Computer vision for potato inspection without singulation

A computer vision system has been developed for grading potatoes into size and shape categories and interfaced with a grader to work at a speed of up to 40 potatoes per second. A feature of the system is that it requires no singulation leading to less complex and hence faster machinery. Lack of singulation makes it difficult for the vision system to discriminate single objects when they can form touching groups. This paper describes the algorithms developed to avoid singulation and to identify each potato reliably and quickly. Test results are presented on the effectiveness of the algorithm. These showed successful operation of the algorithm on 1000 potatoes, most of which were touching with no failures. Details of the computer system are covered in other publications.