基于视频的车流量检测技术研究与实现
发布时间：2018-09-15 19:40
【摘要】：随着人们生活水平的提高,汽车逐渐走进了千家万户。汽车数量的增加,对道路交通的管理带来了挑战,但也促进了智能交通系统的发展。智能交通系统是保证道路交通实时、准确、高效的综合性管理系统,交通信息的提取又是智能交通系统中最为基础的一环。随着数字图像处理技术的发展,基于视频图像处理技术的车辆检测方法日益受到人们的关注,并且也成为了获取交通信息的主要手段。本文主要基于道路监控视频,研究道路车辆通行状况,分析车流信息,为智能交通系统的改善提供基础数据。论文主要完成的工作如下： 1)研究了车辆检测算法,实现了一种结合高斯模型背景建模与区域背景更新的背景差分法。使用混合高斯模型建立初始背景,然后进行背景差分,通过Ostu法获取自适应的阈值来分割前景,更新非前景区域的背景区域。为了消除阴影的影响,文中采用了一种融合LBP与亮度的阴影检测方法,通过LBP检测出候选阴影区域,然后通过亮度指标对阴影区域处理,最后通过连通域处理、区域生长的方法得到最终的阴影区域。实验结果表明,上述车辆检测算法的实时性明显好于混合高斯模型法,检测效果与混合高斯相近,但明显好于多帧平均法：在阴影消除方面,效果明显好于单一使用LBP或亮度的阴影消除算法。 2)在车辆跟踪研究过程中考虑车辆发生粘结的情况,研究并实现了基于轮廓凸包特性的车辆遮挡分割方法,并利用车辆的外接矩形框结合卡尔曼滤波器实现车辆的跟踪和流量提取。 3)研究了夜间车辆检测与跟踪的方法。通过采用基于灰度直方图的自适应阈值的分割法,很好地完成了车灯初始分割,再采用区域生长与形态学处理得到候选车灯,并对车灯进行配对得到车灯对的外接矩形框,最后利用外接矩形框结合卡尔曼滤波器完成夜晚车辆的跟踪与计数。实验结果表明文中算法可以很好地实现夜间车辆的检测与跟踪。 4)设计了车流量检测原型系统,给出了其主要功能模块,并利用该系统对样本视频进行实验,统计车流量数据,通过与虚拟线圈法所得结果进行比较,结果表明该系统具有较高的准确率。
[Abstract]:With the improvement of people's living standards, cars have gradually entered into thousands of households. The increase in the number of vehicles brings challenges to the management of road traffic, but also promotes the development of intelligent transportation system. Intelligent Transportation system (its) is a real-time, accurate and efficient integrated management system for road traffic, and the extraction of traffic information is the most basic part of its. With the development of digital image processing technology, vehicle detection methods based on video image processing technology have been paid more and more attention, and become the main means to obtain traffic information. Based on the video of road surveillance, this paper studies the traffic status of road vehicles, analyzes the traffic information, and provides the basic data for the improvement of intelligent transportation system. The main work of this paper is as follows: 1) the vehicle detection algorithm is studied and a background difference method based on Gao Si model background modeling and regional background updating is implemented. The mixed Gao Si model is used to establish the initial background and then the background difference is carried out. The adaptive threshold is obtained by Ostu method to segment the foreground and update the background region of the non-foreground region. In order to eliminate the influence of shadow, a shadow detection method combining LBP and brightness is adopted in this paper. The candidate shadow area is detected by LBP, then the shadow area is processed by brightness index, and finally by connected region. The method of region growth gets the final shadow area. The experimental results show that the real-time performance of the above vehicle detection algorithm is obviously better than that of the mixed Gao Si model method, and the detection effect is similar to that of the mixed Gao Si method, but it is obviously better than the multi-frame averaging method. The effect is better than the single shadow elimination algorithm using LBP or brightness. 2) the vehicle occlusion segmentation method based on contour convex hull characteristics is studied and realized by considering the case of vehicle bonding in vehicle tracking research. The vehicle tracking and flow extraction are realized by using the external rectangle frame and Kalman filter. 3) the method of vehicle detection and tracking at night is studied. By using the adaptive threshold segmentation method based on gray histogram, the initial segmentation of vehicle lamp is accomplished well, and then the candidate vehicle lamp is obtained by region growth and morphology processing, and the external rectangular frame of the vehicle lamp pair is obtained by matching the vehicle lamp. Finally, the tracking and counting of vehicles at night is accomplished by using the external rectangle frame and Kalman filter. The experimental results show that the algorithm can achieve the detection and tracking of vehicles at night. 4) the prototype system of vehicle flow detection is designed, and its main function modules are given, and the system is used to test the sample video. Compared with the results obtained by virtual coil method, the results show that the system has a high accuracy.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.116
[Abstract]:With the improvement of people's living standards, cars have gradually entered into thousands of households. The increase in the number of vehicles brings challenges to the management of road traffic, but also promotes the development of intelligent transportation system. Intelligent Transportation system (its) is a real-time, accurate and efficient integrated management system for road traffic, and the extraction of traffic information is the most basic part of its. With the development of digital image processing technology, vehicle detection methods based on video image processing technology have been paid more and more attention, and become the main means to obtain traffic information. Based on the video of road surveillance, this paper studies the traffic status of road vehicles, analyzes the traffic information, and provides the basic data for the improvement of intelligent transportation system. The main work of this paper is as follows: 1) the vehicle detection algorithm is studied and a background difference method based on Gao Si model background modeling and regional background updating is implemented. The mixed Gao Si model is used to establish the initial background and then the background difference is carried out. The adaptive threshold is obtained by Ostu method to segment the foreground and update the background region of the non-foreground region. In order to eliminate the influence of shadow, a shadow detection method combining LBP and brightness is adopted in this paper. The candidate shadow area is detected by LBP, then the shadow area is processed by brightness index, and finally by connected region. The method of region growth gets the final shadow area. The experimental results show that the real-time performance of the above vehicle detection algorithm is obviously better than that of the mixed Gao Si model method, and the detection effect is similar to that of the mixed Gao Si method, but it is obviously better than the multi-frame averaging method. The effect is better than the single shadow elimination algorithm using LBP or brightness. 2) the vehicle occlusion segmentation method based on contour convex hull characteristics is studied and realized by considering the case of vehicle bonding in vehicle tracking research. The vehicle tracking and flow extraction are realized by using the external rectangle frame and Kalman filter. 3) the method of vehicle detection and tracking at night is studied. By using the adaptive threshold segmentation method based on gray histogram, the initial segmentation of vehicle lamp is accomplished well, and then the candidate vehicle lamp is obtained by region growth and morphology processing, and the external rectangular frame of the vehicle lamp pair is obtained by matching the vehicle lamp. Finally, the tracking and counting of vehicles at night is accomplished by using the external rectangle frame and Kalman filter. The experimental results show that the algorithm can achieve the detection and tracking of vehicles at night. 4) the prototype system of vehicle flow detection is designed, and its main function modules are given, and the system is used to test the sample video. Compared with the results obtained by virtual coil method, the results show that the system has a high accuracy.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.116
【参考文献】
相关期刊论文 前10条
1 马增强;杨绍普;马新娜;;一种改进的基于轮廓特征拐点的遮挡车辆分离方法[J];北京交通大学学报;2010年05期
2 刘翠微;赵友东;梁玮;;基于时空视频块的背景建模[J];北京理工大学学报;2012年04期
3 刘天键;陈利永;詹e，
本文编号：2244309
本文链接：https://www.wllwen.com/kejilunwen/jiaotonggongchenglunwen/2244309.html
