A two-stage classification method for borehole-wall images with support vector machine.

Area of Science:

• Geology
• Image Analysis
• Machine Learning

Background:

• Geological exploration relies on analyzing borehole images, traditionally done manually, which is time-consuming and prone to errors.
• Existing automated methods struggle with the complexity and variety of borehole-wall images, particularly in coal-rock mass structures.
• Three dominant image types—border, fracture, and intact rock mass—require accurate classification for effective geological analysis.

Purpose of the Study:

• To develop and evaluate a novel two-stage classification approach for borehole images.
• To improve the efficiency and objectivity of geological structure analysis from borehole imagery.
• To accurately differentiate between border, fracture, and intact rock mass images.

Main Methods:

• A two-stage classification strategy was implemented for borehole images obtained via Axial View Panoramic Borehole Televiewer (APBT).
• The first stage utilized texture and grayscale histogram features to classify border images.
• The second stage employed Gabor filters for region segmentation and feature extraction to distinguish fracture from intact rock mass images.

Main Results:

• The two-stage classification system achieved a 95.55% correction rate in the first stage for identifying border images.
• The second stage achieved a 95% correction rate in classifying fracture and intact rock mass images.
• Experimental results demonstrate the effectiveness of the proposed method on real borehole image data.

Conclusions:

• The proposed two-stage classification method significantly enhances the accuracy and efficiency of borehole image analysis.
• This automated approach offers a more reliable alternative to subjective manual examination of geological borehole images.
• The findings contribute to more robust geological exploration through improved image classification techniques.