Research on Semantic Segmentation of High-resolution Remote Sensing Image Based on Full Convolutional Neural Network

Remote sensing data is an important way to reflect the comprehensive information of surface. In this paper, based on the semantic segmentation of high-resolution remote sensing images, a segmentation method based on full convolutional neural network (FCN) is proposed. The method improves the traditional convolutional neural network (CNN) and replaces the final fully connected layer of the CNN network with a convolutional layer. And then optimize the convolution operation by using the matrix expansion technique. The experimental results show that the FCN network with sufficient training and fine-tuning can effectively perform automatic semantic segmentation of high-resolution remote sensing images. The correct segmentation accuracy is higher than 85%, which improves the efficiency of convolution operations.