面向电力施工机器人的图像识别与处理技术研究  

Research on image recognition and processing technology for electric power construction robot

作　　者：龚向阳 杨跃平 张明达 王思谨 江炯 GONG Xiangyang;YANG Yueping;ZHANG Mingda;WANG Sijin;JIANG Jiong(State Grid Ningbo Fenghua Power Supply Company,Ningbo 315506,China)

机构地区：[1]国网宁波市奉化区供电公司,浙江宁波315506

出　　处：《电子设计工程》2023年第5期107-110,115,共5页Electronic Design Engineering

基　　金：国家电网双创项目(5211NB200139)。

摘　　要：针对电力爬杆机器人在攀爬过程中捕捉的视频与图像处理问题,文中开展了相关的技术研究。在对所采集到的数据信息进行预处理的基础上,通过损伤定位确定机器人在电杆上的位置。提取目标图像与样本图像的差异特征,将灰度不同的两类图像进行组合并产生二值图像,以适用于不同灰度值图像之间的识别。对分割后的图像进行缺陷提取,采用形态学方法去除干扰信息、用描述子对区域进行描述,并通过判别函数将特征空间分为不同类别的子空间。实验分析结果表明,该方法的检测准确率可达95%以上,且压缩比较大、处理时间较短,因此其具有一定的工程应用价值。In view of the video and image processing problems captured by the electric pole-climbing robot during the climbing process,related technical researches are carried out in this article. Based on the preprocessing of the collected data information,the position of the robot on the electric pole is determined by damage location. The difference features between the target image and the sample image are extracted,and the two types of images with different gray levels are combined to generate a binary image,which is suitable for identification between images with different gray values. Defect extraction is performed on the segmented image,and then the interference information is removed by morphological method,the area is described by the descriptor,and the feature space is divided into different types of subspaces by the discriminant function. The experimental analysis results show that the detection accuracy of this method can reach more than 95%,and the compression ratio is larger,the processing time is shorter,and it has certain engineering application value.

关 键 词：图像识别 爬杆机器人 特征提取 阈值分割 

分 类 号：TN-09[电子电信]