Identification of Long Bean Seed Varieties Using Digital Image Processing Coupled With Neural Network Analysis

Identification of long bean seed varieties can be used to save plant variety and intellectual property rights. Using digital image processing combined with artificial neural networks (ANN) has a possibility to recognize the seed morphology. The purpose of this research is to identify the image variables that can be used to identify long bean seed varieties so that the best algorithm of artificial neural networks can be arranged and the level of accuracy in expecting the long bean varieties. The samples used in this study were long bean seeds of parade tavi, kanton tavi, branjangan, and petiwi varieties. For each variety, 400 samples were taken for training data and 200 samples for testing data, so the total sample was 2400 long bean seeds. The research stages include image acquisition, image retrieval, image variable estimation, image processing program development, data analysis, ANN training, long bean variety identification program preparation, and program validation. The results showed that ANN with 10 hidden layers is the best model to develop a long bean seed identification. The identification program of long bean seed varieties resulting from the integration of image processing with artificial neural networks has an accuracy of 99.75%.