Study Of Image Measuring System Based On Computer Vision

In abroad and home, the measuring system based on image in computer vision has been researched widely, and get more attention. Currently, the precision and velocity of detecting the micro-size components in industry could not be achieved with traditional detecting system. The measuring system based on computer vision technique could give better performance in the field of detection. This system with the advantages such as non-contact, real-time, high accuracy and automatic operation has a plenty of application because of the ratio of computer hardware and performance increasing and the relative technologies developing.In this paper, the components of the measuring system are introduced and the algorithms for solving problems about precision, speed and stability are discussed. The components of this system include worktable, light source, CCD, video capturing card, grating ruler, motion controlling board and PC. In the course of capturing image of object, the light source is adjusted to get a vivid image; the CCD transforms the optical signals of this image into electrical signals, the video capturing card transfers these signals into the PC. Then, the measurements of geometric parameters of the object are completed using the relative operations such as image processing, geometric computing, motion controlling and ruler reading. The measuring works are finished with simple operations of clicking the mouse, and the operator could be released from complex and hard work.In this paper, the software is developed based on the research about key issues of the measuring system. An algorithm of reducing affection of lens distortion is presented based on the analysis of barrel and pillow distortion. As the basis of measurement to the size and position of object, the edge detection algorithms on image are researched. Based on research of the methods about automatic adjusting of the focus of CCD, this paper proposes a two-stage method that a coarse range of the focus is searched, and then an accurate position is located. In order to complete automatic detection, an algorithm of image matching based on correlation is improved and a two-step matching strategy is discussed in detail.