节点文献

基于视频图像识别的铁路雪情分析及预测方法研究

Railway Snowfall Status Evaluation and Prediction Based on Video Image Recognition

【作者】 朱磊

【导师】 秦勇;

【作者基本信息】 北京交通大学 , 交通运输工程(专业学位), 2016, 硕士

【摘要】 随着我国经济的增长,新建高速铁路的里程也在不断增加,交通事故也随之频发,因此交通安全保障技术越来越成为关注的焦点和研究的热点。雪灾是导致交通事故频发的一个重要因素,会引起铁路供电中断、线路受阻,严重影响了铁路的运输能力和运输安全。由雪灾造成的损失极其巨大,为了能将雪灾的损失降到最低,建立雪灾防护系统显得尤为重要,必须对降雪的信息有一个准确且及时的把握,比如当前雪的平均厚度以及单位时间内的降雪量等,从而实现更好的铁路行车预警。目前,我国铁路的测量主要应用人工测量和激光测量,其中激光测量比较准确而且反馈及时,现如今比较缺乏视频方面的分析雪情技术,遂本文针对雪情分析主要从以下几个方面进行研究:(1)通过图像处理技术对雪情进行分析,根据目前的运动目标检测技术,提出自己改进的基于实时视频图像识别的雪情检测分析方法。一共分为图像预处理、连续五帧法检测雪粒子、基于雪粒子轮廓尺寸的雪粒子提取、开窗选取有效区域、推算实时降雪强度五个部分,实时统计雪粒子密度,通过统计分级进而估计监控点处的实时降雪强度。(2)研究了风吹雪的形成机理,归纳了雪粒子的运动形式主要有蠕移、跃移和悬移,跃移占了绝大部分。风吹雪的堆积主要取决于风速场的大小,之后简单分析了风吹雪的堆积形式。之后分析建立铁路降雪量堆积模型,以铁路降雪量堆积模型作为预测分析积雪的主预测模型,雪的物理性质、当前的降雪量、历史雪害信息等作为模型计算的修正条件,采用定性与定量相结合、计算机仿真与专家经验相结合的方法进行雪情预测分析。最后提出相应的防护措施等。

【Abstract】 With the growth of Chinese economy, the mileage of newly built high-speed railway is increasing, and traffic accidents occur frequently. Therefore, traffic safety technology has become the focus of attention and research. Snow disaster as an important factor will cause traffic accidents. In addition, it will cause railroad power outage and line block. It will seriously affect the railway transportation capacity and transportation safety. In order to minimize the enormous losses caused by snow, the establishment of snow disaster preventing system is very important. Therefore, we must have an accurate and timely information for snow, such as the average thickness of snow and snowfall per unit of time. Only in this way can we perform better railway traffic warning. At present artificial measurement and laser measurement are two main railway measurement methods in China, the latter being accurate and timely. Nowadays the analysis technology of snow condition concerning video is lacking. Hence, this paper examines the issue from the following several aspects:Snow situation is analyzed through image processing technology. According to the moving target detection technologies, I proposed an improved snow detection analyzing method based on real-time video image recognition. The method consists of five parts: image preprocessing, five-consecutive-frame method of detecting snow particles, extraction of snow particles based on particle sizes, open the window to select effective areas, and calculation of real-time snowfall intensity. Statistical results of real-time snow particle density can be drawn, and the real-time snowfall intensity at monitoring points can be estimated.The formation mechanism of snow blown by wind is studied. It is summarized that the movement of snow particles are mainly creep shift, saltation and suspended load, and saltation accounts for the majority. The accumulation of wind-blowing snow depends mainly on the size of the wind field. And the accumulation forms of wind-blowing snow are concluded. Then a railway snow accumulation model is established and used as the main model to forecast and analyze the snow. The physical properties of the snow, present snowfall and history information of snow damage are correction conditions for the calculation of the model. The quantitative and qualitative analysis, and computer simulation combined with expert experience are used to forecast and analyze the snow condition. At last, corresponding protective measures are proposed.

节点文献中: