基于IFGT的人体目标运动blob自动分割

An Automatic Human Moving Blob Segmentation Based on IFGT

为了自动地初始化运动人体的跟踪,采用人体b lob模型,提出了一种基于IFGT的人体目标运动b lob的自动分割方法。该方法通过区域分割预分割运动b lob,然后利用基于IFGT的核密度估计方法更新人体模型中的颜色密度,最后采用似然极大原则分类不同运动b lob。该方法对于直立运动人体的头部、躯干、下身的分类效果较好,计算复杂度较低。

In order to initiate moving human tracking, an IFGT based human moving blob segmentation approach is put forward. At first, moving blobs are segmented by C mean algorithm. Then color density of human model is estimated by IFGT based kernel density estimation method. At last, different moving blobs are classified by maximum likelihood method. The approach has good experiment effect on standing human segmentation and has lower cost.