ANALISIS FITUR TEKSTUR DAUN MANGGA DENGAN FISHER’S DISCRIMINANT RATIO UNTUK PENCAPAIAN FITUR YANG INFORMATIF

Mango tree species recognition system based on the texture of leaves on the previous system gives accuracy up to 88.89%. This indicates that feature selection of research on the mango tree species recognition need to be taken into account. In this research, analysis of mango leaf texture features are used. There are 3 types of features that is: statistics, invariant moment, and the co- occurrence matrix. Methods for analyzing the feature is Fisher's Discriminant Ration (FDR). This method obtained  from  a number of  informative  features,  that  is:  energy (co-occurrence), uniformity (statistics), the third moment (statistics), entropy (co-occurrence), entropy (statistical), and homogeneity (co-occurrence). Performance testing is done by comparing the use of old and new features on the K-Nearest Neighbor method with the value K is 3, 5, 7, and 11. The results showed that the accuracy of the K-NN method with new features reach 0.90, and tend to be better than the old features