Object Based Remote Sensing Using Sentinel Data

Abstract

Identifying changes on the Earth's surface is one of the most fundamental aspects of Earth observation from satellite images. Historically, the predominant form of analysis has measured change at a pixel level. Here, we present a new strategy that conducts the analysis based on objects. The objects are placed inside a random forest regressor. We have tested our approach in Queensland, Australia using Sentinel data. We find that the use of object-based approach either outperforms or is comparable to alternative approaches.