融合颜色特征的核相关滤波器目标跟踪

Kernel Correlation Filter Object Tracking Integrated with Color Features

目标特征表示是视觉目标跟踪领域中的一个热点话题。近年来,核相关滤波器因其时效性在目标跟踪中得到广泛应用,但简单的灰度特征表示难以应对复杂环境中的目标跟踪问题。基于此,融合颜色特征提出了一种实时的目标跟踪方法。实验结果表明,颜色信息可以有效提高目标跟踪的整体性能,该方法能够适应姿态变化、光照变化等多种目标外观变化,平均跟踪速度为62.1帧/s,可以满足实时应用的需求。

Object feature representation is a hot topic in visual object tracking. Recently, kernel correlation filter has found wide application in object tracking with its high efficiency, while the simple pixel representation is not adaptable to tracking in complex scenes. A real-time tracking method integrated with color features is proposed here. Experimental results show that color information can improve the overall performance effectively, and the proposed algorithm can perform superiorly under the condition of different appearance variations, such as pose variation, illumination variation etc. The average tracking speed is 62.1 frames per second, which meets the real-time application requirement.