Kernel Spatial Histogram Target Tracking Based on Template Drift Correction
偏移校正的核空间直方图目标跟踪

Aiming at the limitations of the traditional mean shift, such as invariable kernel bandwidth, inadequate color distribution representation of target and the accumulative tracking errors, an improved tracking algorithm with the following strategies is proposed. The target model and the candidate are described by a modified second-order spatial histogram including color and spatial information, and the similarity between them is evaluated by Bhattacharyya coefficient. According to the target region parameter resulted from template drift correction which can eliminate the tracking errors, the target model can be estimated repeatedly. The tracking region parameters are updated through an affine transform combining corner detection and edge detection. Besides, the target motion is predicted by either Kalman filter or linear filter according to the Kalman residual error. Experimental results show that the proposed algorithm is robust against similarity distraction, scale and orientation variations and short-term occlusion.