The intelligent fault identification method based on multi-source information fusion and deep learning.

Area of Science:

• Geology
• Remote Sensing
• Artificial Intelligence

Background:

• Geological faults are critical structures, traditionally identified using remote sensing imagery (RSI) based on linear features.
• Enhancing morphological features of faults is key for rapid, precise, and intelligent identification.

Purpose of the Study:

• To develop and validate a multi-source information fusion method for intelligent geological fault identification.
• To integrate spectral, topographic, geomorphic, and structural features for improved fault detection.

Main Methods:

• Analysis and fusion of RSI, digital elevation model, and geological map data.
• Extraction of 16 influencing factors across spectral, topographic, geomorphic, and structural domains.
• Application of machine learning for factor importance prediction and Convolutional Neural Network (CNN) for intelligent fault identification.

Main Results:

• The Classification and Regression Trees (CART) model achieved high accuracy (0.993), true positive rate (0.988), and F1-score (0.994).
• Topographic Position Index (TPI), Valley Line (VL), Surface Cutting Depth (SCD), and RSI were identified as crucial factors.
• The CNN model demonstrated strong performance with a Validation Accuracy of 0.990 and Validation Loss of 0.025.

Conclusions:

• Multi-source information fusion significantly enhances geological fault identification.
• Deep learning-based image recognition, particularly CNN, provides an effective tool for intelligent fault mapping.
• The proposed method accurately identifies faults in complex geological areas, as demonstrated in Jinzhai County.