交通视频中的车辆计数方法研究  

作　　者：孙维广[1] 于子钧[1] 

机构地区：[1]沈阳理工大学信息科学与工程学院,辽宁沈阳110159

出　　处：《科技创新与应用》2018年第24期118-120,124,共4页Technology Innovation and Application

摘　　要：对交通视频中的车辆计数方法进行了研究。首先采用背景更新的背景差分法获得无运动目标的背景,并采用帧差法检测是否存在运动目标,以此作为背景更新的依据。然后当前帧与背景图像做差值,并进行二值化提取运动目标。对于光线较暗车辆目标提取不完整和由于光照产生伪目标的情况,采用一种基于虚拟的横向检测线和纵向检测线相结合的检测方法计数。实验表明,本方法可较准确的实现车辆计数,并有较好的实时性。The vehicle counting method in traffic video is studied in this paper. Firstly, background subtraction method of back-ground updating is used to obtain the background without moving objects, and frame difference method is used to detect the exis-tence of moving objects, which is used as the basis of background updating. Then the difference between the current frame and thebackground image is done, and the moving object is extracted by binarization. In the case of incomplete extraction of dim vehicletargets and false targets due to illumination, a detection method based on the combination of horizontal detection lines and longitudi-nal detection lines is used to count the targets. The experiment results show that the method can realize vehicle counting accuratelyand has better real-time performance.

关 键 词：背景差分 二值化 车辆计数 虚拟检测线 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]