基于特征融合的均值偏移目标跟踪算法

An improved mean-shift algorithm for target tracking based on the feature fusion

均值偏移目标跟踪方法采用颜色直方图对所选择的目标区域进行建模,由于颜色直方图是一种对目标特征比较弱的描述,当有遮挡等干扰因素时,算法效果欠佳,为了有效解决均值偏移目标跟踪算法不足而导致目标定位不准的问题,提出了将颜色特征中融入像素点空间位置特征的算法来实现目标跟踪.实验表明该算法能较好地适应复杂背景视频序列,改进了传统均值偏移算法的不足,提高了算法的鲁棒性和准确性.

The mean - shift target tracking method using the color histogram of the selected target region is modeled by the color histogram that is a relatively weak description of the target feature, when there is some occlusion of in- terfering factors, the algorithm is ineffective. To obtain an effective mean - shift tracking algorithm to solve targeting problems, this research proposes to integrate spatial location features with pixel color features in the algorithm to a- chieve target tracking. The experiment shows that the algorithm can better adapt to the complex background of the video sequence, and improves the traditional mean - shift algorithm as well as the robustness and accuracy of the algorithm.