一种基于视觉理解的交通流量统计算法  

A statistics algorithm for traffic flow based on vision understanding

作　　者：王文华[1] 王夏黎[1] 袁绍欣[1] 

机构地区：[1]长安大学信息工程学院,陕西西安710064

出　　处：《微型机与应用》2014年第18期35-38,44,共5页Microcomputer & Its Applications

基　　金：国家自然科学基金资助项目(50978030)

摘　　要：传统的基于视频的交通流量统计算法需要对流量统计区域属性及换算参数进行设置,设置方法复杂繁琐,灵活性差,应用有局限性。针对这一问题,结合人类统计目标数目的思想,提出了一种基于视觉理解的交通流量统计算法,无需设定流量统计区域就能完成流量统计。该算法使用帧间差分法检测出动态目标,通过Camshift与Kalman滤波相结合的方式完成目标跟踪,再通过对跟踪目标运动的分析来对交通流数目进行校正更新,最后获得交通流量数据。实验表明,提出的算法具有较好的统计效果。The traditional statistics algorithm for traffic flow based on video needs settings for the properties of the statistics area and the conversion parameters, which is complex, inferior in flexibility and also has a limitation. This paper proposes an algorithm of counting the traffic flow based on vision understanding, combining human′s thoughts of counting the number of targets, without setting the traffic flow counting area. In the algorithm, the frame difference method helps to detect the moving targets, and the combination of Camshift and Kalman filter can accomplish the tracking. Then we can correct the counting result and provide the traffic flow data as well through the analysis of the targets tracking. The result of the experiment shows that the algorithm functions well.

关 键 词：图像处理 帧间差分 视觉理解 交通流量统计 目标跟踪 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]