基于视频流的地铁人群目标识别

Recognizing the Passenger Number in Subway Station by Video Flow

地铁站内人群数目及分布的在线监测是有效控制和疏散客流,保障地铁安全的重要依据之一.利用站内现有的闭路电视监视系统,通过计算机视觉技术实现人群数目的自动识别是目前国外普遍采用的一种方式.文中提出了一种结合了自适应背景差分和比例自适应模板匹配的算法,用背景减除对图像进行分割,然后再利用比例自适应模板对感兴趣的区域进行搜索匹配,识别目标人群的数目及位置分布.该算法能有效减少传统模板匹配的计算量,提高匹配的准确率,在一定误差范围内可以达到较好的效果.

The number and distribution of passengers in the subway station is one of the key factors in passengers＇ evacuation and subway security. The automatic recognition of passengers＇ number can be realized by digital image process using the current CCTV system in subway station. A method based on background difference and scale adaptive templates matching is proposed in this paper. First images are segmented by background abstraction and then multi-templates matching is operated only in the interesting area. By this method, the calculate capacity of traditional matching can be reduced, and the veracity can be improved. The experimental data indicate that this method can achieve recognition aims within the range of given errors.