Research And Application Of Deep Learning In The Generation Of Rock Thin Section Image

With the rise of deep learning technology,more and more fields have made huge breakthroughs using deep learning technology.Rock slice images are of great significance to the study of petroleum geological characteristics and geological exploration.Due to the limitations of various factors,rock slice images are difficult to obtain,costly to obtain,and low resolution,which limits the researchers' grasp of its information to a certain extent.However,the images generated by the traditional generative confrontation network will have poor visual effects and insufficient super-resolution reconstruction details.In response to these problems,this thesis applies deep learning technology to the field of multi-sample generation and superresolution reconstruction of rock slice images,and proposes an algorithm with better visual effects and objective evaluation indicators.This thesis proposes a single-image generative confrontation network SA-Sin GAN based on the self-attention mechanism.The self-attention mechanism is introduced into the singleimage generative confrontation network,so that the network can learn to pay attention to a specific part of the content during the training process.To improve the texture details of the generated image.The experimental results of multi-sample image generation prove that the SASin GAN algorithm can generate diversified images.The generated multi-sample image has high structural similarity with the real image,but there are differences between the multi-sample images;super resolution The reconstruction experiment results prove that the SA-Sin GAN algorithm can better learn the detailed information of the attention image and improve the quality of the reconstructed image.From the evaluation and analysis of subjective visual effects and objective indicators,the improved algorithm proposed in this thesis can improve the performance of rock slice image multi-sample generation and super-resolution reconstruction.