Deep Learning-Based Building Extraction from Remote Sensing Images: A Comprehensive Review

Lin Luo,Pengpeng Li,Xuesong Yan

doi:10.3390/en14237982

Abstract

Building extraction from remote sensing (RS) images is a fundamental task for geospatial applications, aiming to obtain morphology, location, and other information about buildings from RS images, which is significant for geographic monitoring and construction of human activity areas. In recent years, deep learning (DL) technology has made remarkable progress and breakthroughs in the field of RS and also become a central and state-of-the-art method for building extraction. This paper provides an overview over the developed DL-based building extraction methods from RS images. Firstly, we describe the DL technologies of this field as well as the loss function over semantic segmentation. Next, a description of important publicly available datasets and evaluation metrics directly related to the problem follows. Then, the main DL methods are reviewed, highlighting contributions and significance in the field. After that, comparative results on several publicly available datasets are given for the described methods, following up with a discussion. Finally, we point out a set of promising future works and draw our conclusions about building extraction based on DL techniques.

Highlights

With the rapid development of imaging technology, high-resolution remote sensing (RS) imagery is becoming more and more readily available
We focus on RS-based building extraction using Deep learning (DL) semantic segmentation methods
Compared to other surveys on traditional building extraction methods or RS, this review paper is more devoted to the popular topic of DL semantic segmentation, covering the state-of-the-art and latest work

Summary

Introduction

With the rapid development of imaging technology, high-resolution remote sensing (RS) imagery is becoming more and more readily available. Deep learning (DL), with convolutional neural networks (CNN) [30,31,32,33,34] as its representative, is an automated artificial intelligence technique that has emerged in recent years, specializing in learning general patterns from large amounts of data as well as exploiting the knowledge learned to solve unknown problems It has been successfully applied and rapidly developed in areas such as image classification [35], target detection [36], boundary detection [37], semantic segmentation [16], and instance segmentation [38] in the field of computer vision. There are some reviews on RS image building extraction [39,40,41,42]

Methods

Results

Discussion

Conclusion