A Bearing Roller Surface Defect Detection System Based On Machine Vision

A machine-vision algorithm is used to identify bearing-roller surface defects.The roller cylinder-surface is scanned by a line-focus camera,while the roller end-surface is photographed by an area-focus camera.The raw images are processed by denoising-filter,image segmentation,edge detection,feature extraction,and object recognition in order to identify and classify the roller surface appearance defects.This method has the advantages of high speed,high precision,and the abilities of defects location and defect-types identification,compared with the traditional non-destructive methods,such as the magnetic powder test,the acoustic shock test,and the photoelectric test.This thesis is organized as follows:1.The development status of bearing industry is researched and summarized.Current non-destructive methods and its related technologies for bearing-roller surface defect is discussed and analyzed based on the national development-plan for the bearing industry.The scheme of the detection system for bearing roller surface defect based on machine vision is put forward.Moreover,the feasibility of the program is analyzed,and the main technical parameters is determined.2.The 3-dimentional mechanical structure of the cylindrical(conical)roller surface defects inspection system is modeled by Solidworks.A Twin-screw mechanism is used for cylinder-surface detection to make the roller move linearly,while rotate itself.Besides,the continuity and smoothness are ensured during the motions.The pulley drive is used to ensure the adjustable detection-station in the end-surface detection.The different specifications(shape,size)of the test object can be adapted according to the needs of flexible adjustment by increasing the freedom of structure.3.The detection and control system are developed based on the basic components,such as cylindrical detection system hardware line array camera,bar light,camera link capture card,fiber Optic Sensor,IPC,and so on;End-surface detection system hardware are Area array camera,Ring light,IPC,and so on;Control system hardware mainly include microprocessors,Stepper motor,Linear Motor and motor drives.The microprocessor,which is the main real motor motion control,send a pulse to the Camera link acquisition card coordination detection system.4.The machine vision bar light source and electromagnetic push rod motor are designed.The light source adopts two lenticular surfaces,which is a point light source into a line light source,to ensure the stability of the image for cylinder-surface detection with low cost and high signal to noise ratio.This is different from the market-mainstream bar light source for machine vision.The rod motor is pushed electromagnetically based on voice coil motor principle,to achieve high-frequency reciprocating motion,and to ensure the real-time detection system.5.The main bearing roller surface defects(rust,under wear,abrasion,over wear,crush,etc.)features are analyzed by experiments,compared with a variety of detection methods.The image processing optimization program is established according to the real-time efficiency requirements.The PC software,which integrated the functions of camera control is designed.The software composed of the next crew communication,image acquisition and transmission,database management,and image processing.