【作者】 郭晓东；

【导师】 杨英；

【作者基本信息】 东北大学 ， 车辆工程， 2012， 硕士

【摘要】 基于机器视觉的行人检测系统因其在车辆辅助驾驶系统中的重要应用价值成为当前计算机视觉和智能车辆领域最为活跃的研究课题之一。其核心是利用安装在运动车辆上的摄像机检测行人,并估计出潜在的碰撞危险以保护行人。行人检测是目标检测的一部分,同时又是该领域的一大难点。尽管在计算机视觉领域,目标的检测和跟踪技术已经有了十多年的研究,但其仍然是一个活跃的研究领域。目前还没有一个通用的,精确的,鲁棒的,高效实时的目标检测和跟踪算法。由于人体是非刚性,所处环境复杂,人与人或人与景物之间的相互影响,使得行人的检测和跟踪成为计算机视觉研究领域中最大的一项挑战。本文以单目视觉传感器作为获取车辆前方环境信息主要手段,进行了运动车辆前方行人检测技术的研究。采用统计学习的方法进行行人的识别。首先,对行人的各种特征进行了系统的研究,并提取了行人的边缘特征、纹理特征、梯度特征以及haar特征作为训练分类器的输入；然后运用AdaBoost算法融合SVM算法训练分类器；最后得到了较好的行人分类器。试验结果表明,本文提出的方法能够实现对车辆前方行人的有效识别,对于运动的和静止的行人都有较好的检测效果,系统的实时性也较好。检测出行人之后,本文又利用Kalman滤波的方法预测了行人矩形框各参数的变化,从而实现了对行人的跟踪定位,同时也为单目视觉测距打下了基础。最后,本文研究了安全人车距离预警模型的建立方法。根据视觉传感器检测跟踪的结果,在检测出行人的基础上,改进了安全预警距离模型,确定了一些参数的取值,综合了人、车、路三方面情况,给出了安全人车预警距离。并利用单目视觉技术对行人的距离及运动方向进行估测,为何时预警,何时强制减速,何时强制制动提供了依据。最后综合以上内容,设计了一套完整的行人检测预警系统。更多还原

【Abstract】 Computer vision based pedestrian detection has become one of the hottest topics in the domain of computer vision and intelligent vehicle because of its potential applications in driver assistance systems. It aims at detecting pedestrians appearing ahead of the vehicle using a vehicle-mounted camera, so as to take actions to protect pedestrians in case of danger.Pedestrian detection are part of the target detection, at the same time, pedestrian detection is a major difficult part of target detection. Although the study for object detection and tracking has been more than ten years in computer vision community, it is still an active research area. At present, there is not a target detecting and tracking system which is general, accurate, robust, efficient and real-time. Due to the human body is non-rigid, the scenes are complex, and there are plenty of interactions among the pedestrians or between the pedestrians and the scenes, the human detecting and tracking is one of the most challenges in computer vision.In this paper, a monocular vision sensor is used to get the environmental information in front of the vehicles for people detection technology research. Use the statistical learning methods for the recognition of pedestrian, study the various features of the pedestrian and extract the edge characteristics, texture features, gradient features and haar features of pedestrians as the input of the classifier training. Then use the AdaBoost algorithm and SVM algorithm to train the classifier to get a better pedestrian classification.Test result shows that the proposed method can identify effectively of the pedestrians in front of the vehicles, for pedestrians static and movement both have good results, and the real-time of the system is good.After the pedestrians are detected, this paper takes advantage of the Kalman filter method to predict the changes of the pedestrian rectangle parameters so that the pedestrians dan be tracked. At last, an early warning model of people and vehicles safety distance is established. According to the detection results from the visual sensors, improve the safety distance warning model then determine the values of some parameters, and then comprehensive the conditions of the people, the car and the road to calculate the people and vehicles safety early warning distance. Monocular vision technology is used to estimate the distance and the moving direction of pedestrians so that provide a basis to the car on when to warning, when to decelerate and when to brake. Then a pedestrian detection and warning system is designed更多还原