LAUSR

Abstract. Fagaceae is the second-largest woody plant family in Taiwan and has considerable economic and ecological value. Identifying Fagaceae species is essential for forest managers and forestry technical personnel. In this study, ten Fagaceae species were distinguished using image processing and machine learning. For each species, 100 leaf images were collected using flatbed scanners. The morphological, marginal, color, and textural traits of the leaves were then quantified. A genetic algorithm was next applied to identify the traits that are essential for species identification. Support vector machine classifiers were developed to identify the species using the selected traits as the inputs. The results indicated that the proposed approach had an accuracy of 92.8%. Keywords: Fagaceae, Image processing, Machine learning, Species identification, Trait selection.