Automatic Recognition Algorithm for Clay Slope Surface Cracks Based on Multi-scale Image Fusion

Clay slope surface cracks serve as critical indicators of potential structural instability, necessitating timely and accurate detection for preventive maintenance. This paper proposes an automatic crack detection algorithm based on multiscale image fusion specifically designed for clay slope monitoring. The method employs Gaussian pyramid decomposition to generate multi-resolution representations, followed by scale-specific crack detection using adaptive thresholding and morphological operations. A novel weighted fusion strategy integrates detection results across scales, with weights dynamically determined by crack characteristics and detection confidence. Experimental validation on 1,850 clay slope images demonstrates superior performance with an F1-score of 0.914, outperforming traditional methods by $47.5 \%$ and single-scale approaches by $12.3 \%$. The algorithm exhibits exceptional robustness to environmental variations and computational efficiency suitable for real-time deployment, providing a practical solution for automated clay slope monitoring in geotechnical engineering applications.