基于视频的交通路口车辆计算研究

发布时间：2018-05-01 22:09

本文选题：车辆检测 + 车辆目标分割　；参考：《昆明理工大学》2014年硕士论文

【摘要】：随着城市的发展,车路之间的矛盾日益突出。因此,如何充分利用现有资源,实现智能交通系统,是一个亟待解决的问题。智能交通信号灯控制系统作为智能交通系统的重要组成部分,在减少车辆在路口的总体通行时间,提高通行效率,降低燃油消耗与环境污染方面发挥着不可替代的作用。基于视频的车辆检测与计算问题作为智能交通信号灯控制系统等视觉应用的重要基础,具有十分重要的研究意义与实用价值。目前,路口车辆的测量检测与计算主要通过路面接触式的物理探测装置,诸如压力、电场或磁场等感应装置来检测与计算。但是这类装置的安装维护困难,且易老化,维护成本较高,而且不能解决车辆改道问题。随着视频检测技术的发展,基于视频的非接触式车辆检测与计算方法,由于其检测设备安装、维护简单,维护成本低,且能解决多车道与车辆改道问题,因此,其必将逐步替代传统的车辆检测方法,成为未来的研究重点。然而,基于视频的车辆检测技术仍面临着动态背景干扰、静态车辆目标检测、阴影抑制以及目标遮挡等多种技术难题。为解决以上技术难题,文中提出了以路面分割方式来解决诸如树枝晃动等非背景的动态场景干扰,以基于颜色特征的车辆检测方法来检测静态车辆目标。在通过将阴影区域纳入到路面背景建模的过程中,实现对阴影的抑制。并利用颜色恒常性算法来移除光源颜色渐变对车辆与路面颜色特征造成的干扰。最后,利用改进的车道空间占有率计算模型来度量各车道的车流密度。具体来说,该文的主要工作及创新如下： 1.深入分析各种经典边缘检测算子,以及Hough变换直线检测原理,并利用Hough变换实现了路面RO1分割,以消除非背景目标对车辆检测的干扰。 2.深入分析了车辆与路面颜色在RGB颜色特征空间的分布特征,并将RGB颜色特征转换到新的颜色特征空间,使路面颜色特征分布更加紧凑。在此颜色特征模型下,根据贝叶斯分类器的分类思想,设计了路面与车辆像素的分类规则,将其转化为二值标注图像。最后,利用基于能量最小化图割的图像分割思想分割出了车辆目标。 3.针对光源颜色渐变对车辆与道路颜色特征的影响,提出了多光源非均匀光照场景下的颜色恒常性解决方法。该方法将场景图像划分成若干较小区域,分别对各区域光源颜色进行估计,根据各区域对场景光源颜色的贡献将其合并为一种复合光源颜色作为场景光源颜色的近似估计,再利用对角模型重构图像。该算法打破了现存算法对光源光谱能量均匀分布和物体表面反射属性的依赖,有效地解决了城市交通道路这种复杂场景下的颜色恒常性问题。最后,又将颜色恒常性算法应用于车辆检测算法中,对车辆检测性能作了进一步分析。 4.深入分析了车道空间占有率模型及其存在的问题之后,通过有机地融合图像采集设备与图像物体之间的时空关系,提出了改进的车道空间占有率计算模型来度量交通路口的车流密度。
[Abstract]:With the development of the city , the contradiction between the road and the road becomes more and more prominent . Therefore , how to make full use of the existing resources and realize the intelligent transportation system is an urgent problem . The intelligent traffic signal light control system plays an irreplaceable role as an important part of the intelligent traffic system . The problem of vehicle detection and calculation based on video plays an important role in the visual application such as the intelligent traffic signal lamp control system and the like , and has very important research significance and practical value .

With the development of video detection technology , the detection and calculation method of non - contact vehicle based on video has the advantages of simple installation , simple maintenance and low maintenance cost .

in ord to solve that above technical problem , a road surface segmentation method is proposed to solve the interference of a non - background dynamic scene such as shaking of a branch , and the object of a static vehicle is detect by a vehicle detection method based on the color characteristic .

1 . Analyze all kinds of classical edge detection operators and Hough transform straight line detection principle , and use Hough transform to realize the segmentation of road surface RO1 , so as to eliminate the interference of non - background target on vehicle detection .

2 . The distribution characteristics of vehicle and road surface color in RGB color feature space are deeply analyzed , and RGB color feature is converted to new color feature space , so that the color characteristic distribution of road surface is more compact . Under this color characteristic model , the classification rule of road surface and vehicle pixel is designed according to the classification idea of Bayes classifier . Finally , the image segmentation idea based on energy minimization graph cutting is used to segment the vehicle target .

3 . According to the influence of the light source color gradient on the color characteristics of the vehicle and the road , a method for solving the color constancy in a multi - light source non - uniform illumination scene is put forward . The method divides the scene image into a plurality of smaller areas , and combines the scene image into a composite light source color as an approximate estimate of the color of the scene light source , and then reconstructs an image by using a diagonal model .

4 . After deeply analyzing the space occupancy model and the existing problems , the improved lane space occupancy calculation model is proposed to measure the traffic flow density of the traffic intersection by organically integrating the temporal and temporal relationship between the image acquisition equipment and the image object .

【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.41;U495

【参考文献】