Color Based Feature Extraction and Backpropagation Neural Network in Tamarind Turmeric Herb Recognition

The aim of this paper is finding the optimum image pattern tamarind turmeric herb. So far, in process producing herb, it not constant terms taste and color, which influenced by maturity amount turmeric. Image recognition will use Backpropagation algorithm applied to typical Content-based retrieval systems. main purpose apprehend various parts herb retrieving processing. camera classify product, into 5x5 pixels, take an average RGB value so stable values be obtained each category used as input for algorithm. most suitable fastest from searched a real-time machine. In using two methods, first, train ten data change neuron, layer, momentum, learning rate, last testing with data. results training that hidden layers can recognize 100% inputs, three R, G, B value, neurons first second layers, one output layer parameter Learning rate 0.5 Momentum 0.6. best standard dark yellow 255, 102, 32 up 128, 48.