基于监控视频图像的交通事故车速计算方法研究
发布时间：2018-10-24 20:24
【摘要】：随着汽车产业迅猛发展,汽车保有量也在增加,造成道路交通事故频发,给社会的稳定和人民的生命财产安全带来了很大威胁。在注重预防交通事故发生的前提下,有必要对交通事故鉴定结果的精准性、高效性做进一步的研究。随着监控录像在道路上的广泛应用,众多有价值的道路交通信息被永久储存下来,因而基于监控视频图像完成事故鉴定已成为鉴定人员的重要选择。录像设备、成像系统特性等因素会使视频图像在生成过程中出现空间距离与图像距离不能精确对应的现象。为此,图像空间需进行直接线性变换处理,即摄像机的标定过程。此外,在目标车辆检测与跟踪方面,由于存在阴影干扰、光线变化等,使得检测与跟踪结果不甚理想。总之,基于监控视频图像计算车速的通用方法存在不可消除的缺陷,无法为交通事故鉴定工作提供准确的结果,只能作为辅助证据。以监控视频图像为研究对象,运用摄影学与运动学原理,在摄像机未标定的前提下,应用插值法建立了车速计算数学模型。基于Matlab-GUI环境开发可视化系统平台,从而快速、可靠地计算出交通事故中车辆的速度。根据时间-速度曲线得到目标车辆的运动方程,明确了车辆在交通事故发生前和发生瞬间的运动状态(路径、位置和速度等)。应用插值法确定了目标车辆的距离和时间。对视频图像“近大远小”的透视关系进行了描述,同时设计了相应的验证试验。试验结果表明,在车辆1帧时间内行驶的距离中,可忽略这种透视关系的影响。对视频测速原理进行了描述,将插值法应用到车速计算数学模型的建立中,以达到不对摄像机进行标定也可使计算结果精度满足要求的目的。对比了三种基于插值原理的车速计算数学模型,并对每种模型的适用情况进行了分析。对基于监控视频图像的交通事故车速计算系统进行了设计与开发,开发平台选择Matlab环境。按照交通事故车速计算系统的功能要求,分别进行了系统总体设计、模块设计及工作流程的开发。结合视频测速案例证明了该系统在交通事故车速计算中的适用性。分析了车辆行驶速度、标尺长度与摄像机拍摄角等三个因素对交通事故车速计算系统精度的影响,并提出了减小误差的方法,以便客观公正地反映车辆的实际行驶状态。
[Abstract]:With the rapid development of automobile industry, the number of vehicle ownership is also increasing, resulting in frequent road traffic accidents, to the stability of society and the safety of people's lives and property brought a great threat. On the premise of preventing traffic accidents, it is necessary to make further research on the accuracy and efficiency of traffic accident identification results. With the wide application of surveillance video on the road, many valuable road traffic information are stored permanently, so it has become an important choice for appraisers to complete accident identification based on surveillance video images. Video recording equipment, imaging system characteristics and other factors will make the video image in the process of generating space distance and image distance can not accurately correspond to the phenomenon. Therefore, the image space needs direct linear transformation, that is, the camera calibration process. In addition, in the aspect of vehicle detection and tracking, the detection and tracking results are not satisfactory due to the shadow interference and light change. In a word, the general method of calculating vehicle speed based on video surveillance images has some defects, which can not provide accurate results for traffic accident identification, and can only be used as auxiliary evidence. Based on the principle of photography and kinematics, the mathematical model of speed calculation is established by using interpolation method on the premise of camera uncalibrated. The visual system platform is developed based on Matlab-GUI environment, so that the speed of vehicle in traffic accident can be calculated quickly and reliably. According to the time-velocity curve, the motion equation of the target vehicle is obtained, and the moving state (path, position, velocity, etc.) of the vehicle before and at the moment of the traffic accident is determined. The distance and time of the target vehicle are determined by interpolation method. This paper describes the perspective relation of video image "near, far and small", and designs the corresponding verification experiment. The experimental results show that the influence of this perspective relationship can be neglected in the range of vehicle running in 1 frame time. The principle of video velocimetry is described and the interpolation method is applied to the establishment of mathematical model of speed calculation so as to achieve the purpose of not calibrating the camera and making the accuracy of the calculation results meet the requirements. Three mathematical models of speed calculation based on interpolation principle are compared, and the applicability of each model is analyzed. The traffic accident speed calculation system based on video surveillance image is designed and developed. The Matlab environment is chosen as the development platform. According to the function requirements of the traffic accident speed calculation system, the overall system design, module design and workflow development are carried out respectively. The feasibility of the system in the calculation of traffic accident speed is proved by a video speed measurement case. This paper analyzes the influence of three factors, such as vehicle speed, ruler length and camera shooting angle, on the accuracy of traffic accident speed calculation system, and puts forward a method to reduce the error in order to objectively and fairly reflect the actual driving state of the vehicle.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491.31
[Abstract]:With the rapid development of automobile industry, the number of vehicle ownership is also increasing, resulting in frequent road traffic accidents, to the stability of society and the safety of people's lives and property brought a great threat. On the premise of preventing traffic accidents, it is necessary to make further research on the accuracy and efficiency of traffic accident identification results. With the wide application of surveillance video on the road, many valuable road traffic information are stored permanently, so it has become an important choice for appraisers to complete accident identification based on surveillance video images. Video recording equipment, imaging system characteristics and other factors will make the video image in the process of generating space distance and image distance can not accurately correspond to the phenomenon. Therefore, the image space needs direct linear transformation, that is, the camera calibration process. In addition, in the aspect of vehicle detection and tracking, the detection and tracking results are not satisfactory due to the shadow interference and light change. In a word, the general method of calculating vehicle speed based on video surveillance images has some defects, which can not provide accurate results for traffic accident identification, and can only be used as auxiliary evidence. Based on the principle of photography and kinematics, the mathematical model of speed calculation is established by using interpolation method on the premise of camera uncalibrated. The visual system platform is developed based on Matlab-GUI environment, so that the speed of vehicle in traffic accident can be calculated quickly and reliably. According to the time-velocity curve, the motion equation of the target vehicle is obtained, and the moving state (path, position, velocity, etc.) of the vehicle before and at the moment of the traffic accident is determined. The distance and time of the target vehicle are determined by interpolation method. This paper describes the perspective relation of video image "near, far and small", and designs the corresponding verification experiment. The experimental results show that the influence of this perspective relationship can be neglected in the range of vehicle running in 1 frame time. The principle of video velocimetry is described and the interpolation method is applied to the establishment of mathematical model of speed calculation so as to achieve the purpose of not calibrating the camera and making the accuracy of the calculation results meet the requirements. Three mathematical models of speed calculation based on interpolation principle are compared, and the applicability of each model is analyzed. The traffic accident speed calculation system based on video surveillance image is designed and developed. The Matlab environment is chosen as the development platform. According to the function requirements of the traffic accident speed calculation system, the overall system design, module design and workflow development are carried out respectively. The feasibility of the system in the calculation of traffic accident speed is proved by a video speed measurement case. This paper analyzes the influence of three factors, such as vehicle speed, ruler length and camera shooting angle, on the accuracy of traffic accident speed calculation system, and puts forward a method to reduce the error in order to objectively and fairly reflect the actual driving state of the vehicle.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491.31
【相似文献】
相关期刊论文 前10条
1 周清华;周俊;林必毅;;监控视频流传输的应用研究[J];智能建筑与城市信息;2009年08期
2 谢俊宏;;浅谈监控视频干扰问题的解决[J];海峡科学;2008年08期
3 欧温暖;田绪红;朱同林;;基于教室监控视频的课堂行为计数分析[J];图学学报;2012年02期
4 李勃;陈启美;;基于监控视频的运动车辆行为分析算法[J];仪器仪表学报;2006年S3期
5 梁浩哲;徐树奎;李国辉;张军;;面向监控视频的行为模式挖掘[J];山东大学学报(工学版);2011年06期
6 孙晋明;;远程监控视频无线传输方式的比选设计[J];电气时代;2011年05期
7 姜振凤;迟庆云;;基于DBAN的监控视频数据挖掘[J];河北工程大学学报(自然科学版);2013年02期
8 祝林峰;方思平;张兴明;张平;马旭;;监控视频中的车体特征检验在交通肇事逃逸案件中的应用[J];警察技术;2013年06期
9 胡平;;雾霾天气下高速公路监控视频的一种增强算法[J];公路交通科技;2014年01期
10 闫帅辉;柏正尧;尹立国;杨仁青;高v，
本文编号：2292457
本文链接：https://www.wllwen.com/kejilunwen/daoluqiaoliang/2292457.html
