Research On Spectral Reflectance Measurement Method Based On Digital Still Camera

Digital still camera-based spectral measurement can overcome the limitation of point-based spectral measurement used by spectrophotometer and spectrophotometer,can solve the defect that the spectrophotometer can only perform contact measurement,can avoid the disadvantages of low spatial resolution and poor practical application flexibility of existing spectral cameras,and can enables spectral measurement based on large area and pixel levels.During the digital still camera-based spectral measurement methods,the single RGB image-based one has gradually become a research trend because of its advantages of low equipment cost,fast imaging,flexible operation,convenient practical application and easy realization of universal application.This thesis carried out a systematically study,from the following aspects,on the spectral measurement technology based on a single RGB image of digital still camera.First,the influence of light source on the spectral estimation accuracy.Second,the quantitative evaluation of sample set differences and its optimization method.Third,the optimization of the spectral estimation algorithm.And fourth,the development of the spectral measurement method can be applied to an uncontrolled measurement environment without the support of portable training samples.The related spectral measurement method,with good performance,based on a single RGB image has been developed and tested based on the above researches.Detail contributions of the thesis are listed as follows.First,the influence of light source on the spectral estimation accuracy was investigated based on the investigation of the relationship between color correlated temperature(CCT)and spectral estimation accuracy,and the relationship of spectral power distribution(SPD)of the light source and spectral estimation accuracy.Several metrics to descript the characteristic of light source SPD have been developed to guide the selection and optimization of the light source for the construction of practical spectral measurement system.Second,the influence of the sample set difference on the spectral estimation accuracy was also investigated.The quantitative index of sample set spectral difference was proposed and its effectiveness was verified.And the conclusion of constructing specific training samples for different spectral measurement applications was demonstrated.Furthermore,the sample set optimization method was proposed and verified,which provides theoretical basis for constructing portable color charts that are convenient for practical applications.Third,in order to solve the problems existing in the existing methods,such as deviation from the linear imaging model,improper use of weighted color space for training samples,and lack of adaptiveness in the selection of local training samples,an optimized spectral estimation method based on adaptive local-weighted linear regression was proposed.The novelty of the proposed method is to use a Gaussian weighted linear regression model for spectral estimation,making sure that both the input data and the estimation algorithm are in line with the linear imaging model.At the same time,the adaptive selection of the optimal local training samples is integrated into the estimation method using Gaussian weighting.The weighting matrix is calculated in CIELAB uniform color space.The proposed method has realized the optimization of the spectral estimation matrix solution and improved the spectral measurement accuracy.Fourth,a spectral measurement method based on imaging condition correction was proposed to solve the problem of how to applying the measurement system in an uncontrolled measurement environment.In the proposed method,the standard whiteboard is used to correct the inconsistency in exposure level and type of light source,besides,with the support of light source database and system imaging model,the measurement error caused by light source variability is further adaptively corrected.Although the proposed method realized the application of the measurement system to the uncontrolled environment,it is subject to the limitations of the system's actual sensitivity function and imaging model,and the method needs to be further optimized.