基于图像处理和深度学习的新型巡检无人机应用技术

打开文本图片集

中图分类号：TM726；TP391 文献标志码：A 文章编号：1001-5922（2025）09-0149-04

A new type of inspection unmanned aerial vehicle application technology based on image processing and deep learning

TANG Hongliang，QIU Yang，DING Yeqiang，SU Haohang，ZHAGN Fan （State Grid Zhejiang Electric Power Co.，Ltd.Hangzhou Linping District Power Supply Company，Hangzhou 31110o，China）

Abstract：In order to improve the automatic inspection ability of distribution lines，a new type of unmanned aerial vehicle（UAV）system was constructed，through image recognition technologyand deep learning algorithm model， the automatic inspection of distribution lines was greatly improved，and the anti-shake platform，LVDS transmision technology，anti-interference twisted pair，shielded aluminum plate，DDR FCRAM cache，SLC memory and image processing module were added to the optical sensor.The three aspects of aircraft flight anti-shake，anti-electromagnetic interference and optimization of mobile storage were improved，the UAVinspection ability was improved，the image recognition and diagnosis ofoperation quality were realized byaranging a light neural network in the cloud during he inspection，and the UAVfault diagnosis ability inthe process of automatic inspection of distribution lines was improved.Thesimulationresultsshow thatthe proposed method was highly practical，and the accuracy of fault location in multiple experiments was more than 90% ·

Key words ：unmanned aerial vehicle;transmission lines;fault location ; automatic inspection; Image processing;application

无人机巡检技术，广泛应用于基础设施监测（如电力线路）及环境、灾难救援等[1-2]。（剩余4024字）