Research And Implementation Of Pearl Shape Automatic Sorting System Based On Machine Vision

In order to solve the problem of low automation and intelligent level of pearl sorting,a set of pearl shape sorting system based on monocular dynamic vision is designed for the pearl shape visual recognition.This paper mainly studied the method of obtaining multi surface of pearl by monocular camera and the method of geometric shape detection.According to the research content,realized the pearl detection platform of this system,the main function for image acquisition,data / information processing and network communication,provided the basis for selecting control of the manipulator,and the manipulator control are presented,the key technology to realize the sorting system.In the aspect of solving the pearl image acquisition,the paper designed and implemented a monocular dynamic image acquisition device,which can capture the surface images of pearl at many angles in a short time.In image processing,the shape of the pearl is combined with various shape descriptors Geometric description,according to the national standard "GB / T 18781-2008" step by step to separate the various types of pearls,and calculate the position of pearls in the image;arm control,the camera calibration and coordinate conversion positioning pearl actual three-dimensional coordinates,The use of network transmission protocol and computer communications,control sorting.In the system,single pearl image acquisition takes 0.1s,and the average processing time of each image is 0.063 s.With the high-speed movement of the manipulator,the sorting of pearl can be completed.The mean error rate of the system in shape detection is within 5.38%,compared with the industry's test in the ideal test environment,the average error rate of five categories was 6.60%,an increase of two categories and the average error rate decreased by 1.22%.