Size Classification of Tomato (Lycopersicum Esculentum) Using Machine Vision
Michael T. Dayacap1, Pepito M. Bato1, Omar F. Zubia1, and Arsenio N. Resurreccion1
ABSTRACT
A machine vision system capable of classifying tomatoes according to size was developed. Tomato Sizer Software (ToSS) processed the tomato image by thresholding, boundary mapping, and storing in array the boundary points. The parameters calculated using the boundary point arrays were projected area, diameter, height, shape index, and weight. Tests for precision of values measured by the system at 24 orientation angles from 0 to 345 degrees showed that the system can analyze tomato at any position relative to the camera. Two equations for weight estimation were derived from diameter-weight (Wt1) and area-weight (Wt2) correlations. The weight, as a function of the projected area, obtained the highest accuracy (91.1%). Accuracy of the sizing algorithm was found to be 90%. The average judgment time was 0.2 second.
Keywords: machine vision system, size classification, tomato
1BSAE graduate, Associate Professor, Assistant Professor, and Professor, respectively, of the Agricultural Machinery Division (AMD), Institute of Agricultural Engineering (IAE), College of Engineering and Agro-Industrial Technology (CEAT), University of the Philippines Los Baños (UPLB), 4031 College, Laguna, Philippines.