Research On Image Expansion Model And Algorithm Of Power Equipment Based On Generative Adversarial Network

According to the national artificial intelligence overall plan,the construction of the smart grid big data platform is rapidly advancing,and the core part includes the use of deep learning to implement intelligent detection of power equipment defects.Due to the special nature of various power equipment,it is difficult for professional and technical personnel to conduct on-site surveys and inspections of power equipment for a long time,and it is also hard to arbitrarily collect image data of power equipment.Therefore,the current image data sets of power equipment are hardly enough to meet the requirements of training of the intelligent defect detection model so as to make the progress of scientific research of intelligent detection of electrical equipment defection slow.This thesis starts with the X-ray tensile clamp image data set in the power equipment data set,and carries out research work on the image data set expansion model based on Generative Adversarial Networks(GAN).It also proposes the power equipment image expansion model based on Generative Adversarial Networks algorithm.An image used to extend the image data set of the intelligent detection model of power equipment defects.The main research contents are as follows:(1)Preprocess the original X-ray tensile clip image data set,including data set standardization and data set enhancement,so that the processed image can meet the training of subsequent models.(2)A GAN network model was constructed to generate the X-ray tension clamp image data set.According to the two GAN algorithms introduced,the GAN network structure is designed and implemented,and the experimental environment required for the model operation is established.The two GAN models are trained using the X-ray tensile clamp image image data set.The experimental results finally show that The resulting image is of high quality.(3)A power equipment image expansion model based on Conditional GAN(CGAN)is proposed.Building on the CGAN model,improving the CGAN model for the X-ray tensile clamp image data set and the power equipment image expansion model is constructed so that it can generate X-ray tensile clamp images with specified defection requirements,and the experimental results are analyzed,using by the FrechetInception Distance(FID)to evaluate the performance of the model in this thesis,it is proved that the power equipment image expansion model is better than the previous two models.(4)Use the Convolutional Neural Network(CNN)classification model to evaluate the generated images.A CNN image classification model is constructed,and the generated X-ray tensile clip image and the real X-ray tensile clip image are used as a training set and a test set respectively,and a classification verification experiment is performed and the results of the two classification experiments are compared.Finally,through model performance comparison experiments,it is verified that the images generated by the power equipment image expansion model can effectively expand the power equipment image data set.In this thesis,the generation adversarial network is used for the expansion of X-ray tensile clamp image data sets.A power equipment image expansion model capable of generating specified defects is proposed.Finally,experiments verify that the generated images can be used in power equipment image data sets,The expansion has certain value for subsequent research work.