Road Extraction from High Resolution Remote Sensing Images Based on Vector Field Learning.

Peng Liang,Haolv Shang,Wenzhong Shi,Zhiqiang Liu,Yixing Ding

doi:10.3390/s21093152

Abstract

Accurate and up-to-date road network information is very important for the Geographic Information System (GIS) database, traffic management and planning, automatic vehicle navigation, emergency response and urban pollution sources investigation. In this paper, we use vector field learning to extract roads from high resolution remote sensing imaging. This method is usually used for skeleton extraction in nature image, but seldom used in road extraction. In order to improve the accuracy of road extraction, three vector fields are constructed and combined respectively with the normal road mask learning by a two-task network. The results show that all the vector fields are able to significantly improve the accuracy of road extraction, no matter the field is constructed in the road area or completely outside the road. The highest F1 score is 0.7618, increased by 0.053 compared with using only mask learning.

Highlights

In order to improve the accuracy of road extraction, three vector by vector field learning
In order to improve the accuracy of road extraction, three vector fields, CVF, Road Vector Field (RVF)
Background Vector Field (BVF), are constructed, and road extraction experiments are carried out fields, CVF, RVF and BVF, are constructed, and road extraction experiments are carried by using a two-task network

Summary

Introduction

With the rapid development of remote sensing technology, massive high-resolution images are accessible to the public and provide sufficient and high-quality data for road automatic extraction. Road extraction from high resolution remotes sensing images has become one of the research hotspots in the fields of photogrammetry, remote sensing, computer vision and geographic information science. BVF is a vector field constructed for road background. The vector of each pixel in road area is set to (0,0), and the vector of each background pixel is pointing to its nearest road pixel, namely the nearest zero pixel. By calculating the vectors of all background pixels in turn, the BVF in Figure 4(c) can be obtained. The vector field f can be given by: whole road area by assigning the non-centerline road pixel with the vector of centerline d. The vector field f can be given by: whole road area by assigning the non-centerline road pixel with the vector of centerline d. p by:

Objectives

Methods

Results

Conclusion