Study On Spectral Reflectance Recovery Method Based On Color Characteristic

Spectral reflectance can accurately record color information and realize color reproduction in various scenes.It has been widely used in art reproduction,telemedicine,textile,remote sensing and other fields.The research of spectral recovery method based on digital camera has the characteristics of simple operation,flexibility,strong ability of adaptation and low equipment cost.It overcomes the limitation of spectral data acquisition relying on contact measurement,which has become a research hotspot at present.Due to the neglect of color character in the process of color information acquisition by digital cameras efficient color acquisition and spectral recovery has become a key problem to be solved.This thesis systematically analyzed the redundancy and similarity of color character,systematically studied the spectral recovery method based on camera,and established the spectral recovery related optimization method based on color character.The specific research content is as follows:(1)For the problem that color redundancy and color similarity is ignored in the process of color information acquisition,this thesis introduces the concept of weighting and proposes a color character analysis method based on weight factor.By preprocessing spectral data,the influence of self-information on color character analysis is eliminated.The experimental results show that the calculation accuracy converges when the number of representative samples is small.The proposed method selects the least number of representative samples that compared with other methods,the spectral recovery accuracy of the whole sample is the closest,which shows a better calculation result.(2)For the problem since the existing spectral recovery methods do not consider the redundancy of color character,this thesis introduces the concept of subspace merging and proposes a spectral recovery method based on subspace merging.Each testing sample is regarded as an independent sample subspace,and subspace is merged according to the distance between sample subspaces.The final clustering center point is obtained through multiple iterations,which is used to determine the subspace.Finally,subspace tracking is used to determine the partition where each testing sample is located for spectral recovery.At the same time,the optimal training sample are replaced by the clustering center points as the representative sample for spectral recovery by using the principle of similarity.The proposed method is verified by experiments,and the results show that the proposed method is superior to the existing method in terms of spectral accuracy and chromaticity accuracy.(3)For the problem that existing spectral recovery methods ignore the similarity of color character,this thesis introduces the concept of dynamic clustering and proposes a spectral recovery method based on dynamic partition clustering.Each testing sample is taken as the prior clustering center of the clustering subspace.Different distance modes are selected as the competition basis,dynamic clustering subspace is determined by iterative division.Finally,the competitive way is adopted to determine the testing sample as the prior clustering center of the subspace,which verify the robustness of spectral recovery accuracy in different color Spaces.The effectiveness and universality of the proposed method are verified by the combination of simulation experiment and real experiment.The experimental results show that the proposed method has good performance in the aspect of spectral recovery accuracy.