Key Techniques Of Image Measurement In3D Vision Measuring System

The vision measuring system is the most advanced optical dimensionmeasuring instrument nowadays. Having plenty of outstanding merits such asnon-contact, high reliability, fast and convenient, adaptability and so on, it canfully adapted to the increasingly stringent requirements of advanced manufacturingdimension detection. With the development of image measurement technique, CCDtechnology, computer technology, LED lighting technology and DC/AC servo drivetechnology, the vision measuring system has been developed greatly. Among these,3D vision measuring system represents the direction of future development.The powerful measurement function and high measurement accuracy of visionmeasuring system depend on not only various hardwares but also imagemeasurement technique. The focus of this paper is image measurement technique in3D vision measuring system. Some key techniques and methods of imagemeasurement are deeply researched, the main work are listed as follows:Firstly, the measurement superiority and the basic measurement processes ofvision measuring system are introduced. Principles of scale factor acquirement,geometric elements measurement and3D measurement are described.Secondly, in order to quickly and accurately find measured position ofworkpiece and improve measurement efficiency, map navigation function based onimage matching is proposed. Two map matching methods are investigated. One isbased on optical grating, and the other one is based on optical grating and digitalimage correlation technique. Experimental results are presented to verify thefeasibility of the proposed methods.Thirdly, processes and methods of common2D geometric features detectionare described in detail, and their experimental results are shown. Then, processesand methods of combination method measurement in vision measuring system areintroduced. In order to improve speed and accuracy of circle detection, a newrandomized Hough transform circle detection algorithm based on importancesampling technique is proposed. Experimental results indicate that the algorithmhas high precision, high speed, high identification reliability and strong robustness to noises.Finally, general methods of auto-focusing are discussed. By analyzingfocusing evaluating functions based on digital image processing, a two-stageauto-focusing method is proposed by using Kirsch gradient template and Gaussiancurve fitting. Experimental results shows that the proposed method is effective andheight measurement function is achieved.