基于Faster R-CNN的航拍图像中绝缘子识别  

Faster R-CNN based recognition of insulators in aerial images

作　　者：程海燕[1] 翟永杰[1] 陈瑞[1] CHENG Haiyan;ZHAI Yongjie;CHEN Rui(North China Electricity Power University,Baoding 071003,China)

机构地区：[1]华北电力大学(保定),河北保定071003

出　　处：《现代电子技术》2019年第2期98-102,共5页Modern Electronics Technique

基　　金：河北省自然科学基金资助项目(F2017502016);中央高校基本科研业务费资助项目(2014MS140)~~

摘　　要：为了解决传统绝缘子识别方法存在适用性不强、识别效率低的问题,结合深度卷积神经网络思想,提出一种从电网巡检航拍图像中自动识别绝缘子的方法。应用Faster R-CNN框架,结合电网巡检航拍图像数据库,构建绝缘子识别系统,自动识别航拍图像中的绝缘子,并分析不同模型和参数对识别精确度的影响。实验结果表明,相比于传统航拍绝缘子识别方法,采用深度卷积神经网络对航拍绝缘子进行学习和识别,具有较高的识别准确率和效率,可以很好地识别各种类型的绝缘子,识别性能大幅度提高。In order to solve the problems of poor applicability and low recognition efficiency in traditional insulator recognition methods,a method of auto-recognition of insulators in power grid inspection aerial images is proposed combining with the thought of deep convolutional neural network. The insulator recognition system is established by applying the Faster R-CNN framework and combining with the database of power grid inspection aerial images,so as to automatically recognize the insulators in aerial images,and analyze the influence of different models and parameters on recognition accuracy. The experimental results show that in comparison with the traditional aerial insulator identification method,the proposed method has higher recognition accuracy rate and efficiency by using the deep convolutional neural network to learn and recognize aerial insulators,and can well recognize all kinds of insulators with its greatly-improved recognition performance.

关 键 词：卷积神经网络 深度学习 FASTER R-CNN 航拍图像 绝缘子识别 智能电网 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]