Mangosteen Quality Grading for Export Markets Using Digital Image Processing Techniques

Abstract

Accurate quality grading of mangosteen to meet the needs of consumers is very important for improving the value of the export business. Mangosteen fruit ripens quickly after harvesting, and shipping transportation time is a critical factor. Traditional grading methods by physical visual inspection result in delays and human-induced errors. This paper proposed an automatic grading system of mangosteen fruit that utilizes image processing techniques. The maturity stage, class, and size of mangosteen for the export market are analyzed. There are seven stages of maturity from stage one through to six and the under the mature stage, four classes (extra class, class B, class C, and non-standard class) and seven sizes (Jumbo through to Mini). Skin color, skin defect areas, completeness of calyx integrity are also considered. The preprocessing steps consisted of noise removal using a median filter and image enhancement using the grey level transformation. A combination of the mean intensity of red and green images was used to classify the maturation of the fruit. Areas damaged by yellow latex, cracks, and insect pests were extracted, and calyces were counted for class sorting. The length of the diameter was used for size classification. The thresholding, mathematical morphology, and extended minima transform techniques were also used. The average accuracy of the system was 99.54%, with a high accuracy rate for classifying the premium export grades. Results demonstrated that our proposed system was effective and could be used to improve productivity as an accurate and efficient grading method for mangosteen export.