ABSTRACT

Building detection from very high resolution (VHR) optical remote sensing images, which is an essential but challenging task in remote sensing, has attracted increased attention in recent years. However, despite the many methods that have been developed, an in-depth review of the recent literature on building extraction from VHR optical images is still lacking. In this article, we present a comprehensive review of the recent advances (since 2000) in this field. In total, we survey and summarize 417 articles in terms of the building detection method, post-processing, and accuracy assessment. The building detection methods are categorized into physical rule based methods, image segmentation based methods, and traditional and advanced machine learning (i.e. deep learning) methods. Furthermore, four promising related research directions of building polygon delineation, building change detection, building type classification, and height retrieval from monocular optical images are also discussed. Overall, building detection from VHR optical images is a popular research topic that has received extensive attention, due to its great significance. It is hoped that this review will help researchers to have a better understanding of this topic, and thus assist them to conduct related work.

摘要

超高分辨率（VHR）光学遥感图像的建筑物检测是遥感中一项必不可少但具有挑战性的任务，近年来引起了越来越多的关注。然而，尽管已经开发了许多方法，但仍然缺乏对最近关于从 VHR 光学图像中提取建筑物的文献的深入回顾。在本文中，我们全面回顾了该领域的最新进展（自 2000 年以来）。我们总共从建筑物检测方法、后处理和准确性评估方面调查和总结了 417 篇文章。建筑物检测方法分为基于物理规则的方法、基于图像分割的方法以及传统和先进的机器学习（即深度学习）方法。此外，还讨论了建筑多边形描绘、建筑变化检测、建筑类型分类和单目光学图像高度检索四个有前景的相关研究方向。总体而言，VHR 光学图像的建筑物检测是一个热门的研究课题，因其意义重大而受到广泛关注。希望这篇综述能帮助研究人员更好地理解这个课题，从而帮助他们开展相关工作。