Agarwood chip grading based on color using image processing and artificialintelligence methods

• Main Author: Mad Amin, Mohamad Razi
• Format: Thesis
• Language: English
• Published: 2014
• Subjects: Agarwood (Wood); Grading; Artificial intelligence
• Online Access: http://psasir.upm.edu.my/id/eprint/98038/1/FK%202015%20194%20UPMIR.pdf; http://psasir.upm.edu.my/id/eprint/98038/

Agarwood is the primary material especially in perfume industry. If agarwood with different grades are mixed, the quality will decrease and hence the products prices and quality will decrease. During the grading stages, human knowledge and experience is critically used in decision making. However, human characteristics often tender to fatigue where lead to produce misclassification. The price of agarwood is influenced by the resin content which can be indicated by the colour. The highest grade of agarwood chips has a shining black colour while the lower grade of agarwood chips has a black colour that is alternating with the brown colour. This study was conducted to determine the relationship of agarwoodcolour properties and its related price by adopting the method of artificial intelligence and image processing. Colouragarwood images in Red, Green, Blue, (RGB), Hue, Saturation, Intensity (HIS) and CIE colorimetric space (CIELAB) has been evaluated by comparing the performance of colour pixels classification using Fuzzy C-Means (FCM) method. The performance measurement was done through the evaluation of classification accuracy using 5 cluster validity indices. The result of cluster validity indices shows that CIELAB colorspace with 4 number of cluster proven the most consistent in FCM classification. In the later stage, the use of statistical measurement i.e. Analysis of Variance (ANOVA) and Duncan Multiple Range Test (DMRT) gave a significant relationship when classifying five out of seven grades of agarwood chips used i.e. RM250, RM350, RM800, RM900 and RM2500. Then, the artificial intelligence system using fuzzy logic and neural network concept has been developed and their performance has been compared. The result shows that fuzzy logic system successfully classified 62.8% of overall accuracy while neural network system gives 58% of overall accuracy in grading the agarwood chips. As a conclusion, the proposed system is helpful to the agarwood industry especially in determination of agarwood chips color during the grading process.