Double-channel on-line automatic fruit grading system based on computer vision

The technology of fruit grading based on computer vision was studied and a double-channel on-line automatic grading system was built. The process of grading included fruit image acquiring, image processing and fruit tracking and separating. In the first section, a new approach of image grabbing by employing an asynchronous reset camera was presented. Three images of the different surfaces of each fruit would be collected by rolling the fruits when they passed through the image-capturing area. To acquire clear images, high-frequency fluorescent lamps supplied by three-phase alternating current were used to illuminate. In the image processing section, the diameter and a color model were used to identify the grade of the fruits. Fruits were graded into four grades by size, and two by color. Each fruit identified was tracked and separated by a novel algorithm which was realized with a PLC (Program Logic Controller). The whole grading system was tested with 1000 citrus. It could work stably when the grading capability was twelve citrus per second and the grading level was nine. The on-line grading results indicated that the accuracy of tracking and separating was higher than 99%, and the ultimate grading error was less than 3%.