This research aimed to measure the chlorophyll content of Moringa leaves using machine vision and an optimized artificial neural network (ANN). A total of 480 images were used, 70% as training data and 30% as validation data. Features extraction was used to extract color and textural features. ANN was used as a modeling method, and the filter method was used as a feature selection method to optimize the best ANN input. Sensitivity analysis was done by varying the attribute evaluator in the filter method, as well as the learning function, the activation function, the learning rate, the momentum, the number of hidden layers, and the number of hidden nodes in the ANN. The best ANN structure was 10 input nodes, 30 nodes in the hidden layer 1, 40 nodes in the hidden layer 2, and 1 output node when using a learning rate of 0.1, a momentum of 0.5, the traincgf learning function, a logsig activation function in the hidden layer, and a tansig activation function in the output layer. The correlation coefficient between predicted and real data in the training process was 0.9792 with the training mean square error (MSE) of 0.0100, and the correlation coefficient of the validation process was 0.9794 with the validation MSE of 0.0099.
Copyrights © 2022