Research And Implementation Of Infrared Image Retrieval System For Electrical Equipment Based On Deep Learning

Based on the research of electrical equipment fault diagnosis methods based on deep learning,this paper proposes to use deep learning-based image retrieval technology for intelligent diagnosis of infrared images of electrical equipment in substations,in order to establish a system that can perform fault diagnosis for most equipment in electrical substations Accurate and rapid diagnosis system to meet the requirements of highly intelligent substations in the construction of smart grids.This paper’s work on infrared image retrieval of electrical equipment based on deep learning mainly includes the following aspects:(1)An image retrieval method based on deep learning is introduced into the infrared image retrieval of electrical equipment.By analyzing the characteristics of low brightness and sharpness of infrared images of electrical equipment,research and experiments on the infrared image enhancement methods of electrical equipment,and finally select the Laplacian operator sharpening method to pre-process the infrared images of electrical equipment,making the PSNR of the input image higher.In order to improve the retrieval accuracy,this paper uses VGG16 convolutional neural network to extract features from infrared images of electrical equipment and perform image retrieval.Compared with the retrieval results of experiments performed on CIFAR-10 and Oxford5 k datasets,the final experimental results show that deep learning based electrical equipment infrared image retrieval system is feasible.(2)The image retrieval method based on deep hashing is introduced into the infrared image retrieval of electrical equipment.When the image data greatly increases,converting the features of the image into a binary hash code for indexing can make the retrieval performance even better.In this paper,a layer of fully connected layers is added to the convolutional neural network as a hash layer.A deep hash-based electrical device infrared image retrieval method is designed,and the loss function and the Hamming distance calculation method are improved.Comparing the results of infrared image retrieval based on deep hashing with the experimental results of LSH and ITQ methods,it is concluded that infrared image retrieval based on deep hashing has certain advantages.Experiments on the impact of the number of bits of the hash code on the retrieval results show that the 48-bit hash code is better in the retrieval accuracy of infrared images of electrical equipment.(3)A Flask framework was used to design and implement a web interface for infrared image retrieval of electrical equipment based on deep learning,and conducted experimental tests in the data set.The experimental results show that the deep learning-based electrical equipment infrared image retrieval web interface in this article can retrieve top-N images close to the target image in the database,which meets the basic functional requirements for image retrieval of electrical equipment infrared images.