摘要
为了解决行人交通参数获取困难、精度低等问题,需要开发出一套高精度、高效率和强抗干扰性的行人交通参数提取系统。首先,基于视频流中运动目标时空一致性的原理,采用运动目标梯度方向直方图(HOG)特征提取算法对视频流进行分割;其次,对运动目标特征进行分析并提取特征向量,建立反向传播(BP)神经网络对运动目标进行训练和分类,实现对运动目标的检测和识别;然后,利用MATLAB软件平台,开发了能够对行人流量、速度和时距等数据进行快速处理与准确分析的系统。最后,进行了实例测试。结果表明:检测系统的参数提取精度能达到90%以上,高于现有算法的提取精度。同时,系统有较强抗环境干扰性,提取效率和数据处理功能效果较理想。
For solving the problem of pedestrian traffic parameter difficulties acquisition and low accuracy,it is urgent to develop a pedestrian traffic parameter extraction system with high precision,high efficiency and strong anti-jamming.First,basing on the moving target space-time consistency principle of the video stream,the video stream is segmented by moving the target gradient direction histogram HOG feature extraction algorithm.Secondly,analyzing moving target characteristics,extracting feature vectors,and establishing BP neural network to train and classify the targets,the system can achieve the identification and detection of moving targets;Third,using MATLAB software platform,the system is developed and realized that can carry on quickly processing and accurately analyzing data of the pedestrian flow,the speed and the headway.Finally,the experimental results show that the accuracy of the extraction system can reach more than 90%,which is higher than the existing algorithm.At the same time,the system has strong anti-jamming interference,and has ideal effect extraction efficiency and data processing function.
作者
罗强
吴淑博
臧晓冬
许伦辉
LUO Qiang;WU Shu-bo;ZANG Xiao-dong;XU Lun-hui(School of Civil Engineering,Guangzhou University,Guangzhou 510006,China;School of Civil Engineering and Transportation,South China University of Technology,Guangzhou 510640,China)
出处
《科学技术与工程》
北大核心
2018年第9期343-347,共5页
Science Technology and Engineering
基金
国家自然科学基金(61263024)
广东省教育厅青年创新人才类项目(2015KQNCX121)
广州市属高校科研项目(1201630172)
广东省自然科基金(2014A030310498)资助
关键词
交通工程
行人交通
视频识别
动态跟踪
数据分析
traffic engineering
pedestrian traffic
video recognition
dynamic tracking
data analysis
