Remote Sensing Image Surface Feature Classification Based on VGG-UNet

With the rapid development of aerospace remote sensing technology, the resolution of remote sensing image is constantly improving and the amount of data is constantly increasing. Therefore the requirements for surface feature classification are rising. However, due to the high resolution remote sensing image has the characteristics of strong complexity and rich scene edge details, the traditional image segmentation algorithm can not meet its requirements gradually. In this paper, we will quote the UNet neural network in deep learning to replace the encoder with VGGNet, build a remote sensing image segmentation trainer combined with data enhancement technology and migration learning theory, and utilize WHDLD and BDCI2017 datasets for training and verification to improve the final classification accuracy of remote sensing surface features.