自适应特征融合跟踪在进出人数统计中的应用

传统的Mean Shift算法在背景图像中出现与目标相似物体时容易发生错误,本文将HSV颜色分布和局部二元模式作为观测模型,进行直方图建模,不仅在跟踪过程中为了避免背景干扰,将各个特征融合在一起并通过其与目标模型的相似度动态更新特征的权值,并在此基础构建自动人数统计系统。试验结果表明它能在提取出运动目标后对运动目标进行较为准确的跟踪,并对出入视频场景的人数进行统计,具有较高的正确性和有效性。

An adaptive multi-feature-fused tracking algorithm for people counting application

An improved tracking algorithm is presented to solve the problem that traditional algorithm based on mean shift often causes mistakes if there are similar object appears.HSV color histograms and local binary pattern as the observation model are regards.In order to avoid interference during the tracking,the weights of features are modified according to the similarity between target and model.Then we an automated people counting system is build on that basis.Experimental results show that the proposed algorit...