Research on Landslide Risk Monitoring Method Based on Machine Vision

To address the challenges of daily monitoring in existing landslide monitoring techniques, a landslide risk monitoring method based on machine vision was proposed. Firstly, a slope marker detection algorithm YOLOv5-CS was developed. The algorithm utilized a feature extraction module with rich gradient flow information and a feature fusion module based on neural attention mechanisms to enhance the accuracy of marker detection. Secondly, a slope marker positioning model based on geometric ranging was constructed using camera pitch angle, yaw angle, and imaging model to compute the three-dimensional coordinates of markers. Finally, the landslide risk at monitoring points was evaluated based on the changes in marker coordinates. Results demonstrate that the proposed method can achieve daily landslide monitoring in highway slope scenarios.