Drilling image crack parameter automatic extraction method based on deep learning

Abstract

The invention belongs to the technical field of image recognition and geologic modeling, and particularly relates to a drilling image crack parameter automatic extraction method based on deep learning. The method comprises the steps of establishing a crack image data set, establishing a crack image recognition model, and using the marked crack image data set as a supervision condition; adopting a convolutional neural network to train a crack image recognition model, employing the trained crack image recognition model to detect a crack image to be detected, and converting the extracted related pixel information of the crack into inclination, inclination angle, position, opening and roughness parameters. According to the method, a deep learning technology is introduced into a drilling image data analysis process, and high-level semantic features of cracks are subjected to machine learning through the convolutional neural network, so that automatic extraction of crack information is realized.

Description

technical field [0001] The invention belongs to the technical field of image recognition and geological modeling, and in particular relates to a method for automatically extracting fracture parameters from borehole images based on deep learning. Background technique [0002] Traditional image processing methods are mostly used in the existing research on structural surface recognition of borehole images. For example, gray-level co-occurrence matrix parameters are used to define structural surface features, and at the same time, the strategy of sliding the scanning frame from top to bottom can be used in the borehole image to search for structural surface areas. Moreover, in some processing methods, the extreme value feature of the gray gradient longitudinal projection can also be used as the feature signal of the structural surface, and the structural surface area division can be realized by thresholding and binarizing the synthesized signal. [0003] However, due to the co...

AI Technical Summary

Problems solved by technology

[0003] However, due to the complex and changeable shape and structure of the rock mass in the actual borehole image, there may be large noise in the rock background image near some structural planes, which will interfere with the underlying image features based on grayscale, which not only limits the The generalization ability of such features also reduces the effect of structural surface area division

At the same time, structural surfaces with different inclination angles in the actual fractured rock mass will show different sinusoidal amplitudes in the borehole image, thus corresponding to different structural surface rectangular area heights, and the strategy of fixing the height of the scanning frame needs to be targeted at different drilling images. The height value of the scanning frame is manually intervened in the image, which inevitably reduces the versatility and automation of the algorithm

Images