一种基于计算机视觉的行人流量统计方法  

A pedestrian traffic statistics method based on computer vision

作　　者：肖江[1] 丁亮[1] 束鑫[1] 张文章[1] 

机构地区：[1]江苏科技大学计算机科学与工程学院,江苏镇江212003

出　　处：《信息技术》2015年第8期22-25,共4页Information Technology

基　　金：江苏省自然科学基金(BK20130473);镇江市社会发展科技支撑项目(SH2013015)

摘　　要：提出了一种统计监控视频中行人流量的方法。文中方法利用背景差法提取运动物体,并对运动物体进行跟踪。提取物体的HOG特征,并利用线性SVM判决提取的HOG特征是否符合行人要求,以达到统计视频中行人流量的目的。实验表明在人流量不是特别密集的场景,可以准确地统计视频中的行人数量。A new method for counting pedestrian flow based on video monitoring system was proposed in this paper. This method can adaptively update the background, extracting moving objects using background difference method,and tracking moving objects. Finally extracting HOG features of the object,using Linear SVM judges the HOG feature if it meets the requirements of pedestrians or not,in order to achieve the purpose of counting pedestrian flow. The experiments show that this method can accurately statistics the number of pedestrian if the traffic was not particularly congestion.

关 键 词：HOG特征 SVM 运动物体跟踪 人流量统计 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]