Design And Research Of Ring Blank Feature Classification Vision System

With the continuous development of machining technology,the requirement of automation technology for mass production of products is higher and higher.Driven by the huge market,the production of jewelry,a traditional handicraft,also faces the problem of transformation from pure manual to automation.However,there are many kinds of jewelry styles,and the surface is mostly irregular space curved surface and curve,which brings challenges to the determination of machining reference point,and it is difficult to achieve complete automation in a short period of time.This paper focuses on the technology of "automatic sorting" and takes it as the starting point to improve the processing technology of this kind of products.Using machine vision with deep learning is the main way to realize "automatic classification".By using machines instead of human eyes,not only the production efficiency and automation level of mechanical equipment can be improved,but also the reliability of processing results and the safety of workers can be ensured.Therefore,the research of this paper is based on the establishment of visual classification system for ring feature detection.According to the specific work content and design requirements,this paper proposes a classification scheme characterized by ring blank size and outer contour stone inlay,and designs the overall system of visual classification platform.The ring feature visual inspection platform is designed,and the industrial camera and lens are selected.Combined with the existing sorting mode,the mechanical arm and the mechanical gripper structure of the sorting system are designed,the control scheme is determined,and the required hardware modules are selected.Aiming at the sorting system,the kinematic modeling analysis of the manipulator is carried out and its maximum moving space is determined,which lays the foundation for object grabbing.In order to reduce the load pressure of the manipulator and meet the requirements of lightweight design,the topological optimization modeling analysis of the manipulator is carried out based on ANSYS,and the rationality of the clamping range and the smoothness of the motion process are verified based on ADAMS.Aiming at the size feature of the ring,the camera and pixel ratio value are calibrated,and the circle feature is extracted by traditional image processing and improved Hough circle detection algorithm,and the clamping center of the mechanical gripper and the expanding sleeve type number required for machining machine tool are determined.In view of the ring's outer contour inlay way,a classification standard based on the diamond inlay mode on the ring's outer surface is raised,the data set is constructed and enhanced,the structure of Le Net-5 deep learning model is improved,and the data set is trained by the improved model.The GUI of the system is designed based on Py Qt,which realizes algorithm encapsulation and human-computer interaction.For the scheme proposed above,the algorithm of extracting ring size features is tested,and the parameter optimization experiment of deep learning model is carried out.Finally,the trained classification model of ring shape inlay is obtained.The experimental results show that the detection accuracy,classification accuracy and model classification accuracy of the circle extraction algorithm are very high.At the same time,the human-computer interface can achieve a good encapsulation of the above functional detection.It can be seen that the visual classification system has good performance,smooth operation and basically meets the expected requirements.