DETERMINATION OF CACAO QUALITY USING THE MUMFORD-SHAH FUNCTIONAL ALGORITHM AND IMAGE SEGMENTATION

One of the major issues in the agricultural industry is cocoa quality determination that may be based on subjective sensory assessment techniques that may cause inconsistency and losses after harvesting. The following research will solve the set of issues by taking the Mumford-Shah functional algorithm with the image segmentation, to evaluate the quality of cocoa at various stages of development: un-ripe, ripe, and diseased. The operation of the segmentation is performed on MATLAB simulation, and the images are processed to measure the primary metrics like how many regions they have, the average area, the average perimeter, and the average eccentricity. The outcomes of the segmentation indicate a clear difference in such measures in the three stages of cocoa growth. In the case of unripe cocoa there are 997 regions and the average region area is 10,509.80 pixels, the average perimeter is 18.40 pixels and the average eccentricity is 0.20. In case of ripe cocoa, there are 768 regions with the average area of 13,997.49 pixels, perimeter of 24.57 pixels, and the average eccentricity of 0.21. The diseased cocoa had 1,705 regions with a mean area of 5,936.61 pixels, mean perimeter of 11.68 pixels and with a mean eccentricity of 0.21. The discussion has shown that the Mumford-Shah algorithm offers an accurate means of grading cocoa quality and has been shown to offer major benefits over the conventional sensory assessment procedures, post-harvest losses and quality control in cocoa production.