高速公路恶劣天气及交通状况智能分析系统研究与实现

发布时间：2018-01-13 17:30

本文关键词：高速公路恶劣天气及交通状况智能分析系统研究与实现　出处：《西南交通大学》2014年硕士论文　论文类型：学位论文

【摘要】：恶劣的天气及交通异常状况常常给高速公路的安全带来巨大威胁,高速公路管理部门需要实时地掌握其辖区内各个路段的气象情况及交通状况,以便做出及时准确的应对决策。然而,目前主要依靠传感器等硬件设备提供的数据,经分析得到气象信息和交通信息,安装及维护成本高昂,且使用时间和环境条件可能影响其准确性。本文充分利用已经建立的高速公路监控体系,研究了基于高速公路监控视频的恶劣天气(雾、雨、雪)检测和交通异常状况(拥堵、停车、逆行)检测算法,通过对监控摄像头传入的视频数据的智能分析,得到该路段的天气信息和交通信息。本文算法结合了图像处理、计算机视觉及模式识别等领域的先进技术,对视频图像中的雾、雨、雪的视觉特征进行提取和分析,从而实现恶劣天气的检测和报警,同时对运动车辆进行跟踪,并对其进行行为分析,从而实现交通异常状况的检测和报警。本文分析了雾、雨、雪在监控视频中呈现出的视觉特征,针对雾区的模糊效应提出了基于Canny边缘的雾天检测算法。对于雨和雪的检测,基于监控视频分辨率小及噪声干扰大的特点,本文算法放弃传统的对雨滴和雪花的动态特性进行分析的方法,分析了雨天和雪天在车道上的视觉特征,针对湿滑道路的反光特性和图像钝化的特征,提出了反光度与图像锐度结合的道路湿滑程度评估方法,从检测道路是否湿滑的角度来实现雨天的检测,针对道路上积雪的颜色特征,提出了基于积雪颜色模型的积雪检测方法,从检测道路上是否覆盖积雪的角度来实现雪天的检测。针对交通异常状况的检测,本文首先分析了前景检测和背景重建过程中几种常用的技术,并通过实验对其效果进行对比,经分析,本文算法采用背景差分法与基于统计模型的背景重建与更新方法相结合,实现运动车辆的检测和跟踪,估算出车速、车流量及车辆行驶方向,并运用这些信息进行综合分析和判断,实现高速停车、逆行及拥堵状况的检测。本文使用了丰富的视频数据作为实验数据集,通过对实验结果数据的统计和分析,说明了本文算法在恶劣天气检测及交通异常状况分析中的有效性。基于本文算法实现了具有实用价值的智能分析系统,该系统与视频监控平台兼容,不仅能分析本地历史视频,而且能以多路并行轮询的方式分析在线实时视频,是高速公路信息服务体系中的重要组成部分,功能完整,灵活性高,它的应用将有效提高高速公路的管理能力和应急能力。
[Abstract]:The bad weather and abnormal traffic situation often bring great threat to the safety of freeway. The highway management department needs to grasp the meteorological situation and traffic condition of each section in its jurisdiction in real time. In order to make timely and accurate response to the decision. However, at present, mainly depends on the sensor and other hardware equipment to provide data, through analysis to obtain meteorological information and traffic information, installation and maintenance costs are high. And the use time and the environment condition may affect its accuracy. In this paper, we make full use of the established highway monitoring system, and study the detection algorithm of bad weather (fog, rain, snow) and traffic anomaly (congestion, parking, retrograde) based on highway surveillance video. Through the intelligent analysis of the video data from the surveillance camera, the weather information and traffic information of the section are obtained. This algorithm combines the advanced technology of image processing, computer vision and pattern recognition. The visual features of fog, rain and snow in video images are extracted and analyzed to detect and alarm bad weather, and track and analyze the behavior of moving vehicles. In order to realize the traffic abnormal condition detection and alarm. In this paper, the visual features of fog, rain and snow in surveillance video are analyzed. Aiming at the fuzzy effect of fog region, a fog detection algorithm based on Canny edge is proposed. Based on the characteristics of small resolution and large noise interference, this algorithm gives up the traditional method to analyze the dynamic characteristics of raindrops and snowflakes, and analyzes the visual features of rainy and snowy days in driveway. According to the reflective characteristics of wet slippery road and the feature of image passivation, this paper proposes a method to evaluate the wet slippery degree of road, which combines reflectance and image sharpness, and realizes the detection of rainy day from the angle of detecting whether the road is slippery or not. According to the color characteristics of snow cover on the road, a snow detection method based on snow color model is proposed to detect snow cover on the road. For the detection of traffic anomaly, this paper first analyzes several common techniques in the process of foreground detection and background reconstruction, and compares their effects through experiments. In this paper, the background difference method is combined with the background reconstruction and update method based on statistical model to realize the detection and tracking of moving vehicles, and to estimate the speed, flow and direction of vehicles. This information is used for comprehensive analysis and judgment to detect high speed parking, retrograde and congestion. In this paper, we use rich video data as experimental data set, through the statistics and analysis of experimental data. The effectiveness of this algorithm in the detection of severe weather and the analysis of traffic anomalies is explained. Based on this algorithm, an intelligent analysis system with practical value is implemented, which is compatible with the video surveillance platform. Not only can analyze the local history video, but also can analyze the online real-time video in the way of multi-channel parallel polling. It is an important part of the expressway information service system, with complete function and high flexibility. Its application will effectively improve the ability of expressway management and emergency response.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U495;TP391.41

【参考文献】