CAJ | 학술논문

提出了一种基于视频的车流量自动检测方法,首先用高斯平均法对一定数量视频帧图像进行训练得到初始背景,然后用基于卡尔曼滤波的背景差法提取前景以及背景更新,最后用双虚拟检测线法进行车辆计数。通过对实际道路车辆视频文件进行试验,检测准确率达到90%,满足实际应用要求。
제출료일충기우시빈적차류량자동검측방법,수선용고사평균법대일정수량시빈정도상진행훈련득도초시배경,연후용기우잡이만려파적배경차법제취전경이급배경경신,최후용쌍허의검측선법진행차량계수。통과대실제도로차량시빈문건진행시험,검측준학솔체도90%,만족실제응용요구。
Vehicle flow detection is an important part of ITS.In this paper,we proposed a video-based vehicle flow auto-detection method through experiments.First,we took and trained a certain number of video frame images to get the initial background by RGA.Second,we drew the foreground and updated the background by using the background subtraction based on Kalman filter. Finally,we calculated the vehicles by using double virtual detection line.Through the experiment on actual vehicle video files,we found that the accuracy is up to 90%,which is believed to meet the practical application.