| Rock is a solid aggregate with a stable shape.Formation lithology is the basic research object of petroleum geological rock characteristics,which can accurately reflect the distribution of oil and gas reservoir production capacity.Due to the limitations of various factors in different regions,the acquired rock images often have a lower resolution and cannot clearly display the detailed information of the rock images.Under this demand background,it is necessary to apply super-resolution reconstruction technology to rock images.Although most traditional super-resolution reconstruction methods have fully analyzed the images,due to the lack of data processing capabilities,the reconstruction results still need to be further optimized.The paper attempts to apply deep learning to the super-resolution reconstruction of rock images to explore their feasibility and applicability.In view of the shortcomings of traditional super-resolution algorithms,the paper uses the advantages of convolutional neural networks in image processing to verify the availability of SRCNN super-resolution reconstruction algorithm in rock images.Then,in order to overcome the shortcomings of SRCNN algorithm,the FSRCNN algorithm and the Res-SRCNN algorithm based on the residual network improvement were used to perform super-resolution reconstruction of the rock image,so as to reconstruct a super-quality reconstruction image with better quality.In the experimental part,three different types of rock images are used for super-resolution reconstruction training and verification,and subjective evaluation methods and objective evaluation methods are used to compare and evaluate the reconstructed images.The results prove the effectiveness of the deep learning-based algorithm in the super-resolution reconstruction of rock images,and the Res-SRCNN algorithm based on the residual network model improvement in the deep learning method has better results in the super-resolution reconstruction of rock images. |
PDF Full Text Request