Research On O-ring Size Measurement And Surface Defect Detection System Based On Machine Vision

O-ring is an extremely simple structure but widely used industrial production sealing element.O-rings can be seen everywhere in weapons,ships,aerospace,and petrochemical industries.Moreover,there are not a few major accidents caused by the quality of the sealing ring.Therefore,ensuring the production quality of the O-ring has become a key and cannot be ignored.Nowadays,the inspection method of O-ring seals in my country still relies on manual size measurement to identify defects with naked eyes.However,due to the huge production volume of the sealing ring,the efficiency of pure manual inspection is low,and the inspection is easy to miss and misdetect,it is increasingly unable to meet the production needs of enterprises.Through research,design and verification,this subject has finally completed an online real-time inspection system for O-rings based on machine vision to improve the current status of sealing ring inspection,including the design of the inspection system hardware platform,the design of the software interface,and the size measurement and surface of the sealing ring.Design of defect recognition and classification algorithm.First,optimize the mathematical morphology edge detection algorithm.Based on the original combination of multiple morphological algorithms,new multi-structure elements are added to make the edge detection effect have better sharpness,clear and smooth edges,and better retain the original details of the image.An improved least squares fitting algorithm is proposed to minimize the variance of the radius by iterating the center of the circle,and on this basis,a piecewise fitting circle algorithm is proposed.Effectively eliminate the influence of small error estimates and improve the accuracy of size measurement.It can effectively eliminate the influence of the measurement result caused by the irregular circle,and further reduce the error.The Faster R-CNN model based on the convolutional neural network algorithm is used,and the model evaluation index is formulated.Compared with traditional image recognition and pattern recognition algorithms,not only can achieve 97% defect recognition rate,the recognition rate is greatly improved,and the recognition accuracy of defect categories is also greatly improved.At the same time,it also has real-time detection,and has broader requirements for the detection environment.Designed and built the O-ring visual inspection system hardware platform and software system,and designed and realized the industrial assembly line inspection simulation.Through the scene test experiment of various types of sealing rings,the size of the O-ring sealing ring can be effectively measured and whether there are defects can be identified,and the correct judgment of the quality of the sealing ring can be made.It further verifies the practicability of the designed O-ring visual on-line inspection system.