Computer vision determination of the stem/root joint on processing carrots

Two algorithms for analysing digital binary images and estimating the location of stem/root joints in processing carrots were developed. The first method was based on an analysis of carrot outlines, and was referred to as the slope algorithm. The second method utilized a priori knowledge of diameter relationships in the area near the stem/root joint, and was called the midpoint algorithm. The accuracy, consistency, and operating speeds of the two algorithms were measured to determine the feasibility of using computer vision as part of an automated crown trimming device. Both algorithms were capable of estimating the location of the stem/root joint with an error standard deviation of approximately 5 mm. The midpoint method was considered superior due to its faster execution time and better consistency. A system implementing the midpoint algorithm could feasibly attain speeds exceeding ten carrots per second.