Abstract:Many cities in low- and medium-income countries (LMICs) are facing rapid unplanned growth of built-up areas, while detailed information on these deprived urban areas (DUAs) is lacking. There exist visible differences in housing conditions and urban spaces, and these differences are linked to urban deprivation. However, the appropriate geospatial information for unravelling urban deprivation is typically not available for DUAs in LMICs, constituting an urgent knowledge gap. The objective of this study is to apply deep learning techniques and morphological analysis to identify degrees of deprivation in DUAs. To this end, we first generate a reference dataset of building footprints using a participatory community-based crowd-sourcing approach. Secondly, we adapt a deep learning model based on the U-Net architecture for the semantic segmentation of satellite imagery (WorldView 3) to generate building footprints. Lastly, we compute multi-level morphological features from building footprints for identifying the deprivation variation within DUAs. Our results show that deep learning techniques perform satisfactorily for predicting building footprints in DUAs, yielding an accuracy of F1 score = 0.84 and Jaccard Index = 0.73. The resulting building footprints (predicted buildings) are useful for the computation of morphology metrics at the grid cell level, as, in high-density areas, buildings cannot be detected individually but in clumps. Morphological features capture physical differences of deprivation within DUAs. Four indicators are used to define the morphology in DUAs, i.e., two related to building form (building size and inner irregularity) and two covering the form of open spaces (proximity and directionality). The degree of deprivation can be evaluated from the analysis of morphological features extracted from the predicted buildings, resulting in three categories: high, medium, and low deprivation. The outcome of this study contributes to the advancement of methods for producing up-to-date and disaggregated morphological spatial data on urban DUAs (often referred to as ‘slums’) which are essential for understanding the physical dimensions of deprivation, and hence planning targeted interventions accordingly.

Towards a scalable and transferable approach to map deprived areas using Sentinel-2 images and machine learning

Identifying degrees of deprivation from space using deep learning and morphological spatial analysis of deprived urban areas

Monitoring Slum and Urban Deprived Area in Sub-Saharan Africa Using Geospatial and Socio-Economic Data

Mapping Slums with Medium Resolution Satellite Imagery: a Comparative Analysis of Multi-Spectral Data and Grey-level Co-occurrence Matrix Techniques

Identifying patterns in urban housing density in developing countries using convolutional networks and satellite imagery

Phenotyping urban built and natural environments with high-resolution satellite images and unsupervised deep learning

Local vs. Global: Local Land-Use and Land-Cover Models Deliver Higher Quality Maps

Transfer Learning from Deep Features for Remote Sensing and Poverty Mapping

Population Mapping in Informal Settlements with High-Resolution Satellite Imagery and Equitable Ground-Truth

Granularity at Scale: Estimating Neighborhood Socioeconomic Indicators From High-Resolution Orthographic Imagery and Hybrid Learning

Automated National Urban Map Extraction

Mapping Slums in Mumbai, India, Using Sentinel-2 Imagery: Evaluating Composite Slum Spectral Indices (CSSIs)

Mapping Africa Settlements: High Resolution Urban and Rural Map by Deep Learning and Satellite Imagery

Scalable Machine Learning Approaches for Neighborhood Classification Using Very High Resolution Remote Sensing Imagery

Convolutional neural networks for global human settlements mapping from Sentinel-2 satellite imagery

Deep Learning for Slum Mapping in Remote Sensing Images: A Meta-analysis and Review

Utilizing Satellite Imagery Datasets and Machine Learning Data Models to Evaluate Infrastructure Change in Undeveloped Regions

Lightweight and Robust Representation of Economic Scales from Satellite Imagery

A Cloud-Based Mapping Approach Using Deep Learning and Very-High Spatial Resolution Earth Observation Data to Facilitate the SDG 11.7.1 Indicator Computation

Large-scale School Mapping using Weakly Supervised Deep Learning for Universal School Connectivity

Combining disparate data sources for improved poverty prediction and mapping