计算机与现代化Issue(4)：111-117,7.DOI:10.3969/j.issn.1006-2475.2026.04.015
基于CNN-Transformer的电网设备图像融合算法
CNN-Transformer-based Image Fusion Algorithm for Power Grid Equipment
摘要
Abstract
In the process of power grid equipment inspection,visible images can clearly present the external status and struc-tural details of the equipment,while infrared images can capture the thermal radiation information caused by equipment overload or abnormal operation.Through image fusion algorithms,effective complementation and enhancement of visible and infrared im-age information can be achieved,significantly improving the information richness and target saliency of the images,thereby pro-viding more reliable technical support for equipment anomaly detection and fault diagnosis.However,existing methods struggle to effectively retain the texture feature information of the source images during the fusion process,resulting in a poor quality of the fused image.To address this issue,this paper proposes a CNN-Transformer-based image fusion algorithm for power grid de-vices.The algorithm firstly uses a dual-channel feature extraction module based on convolution neural network(CNN)to realize local image feature extraction.A Transformer-based cascade feature enhancement module is introduced to extract and enhance deep feature.The extracted features are then integrated through feature fusion module to realize the aggregation and reconstruc-tion of the infrared and visible image features.In addition,luminance loss,constructed by luminance probability,is employed to steer the network training process.The experimental results indicate that the proposed algorithm possesses robust feature extrac-tion capabilities,enabling it to fully capture detailed features,texture information,and thermal target features from the source image to generate high-quality fused images.This offers reliable technical support for power equipment inspection,significantly enhancing the accuracy and efficiency of power grid equipment monitoring.关键词
深度学习/图像融合/卷积神经网络/TransformerKey words
deep learning/image fusion/convolutional neural network/Transformer分类
信息技术与安全科学引用本文复制引用
黄文礼,杨可军,汪兆冉,黄贝,曹阳,张义德,吴海峰..基于CNN-Transformer的电网设备图像融合算法[J].计算机与现代化,2026,(4):111-117,7.基金项目
安徽省科技重大专项(202203a05020023) （202203a05020023）
