高速公路视频监控系统中车辆识别与道路状态检测的研究

发布时间：2018-06-09 22:06

  本文选题：运动目标检测 + 阴影消除　； 参考：《北京工业大学》2015年硕士论文

【摘要】：高速公路是城市快速交通的主干线,已成为经济发展的脊梁。高速公路路面状态的安全级别受自然因素影响较大,比如降雨、大雾、降雪与融雪等都将影响高速公路的运营状况。本文以高速公路智能化视频监控系统开发为背景,研究高速公路监控视频中运动目标识别、运动车辆跟踪和道路雨雪状态检测技术,并在基础上完成了应用系统设计和功能模块开发。本文主要工作包括:1)提出了一种新的目标阴影消除方法。为了准确检测运动车辆目标,需要首先进行运动阴影消除。针对传统HSV颜色空间下阴影消除方法的不足,本文提出了一种HSV颜色空间下结合图像区域信息的阴影消除方法。使用该方法可以避免因车辆目标内部颜色特征与阴影区域接近而被误当作阴影,降低目标的误检率。2)提出了一种多目标遮挡状态下的目标跟踪策略。对多车辆目标跟踪问题,本文采用基于目标几何特征结合Kalman预测器的目标跟踪方法,针对多目标出现遮挡的情况,提出了一种以质心跟踪为主,面积信息为辅,根据预测目标与实际目标的个数分组讨论的跟踪策略。3)针对雨雪天气情况下公路路面状态检测预警问题,采用朴素贝叶斯分类器对道路状态进行分类。首先,提取出图像中道路的区域;然后提取道路图像的熵、能量、对比度、平均亮度和平均饱和度等参数作为特征,使用朴素贝叶斯分类器对道路区域状态分类,并对湿滑、积雪较多的道路进行告警。4)设计了高速公路智能视频监控系统。系统可以完成交通视频中路面背景提取、运动车辆检测与跟踪、车辆信息统计、道路雨雪状态检测等功能,并且可以对异常交通情况和恶劣道路状态进行告警处理。
[Abstract]:Expressway is the main line of urban express traffic and has become the backbone of economic development. The safety level of expressway pavement state is greatly affected by natural factors such as rainfall fog snowfall and snowmelt will affect the operation of expressway. Based on the development of intelligent video surveillance system for freeway, this paper studies the technology of moving target recognition, moving vehicle tracking and road rain and snow state detection in expressway surveillance video. And on the basis of the application system design and functional module development. The main work of this paper includes: 1) A new target shadow elimination method is proposed. In order to detect the moving vehicle target accurately, it is necessary to remove the moving shadow first. Aiming at the deficiency of traditional shadow elimination method in HSV color space, a shadow cancellation method combining image region information in HSV color space is proposed in this paper. By using this method, we can avoid being mistaken for shadow because the inner color feature of vehicle is close to the shadow area, and reduce the false detection rate of the target. (2) A target tracking strategy in multi-target occlusion state is proposed. For the multi-vehicle target tracking problem, this paper uses the target tracking method based on the geometric feature of the target and Kalman predictor. In view of the occlusion of the multi-target, a new method is proposed, which is based on the centroid tracking and the area information. According to the tracking strategy. 3) aiming at road pavement condition detection and warning problem in rain and snow weather, naive Bayes classifier is used to classify road state. Firstly, the road region in the image is extracted, and then the parameters such as entropy, energy, contrast, average brightness and average saturation of the road image are extracted as features, and the state of the road region is classified by naive Bayes classifier, and the road is wet and slippery. The highway intelligent video surveillance system is designed. The system can complete the functions of road background extraction, moving vehicle detection and tracking, vehicle information statistics, road rain and snow state detection and so on, and can alarm the abnormal traffic situation and the bad road state.
【学位授予单位】：北京工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;TP391.41;TN948.6
，

本文编号：2000984