Rural Road Extraction from GF-2 Image Based on Deep Learning
In view of the lack of research and application of high-resolution remote sensing images in rural road extraction,the extrac-tion results have some problems such as unclear boundaries and incomplete roads.In this paper,taking GF-2 image as the data source,an improved DeepLabv3+deep learning model is proposed.MobileNetV2 is selected as the feature extraction network,and the coordinate attention module is inserted into the reverse residual block of MobileNetV2,so that the network can capture the information with accurate position.At the same time,channel attention mechanism is added after the spatial pyramid pool,and multi-scale atten-tion mechanism is used to merge the output features,focusing on the features with more information and avoiding using multiple similar feature maps.The experimental results show that the accuracy of the improved DeepLabv3+model is 85.74%,the recall rate is 83.21%,and the F1 score is 0.84.Compared with the original DeepLabv3+model,each precision has been improved to some ex-tent.This study can provide some technical support for high-precision extraction of rural roads.
NETL
NSTL
万方数据
