Abstract

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.

Highlights

  • In 1972, the first Earth Resources Observation Technology Satellite, later renamed Landsat, was launched by the United States

  • We have reviewed the traditional methods before [8], we stick to present a more comprehensive review from a new perspective given the rapid development of deep learning and the numerous articles discussing automatic road extraction employing deep learning models [9]

  • Inspired by [41], Poz and Do Vale [42] introduced road width into the metric function, which can be used for road extraction in high-resolution remote sensing images (HRSI)

Read more

Summary

BACKGROUND

R EMOTE sensing images acquired by airborne or spaceborne sensors are the main resource for the earth surface observation, environment monitoring, objects identification, etc. [1]. The basic processing of HRSI is to extract information for objects classification and recognition according to the spectral features and the shape characters and spatial relations [5]. We will get more than 2.6 million related papers if we search Google Scholar with the keyword “road extraction,” and more than 129 000 articles remain if the publish date is restricted to post 2016 Faced with such a large number of works of literature, a systematic review of road extraction algorithms is valuable for beginners. In order to review the heuristic and data-driven methods systematically, we collected and sieved more 200 papers discussing road extraction from HRSI published in the past two decades. To the best of our knowledge, this article is the first comprehensive review that surveys the road extraction methods based on traditional algorithms and deep learning technology

INTRODUCTION
ROAD FEATURES AND MODEL
Road Features
Road Model
HEURISTIC ROAD EXTRACTION METHODS
Semiautomatic Methods
Automatic Methods
DATA-DRIVEN ROAD EXTRACTION METHODS
Patch-Based DCNN Methods
DeconvNet-Based Methods
Graph-Based Methods
METRICS
COMPARISONS
Comparison of Automatic Methods
Comparison of Data-Driven Methods
Comparison Between Heuristic and Data-Driven Methods
Findings
VIII. CONCLUSION AND PROSPECTS
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call