Image super-resolution reconstruction method for electric power grid inspection robot

• Main Authors: HU YUANHUI, ZHOU LISHA
• Format: Patent
• Language: Chinese; English
• Published: 09.08.2022
• Subjects: ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDEDFOR ELSEWHERE; CALCULATING; CHECKING-DEVICES; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FORADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORYOR FORECASTING PURPOSES; GENERATING RANDOM NUMBERS; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS; REGISTERING OR INDICATING THE WORKING OF MACHINES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE,COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTINGPURPOSES, NOT OTHERWISE PROVIDED FOR; TIME OR ATTENDANCE REGISTERS; VOTING OR LOTTERY APPARATUS

The invention discloses an image super-resolution reconstruction method for an electric power grid inspection robot. The image super-resolution reconstruction method comprises the steps of training an image super-resolution reconstruction network, inputting a natural image into the network, performing up-sampling and reconstruction, outputting a high-definition electric power part image and the like. The image super-resolution reconstruction network comprises a front convolutional layer, a joint information extraction module, a hierarchical information fusion module and an image output module, a CS attention module and an IK attention module are arranged in the joint information extraction module, a feature map is modulated through an attention mechanism, the enhancement effects of the two attention mechanisms are superposed, and the image super-resolution reconstruction is realized. And the network has a good extraction effect on some fine high-frequency features in the low-resolution image. After super-reso