基于循环相关滤波算法的复杂背景目标跟踪方法研究

为了实现在复杂情况中对特定目标的路线跟踪,主要研究了基于循环相关滤波器的目标跟踪算法,并针对复杂环境情况下影响跟踪性能的条件进行分析比较。该算法是一种基于线性分类器的监督学习算法,通过添加空间正则化分量减少边界效应,提取准确目标。在OTB-50和OTB-100数据集上进行算法验证,实验表明,该算法的跟踪结果优于其他算法,在快速移动、背景杂乱、光照变化、遮挡、移出视线和运动模糊6种复杂情况下跟踪准确度比传统算法均高出0.1以上。

Object tracking based on discriminative correlation filters in complex scenes

In order to realize the tracking of a specific target in a complex scene,this paper mainly studies the object tracking algorithm,which based on discriminative correlation filters,analyzing and comparing the conditions that affect the performance in complex scene.This algorithm is a supervised learning algorithm based on linear classifier to extract accurate targets by adding spatial regularization components to reduce boundary effects.The algorithm is verified on OTB-50 and OTB-100.The experiment show that the tracking results of the algorithm are better than other algorithms.The accuracy of the algorithm is higher than 0.1 that of traditional algorithms in six complex scenes:fast motion,background clutter,illumination variation,occlusion,out of view and motion blur.