Intelligent technique for grading tropical fruit using magnetic resonance imaging
Recent application of modern marketing techniques coupled with intelligent agricultural systems of production has transformed small scale farming into large scale, in most part of the world. Characteristically, most of the tropical fruits, such as orange, appeared edible physically but internally su...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IJSER
2013
|
| Subjects: | |
| Online Access: | http://irep.iium.edu.my/35993/1/4._Intelligent_Technique_for_Grading_Tropical_Fruit_using_Magnetic_Resonance_Imaging.pdf http://irep.iium.edu.my/35993/ http://www.ijser.org/ResearchPaperPublishing_July2013_Page1.aspx |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Recent application of modern marketing techniques coupled with intelligent agricultural systems of production has transformed small scale farming into large scale, in most part of the world. Characteristically, most of the tropical fruits, such as orange, appeared edible physically but internally such fruits might be defective based on their tissue and juice. Eventually, these fruits, via the market and undetected, usually get to the consumers who encounter the unfavourable status of the fruits. Our purpose, in this study, is to develop a non-destructive method to predict the status of orange fruits, based on internal quality. Graph of histogram showing the levels of different four colour intensities were acquired and analysed. The features extracted from Magnetic Resonance Imaging (MRI), using any of the two proposed methods, were applied as an input to train artificial neural network (ANN) in order to predict the orange fruit status. Different structures of multi-layer perceptron neural networks with feed-forward and back-propagation learning algorithms were developed using MATLAB. The theoretical background of MRI and artificial neural network (ANN) backpropagation were also explained. At hidden neuron value of 20, search is for backpropagation and number of neurons in the hidden layer to optimize the ANN. Levenberg-Marquardt algorithm (trainlm) gave the best performance fitness out of different types of backpropagation algorithm used with least Mean Square Error (MSE) of 0.0814 corresponding to R-value of 0.8094. This work shows that ANN and MRI have the capability of predicting the internal content and detect defect fruit based on water proton content.
|
|---|