视频多目标遮挡中人体目标跟踪

Human Tracking in Multi-object Occlusion of Video

提出一种视频多目标中跟踪人体目标的有效方法.在单摄像头对固定场景监控中,首先通过彩色空间的降维,根据置信度更新背景模型,消除时间变化引起的阳光,天气等因素对背景图像的影响;然后使用背景减除获得多目标模板,分析多目标遮挡问题,利用人体目标的多子块灰度特征匹配得多目标遮挡中人体目标的跟踪;实验证明这种方法可有效地用于视频的人体跟踪中.

The paper suggests an efficiency method of human tracking in multi-object of video. monitor of single - camera. First, when the dimension of color space is lowered and backgroun In fixing scene d model is updated on likelihood, the method eliminates effect in sunlight change and weather factors. Second, the method is to get multi - object template from background subtracting,to analyze occlusion question of multi-object, to obtain human tracking in multi - object occlusion through multi - block intensity feature of human object. The experiment shows the method is good performances for human tracking in video.