机器视觉的车辆安全检测技术研究

发布时间：2018-04-14 08:34

本文选题：机器视觉 + SIFT特征点　；参考：《西安科技大学》2015年硕士论文

【摘要】：随着工业化的快速发展,人民生活水平的不断提高,私家车也步入普通家庭,公路交通的压力也随之而来。交通压力的剧增导致了车祸的不断增加,不但造成了人身伤害与财产的损失也使得家庭陷入巨大的困境。如何让驾驶车辆成为安全的出行方式,就成为人们追求的目标,所以对于车辆安全检测技术的研究就具有更为重要的意义。本文重点研究车辆安全距离的测量与车辆目标的追踪,安全距离是车辆行驶中保证人身安全的必要条件。目标车辆追踪以及距离测量客观的反映了当前的路况信息,该辅助技术能够有效减少交通意外的发生。论文深入分析了国内外基于视觉车辆安全检测技术的研究现状和发展。介绍了车辆主动安全系统的结构并对车辆测距以及车辆追踪进行深入的研究,重点在于安全距离的测量以及车辆目标追踪算法的研究。针对车辆测距系统,分析对比了常见的测距方法,并研究了它们各自在实际工程环境应用中的缺陷,最后采用基于SIFT特征点匹配的测距方法,在SIFT特征点匹配过程中构造新的匹配约束函数,实验结果验证该测距方法可以减少匹配时间,同时也可以减少极少数的特征点的跳变而造成的误差。针对目标车辆追踪,分析对比了各种车辆目标追踪方法,提出了将矩不变与基于更新固定模板相结合的目标追踪算法,并采用Matlab软件以及OPENCV进行编程仿真实验,通过实验可知,本文算法可以有效的改善车辆运行过程中尺度、旋转、遮挡等原因造成的追踪丢失现象,进一步验证了车辆安全距离测量以及车辆追踪的可实施性和准确性。
[Abstract]:With the rapid development of industrialization, the improvement of people's living standards, private cars have entered into ordinary families, the highway traffic pressure to follow. The traffic pressure surge has led to increasing traffic accident, not only caused personal injury and property loss also makes the family into a huge dilemma. How to make the vehicle become safe travel on the way to become a goal, so for the study of vehicle safety detection technology has a more important significance. This paper focuses on the measurement and vehicle vehicle safety distance tracking, safety distance is a necessary condition to ensure the safety of the people in the vehicle. The vehicle tracking and target distance measurement objectively reflects the current the traffic information, the auxiliary technology can effectively reduce the occurrence of traffic accidents. This paper analyzed the domestic and international vision based vehicle safety inspection The research status and development of technology. This paper introduces the structure of vehicle active safety system and the vehicle location and vehicle tracking in-depth research, focus on safety distance measurement and tracking algorithm of vehicles. The vehicle distance measuring system, analysis and comparison of the ranging method is common, and to study their own defects in the actual application environment in the project, the SIFT ranging method based on feature points matching, the SIFT feature point matching process, construct new constraint function, the experimental results verify that the measurement method can reduce the matching time, but also can reduce the error of feature points a few jump caused by the target vehicle tracking, comparative analysis the tracking method of various vehicle targets is proposed based on invariant moments and update the fixed template combination of target tracking algorithm, and uses the Matlab software to OPENCV programming and simulation experiment, experimental results show that the algorithm can effectively improve the vehicle scale, operation process and other reasons caused by rotation, occlusion tracking loss phenomenon, further validation can be implemented and the accuracy of tracking the vehicle safety and vehicle distance measurement.

【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;U463.6;TP391.41

【引证文献】