A Semantic Segmentation Method for Remote Sensing Images based on Deeplab v3
Zhaoyong Qian, Yuhua Cao, Zengkai Shi, Luyi Qiu, Chenguang Shi · 2021
As a basic technology for image analysis and scene understanding, image semantic segmentation is widely used in the field of remote sensing images. It can better help humans understand the world’s scenes and analyze potential changes from the top view of the earth, which has high practical value and development prospects. In this paper, a semantic segmentation method for high-resolution remote sensing images based on DeepLab v3[9] is proposed. The network structure of this method uses the atrous spatial pyramid pooling (ASPP) to extract the multi-scale feature information of high-resolution remote sensing images. Besides, the experimental part of this paper is evaluated on the public dataset of remote sensing images named ISPRS Vaihingen. Experimental results show that the proposed method can achieve an average accuracy of 81.72% on the ISPRS Vaihingen dataset. Therefore, this method can be used as an automated tool for remote sensing semantic segmentation.