Abstract
In this paper, the development of a novel quality assessment system for Bombyx mori L. cocoons is presented, which offers significant advantages over the conventional manual method (subjective, tests only few sample cocoons, involves health hazards) in terms of labor friendliness, accuracy, speed and running cost. This system consisted of a conditioned illumination unit, image acquisition and processing unit realized with a smart camera. The camera acquired the images of cocoons and by image processing algorithms (morphological operation, image enhancement, and ellipse fitting), quantitative measurements of size, shape and stain color were accomplished and automatically classified each cocoon into four defective categories and good cocoons. The system not only highlighted each category on camera screen but also displayed statistical information such as counts of cocoons in each category and overall defect percentage. In addition to that, the system was programmed to alert the user when the defect percentage exceeded a particular threshold value. The results showed that the system was capable of assessing 96 cocoons per second acquired within a single frame. It showed 100% accuracy on a sample size of 137 cocoons. To expose whole cocoon surface, they were rolled over a slope at a speed of eight rotations per second, while the system captured and processed the video of the whole surface. This process enabled in meeting the same level of quality assessment standard and counting accuracy as that of manually exposing the defective areas to the field of view when acquired in a single image.
