THE APPLICATION OF SUPPORT VECTOR MACHINES AND RANDOM FORESTS FOR INTERPRETING GARLIC PLANTED FIELDS

摘要

Seven types of crops have been named as sensitive crops by the Taiwan Council of Agriculture because of the crash of price caused by unstable production. Among these sensitive crops, the garlic is the most important one in Taiwan and the garlic planted fields in Yun-Ling County are more than 85% of the whole country. The machine learning algorithms traditionally have been used for medical, accounting, and information analysis. Recently, algorithms, such as support vector machine (SVM) and random forests (RF), has been utilized in the remote sensing application and acquires promised results. Therefore, this research applies the maximum likelihood (ML), SVM and RF classifiers to interpret garlic planted fields in Sihhu Township of Yunlin County. The very high spatial resolution Pleiades imagery, dated January 12, 2014, is used in this research. The classification accuracies of using these three classifiers are compared to determine their efficiencies in interpreting garlic planted fields.