Research On Moving Target Tracking Based On The Improved CAMshift Algorithm

With the rapid development of computer technology,moving target detection and tracking technology has attracted more and more attention in the research field of both academia and industry.Real time detection and tracking techniques could help make instant response intelligently.Based on the above condition,research on moving target detection and tracking is discussed in detail.Besides,some common algorithms in this field have been tested and improved,what's more,the automatic tracking algorithm based on multiple features is proposed.In dealing with the moving target detection problem,by way of theoretical analysis and experimental comparison among models in terms of frame difference,optical flow and background subtraction,the ViBe algorithm,which belongs to the background subtraction method,is proved to outperform the other two models.However,in the case of the original frame containing the moving target,due to the single frame modeling strategy,traditional ViBe algorithm tends to fail to detect target accurately,thus resulting in the improved model by incorporating ViBe algorithm into multi-frame modeling strategy.When dealing with dynamic background,the improved method could well avoid incorrect detection resulted from conservative update strategy.The model calculates the dispersion among the sample points in the background firstly.Then the sample point with the highest degree of dispersion would be replaced by the new pixel that matches the model,so as to improve the ability of ViBe to adapt to the dynamic background.Finally,the foreground shadow is detected and filtered to obtain a complete and pure moving target.In the process of moving target tracking,Meanshift and CAMshift are discussed from the perspective of theory analysis and implementation process.Judging from the experimental results,CAMshift is able to achieve ideal tracking effect and well adapt to the size change of moving target,but limited to the simple environment.In this paper,the automatic algorithm based on multi-feature fusion is proposed to tackle the shortcomings of traditional CAMshift.Firstly,CAMshift,tracking algorithm based on a single color channel(Hue-component),is not robust enough to outperform other algorithms.Therefore,an improved CAMshfit model by combining Hue-component and Saturation-component is proposed.Experimental results with the improved model fully demonstrate the superiority of the improved CAMshift in dealing with complex background.Besides,to make up for the disadvantage of traditional CAMshift,which only relies on the single color feature for tracking,reflecting in inability of tracking the target in the scene where target and background are similar.Based on the color information,Speeded Up Robust Features(SURF),which could well describe the textural information,and has the advantage of being invariant to shift,rotation and scaling,is added to improve the performance.The new tracking algorithm based on multiple features is proposed after comprehensive comparison of the performance between two algorithms.We choose SURF to rematch the target when traditional CAMshift loses the target in tracking.Then CAMshift algorithm is activated again based on the centroid points of target collected by SURF.Furthermore,the ViBe algorithm is incorporated to enable automatic tracking.The improved ViBe algorithm detects moving target and filter out the shadow,then the pure foreground region is extracted as the initial target region.The automatic multiple features algorithm can not only make up for the original one in manual selection of the region of interest in traditional CAMshift,but also improve the detecting accuracy.Rigorous experiments are carried out to demonstrate the effectiveness of the multi feature fusion algorithm.Besides,real time tracking could also be achieved.