Overhead line robot obstacle recognition method

Abstract

The invention discloses an overhead line robot obstacle identification method, wherein the method comprises the steps: performing deployment based on an interlayer compression formula with strict closed form guarantee, and performing re-learning to keep responses between an original network and a compression network consistent. According to the method, a new compression scheme based on low-order decomposition is provided, a convolution layer is accelerated at the same time, and a completely connected layer is decompressed through accurate closing. For further reducing precision loss caused bythe compression scheme of order decomposition under a high compression rate, the invention further provides an effective knowledge transfer scheme, and output and intermediate response of an originalnetwork are aligned with explicit expression of the compression network. The proposed knowledge transfer scheme format operates non-linear transformation functions within and between all layers and minimizes 'local' and 'global' reconstruction errors in a unified manner.

Description

Technical field[0001]This application relates to the technical field of electrical equipment, and in particular to an obstacle recognition method for an overhead line robot.Background technique[0002]The operating status of high-voltage overhead transmission lines directly affects the distribution status of the power system and plays a key role in the safe and stable operation of the power system. Therefore, regular inspections of the operating conditions of high-voltage overhead transmission lines are required. In recent years, the development of mobile robot technology has provided new ideas for the inspection of overhead power lines. The inspection method of high-voltage overhead transmission lines has been transformed from manual inspection to special robot inspection.[0003]Overhead line inspection robots (hereinafter referred to as overhead line robots) are used to detect and repair line defects. In actual use, when overhead line robots crawl along transmission lines at a certai...

AI Technical Summary

Problems solved by technology

[0005] This application provides an obstacle identification method for overhead line robots to solve the problem that the existing convolutional neural network model for identifying obstacles on high-voltage overhead transmission lines consumes a large amount of processor running memory and calculations during the training process. Amount, affecting the rapid update of the convolutional neural network model, thus affecting the identification efficiency of the overhead line robot to obstacles

Images