Determination of the Volume of Unboiled Arecanut Kernels using Image Processing

Abstract Arecanut cultivation, though widespread in India, accounting for 50% of the global output, relies on visual attributes such as colour, texture, size, shape, and tactile traits like softness for quality assessment. The paradigm shift towards alternate industrial and medical applications emphasises the need for grading based on physical and mechanical properties. Real-time measurement of crucial physical parameters, particularly the volume of kernels, poses a considerable challenge that is not yet effectively addressed. The study focused on estimating the volume of unboiled arecanut kernels cultivated in Goa, India, utilising real-time image processing with segmentation. The devised system employed two cameras positioned at right angles to determine the diameters and thickness of the kernels. The research findings disclosed an average difference of 2.67% between volume estimations by image processing and the water displacement method. The paired t -test results revealed no significant difference ( P >0.05) between the volumes obtained from the two methods. Additionally, Spearman’s rank correlation coefficient demonstrated a very strong correlation between the two measurement techniques. The proposed method overcoming the limitations of the ellipsoidal method achieved 97.33% accuracy with the potential for real-time measurement of the kernel volumes, providing a basis for automated quality grading based on density and porosity.