Socio Economic Analysis of India with High Resolution Satellite Imagery to Predict Poverty

Abstract

Eradicating poverty is the numero uno objective of the United Nations for sustainable development of the world by 2030. But, in order to develop a feasible, targeted solution to this problem, an exact poverty map is required. In India, especially in rural areas, there is a dearth of reliable and frequent data related to indicators of poverty line as the national statistics division of the country releases data only once in five years. In this paper, we look at an alternative to the slow, ineffective collection of data on ground: mapping poverty from outer space using medium and high-resolution satellite imagery. Using both satellite imagery and survey data for the rural areas of India, we review how machine learning tools like convolutional neural networks have been harnessed to efficiently identify image features that help us effectively predict socio-economic indicators of poverty. We also explore how these methods offer promising means for policy makers to tackle poverty at the grassroot level and a potential for application across several domains of science.