Abstract: For solving the problem that tracking algorithms with fixed tracking windows can not track an object which changes size during motion efficiently, this paper proposes a tracking method with an adaptive tracking window based on maximum average posterior probability indicator. On the basis of visual tracking algorithm on maximum posterior probability, this paper analyzes the posterior probability indicator of moving target when the tracking window largens or diminishes, and according to the average indicator in these windows, this method decides the update strategy of tracking window. The experimental results manifest that the method proposed in this paper can adjust the tracking window of rigid objects and nonrigid objects well, and track them steadily.

Key words: visual tracking, average posterior probability indicator, adaptive window

摘要： 针对跟踪窗口固定的跟踪算法不能有效地跟踪存在明显尺度变化的目标这一问题，提出了一种基于最大平均后验概率指标的自适应窗口目标跟踪算法。在最大后验概率视觉跟踪算法基础上，分析了运动目标窗口增大以及减小时的后验概率指标，根据这些窗口内的平均指标量来决定目标跟踪窗口的更新策略。实验结果表明，该算法对刚体及非刚体运动目标的跟踪窗口都能较好地进行调整，实现稳健的跟踪。

关键词: 视觉跟踪, 平均后验概率指标, 自适应窗口