LAUSR

Abstract. Carrot grading is a labor intensive, time-consuming process and is usually performed manually in practical manufacturing. Manual inspection poses many problems in maintaining consistency and guaranteeing the detection efficiency. To improve the grading efficiency and achieve automatic detection, we developed an automated carrot sorting system using machine vision technology. The system consisted of an image processing system, an image acquisition system, a roller conveying system and a control system. It first picked out carrots with surface defects and then graded the qualified carrots by length. We proposed detection methods for three kinds of surface defects: misshapen, fibrous root, and surface crack. Given the fact that the regular carrot is convex-shaped, convex polygon was used to detect carrot shape. We proposed concave point method to detect the fibrous root and adopted Hough Transform to detect surface crack. Experimental results showed that the proposed methods could not only achieve satisfying detection accuracy but also high efficiency. The accuracy of curvature, fibrous root, and surface crack were 95.5%, 98% and 88.3%, respectively. The proposed methods and constructed sorting system could meet the demand of carrot grading and sorting.