Classification of canned pineapple based on first order colour statistics

• Main Author: Sharmiza, Kamaruddin
• Format: Thesis
• Language: English
• Published: 2016
• Subjects: TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://umpir.ump.edu.my/id/eprint/17658/1/Classification%20of%20canned%20pineapple%20based%20on%20first%20order%20colour%20statistics.pdf; http://umpir.ump.edu.my/id/eprint/17658/

Automatic systems for various applications in agriculture have been applied using an image processing method. The implementation of this system produces a low cost system, efficient, fast and reliable product inspection. Grading the colour of canned pineapple is one of the specifications in Malaysian Standard of Canned pineapple before export. There are a few studies on fresh pineapple using computer vision, but not for canned pineapple. Today, canned pineapple colour fully depends on experienced workers to grade according to the standard of MPIB. This may lead to the misclassification standard due to human factors such as fatigue, emotional condition, personal preference and honesty. The objective of this research is to use image processing technique that can replace human in grading the standard colour of canned pineapple in optimizing the quality of canned pineapple being export. In our literature review, we found that grading and sorting colour has been popular with various methods applied. The image acquisition taken at MPIB laboratory are using controlled environment to prevent the outside illumination. Otsu's method has been chosen as image filtering technique during pre-processing image since it gives the best results in eliminating almost unwanted pixels in the canned pineapple image. In order to smooth the Region of Interest (ROI), morphological operations using dilation and erosion were used. The disk shape of a structuring element with radii of nine (9) and ten (10) were used in this operation. Multiplying the ROI with original image was the last step before features extraction. First order statistic such as minimum,1 maximum, mean and standard deviation were calculated for each RGB, HSY, and CIELAB. 100% classifications for Standard 15 were obtained using standard deviation and 93 .3% for Standard 16 in HSY colour space. HSV colour space using hue were chosen as a method of classification in this research since it was simple for transformation of device-dependent RGB models and it was appropriate for human sight.