基于边缘的自适应相关跟踪算法

Adaptive Correlation Tracking Algorithm Based on Edge Detection

针对相关跟踪实时性差和误差累积问题,提出了基于边缘检测的自适应相关算法。采用修正后的灰度差的绝对平均值(MAD)相关匹配克服了对噪声敏感的缺点;利用目标的边缘信息确定模板中参与匹配的区域,并采用粗精结合的搜索策略,加快了匹配速度;采用模板加权更新和中心修正有效地消除了相关运算累积误差的影响。

This paper describes adaptive correlation tracking based on edge detection to overcome the time-consuming and cumulated errors in correlation tracking. The corrected MAD (maximum absolute difference) distance method is not sensitive to noise. Adjusting the reference image according to the detected edge of the object and adopting a rough-precise searching strategy reduce the needed time. On the basis of similarity of matching, a weighted updating of the reference template is applied.