视频车辆检测与预警算法的研究与DSP实现

发布时间：2018-06-25 10:07

本文选题：车辆检测 + 预警　；参考：《东北大学》2014年硕士论文

【摘要】：近年来,随着交通的发展和车辆的持续增多,交通安全事故频频发生,安全驾驶已成为日益突出的问题。智能交通系统(Intelligent Transportation Systems, ITS)与车辆辅助驾驶系统(Driver Assistance Systems, DAS)成为国内外研究的热点。其中,前方车辆的实时检测技术是车辆辅助驾驶系统中的关键内容之一,对于实现停车辅助、碰撞避免等具有十分重要的意义。本文设计了一种基于车牌的动静态车辆检测算法,开发了实时车辆距离预警系统。综合利用车辆的车牌特征、纹理特征、灰度对称性特征以及统计特征来检测前方车辆,并将该算法在DSP应用板中编程实现。在算法研究方面,本文利用实际生活中最为常见的蓝色和黄色车牌在YUV颜色空间的聚类性,实现了车牌提取,利用腐蚀、膨胀的开闭运算等形态学处理方法对车牌图像进行了降噪和修复,并结合连通域标记方法实现了对多个车牌的标记和定位,利用基于知识的纹理及灰度对称性特征完成对非车辆区域的排除。为解决车辆遮挡问题,本文设计了构造车辆局部特征子空间的方法,并采用二维主成分分析(Two-Dimensional Principal Pomponent Analysis,2DPCA)与最小距离分类器进行车辆的验证,实现车辆检测；最后,本文建立了车距预警模型,结合检测算法,实现了近距离车辆的警示,并对警示车辆的左、右侧方位进行判别。本文首先在MATLAB平台上完成了视频车辆检测算法的研究和仿真测试。之后,在硬件实现方面,分析了系统硬件需求,选取ICETEK-DM6437-B评估板作为硬件平台,选用CCD摄像头作为视频采集工具；其次,在车牌提取与初步定位阶段,将图像进行抽行抽列处理,减少了DSP处理的数据量,提高系统检测效率；最后,将算法移植到CCS集成开发环境中,利用C语言编程并进行了代码的优化,完成视频车辆的检测与车距预警。实验结果表明,本文设计的车辆检测与预警算法检测率可达90%以上,具有一定的鲁棒性和实用性。经过代码优化,效率得到了显著提升,每秒钟能完成大约三帧数据的处理,具有较好的实时性。并且该算法能有效的检测存在遮挡的车辆,减小光照等外界条件的影响。
[Abstract]:In recent years, with the development of traffic and the continuous increase of vehicles, traffic safety accidents occur frequently, safety driving has become an increasingly prominent problem. Intelligent Transportation Systems (its) and driver Assistance Systems (Das) have become the focus of research at home and abroad. Among them, the real-time detection technology of the vehicle in front is one of the key contents in the vehicle assisted driving system, which is of great significance for the realization of parking assistance and collision avoidance. In this paper, a vehicle detection algorithm based on vehicle license plate is designed, and a real-time vehicle distance warning system is developed. Vehicle license plate features, texture features, gray symmetry features and statistical features are used to detect the vehicle in front of the vehicle, and the algorithm is implemented in DSP application board. In the aspect of algorithm research, this paper makes use of the clustering of blue and yellow license plates in YUV color space, which is the most common in real life, and realizes license plate extraction and corrosion. The image of license plate is de-noised and repaired by morphological processing, such as expansive open and close operation, and the multiple license plates are marked and located with the method of connected domain marking. The knowledge-based texture and gray-scale symmetry feature are used to eliminate the non-vehicle region. In order to solve the problem of vehicle occlusion, a method of constructing vehicle local feature subspace is designed, and the Two-dimensional Principal component Analysis (2DPCA) and minimum distance classifier (2DPCA) are used to verify the vehicle to achieve vehicle detection. In this paper, the warning model of vehicle distance is established, combined with detection algorithm, the warning of close distance vehicle is realized, and the left and right azimuth of warning vehicle are distinguished. Firstly, the research and simulation test of video vehicle detection algorithm are completed on MATLAB platform. Then, the hardware requirements of the system are analyzed, ICETEK-DM6437-B evaluation board is selected as the hardware platform, CCD camera is selected as the video capture tool. Secondly, the image is extracted and processed in the stage of license plate extraction and initial location. Finally, the algorithm is transplanted into the CCS integrated development environment, and the code is optimized by C language to complete the video vehicle detection and distance warning. The experimental results show that the detection rate of the vehicle detection and early warning algorithm designed in this paper can reach more than 90%, which is robust and practical. After code optimization, the efficiency is greatly improved, about three frames of data can be processed per second, and it has good real-time performance. And the algorithm can effectively detect the presence of occlusion vehicles, reduce the impact of light and other external conditions.
【学位授予单位】：东北大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U495;TP391.41

【相似文献】