Aprendizaje automatizado basado en transformadas wavelet para la segmentación de hojas en imágenes infrarrojas

Abstract

In this work we present an algorithm to segment leaves in NIR images captured inside a growth chamber. The proposed method uses a features vector composed by four elements extracted from different levels of wavelets transform. We have tested three classifiers: knearest neighbour (kNN), Naive Bayes classifier (NBC) and Support Vector Machine to determine the optimal machine learning algorithm to carry out the leaf segmentation. Method developed has been validated by means of the Receiver Operating Characteristic (ROC) curve and it has obtained a maximum performance of 98.33% in the leaf segmentation using SVM classifier.