Research On Key Technology Of Printing Color Measurement System Based On Spectrum

With the continuous development of technology and social economy,people have higher and higher requirements for product printing colors.Especially in the printing and packaging industry,the color quality of printing products has become the important indicator of product quality evaluation.Through the accurate measurement of the printed product color can improve the printing color reproduction accuracy,and reduce the phenomenon of same color but different spectrum.At present,there are few studies on the color measuring instrument of printed products in China.Compared with foreign countries,the test accuracy is low,the volume is large,and the color RGB data collection is easily affected by the light source,ambient temperature and the color measuring instrument itself,so it is difficult to ensure the consistency of the color indicators for multiple batches.For this reason,the study of the printing color measurement system with good adaptability and high measurement accuracy has important practical significance.With the continuous deepening of research,to obtain accurate information on printing colors,the application of spectral technology to analyze printing colors has become a hot spot in printing quality inspection and analysis.In this study,based on the color measurement theory,spectral technology is applied to printing color measurement,and a set of printing color measurement system based on spectrum is developed.The main research contents are as follows:(1)Optical path structure design based on flat-field concave grating.Based on the measurement principle of the optical path system,the aberration model was built by defocus,astigmatism,meridional coma,sagittal coma and correction methods of the system,and the system structure parameters are optimized by using the genetic algorithm.Considering that the result of the genetic algorithm used is easy to fall into the local optimal solution,the optimization parameters are optimized twice through the evaluation function in the ZEMAX software,and the minimum aberration is reduced from2.2911×10^-10mm to 1.3882×10^-10mm.The grating resolution is less than or equal to 3nm,and the optical path structure shows better optical performance.(2)System hardware circuit design.In order to achieve effective collection of sample spectral information,the measurement system uses TCD1304DG linear array CCD as the photodetector and STM32F103C8T6 as the main control chip.The linear array CCD drive circuit design,signal conditioning,USB communication and host computer acquisition interface design are completed.(3)Research on preprocessing methods of spectral information.In order to solve the problem that there is many interference information such as stray light and noise in the collected spectral information,and the noise removal ability of common filtering methods is limited in the frequency domain,this study uses the wavelet transform,median filtering,and composite filtering methods combining wavelet transform and median filtering to compare the filtering effect of noisy signals.The test results show that the combined filtering method of median filter and wavelet transform can effectively suppress the mixed Gaussian white noise and impulse noise in the process of spectral signal acquisition,and the original light source signal is restored well.(4)System measurement algorithm research and test analysis.Aiming at the problem of more redundant data in the ink content spectrum information of the collected printing primary colors cyan,magenta,and yellow samples,the Successive Projections Algorithm(SPA)and Competitive Adaptive Reweighted Sampling(CARS)algorithm are proposed to extract characteristic wavelengths,thereby reducing the amount of modeling data.Based on the selected characteristic wavelengths,to achieve accurate measurement of the printing primary colors ink content.The color measurement prediction models were built by the Partial Least Squares Regression(PLSR)and Support Vector Regression(SVR)algorithms,and compared with PLSR and SVR prediction models built by full-wavelength spectrum reflectance data.The test results show that SPA extracts 7characteristic wavelengths from the spectral signals of the cyan,magenta,and yellow inks,while CARS algorithm can retain more characteristic wavelengths,and the number of modeling wavelengths can be reduced by 82.75%,69.25%,and 90.50%,respectively.The method of extracting characteristic wavelengths reduces the number of modeling wavelengths and simplifies the prediction model.The prediction effects of the built SVR models are better than those of the PLSR models.Among them,the CARS-SVR algorithm has the best prediction effect.The coefficient of determination(R²)and Cross-Validation Root Mean Square Error(RMSECV)are 0.9931,0.9966,0.9936 and 0.0256?0.0182?0.0249,respectively.The system shows better measurement accuracy.