| Abstrakt: |
The goal of this study is to determine the best picture pattern for the tamarind turmeric herb. So far, the taste and color of tamarind turmeric herb have not been consistent, as they are impacted by maturity, and the amount of Turmeric used. The error backpropagation technique, which is commonly used in Content-based image retrieval systems, will be used to recognize image patterns. The main goal is to capture various portions of the tamarind turmeric herb during the extraction procedure. The camera is used to classify the tamarind turmeric herb product, process it into 5x5 pixels, and average the RGB value to obtain stable RGB values in each category, which are then fed into the Error Backpropagation algorithm. The most appropriate and fastest Error Backpropagation algorithm procedure will be found and implemented in a real-time computer. The first way will be to train the algorithm with ten data by changing neurons, layer, momentum, and learning rate, and the second technique will be to test the algorithm with ten data. The results of the training and testing procedure show that the two hidden layers can recognize 100% of inputs, with three input layers for R, G, and B values, ten neurons in the first and second hidden layers, and one output layer with Learning rate 0.5 and Momentum 0.6 as a parameter. Dark yellow is the best image pattern standard for tamarind turmeric herb, with RGB values in the range from 255, 103, 32 to 255, 128, 48. [ABSTRACT FROM AUTHOR] |
