Artificial neural vision and digital image processing machines are alternative methods that can be done to identify and evaluate the diversity of rice varieties. In contrast to direct observational methods that have high levels of subjectivity and chemical methods (PCRs) that are both destructive and expensive, neural network-based visioning machines offer rapid, practical, inexpensive and accurate identification and evaluation systems, and are non-destructive. This paper discusses artificial neural network based vision technology as an alternative technology for the identification of South Kalimantan swamp varieties based on morphological features, ie area, perimeter, major axis, minor axis, circularity, aspect ratio, roundness and feret for each seed sample rice. In this paper, the identification system of rice seed varieties using Radial Basis Probabilistic (RBP) neural network with optimization of hidden center weight using Orthogonal Least Square (OLS) algorithm. From the learning process, the training performance is 88,329% and the testing performance is 88,2091%, with the success rate in the training process from each variety of Bayar Papuyu, Bayar Putih, Yellow Seed, White Seed, Ketan, Siam Gadis, Siam Unus and Karan Hamlet each of 100%; 92.59%; 88.89%; 92.59%, 92.59%, 81.48, 100%; and 100%. For the testing process the success rate of each variety is 100%; 87.50%; 88.89%; 100%, 88.89%, 88.89, 100%; and 100%.
Copyrights © 2013