采用互补特征的核相关滤波目标跟踪算法

为了改善跟踪算法的性能,提出一种自适应加权的融合颜色特征和方向梯度直方图(HOG)特征的多核多通道的相关滤波跟踪算法.针对核相关滤波算法特征单一的问题,采用互补特征核空间描述目标,并根据互补特征响应值的大小,自适应为互补特征核空间分配权重、更新模型,提高算法的鲁棒性.实验结果表明:所提出的算法不仅能在一定程度上处理目标外观变化问题,而且完全满足跟踪场景的实时需求.

Target Tracking Algorithm Using Complementary Features of Kernelized Correlation Filter

In order to improve the performance of tracking algorithm, a correlation filter tracking algorithm with multi-kernel and multi-channel using an adaptive weighted fusion method based on color feature and histogram of oriented gradient(HOG)feature is proposed. As kernelized correlation filter can extract few features, this algorithm presents target appearance by using complementary kernel features. According to the magnitude of the response values of the complementary features, the weights of the complementary kernel features and updating model are adaptively assigned, improving the robustness of the algorithm. The results of experiments show that the proposed algorithm not only can handle changes of object’s appearance, but also completely meet the tracking demand of real-time scenario.