A Combination of Convolutional and Graph Neural Networks for Regularized Road Surface Extraction

Abstract

Road surface extraction from high-resolution remote sensing images has many engineering applications; however, extracting regularized and smooth road surface maps that reach the human delineation level is a very challenging task, and substantial and time-consuming manual work is usually unavoidable. In this paper, to solve this problem, we propose a novel regularized road surface extraction framework by introducing a graph neural network (GNN) for processing the road graph that is preconstructed from the easily accessible road centerlines. The proposed framework formulates the road surface extraction problem as two-sided width inference of the road graph, and consists of a convolutional neural network (CNN)-based feature extractor and a GNN model for vertex attribute adjustment. The CNN extracts the high-level abstract features of each vertex in the graph as the input of the GNN, and also the road boundary features that allow us to distinguish roads from the background. The GNN propagates and aggregates the features of the vertices in the graph to achieve global optimization of the regression of the regularized widths of the vertices. At the same time, a biased centerline map can also be corrected based on the width prediction result. To the best of the authors' knowledge, this is the first study to have introduced a GNN to regularized human-level road surface extraction. The proposed method was evaluated on four diverse datasets, and the results show that the proposed method comprehensively outperforms the recent CNN-based segmentation methods and other regularization methods in the intersection over union (IoU) and smoothness score, and a visual check shows that a majority of the prediction results of the proposed method approach the human delineation level.

Extraction of Roads Using the Archimedes Tuning Process with the Quantum Dilated Convolutional Neural Network

“…In addition, a lightweight encoder-decoder network is used in order to enhance the accuracy of road boundary extraction. Yan et al [38] proposed an innovative approach to road surface extraction, incorporating a graph neural network (GNN) that operates on a pre-existing road graph composed of road centerlines. The suggested method exploits the GNN approach for vertex adjustment and employs CNN-based feature extraction to define road surface extraction as a two-sided width inference problem of the road graph.…”

A holistic framework for forestry and rural road detection based on satellite imagery and deep semantic segmentation

“…State-of-the-art semantic segmentation algorithms using Deep Convolutional Neural Networks such as DeepLab, 2 U-NET 3 and others have been used for this task. Yan et al 4 combined a Convolutional Neural Network (CNN) with a Graph Neural Network (GNN) to improve performance. Specifically, the CNN was used to extract semantic features, while the GNN was used to apply reasoning on these features and determine the final road map.…”

ViGEO: an Assessment of Vision GNNs in Earth Observation

“…More recently, the research community started investigating the application of transformer-based models in the EO field, confirming better performances than CNNs in case of high data availability [10]. Few applications of vision GNNs are present in literature, limited to small datasets [34], [35]. To the best of our knowledge, this paper introduces the first evaluation of a vision GNN, namely ViG [15], to a largescale remote sensing multilabel benchmark and its adaptation to the EO domain.…”