基于局部稀疏表示的模板匹配跟踪算法研究

Research on template matching tracking algorithm based on local sparse representation

在目标跟踪过程中,局部稀疏表示跟踪算法(LSRTA)在目标被遮挡面积过大、运动过快或形变量过大等情况下,跟踪过程中会发生目标偏移现象。针对这个问题,在LSRTA算法基础上,融入surf、flann和knn相结合的模板匹配算法,提出了基于局部稀疏表示模板匹配算法(LSRTMTA)以解决目标跟踪偏移问题。在LSRTA算法跟踪过程中,通过不断地计算新模板与当前模板的匹配值来判断是否发生偏移。当目标发生偏移时,停止LSRTA算法的跟踪,通过模板与帧图像之间匹配来重新确定目标位置;当确定目标位置后,再次进行LSRTA算法的跟踪。实验结果表明:该算法不仅保留了LSRTA算法的优点,还具有自动修正目标偏移的功能,改善了跟踪效果,增强了目标跟踪过程的容错性。

In the target tracking process,when the target is sheltered excessively,moves fast is deformed largely,the target deviation phenomenon will occur to local sparse representation tracking algorithm(LSRTA)in the target tracking process.To solve this problem,based on the LSRTA,the template matching algorithm which combines surf,flann and knn algorithm was integrated,and a template matching tracking algorithm(LSRTMTA)based on local sparse representation was proposed to solve the target offset phenomenon.In the LSRTA tracking process,the offset is judged through calculating the matching value of the new template and the current template.When the target is offset,the LSRTA tracking stops.The target is repositioned by matching between the template and frame image.When the target position is determined,the LSRTA tracking starts again.Experimental results show that LSRTMTA not only retains the advantages of LSRTA,but also has the function of automatic correction of target offset,improve tracking effect and increases fault tolerance in the target tracking process.