This study integrates unmanned aerial vehicles (UAVs) and deep learning techniques to improve the efficiency of monitoring soil and water conservation facilities in mountainous regions. Traditional inspection methods face challenges such as difficult accessibility, high labor demands, and inefficiency, particularly in vast and remote areas. To address these issues, UAVs were deployed to capture high‐resolution aerial images, providing extensive coverage and overcoming terrain constraints. A deep learning model, based on the ResNet‐50 architecture, was developed to automatically classify anomalies and offer maintenance recommendations. The model demonstrated high accuracy, achieving 0.966 with regularization methods and 0.959 using neural networks. The study further employed the hue, saturation, and brightness (HSB) color space for anomalous area estimation, achieving precise segmentation with a mean error rate of approximately 7.1%, as verified against 3D model‐based ground truth. This novel approach enhances slope failure analysis, offering more accurate anomaly detection and area estimation. By integrating UAVs for data acquisition and deep learning for automated analysis, the proposed system significantly reduces human labor costs, improves inspection efficiency, and provides a robust and reliable solution for anomaly detection and analysis. These findings contribute practical insights and innovative methodologies to the field of soil and water conservation management.
Pan et al. (Wed,) studied this question.