Cemetery Detection Using Satellite Images in Google Earth Engine

Abstract

This work proposes a methodology to identify areas which may contain cemeteries by generating a grid over the terrain and then classify this grid with an ensemble of learning methods. The proposal is to characterize known cemeteries from satellite images information and building the training data. The learning methods are the Google Earth Engine's implementation of Random Forests, Support Vector Machines and Gradient Tree Boost Classifier. Results show that using an ensemble of these learning algorithms can decrease the number of false positives in the classification. Furthermore, little or no tuning of the classifiers is required for the ensemble to achieve good accuracy.