Abstract. At present, the manual grading of soybean seeds is both time consuming and laborious, and detecting the full-surface information of soybean seeds using an existing automatic sorting machine is difficult. To solve this problem, an on-line omnidirectional inspection and sorting system for soybean seeds was developed using embedded image processing technology. According to the principles employed by the system, the surface friction properties and full-surface information such as the shape, texture and color of soybean seeds were adopted in the study. Soybean seeds were inspected and sorted using their full surface information in combination with the embedded image processing technology. Split, worm-eaten, gray-spotted, slightly cracked, moldy and normal soybeans were used to test the system. According to the test results, the optimum design parameters of the preliminary sorting device based on the friction properties were a tilting angle of 12° and a linear velocity of 0.4 m/s. Furthermore, the optimum design parameters of the directional integrated device were a tilting angle of 19° and a linear velocity of 0.45 m/s. The sorting speed was 400 soybeans per minute with 8-channel parallel transmission. The average sorting accuracies were 99.4% for split soybeans, 98.5% for worm-eaten soybeans, 98.5% for gray-spotted soybeans, 97.7% for slightly cracked soybeans, 98.6% for moldy soybeans, and 98.9% for normal soybeans. The overall results suggest that the system can potentially meet the needs of the rapid inspection and automatic sorting of soybean seeds and provide references for research on the alternating rotational motion of granules and on-line collection of full-surface information. Keywords: Embedded image processing technology, Full surface, Granules, Inspection, On-line, Sorting, Soybean seeds.
Development of On-line Resin Pellet External Testing and Automatic Selecting Control Systems by using Image Processing Technology.
