Identifikasi Jenis Aglaonema Menggunakan Probabilistic Neural Network

Abstract

Aglaonema is an ornamental plant that is quite popular in Indonesia. It is estimated that there are nearly 8000 species Aglaonema in the world both native and hybrid. The many types of Aglaonema in the world causes difficult in identifying some types of Aglaonema. This research attempts to identify the type of Aglaonema based on the image using Probabilistic Neural Network (PNN). The data used in this study have 900 images of leaves, which consists of 30 types of Aglaonema. Each image of Aglaonema that will be identified by the system will first be subjected to two stages, texture feature extraction and colour feature extraction. The texture feature extraction used Local Binary Pattern Variance (LBPV) and Co-occurrence Matrix while the colour feature extraction used Histogram-162 (HSV-162). We perform two type of experiment, one where we uses each feature seperately and another where the two features are combine. For classifier we used Probabilistic Neural Network (PNN). The results indicates that the combination of Co-occurrence Matrix with HSV-162 yield a better accuracy compact two when the feature are used seperately. On the other hand, the combination of Local Binnary Pattern Variance (LBPV) and HSV-162 does not yield an increase in accuracy, however the accuracy on this case is better than the accuracy of the combination between Co-occurrence Matrix and HSV-162. The highest accuracy is obtained in the case of the Local Binnary Pattern Variance (LBPV) and HSV-162 with the value of 55.56%