Machine Learning-Powered UAV Imaging for Landslide Crack Identification

Landslides pose a significant threat to both human lives and economic well-being. Studying landslide behavior involves monitoring accumulated displacement and ground crack evolution. To this end, the integration of scientific recognition methods with high-resolution spatial–temporal imagery plays a crucial role in mitigating landslide disasters. In this study, we introduce a machine learning-based technique for detecting surface cracks using high-resolution images from unmanned aerial vehicles (UAVs). Our research focuses on a representative landslide site, Ba Zimen, situated in Zigui County along the Three Gorges Reservoir in China. A UAV captured optical images of the Ba Zimen landslide, which were subsequently divided into four datasets based on crack characteristics (Dataset 1: 52, Dataset 2: 53, Dataset 3: 49, and Dataset 4: 56). We employed two advanced machine learning technologies, namely the Support Vector Machine (SVM) and Artificial Neural Network (ANN), for landslide crack recognition. The optimal machine learning algorithm was selected based on the results. The outcomes revealed that SVM outperforms the ANN algorithm. SVM achieved an average recognition accuracy of 70.55%, whereas ANN yielded a recognition accuracy of 66.72%. Therefore, the amalgamation of high-resolution UAV imagery and machine learning algorithms holds great promise for landslide crack identification.