Mapping Informal Settlements in Developing Countries using Machine Learning and Low Resolution Multi-spectral Data

Abstract

Informal settlements are home to the most socially and economically vulnerable people on the planet. In order to deliver effective economic and social aid, non-government organizations (NGOs), such as the United Nations Children's Fund (UNICEF), require detailed maps of the locations of informal settlements. However, data regarding informal and formal settlements is primarily unavailable and if available is often incomplete. This is due, in part, to the cost and complexity of gathering data on a large scale. To address these challenges, we, in this work, provide three contributions. 1) A brand new machine learning data-set, purposely developed for informal settlement detection. 2) We show that it is possible to detect informal settlements using freely available low-resolution (LR) data, in contrast to previous studies that use very-high resolution (VHR) * Both authors contributed equally to this research. satellite and aerial imagery, something that is cost-prohibitive for NGOs. 3) We demonstrate two effective classification schemes on our curated data set, one that is cost-efficient for NGOs and another that is cost-prohibitive for NGOs, but has additional utility. We integrate these schemes into a semi-automated pipeline that converts either a LR or VHR satellite image into a binary map that encodes the locations of informal settlements.

Satellite-Based Mapping of Urban Poverty With Transfer-Learned Slum Morphologies

“…In [23], [24], nighttime light intensities were used as a proxy for poor urban areas to transfer learn a CNN to high-resolution remote sensing data. In [25]- [27], fully convolutional neural networks were used to map slums on either high-resolution or very high-resolution data, while [18], [28] used different transfer learning techniques to map slums between different satellite sensors in the same city and between geographically separated cities, respectively. The authors concluded that not only more data, but also a novel deep learning architecture and more rigorous regularization is necessary for robust segmentation of slums on a large scale.…”

Big Data and Energy Poverty Alleviation

“…The latest studies looked at daylight images that can help distinguish different levels of economic well-being in developing countries [91][92][93][94]. Even though there is access to satellite imagery for the entire globe, many of these raw data are not in a usable format for machine learning frameworks, making it difficult to extract actionable insights at scale [92,95]. The earliest approach that combined machine learning with satellite imagery and crowdsourced assistance to identify poor villages in Kenya was presented by two NGO Give Directly volunteer researchers in 2014 [76,96].…”

“…The very high cost of VHR imagery per image or collection of images motivated Perez et al [99] and Gram-Hansen et al [95] to carry out their research by using freely available multi-spectral imagery, which is typically of a much lower resolution. In particular, Perez et al [99] trained CNN models on free and publicly available multispectral daytime satellite images of the African continent from the Landsat 7 satellite, which has collected imagery with global coverage for almost two decades.…”

“…Although the Landsat images are of lower resolution than the Google Static Maps images used by Jean et al [91], the authors achieved equivalent or better results than the previous benchmarks. Furthermore, Gram-Hansen et al [95] introduced a brand new machine learning dataset, purposely developed for informal settlement detection and they showed that it is possible to detect informal settlements using freely available low-resolution (LR) data, in contrast to previous studies that use very-high resolution (VHR) satellite and aerial imagery.…”

Predicting Poverty Index using Deep Learning on Remote Sensing and Household Data

“…In sequence of a series of such contributions, one is to identify the location for the informal settlement of vulnerable people using even low resolution images. They used back propagation algorithm with weights trained on PASCAL dataset and demonstrated classification schemes on their curated datasets [13].…”