Research On Key Technology Of Color Measurement In Visual Inspection System

The online printing quality detection based on machine vision system has become a trend,color is an important index to evaluate the quality of printing products,but the image by digital camera can only obtain in RGB color space in the machine vision detection system,the same measuring target in different imaging environment RGB values are not the same.In the printing color quality detection process,color measurement technology is used to get the real color values.This thesis uses the spectral characteristics of surface to achieve the color measurement.At the same time,based on the requirements of color measurement controllable,this thesis presents the method of device-independent color space sample design to limit color estimation error,so the machine vision inspection system control problem can be solved.This thesis presents a spectral reflectance reconstruction algorithm based on R matrix,which can achieve the conversion between different color spaces.The method under twelve channels line scanning CCD imaging system,using the principal component analysis and R matrix to analyze and verify the theory,and improved R matrix by using principal component.Then the spectral information is converted to chroma system,and estimate the color quality through the accuracy of the spectral reflection reconstruction and and the chromatic aberration.The experiments show that the improved R method can achieve high accuracy in both color and spectrum,so it can be used for industrial printing color quality detection practical.This thesis presents the design method of CIE 1976 L*a*b*color space uniform segmentation.This method transform CIE 1976 L*a*b*color space to color lightness,hue and saturation in polar coordinates LCH form,based on the lightness,saturation,hue equal segmentation,color estimation error is defined between the sampling points uniformly.Training samples are used to test the influence of color measurement accary,and color space conversion model established by BP neural network method and the support vector machine method.The experiments show that training samples based on CIE 1976 L*a*b*color space uniform segmentation,can control the color measurement accuracy and improve the validity of color measurement results.