摘要
为了利用深度学习实现对输电线路的多目标识别以及多种故障的检测,以Faster RCNN(faster regions with convolutional neural networks features)网络为算法框架进行无人机图像的数据挖掘,针对输电线路的6种目标检测任务,提出了3种改进策略,分别为自适应图像预处理算法,基于面积的非极大值抑制算法,以及切分检测方案。研究结果表明:所提改进算法能够利用挖掘的数据对故障进行准确定位与识别,实现对复杂背景下航拍图像中多目标的故障检测,也可类推至其他类似多目标应用场景。论文研究可为多目标的检测和识别提供参考。
In order to use the deep learning to achieve multi-target recognition of transmission lines and multiple fault detection,adopting the"faster regions with convolutional neural networks features"(Faster RCNN)network as an algorithm framework,we mined data of drone images,and proposed three improvement strategies for 6 target detection tasks of transmission lines,namely the adaptive image preprocessing algorithm,the area-based non-maximum suppression algorithm,and the segmentation detection scheme.The research results show that the improved algorithm in this paper can accurately locate and identify faults by use of mined data and realize multi-target fault detection in aerial images under complex backgrounds,and can be analogized to other similar multi-target application scenarios.The research can prove a reference for the detection and recognition of muti-target.
作者
白洁音
赵瑞
谷丰强
王姣
BAI Jieyin;ZHAO Rui;GU Fengqiang;WANG Jiao(Nanrui Group Co.,Ltd.,Beijing 100192,China;Beijing Kedong Electric Power Control System Co.,Ltd.,Beijing 100192,China)
出处
《高电压技术》
EI
CAS
CSCD
北大核心
2019年第11期3504-3511,共8页
High Voltage Engineering
基金
国家电网公司科技项目(人工智能驱动的电动汽车智慧出行服务关键技术研究、设备研制和工程示范)(52020118000G)~~
作者简介
通信作者:白洁音,1976-,男,硕士,高工,主要研究方向为计算机科学与技术,E-mail:964012015@qq.com。
