基于 Kalman 滤波和直方图匹配的双目视觉跟踪

Visual tracking is a very important problem in the computer vision. Most of the studies focus on the monocular camera target tracking. the object tracking methods based on monocular camera would lose the depth information, which will cause much more difficult to track the object when faced with occlusion, shadow and so on. This paper proposes an algorithm of binocular vision tracking based on Kalman filter and histogram matching. This algorithm firstly uses Kalman filter to predict the position of the target. Then histogram matching is used to search the object in the local region. At last epipolar constraint will be adapted to verify the tracking result at last, which realizes a complete prediction-matching-verification tracking process. The experiments show the algorithm could effectively reduce tracking failure and get better tracking performance compared with the Mean Shift algorithm.