Holistically-Nested Structure-Aware Graph Neural Network for Road Extraction

TL;DR

This paper introduces a multi-task graph neural network that jointly detects roads and their borders from satellite images, improving the delineation of complex road structures by leveraging holistic topological information.

Contribution

The novel GNN architecture simultaneously detects roads and borders, capturing road topology and structure to enhance extraction accuracy over existing CNN-based methods.

Findings

Improved road border delineation accuracy

Enhanced road connectivity recognition

Better performance on challenging datasets

Abstract

Convolutional neural networks (CNN) have made significant advances in detecting roads from satellite images. However, existing CNN approaches are generally repurposed semantic segmentation architectures and suffer from the poor delineation of long and curved regions. Lack of overall road topology and structure information further deteriorates their performance on challenging remote sensing images. This paper presents a novel multi-task graph neural network (GNN) which simultaneously detects both road regions and road borders; the inter-play between these two tasks unlocks superior performance from two perspectives: (1) the hierarchically detected road borders enable the network to capture and encode holistic road structure to enhance road connectivity (2) identifying the intrinsic correlation of semantic landcover regions mitigates the difficulty in recognizing roads cluttered by regions with similar appearance. Experiments on challenging dataset demonstrate that the proposed architecture can improve the road border delineation and road extraction accuracy compared with the existing methods.