Building Footprint Identification
The Issue
Countries in emerging economies generally do not have reliable civil/health records, nor periodic, extensive or updated population and housing censuses. This can mean millions of children are left out and forgotten - unable to access the healthcare, education and protection they require. We want to make these invisible children visible.
Why Does it Matter?
Article 24 of the Convention on the Rights of the Child states that every child has the right to the best possible health. Similarly, the UN Sustainable Development Goals have a specific focus to ensure healthy lives and the promotion of well-being for all ages. However, with unreliable census data and health records, it is difficult to maintain an accurate understanding of the number of children in each community. This can heavily impact service planning and delivery, leaving some children at risk of missing out on life-saving vaccinations and treatment.
Having a better understanding of the community - its size, number of buildings and characteristics - will allow UNICEF to tailor interventions and services more accurately and reach as many children as possible.
Our Project
UNICEF is currently working in Chad and Mozambique to improve target population estimates at the local level in order to better support health services planning and delivery. We want to explore if we can create a model that is better able to distinguish and systematise building types using satellite images, in turn providing better populations estimates at a local level. This will help to support health services and delivery, with a goal to initially inform immunisation programming, but with scope to be transferable to other programmes in the future.
The primary work package will involve performing an extensive literature review on population estimates and building footprint identification, as well as analysing the existing databases on estimated population density and building footprints to see if they are sufficient for UNICEF’s needs. We will then assess the performance of existing methods, both from reported performance in the literature and from running available tools on data that is readily accessible, to explore if we have adequate and relevant data for more extensive machine learning approaches. This will enable us to identify any gaps in the data and methods before we move to further project phases.
Our end goal is to build software that uses satellite images and survey data to fine tune an existing predictor and apply it to non-surveyed areas. We also aim to develop software that extracts objects from satellite imagery -such as buildings and tents – by either using GPS locations or ‘dot annotation’ methods.
We will automate and further improve existing building identification algorithms by more narrowly focusing on classifying rooftop structures. Our output will give users information such as the number of buildings, house counts and the area of rooftops alongside information available from local registries. This will enable better estimates of local populations and a more accurate understanding of where children at risk are located.