面向电力无人机巡检图像分析处理的自动化深度学习系统:架构设计与关键技术OACSTPCD

Current models for analysing images captured by Unmanned Aerial Vehicles in transmission line inspection face limitations in their applicability,high development costs,and long development cycle.This paper proposes a new automated deep learning system,with the key principles of system design being generalisability,scalability,and automation.The literature review of related technological advances and the system architecture design are presented.Experimental results with our prototype system show that the automated model constructed by the system achieved Mean Average Precision values of 91.36%and 86.13%,respectively,in identifying insulator explosions and bird nests on inspection images,demonstrating that the system design is sound,and the architecture is feasible.