Study on Physical Feature Extraction of Fermented Cocoa Bean with Digital Image Processing

The study aimed to develop software for estimating the size of the fermented cocoa bean. The size of cocoa bean is one of the quality attributes needed in secondary cocoa processing, especially those related to the activities of sorting and preparing raw materials. This system was developed by adopting digital image processing techniques, starting from sample preparation and followed by capturing the image of cocoa bean samples, converting the color image into monochrome, segmentation, improving image results with a combination of dilation and erosion processes, to the feature extraction as the final stage. The extracted features of the beans' size are length, width, circumference, and area of the samples, all measured in pixel unit. The features, called as estimated size, were then regressed to the real samples' size, called as physical features, measured physically in the laboratory. The obtained regression formula showed a strong relationship between the estimated and physical features. The strong correlation between the two features enables the development of a rapid estimation of the physical features of cocoa beans based on digital image processing.