在线特征选择的目标跟踪*

为提高目标与背景对比度低、相似物体干扰等复杂环境下目标跟踪的效果，提出将在线学习选择最优颜色特征嵌入跟踪算法中，以改善跟踪的稳定性。以当前时刻目标的区域为目标区域，利用卡尔曼滤波预测目标的下一时刻位置，在卡尔曼滤波预测的位置为中心取某一区域作为背景区域进行在线特征选择作为下一时刻的跟踪特征，以卡尔曼滤波预测的位置为初始位置利用Mean-shift搜索目标位置，此位置作为量测进行卡尔曼滤波校正。通过实验表明，该方法在目标与背景的对比度低、相似物体干扰等复杂环境下极大地改善了跟踪的稳定性。

Online select feature target tracking

In order to improve robustness of tracking system under complex surrounding such as low-contrast, similar to object interference, this paper proposed that online select optimal color feature mechanism was embedded in target tracking algorithm. The target region of current moment was seen as the target region, employed Kalman filter predict the target position in next moment, and selected a region in the predicted position as the background region. Employ this two region online learning and select the optimal color feature as tracking feature of the next moment. The Kalman filter predicted position as initial position,utilized Mean-shift search the target position.Used this searched position as measurement correct the Kalman filter.Experimental results show that this method can greatly improve the robustness of target tracking method under complex surrounding such as low-contrast, similar to object interference.