Abstract: In the video sequence image tracking, the scale and motion direction of the tracking object relative to background are changing in real time, the traditional tracking algorithm will cause deviation of the scale and direction, leading to tracking drift, even bring about the failure of tracking, so an adaptive scale and direction object tracking method is proposed. In the framework of Kalman filtering, the object motion is modeled by transforming the minimum bounding rectangle information of the moving object into Kalman filter parameters. The two step block matching search method based on the minimum bounding rectangle is used to position the center of the object and then the incremental search matching method is adopted to update the object scale and direction according to the criterion of optimal scale and direction. By dynamically evaluating the confidence of different object models in different tracking scenarios, the object model is updated dynamically. Using public video sequence to test, the experimental results verify the effectiveness of the proposed method.

Key words: object tracking, scale adaptation, direction adaptation, block matching

摘要： 针对在视频序列图像目标跟踪中，跟踪目标尺寸和跟踪目标相对背景运动的方位角都在实时变化，常规目标跟踪算法会引起尺度和方向定位偏差，导致跟踪漂移，甚至跟踪失败问题，提出鲁棒的目标尺度和方向自适应的跟踪方法。在Kalman滤波框架下，通过将运动目标的最小外接矩形信息转化为Kalman滤波参数，对目标运动进行建模。采用基于最小外接矩形的两步块匹配搜索方式实现对目标的中心定位，然后采用增量式搜索匹配方法根据最优尺度和角度的判别条件修正目标尺度和方向角度。通过动态评估不同目标模型在不同跟踪场景中的置信度，对目标模型进行动态更新。使用公用视频图像序列测试，实验结果验证了该方法的有效性。

关键词: 目标跟踪, 尺度自适应, 方向自适应, 块匹配