A Deep Learning-Based Solution for Large-Scale Extraction of the Secondary Road Network from High-Resolution Aerial Orthoimagery

Abstract

Secondary roads represent the largest part of the road network. However, due to the absence of clearly defined edges, presence of occlusions, and differences in widths, monitoring and mapping them represents a great effort for public administration. We believe that recent advancements in machine vision allow the extraction of these types of roads from high-resolution remotely sensed imagery and can enable the automation of the mapping operation. In this work, we leverage these advances and propose a deep learning-based solution capable of efficiently extracting the surface area of secondary roads at a large scale. The solution is based on hybrid segmentation models trained with high-resolution remote sensing imagery divided in tiles of 256 × 256 pixels and their correspondent segmentation masks, resulting in increases in performance metrics of 2.7–3.5% when compared to the original architectures. The best performing model achieved Intersection over Union and F1 scores of maximum 0.5790 and 0.7120, respectively, with a minimum loss of 0.4985 and was integrated on a web platform which handles the evaluation of large areas, the association of the semantic predictions with geographical coordinates, the conversion of the tiles’ format and the generation of geotiff results compatible with geospatial databases.

Development of Nationwide Road Quality Map: Remote Sensing Meets Field Sensing

“…Remote sensing optical data obtained from UAVs or airborne/spaceborne platforms are mainly used for the extraction of roads, as described in several previous studies [ 18 , 19 , 20 , 21 , 22 ]. Meanwhile, synthetic aperture radar (SAR) data, especially high-resolution X-band SAR datasets, focus on the quality of the road.…”

Generative Learning for Postprocessing Semantic Segmentation Predictions: A Lightweight Conditional Generative Adversarial Network Based on Pix2pix to Improve the Extraction of Road Surface Areas

“…Many resulting segmentation masks presented discontinuities, and the connection points were often overlooked, resulting in road segments that were unconnected. We also identified higher rates of "false positive labels in areas where the materials used in the road pavement have a similar spectral signature with their surroundings, or areas where geospatial objects with similar features are present (such as dry riverbeds, railroads, or irrigation canals) and higher rates of false negatives in sections where other objects cover large portions of the roads were covered" (page 13 in [8]). Similar problems are still observed in recent works dealing with the road extraction from high-resolution aerial imagery-improving the road extraction task is an active area of research [11][12][13][14].…”

“…Land 2021, 10, 79 2 of 15 In one of our previous works [8], we studied the appropriateness of using state-of-theart segmentation models for extracting the surface areas of secondary roads and conducted a large-scale evaluation on unseen areas, obtaining IoU and F1 scores of 0.5790 and 0.7120, respectively (with 97.87% of the samples being correctly classified). However, even the best performing state-of-the-art segmentation model (U-Net [9] as base architecture with SERes-NeXt50 [10] as backbone network) displayed the problem of inaccurate extraction.…”

“…To overcome the deficiencies observed in our previous work [8], we developed a postprocessing technique based on image-to-image translation [15] concepts to operate over the initial semantic segmentation predictions and improve the road surface extraction for automatic mapping purposes. In this work, we apply generative learning techniques and propose a conditional Generative Adversarial Network (cGAN) architecture based on Pix2pix [15], greatly improved for computational efficiency (92.4% decrease in the number of parameters in the generator G network and 61.3% decrease in the discriminator D network, when compared to the original Pix2pix) to improve the initial road surface area extraction.…”

Urban Road Change Detection using Morphological Processing

“…The solution is based on hybrid segmentation models trained with high-resolution remote sensing imagery divided in tiles of 256 * 256 pixels and their correspondent segmentation masks, resulting in increases in performance metrics of 2.7-3.5% when compared to the original architectures, 17 October 2020. [4] The road extraction process can perform one or more operations, such as image segmentation (classification method), line segments of a certain width (Huff transforms and edges detector), contours of objects based on contours, merging associated with removing droplets. Road segments (morphological work), similarity to road patterns, etc.…”