基于塔型结构的快速相关跟踪算法

A fast correlation tracking algorithm based on pyramid structure

提出一种能自适应的确定相关匹配搜索范围的快速相关跟踪算法。该算法结合两点线性预测器和图像塔型结构的多分辨力分析方法,综合图像在不同尺度下的位置信息,自适应地计算出相关匹配的搜索范围,减小了相关运算的计算量,同时确定了有效的帧间相关置信度估计和模板更新策略。实验证明此算法计算速度快,能持久稳定地跟踪运动目标。

A fast correlation tracking algorithm which can adaptively determine the search range in correlation matching is proposed. Combining two-point linear predictor with multi-resolution analysis method for image pyramid structure and integrating the image position information under the different sizes, the algorithm adaptively calculates the search range in correlation matching, reduces the calculation quantity in correlation operation, determines confidence estimation for the effective interframe correlation and refreshing policy for template. Experiments demonstrate that speed of the algorithm is fast and can continuously track moving target for a long time.