输电线路障碍物检测识别研究

发布时间：2018-06-17 08:55

本文选题：除冰机器人 + 障碍物检测　；参考：《湖南大学》2016年硕士论文

【摘要】：高压输电线路覆盖冰雪往往会导致输电线路不能正常稳定地工作,很多国家曾经都因输电线路上的积雪引发的严重供电系统瘫痪事故而蒙受了重大的经济损失。因此,如何安全高效地清除输电线路上的积雪是如今很多国家电力系统迫切需要解决的问题。随着机器人技术的蓬勃发展,利用除冰机器人代替传统除冰方法成为当前解决输电线路覆冰的一个重要研究课题。除冰机器人的工作环境为覆盖了冰雪的高压输电线路,机器人需要在输电线路上完成行走、越障、除冰等动作。机器人行走过程中,如何高效准确地检测识别输电线路上的障碍物是保证机器人完成除冰工作的必要前提,也是目前除冰机器人技术研究领域中的核心课题。因此,本文针对220KV高压输电线路上的障碍物特征,着重研究了基于视觉的障碍物检测识别研究。本文首先介绍了高压输电线路覆盖冰雪对人们工作生活带来的危害。在对传统的除冰方法的总结的基础上,提出了利用除冰机器人作为新型除冰方式的意义。接着,总结了国内外巡线机器人和除冰机器人的研究现状以及除冰机器人障碍物检测识别方法的研究成果和发展方向。结合了线路结构特点的基础上,主要介绍了三臂式除冰机器人本体结构及其视觉系统,同时介绍了输电线路上障碍物的类型。针对输电线路上的障碍物的检测识别课题,本文做了以下研究：1.基于输电线路上障碍物检测的繁琐性,本文根据图像的边缘特征,提出基于包围盒的检测算法,通过计算包围盒内完全包含的图像边缘数检测障碍物。首先,对采集到的障碍物图像进行了常规的图像预处理,图像的噪声干扰减少、可读性增强；接着利用基于结构的随机森林算法提取图像边缘；然后通过定义包围盒评分函数对包围盒进行评分判定包含障碍物的包围盒。2.针对障碍物识别的复杂性,提出了基于卷积神经网络的障碍物识别算法。首先介绍了卷积神经网络的发展背景以及历史,接着根据卷积神经网络的基本结构提出了适用于障碍物识别的网络架构,最后利用前向传播以及反向传播相结合训练该网络,实现障碍物的分类识别。
[Abstract]:High voltage transmission lines cover with snow and ice often lead to transmission lines can not work normally and stably. Many countries have suffered great economic losses due to serious power supply system paralysis caused by snow on transmission lines. Therefore, how to remove snow from transmission lines safely and efficiently is an urgent problem in many countries. With the rapid development of robot technology, using deicing robot instead of traditional deicing method has become an important research topic to solve the problem of transmission line icing. The working environment of deicing robot is a high voltage transmission line covered with ice and snow. The robot needs to complete walking, obstacle surmounting and deicing on the transmission line. In the process of robot walking, how to detect and identify obstacles on transmission lines efficiently and accurately is a necessary prerequisite for robot to complete deicing work, and it is also the core subject in the field of deicing robot technology. Therefore, this paper focuses on the obstacle detection and recognition based on vision for the obstacle characteristics of 220 KV HV transmission line. This paper first introduces the harm of high-voltage transmission line covering snow and ice to people's working life. Based on the summary of the traditional deicing method, the significance of using deicing robot as a new deicing method is put forward. Then, the research status of line patrol robot and deicing robot at home and abroad is summarized, as well as the research results and development direction of obstacle detection and identification method for deicing robot. Based on the characteristics of the line structure, this paper mainly introduces the body structure and vision system of the three-arm deicing robot, and also introduces the types of obstacles on the transmission line. Aiming at the problem of obstacle detection and identification on transmission lines, this paper does the following research: 1. 1. Based on the complexity of obstacle detection on transmission lines, an algorithm based on bounding box is proposed to detect obstacles by calculating the number of image edges contained in the bounding box. Firstly, the obstacle image is preprocessed, the noise interference is reduced, the readability is enhanced, and the edge of the image is extracted by using the structure based random forest algorithm. Then the bounding box is evaluated by defining the bounding box scoring function to determine the bounding box containing obstacles. 2. Aiming at the complexity of obstacle recognition, an obstacle recognition algorithm based on convolution neural network is proposed. Firstly, the development background and history of convolutional neural network are introduced. Then, according to the basic structure of convolutional neural network, a network architecture suitable for obstacle recognition is proposed. Finally, the network is trained by combining forward propagation and back propagation. The classification and recognition of obstacles are realized.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TM75;TP391.41

【相似文献】