APPLYING MACHINE LEARNING ALGORITHMS AND WORLDVIEW-2 SATELLITE IMAGERY TO CLASSIFY CROP TYPES

摘要

Remote sensing techniques can be used to classify different crop types quickly and easily over a large area. Recently, several scholars have adopted the machine learning algorithms, such as support vector machine (SVM) and random forest, to classify various land use/land covers. The purpose of this study is to utilize the machine learning algorithms and the traditional maximum likelihood classifier to classify corn, peanuts, green manure and other categories. The effectiveness of each classifier for interpreting different crop types is evaluated. The scope of the study area is located at the Tuku township of Yunlin counties, range of Taiwan high-speed rail on both sides 1.5 km, total areas approximately 2700 ha. World View-2 images, dated October 2014 and November 2014, are adopted in the process.