基于快速傅里叶变换的局部分块视觉跟踪算法

Local Patch Tracking Algorithm Based on Fast Fourier Transform

针对视觉跟踪中目标表观变化、局部遮挡、背景干扰等问题,该文提出一种基于快速傅里叶变换的局部分块视觉跟踪算法。通过建立目标分块核岭回归模型并构建循环结构矩阵进行分块穷搜索来提高跟踪精度,利用快速傅里叶变换将时域运算变换到频域运算提高跟踪效率。首先,在包含目标的初始跟踪区域建立目标分块核岭回归模型;然后,提出通过构造循环结构矩阵进行分块穷搜索,并构建目标分块在相邻帧位置关系模型;最后,利用位置关系模型精确估计目标位置并进行分块模型更新。实验结果表明,该文算法不仅对目标表观变化、局部遮挡以及背景干扰等问题的适应能力有所增强,而且跟踪实时性较好。

In order to solve the problems of appearance change, local occlusion and background distraction in the visual tracking, a local patch tracking algorithm based on Fast Fourier Transform（FFT）is proposed. The tracking precision can be improved by establishing object＇s patch kernel ridge regression model and using patch exhaustive search based on circular structure matrix, and the efficiency can be improved by transforming time domains operation into frequency domains based on FFT. Firstly, patch kernel ridge regression model is constructed according to the initialized tracking area. Secondly, a patch exhaustive search method based on circular structure matrix is proposed, then the position model is constructed in adjoining frame. Finally, the position of the object is estimated accurately using the position model and the local patch model is updated. Experimental results indicate that the proposed algorithm not only can obtain a distinct improvement in coping with appearance change, local occlusion and background distraction, but also have high tracking efficiency.