Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
May 13, 2026ISPRS International Journal of Geo-InformationOpen Access

Topology-Aware Road Extraction from Remote Sensing Images Using Deep Learning and Graph-Based Connectivity Refinement

View Full Paper
Ask AI
Bookmark
Share

Authors

ZTZixuan TengZZZezhong ZhengXSXiangyang Sun

Discussion

Loading...

Member takes

Overview

Randomized trial demonstrates improved road connectivity in remote sensing images, suggesting practical applications for urban planning.

Key Points

  • The aim is to develop a method that enhances road extraction from remote sensing images by addressing topological fragmentation.
  • Employs Pyramid Scene Parsing Network (PSPNet) for initial road probability maps.
  • Integrates a graph-based connectivity refinement strategy using a multi-source cost function and direction-aware Dijkstra algorithm.
  • Applies dynamic road width restoration to convert refined skeletons into consistent road entities.
  • Conn metric improved by 0.1989 on CHN6-CUG and 0.3055 on DeepGlobe datasets.
  • Marginal decrease in MIoU of 1.07% on CHN6-CUG and 0.45% on DeepGlobe.
  • Method effectively restores structural continuity for reliable road network generation.

Cite This Study

Teng et al. (2026) studied this question.

synapsesocial.com/papers/6a04153d79e20c90b4445066https://doi.org/10.3390/ijgi15050208
View Full Paper
Ask AI
Bookmark
Share