Research On Key Technology Of Bearing Cylindrical Roller Size Measurement Based On Machine Vision

Cylindrical roller bearings are mainly load-bearing in mechanical equipment.The difference of roller dimension accuracy directly affects the mechanical properties and life of bearings.At this stage,the cylindrical roller size measurement is mainly manual,and the detection efficiency is low and it is difficult to satisfy the accuracy requirements of the bearing.Therefore,designing a multi-dimensional measurement system for bearing cylindrical rollers is an urgent problem to be resolved.Aiming at this problem,this paper uses a combination of theoretical analysis,image system design,software algorithm design and physical experiment to study the fundamental technology of cylindrical roller size measurement.Research on image preprocessing of cylindrical roller sizes measurement.First,according to the cylindrical roller characteristic analysis and measurement principle,the cylindrical roller size measurement image acquisition imaging and illumination system research are carried out to avoid edge annihilation due to improper image acquisition;secondly,the beating during the cylindrical roller acquisition process is studied according to the measurement principle.Establish an unfolding model to analyze the relationship between the rotational angular velocity of the cylindrical roller and the runout displacement,and determine the optimum unfolding angular velocity of the cylindrical roller;finally,conduct a preprocessing study on the image distortion during the acquisition process,and provide accurate roller dimensions for subsequent identification and positioning image.Research on image edge recognition and positioning algorithm of cylindrical roller.Based on the above-mentioned reprocessed image,there will still be the problem of edge blurring and distortion.First,conduct the study of cylindrical roller edge coarse positioning algorithm to eliminate interference factors.Second,study the sub-pixel edge subdivision algorithm,combined with the best edge fitting algorithm to solve the image edge For information loss and discontinuity issues,accurate image edge contours are obtained,and the pixel size of the length and diameter of the cylindrical roller is determined and extracted.Create a physical prototype of the cylindrical roller size image acquisition system,design a size measurement software system and write related programs,and calibrate the obtained measurement image size.Select 10 samples for length and diameter measurement,and each sample was collected 20 time and compared with the standard roller size.Combine MATLAB programming to construct LSTMNN neural network prediction model.Predict the measurement data size 100 time to judge whether the measurement accuracy of the system meets the design requirements.This paper studies the key technology of bearing cylindrical roller dimension measurement based on machine vision,which provides a theoretical basis for the application of cylindrical roller multi-size non-contact measurement equipment.