摘要
视频分析在客流统计上的应用是当前研究的热点及难点,常见的视频计数实现过程主要是依靠对象分割、目标跟踪、目标识别等尚未成熟的算法,尤其要实现目标的准确跟踪仍非常困难。为避免目标跟踪算法对计数精度的影响,文章在对象分割算法的基础上统计目标像素,再采用提出基于加权步长的波峰识别搜索算法对像素统计曲线进行识别,从而实现对乘客目标计数;最后以长途大巴车视频客流计数问题为背景,结合提出的计数算法设计一套完整的视频客流计数系统。实验结果表明,文中设计的客流计数系统能够对长途大巴车乘客有秩序上(下)车的视频序列准确计数,计数精度达到90%以上,并具有超载报警和智能结算等功能。
Video analysis on the passenger flow statistics is current research hotspot and difficulty, and the common video counting implementation process is mainly rely on the object segmentation, target tracking and target recognition which are not yet mature algorithms. Especially, it is very difficult to track the target accurately. In order to avoid the influence of target tracking algorithm on the counting accuracy, in this paper, the video of counting algorithm statistics target pixels that are based on the object segmentation algorithm.For achieving the passengers counting, the peak identification search algorithm which is based on weighted step wave is proposed, in order to identify the statistical curve of pixels. Finally, a set of complete video passenger flow counting system which is under the background of the long distance bus video passenger flow counting problem and combining with the count of algorithm are designed. Experimental results show that the design of passenger flow counting system could count accurately on the video sequence of the long distance bus in which passengers orderly on (off) the car, and the counting precision is above 90%, also functions as overload alarm, intelligent settlement and so on are added.
出处
《现代交通技术》
2015年第6期76-80,共5页
Modern Transportation Technology
关键词
客流计数
目标检测
波峰识别
视频分析
passenger counting
target detection
peak identification
video analysis