基于机器视觉的平面透明元件测量方法研究

打开文本图片集

Abstract： The article proposes a machine vision based method for measuring planar transparent components to address issues such as unclear features and low measurement accuracy in large fields of view.This method mainly includes two parts： two-dimensional information measurement of planar transparent elements and height information measurement. Adding feature backgrounds to enhance the texture information of planar transparent elements，and based on this，an image registration method using SURFcorner detection is proposed to achieve high-precision image stitching in a large field of view environment. An improved binarization method is also proposed to remove feature backgrounds and restore image information. Combined with contour detection and rectangular contour recognition techniques，two-dimensional information of planar transparent elements is obtained. At the same time， a laser projection device is introduced to achieve high-precision detection of laser ectopic imaging on the upper and lower surfaces of planar transparent elements through grayscale centroid method and improved bilateral filtering method，and obtain height information of planar transparent elements based on the principle of refraction. The experimental results show that the accuracy of two-dimensional detection is5 microns，with arelative error kept below 0.02% .Theaccuracyofheight detection is 50 microns， with a relative accuracy kept below 0.78% ，achieving high-precision measurement of planar transparent components in a large field of view.

Keywords： machine vision; transparent component; image processing; grayscale centroid method

0 引言

多色成像与无缝光谱巡天项目中的无缝光谱组件主要任务是完成巡天观测任务中的无缝光谱观测。（剩余9723字）