基于相机RGB值的光谱反射率重建方法研究

Research on spectral reflectance reconstruction methodbased on camera RGB value

光谱反射率重建技术可实现背景颜色数据准确校正和复制还原,以此为基础可设计实施高融合迷彩,为机动装备在不同背景环境下提供有效防护屏障,具有重要价值。参照光谱成像技术特点,根据颜色形成原理和三刺激值定量计算公式,通过控制单一变量法设计实验方案,提出加权平均光谱反射率重建方法(TWA),利用相机和光谱仪分别获取训练样本的成像系统响应值(RGB)和光谱反射率数据,对所提取的样本RGB值和光谱反射率数据分别采用传统最小二乘法(OLS)、正则化最小二乘法(RLS)、偏最小二乘法(PLS)、主成分分析法(PCA)和本文中提出的加权平均法(TWA)进行光谱反射率重建训练,获得光谱重建矩阵,最后利用光谱重建矩阵对测试样本响应值进行光谱反射率重建,并对实验结果进行分析,对比不同方法的光谱反射率重建效果。研究结果表明,加权平均法在光谱均方根误差SRMSE、平均拟合优度系数GFC、平均色差等指标上效果更优,重建所得光谱反射率精度更高,颜色校正效果更佳。

The spectral reflectance reconstruction technology can realize the accurate correction and replication of background color data,and on this basis,high-fusion camouflage can be designed and implemented,which provides an effective protective barrier for mobile equipment in different background environments,which is of great value.In this paper,referring to the characteristics of spectral imaging technology,according to the principle of color formation and the quantitative calculation formula of tristimulus value,the experimental scheme is designed by controlling the single variable method,and a weighted average spectral reflectance reconstruction method is proposed(TWA),and the imaging system response value(RGB)and spectral reflectance data of the training samples are obtained by camera and spectrometer,and the traditional least squares method(OLS),regularized least squares method(RLS),partial least squares method(PLS),and the extracted sample RGB value and spectral reflectance data are respectively used by traditional least squares method(OLS),regularized least squares method(RLS),partial least squares method(PLS),respectively.Principal component analysis(PCA)and weighted average method(TWA)proposed in this paper were trained to obtain spectral reflectance reconstruction matrix,and finally the spectral reflectance reconstruction matrix was used to reconstruct the spectral reflectance of the test sample response value,and the experimental results were analyzed to compare the spectral reflectance reconstruction effects of different methods.The results show that the weighted average method proposed in this paper has better effects on spectral rms error SRMSE,average goodness-of-fit coefficient GFC,mean color difference and other indicators,and the spectral reflectance accuracy of the reconstruction is higher and the color correction effect is better.