|
MODEL FOR PREDICTING AND CLASSIFYING DURIAN FRUIT BASED ON DEFECTS USING NEURAL NETWORK
Amin Rejo1), Suroso1), Hadi K. Purwadaria1), I Wayan Budiastra1), Yul Y. Nazaruddin2)
1) Dept. of Agricultural Enginering, Bogor Agricultural University (IPB), Indonesia
2) Dept. of Engineering Physics, Bandung Institute of Technology (ITB), Indonesia
This study was aimed to develop the model to predict the defects of durian based on its physical and chemical characteristics by using the neural network. The diameter, mass, volume and acoustic characteristics measurement was fed into the model as the inputs, which provided the levels of defects as the output. Data training were tested to models of neural network with various nodes in the hidden layer, i.e., 4, 6, and 8 nodes. The results recommended the use of 3 nodes in the hidden layer that would provide the highest accuration of 75.76 90.90% in classifying the durian based on its non-defects and 61.76 94.11 on its defects.
Keywords: neural network, durian, defects, mass, volume, zero moment power
Session slot T-Tu-E21: Posters of Agricultural, Biological and Environmental Systems/Area code 4a : Modelling and Control in Agricultural Processes
 |
