基于视频的多信息融合公交客流采集方法研究  

Research on head counting method based on improved hough transform and multi-feature

作　　者：苏彦莽[1] 王彦[2] 郭志涛[1] 闫林[1] 

机构地区：[1]河北工业大学信息工程学院,天津300401 [2]天津职业大学电子信息工程学院,天津300410

出　　处：《河北工业大学学报》2013年第4期23-27,共5页Journal of Hebei University of Technology

基　　金：天津市应用基础及前沿技术研究计划(11JCYBJC00200);河北省高等学校科学技术研究青年基金(Q2012012)

摘　　要：公交车辆的客流信息采集对于公交公司的营运、调度有着十分重要的意义.为此,提出了一种基于视频的多信息融化客流统计算法.该算法首先通过背景法提取目标,采用改进的Hough变换实现乘客头部的快速检测,继而通过人头轮廓的类圆特征以及颜色特征相融合实现目标定位,最后采用运动目标连续性匹配策略对乘客运动进行跟踪,从而实现客流的准确统计.实验结果表明,该方法可快速统计出通过目标区域的人数,时效性强,适合公交领域应用.It is very important for bus company's operations and scheduling to get the transit vehicle passenger information. Therefore, a new video-based traffic statistics algorithm with multi-information is proposed. Firstly, the target is extracted by the background method. Secondly, the transform passenger head is detected rapidly with the improved Hough algorithm, and the target locating is realized by the class round feature of head contour and the feature of head color.Finally, using the outline feature and color feature to identify head, regarding the circle of head as the moving targets and using the continuity of moving targets to do the head match and track and then implementing the accurate statistics. Large numbers of experiment results show that this method can accurate count the continuous passenger, have strong timeliness and can rapidly count the number of people passing the target area.

关 键 词：HOUGH变换 多特征 人头检测 人数统计 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]