Neighbor-aware informal settlement mapping with graph convolutional networks

TL;DR

This work addresses the challenge of detecting informal settlements by moving beyond independent cell classification to a graph-based approach that leverages local spatial context. Specifically, each target cell is enclosed with its neighboring cells into a local graph and processed by a lightweight Graph Convolutional Network, using a grid derived from a 150 m resolution and features from Sentinel-2, the Copernicus DEM, and OpenStreetMap roads. On a Rio de Janeiro case study with spatial cross-validation across five zones, the graph-based method consistently outperforms single-cell baselines and simple neighbor feature concatenation, achieving up to a 17-point gain in the Kappa coefficient. These results demonstrate the value of encoding spatial structure for fine-grained urban land classification and point to future opportunities such as learning the graph structure, incorporating temporal dynamics, or using attention mechanisms to further improve transferability.

Abstract

Mapping informal settlements is crucial for addressing challenges related to urban planning, public health, and infrastructure in rapidly growing cities. Geospatial machine learning has emerged as a key tool for detecting and mapping these areas from remote sensing data. However, existing approaches often treat spatial units independently, neglecting the relational structure of the urban fabric. We propose a graph-based framework that explicitly incorporates local geographical context into the classification process. Each spatial unit (cell) is embedded in a graph structure along with its adjacent neighbors, and a lightweight Graph Convolutional Network (GCN) is trained to classify whether the central cell belongs to an informal settlement. Experiments are conducted on a case study in Rio de Janeiro using spatial cross-validation across five distinct zones, ensuring robustness and generalizability across heterogeneous urban landscapes. Our method outperforms standard baselines, improving Kappa coefficient by 17 points over individual cell classification. We also show that graph-based modeling surpasses simple feature concatenation of neighboring cells, demonstrating the benefit of encoding spatial structure for urban scene understanding.

Figures (3)

• Figure 1: Grid-Based Partitioning of the Urban Space and an Example of Graph Constructed around a Specific Cell to Classify
• Figure 2: Our Graph-Based Classification Approach
• Figure 3: Study Area in Rio de Janeiro Divided into Five Distinct Urban Zones used for Cross-Validation