Research and Analysis of Image Enhancement Algorithm in the Classification of Rock Thin Section Images

摘要

The structure of rock flakes is complex and difficult to be classified accurately. The proposed method to solve the problem is to use an image enhancement algorithm to enhance the rock slice image. In the study, the neural network ResNet50, which has a significant effect on fine-grained classification, was used to construct the rock cast thin section image classifier, and three image enhancement algorithms, CutOut, MixUp, and CutMix, were used to enhance the rock thin section image. The rock slice images used in the data set are from Ordos, and are divided into five categories according to the size of the rock. The experimental result obtained was that the CutOut algorithm performs well on the data set, and the accuracy of the classifier was as high as 93.39%, which is 1.3% higher than the result of only using ResNet50 for classification. The experimental results show the effectiveness of the image enhancement algorithm in the classification of rock slice images.