Dissertation > Excellent graduate degree dissertation topics show

Dectection of Surface Defects on Steel Balls Using Machine Vision

Author: YangYang
Tutor: ZhouJianMin
School: East China Jiaotong University
Course: Mechanical Manufacturing and Automation
Keywords: Ball Surface defects Combined filter Geometric characteristics Moment invariants BP neural network
CLC: TP274.4
Type: Master's thesis
Year: 2009
Downloads: 239
Quote: 1
Read: Download Dissertation

Abstract


Steel ball surface quality has a crucial impact on the precision of the bearing, sports performance, and other aspects of life, but most production companies during the ball surface detection is still artificial visual-based, both spend a lot of manpower and resources, they can not ensure the reliability of detection. Therefore, an urgent need for an efficient, low-cost technology to detect Ball appearance. Machine vision is an emerging technology along with the development of computer technology. Application of machine vision technology to the steel ball surface defect detection, according to the characteristics of the defect image acquisition, image processing and to extract the defect characteristics, and finally the use of effective characteristics of defect classification Ball automatic detection of surface defects . Steel ball surface defects of various shapes and diverse types. According to several common types of defects, develop the overall plan of the steel ball surface defect detection system, build the hardware platform and visual inspection system for steel ball surface defect. Metal ball surface will produce a strong reflective imaging characteristics, design of image acquisition when the lighting system, to solve the problem of light can not be irradiated uniformly, and a clear image of the ball surface. Focusing on the steel ball surface defect image acquisition to discuss machine vision detection algorithm and simulation experiments. By experiments comparing the advantages and disadvantages of the traditional smoothing techniques include median filtering, Gaussian filtering, anisotropic diffusion filtering method, and on this basis, the eight anisotropic diffusion filtering and median filtering combined design portfolio filter, both effectively eliminate Gaussian noise and impulse noise protection edge. Then with threshold gradient operator image sharpening, iterative threshold algorithm for image binarization image segmentation. Segmentation defects connected domain recursive mark, and then simply connected domain and multi-connected domain the contour tracking technology applied to the defect area chain code, extracted the perimeter of the area of ??the defect area, center of gravity, roundness, fine length , rectangles and other geometric characteristics, and defect regions rotate, retractable invariant Hu moments, the shape of moment invariants characteristics. According to the experiment, select Remove the roundness of the many features of slenderness, rectangles and shape of the moment invariants four characteristics as subsequent surface defect identification classification features. The ball surface defect classification method, designed to meet the ball surface defects classification based on BP neural network classifier identification requirements. Steel ball surface defect identification classification than the template matching method has a higher accuracy rate proved by experiments, BP neural network, the average recognition rate of 87.5%, and also analyze the reasons that may affect the correct recognition rate. System design a series of experiments and test results show that the system has the speed, the overall detection effect, good stability, to meet the requirements of the steel ball surface defect detection, has high application value.

Related Dissertations

  1. The Theoretical and Experimental Research on Preload and Supporting Stiffness of a Wire Race Ball Bearing,TH133.331
  2. Research on Feature Extraction and Classification of Tongue Shape and Tooth-Marked Tongue in TCM Tongue Diagnosis,TP391.41
  3. The Research on Paper Currency Classification Method Based on Harr-Like Feature and Minimal Ball Including Samples,TP391.41
  4. Study on Taste Characteristic of Taste Peptide Enzymatic Production from Oyster Base on A Neural Network Method,TS254.4
  5. Research on Control Algorithm of Ball&Plate System,TP13
  6. Lung cancer and benign lung tumor CT misdiagnosis and control study,R734.2
  7. Preparation and Characterization of Luminescent Material SrWO4: Eu3+,TB383.1
  8. Analysis and Research about the 17th, 18th, 19th Football World Cup’s Goal Feature,G843
  9. Preparation and Properties of ZAO Target,TB383.1
  10. Preparation and Characterization of Rare-earth Luminescent Material CaWO4: Eu3+,TB383.1
  11. Study on PMT Preparation, Structure and Performance of NiCoMnIn Alloy Foam,TG139.6
  12. Preparation of Silicon-based Alloy Anode Material Andelectrochemical Properties,TM912
  13. Table Tennis Training Changchun Reserve Investigation and Development Strategies,G846
  14. The Applicability of Data Mining Technology for Environmental Protection Management System,TP311.13
  15. Mechanical Engineering Control Network Laboratory Research and Experimental Design,TH122
  16. Needle Coke Powder Molding Process and Its Impact on the Performance of Carbonization Products,TQ127.11
  17. The Study on Sensitivity of Driver’s Dhysiological Indices on Travel Fatigue,U491.254
  18. Study on Hand Gesture Recognition Based on Machine Perception,TP391.41
  19. Modeling and PID Control of Air Ball System,TP273
  20. Study on the Methods of Risk Management in Engineering Project,F224
  21. Ag / CTS imprinted gel Preparation and properties of the ball,TQ424

CLC: > Industrial Technology > Automation technology,computer technology > Automation technology and equipment > Automation systems > Data processing, data processing system > Centralized testing and roving detection system
© 2012 www.DissertationTopic.Net  Mobile