Road Extraction Methods in High-Resolution Remote Sensing Images: A Comprehensive Review

• Published in: IEEE journal of selected topics in applied earth observations and remote sensing Vol. 13; pp. 5489 - 5507
• Main Authors: Lian, Renbao, Wang, Weixing, Mustafa, Nadir, Huang, Liqin
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Adversarial networks; Artificial neural networks; Computational modeling; Convolution; Convolutional networks; Convolutional neural networks; Data; Data mining; Data-driven; Data-driven methods; decadal variation; Deep learning; Deep neural networks; Extraction; Extraction method; Feature extraction; Heuristic; Heuristic methods; heuristics; High resolution; High resolution remote sensing images; Hot research topics; image analysis; Image edge detection; Image processing; Image resolution; Machine learning; Neural networks; Number of methods; Remote sensing; remote sensing image; Resolution; road extraction; Road extraction method; Roads; Roads & highways; Roads and streets
• ISSN: 1939-1404; 2151-1535; 2151-1535
• DOI: 10.1109/JSTARS.2020.3023549

Road extraction from high-resolution remote sensing images is a challenging but hot research topic in the past decades. A large number of methods are invented to deal with this problem. This article provides a comprehensive review of these existing approaches. We classified the methods into heuristic and data-driven. The heuristic methods are the mainstream in the early years, and the data-driven methods based on deep learning have been quickly developed recently. With regard to the heuristic methods, the road feature model is first introduced, then, the classic extraction methods are reviewed in two subcategories: semiautomatic and automatic. The principles, inspirations, advantages, and disadvantages of these methods are described. In terms of the data-driven methods, the road extraction methods based on deep neural network, particularly those based on patched convolutional neural network, fully convolutional network, and generative adversarial network are reviewed. We perform subjective comparisons between the methods inner each type. Furthermore, the quantity performances achieved on the same dataset are compared between the heuristic and data-driven methods to show the strengthening of the data-driven methods. Finally, the conclusion and prospects are summarized.