Research on Intelligent recognition of Rock samples based on Deep Learning method

Rock sample recognition plays an important role in oil and gas exploration mineral resources exploration. At present, the main rock identification methods are logging, seismic, gravity magnetism, geochemistry, hand samples thin slice analysis so on. In particular, mainstream section requires professionals to analyze slices manually, but manual detection preparation of take a long time. order reduce workload shorten time period, image depth learning method is used establish model. After cuttings core collected by industrial camera mud logging site, real-time automatic classification evaluation gas-bearing properties realized. this paper, file .JPG format cut discard information that does not contain at edge. Then it was divided into 4* 4, 6*6, 8*8, 10*10 segments. enhancement, transfer train neural network using Desnet201 built-in library torch, selected as our framework. last softmax layer changed multi-layer perceptron with three hidden layers, improved model identify types paper.