Estimating GeoJSON Coordinates using Image Processing to Improve Census Credibility

Abstract

Census is the process of gathering, analyzing, compiling, and spreading social, cultural, demographic, and economic data relating to all the people in a country. A census gives a statistically accurate view which is important to fill the gaps in the system. Enumerators are majorly responsible for the credibility of the census, and to maintain its reliability, it is important to monitor their location to confirm no random form fills. However, there is a lack of GeoJSON data for small, remote villages, districts, and talukas. This hinders the monitoring process. Hence, we devise a method to retrieve GeoJSON data from an image of the map and the border GeoJSON of the parent map in the hierarchy, using computationally efficient image processing. The proposed pipeline involves a 5 step process that includes preprocessing, extracting boundary coordinates, determining the scaling factor, inner boundary localization, and mapping. The results are computed by comparing the areas of the predicted and actual polygons of the retrieved regions whilst confirming that there is a massive overlap between the two polygons. An error rate of 4.87% is achieved (95.13% accuracy).