基于距离信息的Mean-Shift跟踪算法

针对图像制导目标跟踪系统在跟踪过程中，成像视角和距离的变化带来的跟踪漂移问题，提出了利用距离信息动态更改Mean-Shift算法跟踪窗口尺度的算法。依据初始跟踪时选定的目标模板，建立目标的灰度特征模型，确定初始跟踪窗口尺度；在跟踪过程中依据距离信息来动态更新跟踪窗口尺度，保证在跟踪系统逐渐接近目标的过程中跟踪窗口能够完全包括或者绝大部分包含目标；在每一帧跟踪收敛后利用Bhattacharyya相关系数对目标模板进行非线性更新。以Vega产生的模拟飞行视频数据进行了算法仿真，结果表明：该算法能够适应目标不断膨胀的情况，在很大程度上降低跟踪漂移。

Mean-Shift Tracking Algorithm Based on Distance Information

Aimed at the tracking drift bought by imaging perspective and distance changes in the tracking process of image-guided target tracking system, a tracking algorithm using the distance information to dynamically change the scale of tracking window of the Mean-Shift algorithm was proposed. Firstly, the targets gray feature model was established according to the selected initial tracking template, and the initial scale of tracking window of Mean-Shift was determined. Then, the window’s scale was updated according to the distance information to ensure the tracking window contain the target completely or partially while the tracking system gradually closed to the target. Finally, Bhattacharyya correlation coefficient was used to update the target template nonlinearly after the tracking convergence of each frame. This paper uses the flight video data generated by Vega to simulate the algorithm, and the result shows that it can adapt to the expanding of the goal, and reduce the tracking drift in a large extent.