基于改进ECO-HC的目标跟踪方法

Target tracking method based on improved ECO-HC

基于相关滤波的跟踪方法帧率高但容易发生跟踪漂移导致目标丢失。针对此问题,本文以相关滤波方法ECO-HC为基线,提出了一个用于描述图像场景变化幅度的参数D,根据实时计算该参数的值并以不同的间隔数更新模型,实现自适应更新模型在使算法以较高帧率运行的同时兼具良好的准确率,此效果在图像光照变化复杂、目标变形或被遮挡的情况下更加明显。实验结果表明,ECO-HC嵌入本文方法后,在光照变化的情况下成功率提高了1.6%,准确率提高了2.2%;在目标离开视场的情况下成功率提高了3.7%,准确率提高了3.4%。在本文实验环境中平均帧率达到60 FPS。

The tracking method based on correlation filtering has fast frame rate but is prone to tracking drift,resulting in target loss.Aimming to this problem,taking the correlation filtering method ECO-HC as the baseline,this paper proposes parameter D to describe the change amplitude of the image scene.According to real-time calculation of the value for this parameter and updating the model with different intervals,the adaptive updating model is realized,so that the algorithm runs at a higher frame rate and has a good accuracy.This effect is more obvious in the case of complex image illumination changing,target deformation or occlusion.Experimental results show that success rate and accuracy rate of ECO-HC embedded in this method are improved by 1.6%and 2.2%under the condition of changing illumination.When the target leaves the field of view,the success rate is improved by 3.7%,and the accuracy is improved by 3.4%.In this experiment environment,the average frame rate reaches 60 FPS.