输电线路中关键部件图像识别及异常检测方法研究

发布时间：2018-03-15 08:13

本文选题：绝缘子　切入点：杆塔　出处：《华北电力大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着智能电网和电力系统自动化的发展,计算机视觉技术越来越多地应用在电力设备的智能巡检和在线监测中。输电线路中的关键部件(绝缘子、输电杆塔、输电线等)是故障频发元件,绝缘子自爆、破损、异物等故障,输电杆塔鸟巢、异物等故障,输电线断股、异物、雷击闪络等故障严重威胁着输电线路的安全可靠运行。因此,定期监测输电线路关键部件状况,及时发现故障至关重要。通过对输电线路巡检采集的图像数据进行分析处理,从而发现输电线路故障已成为近几年的研究热点。本文主要围绕航拍图像中绝缘子、输电杆塔、输电线的识别及绝缘子自爆故障、输电杆塔鸟巢故障、输电线异物搭挂故障的检测进行研究,论文的主要内容如下:首先,提出了一种基于显著性检测与形态学的绝缘子识别及自爆故障检测方法。利用融合多特征的显著性算法定位绝缘子;对绝缘子定位结果区域进行OTSU二值分割提取绝缘子细节,然后进行形态学处理,实现自爆故障检测。其次,提出一种融合角点、直线、颜色和形状特征的输电杆塔识别和杆塔中鸟巢检测方法。通过LSD线段检测和Harris角点检测方法分别提取图像中的直线段和角点,通过融合处理和形态学处理后实现输电杆塔初定位结果,然后通过提取HOG特征训练SVM分类器实现输电杆塔的终定位。对于输电杆塔中的鸟巢故障,采用融合颜色特征、形状特征实现准确检测。随后,提出一种基于直线检测和平行性的输电线提取和异物搭挂检测方法。通过hough直线检测和平行性判定提取输电线,基于不变矩特征和adaboost算法在输电线区域检测是否存在搭挂的异物。最后,对本文工作进行了总结,并指出了需要进一步开展的研究工作。
[Abstract]:With the development of smart grid and power system automation, computer vision technology is more and more used in intelligent inspection and on-line monitoring of power equipment. Transmission line) is a fault frequency component, insulator self-detonation, breakage, foreign body fault, transmission tower bird's nest, foreign body fault, transmission line broken wire, foreign body, lightning flashover and other faults seriously threaten the safe and reliable operation of transmission line. It is very important to monitor the condition of the key parts of transmission line regularly and find the fault in time. By analyzing and processing the image data collected by the transmission line inspection and inspection, It is found that the fault of transmission line has become a hot research topic in recent years. This paper focuses on the identification of insulators, transmission towers, transmission lines and insulator self-detonation faults, and the bird's nest fault of transmission towers. The main contents of this paper are as follows: first, A method of insulator identification and self-detonation fault detection based on salience detection and morphology is proposed. The location of insulator is based on the salience algorithm of fusion multi-feature, and the details of insulator are extracted by OTSU binary segmentation to the location result area of insulator. Then the morphological processing is carried out to realize the fault detection of self-explosion. Secondly, a fusion corner, a straight line, is proposed. The color and shape features of the transmission tower and the bird's nest detection in the tower are identified. The straight line and corner in the image are extracted by LSD line segment detection and Harris corner detection, respectively. After fusion and morphological processing, the initial location results of transmission tower are realized, and then the final location of transmission tower is realized by extracting HOG feature training SVM classifier. For the bird's nest fault in transmission tower, the fusion color feature is adopted. Then, a method of line detection and parallelism detection for power transmission line and foreign body hanging detection is proposed. The line detection and parallelism detection are used to extract transmission line by means of hough line detection and parallelism judgment. Based on the moment invariant feature and adaboost algorithm, the paper detects whether there is a hanging foreign body in the transmission line area. Finally, the work of this paper is summarized, and the further research work is pointed out.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM755

【参考文献】