基于视觉技术的道路交通信息提取方法研究

发布时间：2019-02-16 10:34

【摘要】：为了解决城市交通快速发展所衍生的各种问题,智能交通系统成为国内外研究的重点。全面、准确、实时的交通信息能够为智能交通系统建设提供数据支持,是交通疏导、路网规划以及行人制定出行路线的决策依据,如何快速、有效的进行交通信息提取是制约智能交通系统发展进程的关键因素。本文重点针对交通信息的提取方法展开研究。与传统的交通信息提取方法相比,基于计算机视觉技术的交通信息提取方法因其具备设备安装维护方便、成本低等优点,成为了智能交通领域的热门课题。但由于光照产生的车辆阴影以及车辆检测中的鬼影会极大地降低检测精度,信息提取方法中常用的虚拟线圈大多需手动设置且参数难以确定等问题,使基于视觉技术的交通信息提取方法应用仍有一定局限性。本文针对以上缺陷与不足展开研究,主要成果如下:(1)针对车辆检测中的阴影问题,提出了一种基于主成分分析法的交通视频车辆阴影消除算法。该算法具有较高的鲁棒性,对交通场景无特殊要求,无需前期训练和人工干预;引入主成分分析法,极大的降低了运算复杂度;与传统阴影消除算法相比较,本文算法阴影消除综合指数提高10%以上,运算效率提高30%以上。(2)针对车辆检测中的鬼影问题,以实时性较高的ViBe算法为基础,提出了一种V-ViBe算法。该算法通过构造“虚拟”背景图像,改变传统ViBe算法初始背景模型建立方法,从源头上抑制鬼影的产生;利用形态学相关知识完善检测目标;实验表明,本文算法精确度、召回率、误检率等六项性能指标均优于原ViBe算法。(3)在信息提取阶段,利用车道线颜色在(4(7空间的突出特征以及霍夫变换原理提取车道线;根据车道线在图像中的形变系数设置与车道形状相吻合的虚拟线圈;结合本文车辆检测算法,提取交通监控视频交通信息参数。
[Abstract]:In order to solve the problems arising from the rapid development of urban traffic, intelligent transportation system has become the focus of research at home and abroad. Comprehensive, accurate and real-time traffic information can provide data support for the construction of intelligent transportation system. It is the basis for traffic dredge, road network planning and pedestrian route decision. The effective extraction of traffic information is a key factor restricting the development of intelligent transportation system (its). This paper focuses on traffic information extraction methods. Compared with the traditional traffic information extraction method, the traffic information extraction method based on computer vision technology has become a hot topic in the field of intelligent transportation because of its advantages of convenient installation and maintenance of equipment and low cost. However, because of the vehicle shadow caused by illumination and the ghost image in vehicle detection, the detection accuracy will be greatly reduced. Most of the virtual coils commonly used in information extraction methods need to be manually set and the parameters are difficult to determine. The application of traffic information extraction method based on visual technology is still limited. The main achievements of this paper are as follows: (1) aiming at the shadow problem in vehicle detection, this paper proposes a shadow cancellation algorithm for traffic video vehicles based on principal component analysis (PCA). The algorithm has high robustness, no special requirements for traffic scenes, no pre-training and manual intervention, the introduction of principal component analysis (PCA) greatly reduces the computational complexity. Compared with the traditional shadow cancellation algorithm, the comprehensive index of shadow cancellation in this algorithm is increased by more than 10%, and the computational efficiency is increased by more than 30%. (2) aiming at the ghost image problem in vehicle detection, the algorithm is based on the real-time ViBe algorithm. A V-ViBe algorithm is proposed. By constructing a "virtual" background image, the algorithm changes the original background model of the traditional ViBe algorithm to suppress the generation of ghost images from the source, and uses morphological knowledge to perfect the detection target. Experimental results show that the performance of this algorithm is better than that of the original ViBe algorithm. (3) in the stage of information extraction, the performance of this algorithm is better than that of the original ViBe algorithm. Using the prominent feature of lane color in (4) (7) space and the principle of Hough transform to extract lane line; According to the deformation coefficient of the lane line in the image set the virtual coil which coincides with the shape of the lane. Combined with the vehicle detection algorithm in this paper the traffic information parameters of the traffic surveillance video are extracted.
【学位授予单位】：山东理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U495

【参考文献】