Intelligent Landslide Extraction Method Based on Deep Learning
The aim of this paper is to extract landslides quickly,efficiently and intelligently to improve the efficiency of landslide dynamic monitoring.We explore the application of the YOLO V8 model in automatic interpretation of high-resolution remote sensing landslide image.By establishing landslide interpretation samples in different regions,we constructed a recognition sample library oriented to geohazard monitoring.Based on the YOLO V8 deep learning network structure and pre-training model,we constructed an intelligent interpretation model for large image landslides to rapidly and automatically identify and extract landslides.Findings indicate that the landslide results extracted by the model are highly overlapped with the manually interpreted landslide extraction boundaries,with basically the same shape and good segmentation accuracy.On the validation set,the mAP50 reached 0.995,the mAP50-95(M)was 0.716 18,and the recall rate was 0.904 24.Compared with the traditional convolutional neural network(CNN),this model has higher accuracy and faster processing speed,providing technical support for landslide dynamic monitoring.
NETL
NSTL
万方数据
