基于改进Densenet的输电线路杆号牌识别模型  

A Recognition Model for Transmission Line Pole License Plates Based on Improved Densenet

作　　者：周坤 李鹤健 李海山 Zhou Kun;Li Hejiang;Li Haishan(Yunnan Power Grid Co.,Ltd.,Dali Power Supply Bureau,Dali 651000,Yunnan,China)

机构地区：[1]云南电网有限责任公司大理供电局,云南大理671000

出　　处：《云南电力技术》2024年第6期50-54,70,共6页Yunnan Electric Power

摘　　要：针对输电线路杆号牌在树木遮挡、以及部分运维人员无人机控制技术欠佳的情况下搜寻困难,导致精细化无人机巡检质量较差、效率较低的问题,本文提出基于改进Densenet的输电线路杆号牌识别模型。该模型首先采取模板匹配算法对航拍图像中的杆号牌位置进行定位,以区域识别的模式代替传统全局遍历识别,从而降低背景干扰、提高识别效率;然后,本模型采取特征预置剔除干扰和特征预置识别目标的方法,对Densenet模型的dense block进行了改进,改进后的Densenet模型通过分析杆号牌所在的区域,进行杆号牌识别标准和信息提取,以便于巡线人员高效维护杆号牌信息。实验结果表明,本文模型在开阔地带、树丛遮挡等情况下,对杆号牌文字、数字识别准确率相对于Yolo v7、Crnn、Densenet模型均有显著提升。In response to the difficulties in searching for transmission line pole license plates due to tree obstruction and inadequate drone control technology by some operation and maintenance personnel,resulting in poor quality and efficiency of refined drone inspections,this paper proposes an improved Densenet based recognition model for transmission line pole license plates.The model first adopts a template matching algorithm to locate the position of pole plates in aerial images,replacing traditional global traversal recognition with regional recognition mode,thereby reducing background interference and improving recognition efficiency;Then,this model adopts the methods of feature pre setting to eliminate interference and feature pre setting to identify targets,and improves the dense block of the Densenet model.The improved Densenet model analyzes the area where the pole number plate is located,identifies the pole number plate recognition standards and extracts information,in order to facilitate efficient maintenance of pole number plate information by patrol personnel.The experimental results show that the model proposed in this paper has significantly improved the accuracy of recognizing pole and license plate text and numbers compared to Yolo v7,Crnn,and Densenet models in open areas,tree cover,and other situations.

关 键 词：杆号牌 模板匹配 特征预置 Densenet 

分 类 号：TM74[电气工程—电力系统及自动化]