Exploiting deep learning and volunteered geographic information for mapping buildings in Kano, Nigeria

Abstract

Buildings in the developing world are inadequately mapped. Lack of such critical geospatial data adds unnecessary challenges to locating and reaching a large segment of the world’s most vulnerable population, impeding sustainability goals ranging from disaster relief to poverty reduction. Use of volunteered geographic information (VGI) has emerged as a widely accepted source to fill such voids. Despite its promise, availability of building maps for developing countries significantly lags behind demand. We present a new approach, coupling deep convolutional neural networks (CNNs) with VGI for automating building map generation from high-resolution satellite images for Kano state, Nigeria. Specifically, we trained a CNN with VGI building outlines of limited quality and quantity and generated building maps for a 50,000 km2 area. Resulting maps are in strong agreement with existing settlement maps and require a fraction of the manual input needed for the latter. The VGI-based maps will provide support across multiple facets of socioeconomic development in Kano state, and demonstrates potential advancements in current mapping capabilities in resource constrained countries.

Mapping Human Settlements with Higher Accuracy and Less Volunteer Efforts by Combining Crowdsourcing and Deep Learning

“…For the case of automated village boundary detection Gueguen et al (2017) [8] report an average precision of around 70% and a sensitivity of around 84%. Yuan et al (2018) [30] compare a deep learning-based approach against GUF and GHSL for Kano city (Nigeria) and reach similar results. Their building extraction algorithm performs with a precision of 72% and a sensitivity of 70%.…”

“…Regarding land cover mapping, several authors propose workflows based on deep neural networks with a focus on urban areas [27,28]. Furthermore, building footprint extraction based on deep learning has been a central research topic in recent years [29,30].…”

“…The transferability of deep learning models constitutes a key challenge towards global scale data products, e.g., for human settlements. For instance, Yuan et al (2018) [30] highlight difficulties in extracting footprints for buildings in rural areas which differed significantly from the footprints presented in the training set. Missing benchmark data sets tailored to remote sensing tasks make it difficult to compare the growing number of deep learning algorithms [24].…”

Principles and Applications of the Global Human Settlement Layer as Baseline for the Land Use Efficiency Indicator—SDG 11.3.1

“…This new generation of information support researchers and policymakers in the understanding of the new geographies of human settlements in the era of a predominantly urban society. In particular, the finer spatial resolution of sensors contribute to the improvement of the detection of small roofed surfaces and urban features extraction [72], reducing omission and commission errors [64,73], and improving settlement mapping in territories subject to intense and fast transformations (especially in Africa and Asia).…”

The impact of seasonality on multi-scale feature extraction techniques

“…The 240 classified maps for Charlotte were compared on a pixel by pixel basis against LiDAR (Light Detection and Ranging) derived building features with a spatial resolution of 0.5 meters (Government 2013). The 240 classified maps for Kano were compared on a pixel by pixel basis with building features produced by deep convolutional neural networks (CNNs) with a spatial resolution of 0.5 meter (Yuan et al 2018). These CNNs utilized volunteered geographic information building outlines as training and generated a data set with impressive validation results that could be considered comparable to those of LiDAR building features.…”