基于椭圆对数极坐标变换的尺度变化目标跟踪算法

Algorithm of scale-variant objects tracking based on ellipse log-polar transform

针对传统对数极坐标变换局限于跟踪圆形或类圆形尺度变化目标这一问题,提出一种基于椭圆对数极坐标变换域下目标匹配的尺度变化目标跟踪算法。算法利用Mean Shift进行空间定位,确定目标的形心,通过椭圆对数极坐标变换域中目标和候选的最大相关匹配系数来确定目标的尺度参数。实验结果表明:该文算法在目标小形变和光照变化条件下,跟踪误差较小,尺度跟踪准确率更高,具有较好的鲁棒性。

To solve the problem of the traditional logpolar tracking method which could only catch up with circular and quasicircular objects with scale changing, we introduced ellipse log polar transform（LPT） to estimate the target＇s scale parameters within the framework of Mean Shift tracking. Experimental results demonstrate that the composite algorithm has lower track ing error and better tracking accuracy rate on the condition of small deformation and light in tensity changes. Comspoured with the traditional, it has a better robustness.