The Development of Soybean Grading Algorithm Using Image Processing and Artificial Neural Network

Abstract

Grading method influences the homogeneity of each grade of such product and a good grading method provides producer and consumer with standard for the price of the product. The objective of this research was to develop a computer program of image processing and artificial neural network to identify the quality of fresh soybean into four classes namely SQ (standart quality), SG (second grade), TG (third grade), and RJ (reject) using image processing and artificial neural network. The total samples were 2500 fresh soybean produced by PT. Mitra Tani Dua Tujuh Jember. Soybean image was analyzed to get six quality parameters whose match with soybean quality criteria namely pod length, pod area, perimeter, defect area, index of red color, and index of green color. Those six quality parameters will be used as inputs of the artificial neural network (ANN). Six variations of ANN were developed for ANN training purposes (2000 data). The weights of the selected ANN architecture was used to identify the quality class of testing data (500 data), then integrated with image processing program so the program could identify soybean quality class automatically. The quality parameter used in this research has relevancy with soybean quality criteria. The selected architecture of the ANN was the one with 20 nodes hidden layer in which normalization input data representation with zero mean and standard deviation equals one. The accuracy of image processing program observed 81, 4 percent based on the 500 testing data.