基于环形像素点匹配的快速图像跟踪算法

相关跟踪法能有效地跟踪目标,但非常费时。序贯相似检测法结合了粗-精搜索策略,在跟踪性能和时间上都有很大改进,但在处理时间上还有进一步提高的可能。基于特征点的跟踪方法在跟踪时运算量小,但提取目标的特征时需要大量运算,且特征点的提取往往不理想。针对以上不足,论文提出了基于环形像素点匹配的快速图像跟踪算法,该算法采用环形像素点为特征点,在借鉴序贯相似检测法的基础上,对匹配区域先进行粗搜索以确定待匹配窗口,再进行精搜索以确定正确的匹配窗口,最后利用相关跟踪法中的归一化相关系数确定模板的更新策略。实验结果表明,该算法不仅能稳定地跟踪目标,与相关跟踪法和序贯相似检测法相比,在处理时间上具有明显的优越性。

Fast Image Tracking Method Based on Circle Pixels Matching

Correlation-based method can track target effectively,but it has a high computational cost.Serial similar detection algorithm uses coarse-to-fine search strategy,so it has a good improvement in tracking performance and time cost,but it still has a potential to decrease computational cost.Feature-based method has a little time cost in tracking target,but it needs a lot of time to extract the target features,and the extracted features are always not well.Aimming at above defects,a fast image tracking method is presented in this paper.This method uses circle pixels as feature points.By using the idea of serial similar detection algorithm,this method adopts coarse-to-fine strategy to search the correct target position.At last,normalize cross -correlation function is used to decide the template change strategy.Simulation results show that this method can not only track targets stably,but also has superiority in computational cost compared with correlation-based method and serial similar detection algorithm.