Rock lithology classification method suitable for hydropower station geological survey

• Main Authors: XIN CUN, LIU XIAODONG, GUO YUAN, JIA XINHUI, SONG XUAN, HUANG YONG, LIU QINGYI, KOU YIDAN
• Format: Patent
• Language: Chinese; English
• Published: 21.11.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The invention provides a rock lithology classification method suitable for hydropower station geological survey, and solves the problems that in the field of hydropower station survey, rock images have the characteristics of multiple types and small image quantity, the quantity of partial types of samples is deficient, and the sample distribution is very non-uniform, so that the existing rock lithology model training based on machine learning is insufficient, and the accuracy of rock lithology classification is poor. And the rock lithology classification precision is low. According to the method, a twin network architecture is adopted, and the network is trained by constructing the sample pair, so that the problem of low lithology classification precision caused by uneven distribution of rock image samples is effectively solved; meanwhile, on the basis of a twin network architecture, a residual network learning module is added, and a twin-residual neural network model is established, so that the learning abil