Delineation of Road Networks Using Deep Residual Neural Networks and Iterative Hough Transform

System overview. The input RGB image is processed using SegNeXt and results in a grayscale classification image of road and non-road pixels. The classification image is then divided into patches which are further processed. The refinement process involves an iterative application of patch-based Hough transforms which results in a set of extracted lines. Erroneously extracted lines resulting from misclassification are removed, and nearby lines are either connected (if not parallel) or suppressed (if parallel). The result is a set of vectors representing the road network in the input image shown in yellow overlaid on the input image.

PAPER

SOURCE CODE

SUPPLEMENTAL MATERIAL

Abstract

In this paper we present a complete pipeline for extracting road network vector data from satellite RGB orthophotos of urban areas. Firstly, a network based on the SegNeXt architecture with a novel loss function is employed for the semantic segmentation of the roads. Results show that the proposed network produces on average better results than other state-of-the-art semantic segmentation techniques. Secondly, we propose a fast post-processing technique for vectorizing the rasterized segmentation result, removing erroneous lines, and refining the road network. The result is a set of vectors representing the road network. We have extensively tested the proposed pipeline and provide quantitative and qualitative comparisons with other state-of-the-art based on a number of known metrics.

Keywords

road network extraction, residual neural networks, semantic segmentation

System Overview

The input to our system is an RGB image which is fed forward into a deep autoencoder network (SegNeXt) with aggregated residual transformations. The network outputs a semantic segmentation of the image in the form of a grayscale image in which each pixel is classified into a road or non-road classes. The classification image is then divided into patches which are further processed. During the refinement process, an iterative patch-based Hough transform is applied. Extracted lines are tracked from one patch to the other. Erroneously extracted lines resulting from misclassification are removed, and nearby lines are either connected (if not parallel) or suppressed (if parallel). The result is a set of vectors representing the road network in the input image. summarizes the system overview.