Recognizing matured cinnamon tree using image processing techniques

摘要

Cinnamon cultivation is the main income source of a set of areas in Sri Lanka. Peeling cinnamon is a complex task of the cinnamon harvesting process after identifying the matured cinnamon trees. Expertise knowledge is essential to identify matured trees using the traditional method. Otherwise, it may cause the wastage of cinnamon, by cutting immature cinnamon trees. This research addresses the automated system to recognize the matured cinnamon trees using image processing techniques which can be used to identify the matured cinnamon trees without any expert knowledge. All the trees that are selected to this study are more than three year old. Image preprocessing, algorithm selection and use the Speeded Up Robust Features (SURF) features to extract data from leaves and at last, prediction of the maturity level of cinnamon trees using Support Vector Machine (SVM) classifier are the main phases of this study. Hundred cinnamon trees were tested from two different farms and the system performed 68.0% accuracy for matured trees and 86.0% accuracy for immature trees.