基于图像处理的接触网吊弦和受电弓滑板的检测与识别

发布时间：2018-03-20 14:41

本文选题：接触网吊弦　切入点：受电弓滑板　出处：《西南交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来,随着铁路运营量不断增加以及社会各界对能源利用率的重视,电气化铁路凭借其牵引功率大、节能环保、能大幅度提高运输能力和速度,并具有技术、经济、环保方面的优点,成为了各国铁路优先发展的铁路牵引动力方式。而我国高速铁路在近几年的飞速发展也使得电气化铁路在现已运营的所有铁路中的比例越来越大。在铁路大幅提速的背景下,交通的安全性日益受到关注,成为铁路运营的重要议题。而传统的依靠人力来进行的巡检方式不仅效率低、成本高、检测周期长,已经不能满足人们的要求,故采用图片处理技术的智能巡检方法越来越受到人们的重视。本文的研究工作是按照6C系统中的受电弓安全巡检装置和接触网安全巡检装置的技术规范来展开的,本文算法以受电弓巡检图片及接触网巡检图片为实验数据,利用图片处理的技术手段来实现对巡检图片内的敏感设备进行智能识别,通过实验测试也论证了本文所提算法的有效性。本文的主要工作及创新内容包括以下几个方面:在接触网吊弦检测的过程中,研究采用了基于海森矩阵的Ridge Filter对图片进行过滤,然后对过滤后的图片进行霍夫曼直线检测,通过设定阈值的方法对所有直线进行筛选,排除不符合条件的直线,则剩下的直线即为代表吊弦的直线,实现了对接触网吊弦的准确识别。在受电弓滑板检测中,首先研究采用了 HOG特征与广义霍夫曼变换相结合的方法来对受电弓滑板进行识别;该方法不仅有效去除了车顶复杂的钢架结构对识别的干扰,还较为理想的提高了程序的检测效率。在受电弓滑板检测中,采用了翻转检测的方法,该方法在实验数据较少的情况下有效的增加了训练样本数量,使得分类器得到了较为有效的训练,提高了程序检测的准确率。最后,对目前收集到的受电弓巡检图片及接触网巡检图片进行了实验测试,本文算法具有较好的适用性,得到了较为理想的识别率,验证了本文所采用的两种方法均具有一定的有效性。
[Abstract]:In recent years, with the increasing of railway operation and the importance of energy utilization, electrified railway, with its large traction power, energy saving and environmental protection, can greatly improve the transportation capacity and speed, and has the technology and economy. The advantages of environmental protection, In recent years, the rapid development of high-speed railway in our country also makes the proportion of electrified railway in all the railways in operation more and more. Under the background of the rapid increase of railway speed, The safety of traffic has been paid more and more attention to, which has become an important issue in railway operation. However, the traditional way of patrol and inspection based on manpower is not only low efficiency, high cost, long detection period, but also can not meet the requirements of people. Therefore, more and more attention has been paid to the intelligent inspection method using image processing technology. The research work in this paper is carried out according to the technical specifications of pantograph safety patrol device and catenary safety patrol device in 6C system. In this paper, the pantograph and catenary images are used as experimental data, and the technology of image processing is used to realize the intelligent recognition of sensitive devices in the inspection pictures. The effectiveness of the proposed algorithm is also demonstrated through experimental tests. The main work and innovative contents of this paper include the following aspects: in the detection process of catenary hoisting string, the Ridge Filter based on Hessen matrix is used to filter the images. Then the filtered images are detected by Hoffman line, and all the lines are screened by setting a threshold, and the non-conforming lines are excluded, and the remaining lines are the lines representing the hanging string. In the pantograph slide detection, the HOG feature and the generalized Hoffman transform are used to identify the pantograph slide plate. This method not only effectively removes the interference from the complex steel frame structure of the roof, but also improves the detection efficiency of the program. In the pantograph slide detection, the flipping detection method is adopted. This method can effectively increase the number of training samples in the case of less experimental data, make the classifier get more effective training, and improve the accuracy of program detection. Finally, Experiments are made on the pantograph and catenary images collected at present. The algorithm in this paper has a good applicability and an ideal recognition rate. The validity of the two methods is verified.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U226.8;U269.6;TP391.41

【相似文献】