Land Cover Classification using Google Earth Engine's Object-oriented and Machine Learning Classifier

Google Earth Engine (GEE) provides users access to massive amounts of geospatial data that have been processed, allowing them to implement and combine pixel-based (PB) and object-oriented (OO) land cover classification approaches with machine learning algorithms. The goal of this study is to implement and test an OO method using superpixel segmentation based on Simple Non-Iterative Clustering (SNIC) algorithm to classify spatial clusters, and three commonly used machine learning algorithms to perform the final classification [Random Forest (RF), Classification and Regression Trees (CART), and Gradient Boost (GB)] for applied in rice paddy field in Lao Luang Sub-district, a part of Thung Kula Rong Hai, Thailand. Our data showed that the G B classifier gave the highest accuracy and almost perfect level of agreement, while the RF gave lesser results and excellent level of agreement. Unexpectedly, CART produced different results for PB (Good level of agreement) and OO (Excellent level of agreement). In this study, the OO was more efficient than that of the PB in terms of overall efficiency.