As a part of a whole machine-vision system, an image processing
technique was developed and elaborated to classify potatoes
based on shape analysis. The experimental machine vision system
was designed and constructed to test the image processing algorithms’
ability to detect misshapen potatoes. Algorithms were
developed to acquire and process images of potatoes in real time
while they travel over a high speed conveyor belt. Stepwise linear
discriminant analysis was used to determine which shape parameters
were most responsible for detecting misshapen potatoes.
Roundness and extent as size–shape parameters in addition to
other four parameters calculated from Fourier expansion of the polar
signatures of potato boundary were found to be effective in
describing potato shapes. The in-line evaluation of the proposed