基于机器视觉的前方车辆检测与测距系统设计

发布时间：2018-05-28 00:30

本文选题：智能车辆 + 自适应阈值分割　；参考：《哈尔滨工业大学》2015年硕士论文

【摘要】：随着我国高速公路建设的飞速发展和人均汽车占有量的日益增加,交通安全方面的问题已经成为不可忽视的问题。因此,能够对前方车辆进行实时检测和测距是车辆安全行驶和自主导航的重要措施,也是智能交通系统研究领域的热点之一。虽然目前对车辆的检测有很多经典的检测算法,如,帧差法,光流法等,但是这些检测算法不太适合检测前方运动中的车辆。论文根据前方车底阴影的特性提出了一种有效的检测算法,并且通过比较基于不同传感器的测距方法的优缺点选择了基于机器视觉(单双目)的测距方法,实现对前方最近车辆距离的测量,为驾驶员提供准确的辅助驾驶信息。论文主要完成了以下工作:(1)介绍了各种不同方式的车辆检测算法,分析各种算法的适用场景及优缺点。阐述了基于红外线、超声波等传感器前方车辆测距的弊端,突出了机器视觉测距技术的应用前景。(2)根据车辆底部始终稳定的存在阴影这一特性,采用两次自适应阈值分割算法把车辆从复杂的实际背景下分割出来,并生成车辆假设区域。根据车辆尾部的对称性,采用sobel算子的边缘检测计算其对称性测度,以便过滤掉已生成的虚假车辆,保留真实车辆。(3)研究摄像机成像原理,利用张正友平面标定法对摄像机内部参数进行标定,然后分别采用单目视觉测距法和双目视差测距法对前方车辆实时测距,并且分析比较两种方式的测量结果。(4)在Lab VIEW平台上开发基于机器视觉的前方车辆检测与测距系统软件,利用Lab VIEW的计算机视觉库完成系统各个模块的开发,并在实际的道路环境中进行系统实验,实验结果表明,本系统能够较准确的检测到前方车辆,并且两车之间的距离最近可以测到10米以内,最远可以测到50米左右,满足安全车距50米的要求。
[Abstract]:With the rapid development of highway construction and the increase of per capita vehicle possession, traffic safety has become a problem that can not be ignored. Therefore, it is an important measure for vehicle safety and autonomous navigation to be able to detect and range the vehicle in real time. It is also one of the hot spots in the research field of Intelligent Transportation system (its). Although there are many classical detection algorithms for vehicle detection, such as frame difference method, optical flow method and so on, these detection algorithms are not suitable for detecting vehicles in the front motion. In this paper, an effective detection algorithm is proposed according to the characteristics of the shadow of the front car bottom, and by comparing the advantages and disadvantages of the ranging methods based on different sensors, the ranging method based on machine vision (mono-binocular) is selected. To realize the distance measurement of the nearest vehicle in front and provide the accurate auxiliary driving information for the driver. The main work of this paper is as follows: 1) introduce various vehicle detection algorithms, analyze the applicable scenarios, advantages and disadvantages of the algorithms. Based on infrared, ultrasonic and other sensors, the disadvantages of vehicle ranging in front of vehicle are expounded, and the application prospect of ranging technology of machine vision is highlighted. (2) according to the characteristic that the shadow always exists in the bottom of the vehicle, The two-time adaptive threshold segmentation algorithm is used to segment the vehicle from the complex background and generate the vehicle hypothesis region. According to the symmetry of the vehicle tail, the symmetry measure is calculated by using the edge detection of the sobel operator in order to filter out the generated false vehicle and retain the real vehicle. The camera internal parameters are calibrated by means of Zhang Zhengyou plane calibration method, and then the single vision ranging method and the binocular parallax ranging method are used to measure the real time range of the vehicle in front. The software of front vehicle detection and ranging system based on machine vision is developed on the platform of Lab VIEW. The computer vision library of Lab VIEW is used to complete the development of each module of the system. The experimental results show that the system can accurately detect the front vehicle, and the distance between the two vehicles can be measured within 10 meters recently, and the farthest distance can be measured to about 50 meters. Meet the requirements of safety vehicle distance of 50 meters.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;U463.6;TP391.41

【参考文献】