The Classification of Freycinetia Based on Stomata Anatomic Image Using K-Nearest Neighbor and Artificial Neural Network

Abstract

The classification of Freycinetia is conducted to portray the potential result and function of nutfah plasma spreader. This research is purposed to classify a variety of Freycinetia based on stomata anatomic image by K-Nearest Neighbor (K-NN) and Artificial Neural Network (ANN). The Variety of Freycinetia classification can be described by using two different ways that are morphology and anatomic characteristic. Stomata anatomic image is used to support morphology characteristic in the identification process, especially in uncompleted morphology image specimen and sample. There are two feature extraction techniques in this research. First, is to pick up the color and gray scale element in the image. The scale element has a red, green and blue color (RGB). The grayscale has an entropy, contrast, energy, homogeneity, gray scale and deviation standard. Second, is a wavelet decomposition technique which has a function for decreasing a size of the image without loosing its important element. The taken element becomes a coefficient image of w-entropy for each wavelet level. Ninety six data is used in this research, it contains of four kinds of Freycinetia. They are Freycinetia angustifolia, Freycinetia imbricata, Freycinetia javanica, and Freycinetia sumatrana. The result shows an accuracy level for K-NN according to RGB and gray scale is 86.46% and according to wavelet decomposition is 96.88%. Whereas, the accuracy level of ANN based on RGB color and gray scale element is 94.79% and according to wavelet decomposition is 9 9%. The classification result of plant variety, especially Freycinetia based on stomata image with high accuracy, will be an alternative tool to identify taxonomy type.