PERTANIKA JOURNAL OF SCIENCE AND TECHNOLOGY

 

e-ISSN 2231-8526
ISSN 0128-7680

Home / Regular Issue / JST Vol. 27 (3) Jul. 2019 / JST-1526-2019

 

Intelligent Non-Invasive Sensing Method in Identifying Coconut (Coco nucifera var. Ebunea) Ripeness Using Computer Vision and Artificial Neural Network

Yusuf Hendrawan, Aisyiyah Amini, Dewi Maya Maharani and Sandra Malin Sutan

Pertanika Journal of Science & Technology, Volume 27, Issue 3, July 2019

Keywords: Artificial neural network, coconut ripeness, computer vision, texture analysis and 0.107265 with actual value correlations and predictions as measured by R-training and R-testing

Published on: 24 July 2019

The use of coconut in the food industry is determined by the condition of its fruit ripeness level, which is very difficult to be conducted. The suitable non-invasive sensing method is the application of computer vision. The purpose of this study is to identify the coconut ripeness level based on several parameters of fruit volume, coconut flesh thickness, and coconut flesh weight by using Artificial Neural Network (ANN) modeling. The best ANN model resulted in 14 inputs consisting of color, texture, shape and size parameters. Color features include: Red, Green, Blue, Hue, Saturation, and Intensity. Textural features include: contrast, correlation, energy, homogeneity. The shape and size parameters include area, perimeter, eccentricity, and metric. The best ANN structure consisted of 14 inputs, one hidden layer with 100 nodes, and 3 outputs of coconut ripeness (indicated by coconut water volume, coconut flesh thickness and wet weight of coconut flesh). The best ANN model produced the smallest Mean Squared Error (MSE) training and MSE testing values of 0.002155 respectively of 1 and 0.90331. Thus, computer vision and ANN models can be utilized to predict the coconut ripeness level.

ISSN 0128-7680

e-ISSN 2231-8526

Article ID

JST-1526-2019

Download Full Article PDF

Share this article

Recent Articles