribed rectangle of the target’s largest contour is filled up by removing the shadow of the target.The minimum circumscribed rectangle of the target to be tracked is used as the initial target rectangle box of MeanShift tracking,and the target is tracked.