Artificial neural network and shortwave near-infrared light in pineapple internal quality classification

• Main Author: Jam, Mohamad Nur Hakim
• Format: Thesis
• Language: English; English; English
• Published: 2018
• Subjects: QA Mathematics; QA71-90 Instruments and machines; QA76.75-76.765 Computer software
• Online Access: http://eprints.uthm.edu.my/7487/1/24p%20MOHAMAD%20NUR%20HAKIM%20JAM.pdf; http://eprints.uthm.edu.my/7487/2/MOHAMAD%20NUR%20HAKIM%20JAM%20COPYRIGHT%20DECLARATION.pdf; http://eprints.uthm.edu.my/7487/3/MOHAMAD%20NUR%20HAKIM%20JAM%20WATERMARK.pdf; http://eprints.uthm.edu.my/7487/
• Institution: Universiti Tun Hussein Onn Malaysia

The determination of the internal quality of pineapples is important for grading the fruit quality. The common procedure to determine the internal quality of the fruit is using a digital refractometer that are destructive, expensive, and time-consuming. In this research, a non-destructive approach was designed to classify the internal quality of pineapples using Artificial Neural Network (ANN). The device applied diffuse reflectance mode using five Near-Infrared (NIR) light with different wavelengths that were in the range of 780 nm and 940 nm. A photodiode sensor (OPT101) was used to collect the reflected light from pineapples non-destructively. Firstly, the reflected NIR light from the top, middle, and bottom sections of a pineapple were non-destructively acquired using a designed NIR device. After that, the Soluble Solids Content (SSC) of the pineapple was conventionally acquired using a digital reflectometer. Next, the relationship between the NIR light and the SSC of pineapples was established using ANN. The potential outliers were identified and excluded using box-plot and leave-one-out cross-validation. The effects of the random seed and the hidden neurons of ANN were investigated to optimize the classification accuracy. K-fold cross-validation was used to validate the performance of the classification accuracy. The results indicate that the top, middle, and bottom of pineapples can be categorized using the proposed NIR approach with the classification accuracy of 76%, 85%, and 77.78% for training, testing, and K-fold cross-validation analysis, respectively. Next, when the proposed method was tested using new pineapples, findings indicate that the predictive model is capable of achieving 77.62% accuracy without any outlier detection. Thus, the combination of ANN and shortwave NIR light is able to rapidly, and non-destructively classify the internal quality of pineapples with satisfactory accuracy