基于色匹配函数的卫星影像真彩色合成研究

True Color Synthesis of Satellite Images based on Color Matching Functions

色匹配函数通过构建地物反射光的光谱分布与颜色三刺激值的关系可将任意波段的信息映射到三原色基准值上,经过色彩空间的转换后重建多光谱遥感影像。针对大多数多光谱传感器在可见光内波段数量有限、波段通道较窄且波段间隔不均匀,直接在相邻波段间插值会使色匹配积分过程出现较大误差的问题,结合遥感数据的模拟和反向传播(Back Propagation,BP)神经网络,以可见光范围内通道设置丰富的近岸高光谱水色成像仪(Hyperspectral Imager for the Coastal Ocean,HICO)经色匹配函数和色彩空间转换后得到的R、G、B三刺激值作为网络输出值,其波段重建后得到的目标传感器的模拟波段作为输入值,训练得到适用于Landsat-8 OLI、Terra MODIS、Himawari-8 AHI传感器的真彩色影像合成模型。计算均值、标准差、平均梯度和信息熵这4类客观评价参数,并结合对真彩色影像和直方图的主观分析,结果表明该方法能丰富和扩展有限的波段信息,提高影像的清晰度、色彩饱和度以及包含的信息量,校正了三波段合成影像存在的色偏,解决了在原始数据波段数受限的情况下通过简单插补波段带入色匹配函数进行积分计算产成误差的问题。

The color matching function can map the information of any band to the three primary color reference values by constructing the relationship between the spectral distribution of the reflected light of the ground object and the color three stimulus values,and reconstruct the multispectral remote sensing image after the conversion of the color space.In view of the limited number of bands in visible light,narrow band channels and uneven band intervals of most multispectral sensors,direct interpolation between adjacent bands will lead to large errors in the color matching integration process,combined with remote sensing data simulation and back propagation(BP)neural network,the R,G and B three stimulation values obtained by the color matching function and color space conversion of the Hyperspectral Imager for the Coastal Ocean(HICO)in the visible range channel are used as the output value of the network,and the simulated band of the target sensor after band reconstruction is used as the input value.The true color image synthesis model suitable for Landsat 8 OLI,Terra MODIS and Himawari-8 AHI sensors was trained.By calculating the four objective evaluation parameters of mean,standard deviation,mean gradient and information entropy,and combining with the subjective analysis of true color images and histograms,the results show that the proposed method can enrich and expand the limited band information,improve the clarity,color saturation and the amount of information contained in the image,correct the color deviation existing in the three-band composite image,and solve the problem of integral calculation error by inserting the color matching function into the interpolation band under the condition of limited band number of original data.