关键词: 时空上下文, Adaboost算法, Kalman滤波, 视觉跟踪

Abstract: Aiming at the problem that the first frame needed to be manually selected and the subsequent tracking interference was caused by selection deviations in the spatiotemporal context tracking algorithm, we proposed to use Adaboost algorithm to detect the first frame and introduce Kalman prediction mechanism to assist spatiotemporal context tracking algorithm. When occlusion, jitter and other problems occurred, it could ensure the stability of tracking and improve the robustness of the  algorithm. Comparative experiments were carried out on three public data sets, such as Shelter1. The experimental results show that the proposed algorithm can realize the automatic detection function of the first frame, and the robustness 
and tracking effect of the subsequent tracking algorithm are also significantly improved.

Key words: spatiotemporal context, Adaboost algorithm, Kalman filter, visual tracking