多光谱图像的光谱反射率重建技术研究

发布时间：2018-03-21 02:27

本文选题：光谱重建　切入点：光谱图像系统　出处：《华东交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：自然界中的颜色丰富多彩,随着人们对颜色的认识越来越深,颜色的复制也显的越发重要。而对于传统的颜色复制来说,可能会由于同色异谱色的存在而使得不同的原颜色通过复制过程而得到同一个颜色。而光谱反射率作为物体表面颜色的属性具有唯一性,即不同颜色的光谱反射率是不同的。传统的光谱反射率测量方法即是采用分光光度计测量,然而分光光度计一次可测量范围很小,并且只能对平面进行测量,使得我们只能寻求其他方法来测量物体表面的光谱反射率。近年来,通过CCD数码相机和滤光片组结合组成的多光谱图像系统来测量光谱反射率已成为研究的重点。本文通过复合偏最小二乘法重建光谱反射率,将预测色块与训练集色块距离进行比较,对于不同距离的色块采用多主成分数偏最小二乘法和普通偏最小二乘法进行不同的运算,最终得到反射率预测结果,并且与普通偏最小二乘法(PLS)作比较。实验结果显示,复合偏最小二乘法的重建结果优于传统偏最小二乘法。对于多光谱图像系统来说,滤光片数量不同会得到不同通道数的多光谱图像。但是对于多光谱图像系统的滤光片数量,现在还没有统一的规定,并且滤光片数量对于光谱反射率重建精度的影响也不得而知。鉴于此,我们在本文中采用不同数量的滤光片结合CCD相机组成不同的多光谱图像系统。并采用偏最小二乘法和基向量法结合神经网络的方法重建光谱反射率。通过实验结果可知,对于偏最小二乘法和基向量结合神经网络法来说,滤光片数量的增加,光谱反射率的重建结果精度越高。
[Abstract]:Color is rich and colorful in nature. As people know more and more about color, color reproduction becomes more and more important. And for traditional color reproduction, It is possible that different primary colors can be obtained by copying the same color because of the existence of the same color and heterospectral color. The spectral reflectivity is unique as the attribute of the color on the surface of the object. That is, the spectral reflectivity of different colors is different. The traditional method of spectral reflectance measurement is to use a spectrophotometer, but the spectrophotometer can be measured in a very small range, and can only be measured on a plane. So that we can only find other ways to measure the spectral reflectivity of the surface of an object. The measurement of spectral reflectivity by a multispectral image system composed of CCD digital camera and filter set has become the focus of research. In this paper, the spectral reflectivity is reconstructed by compound partial least square method. The distance between the predicted color block and the training set color block is compared. The multiple principal fractional partial least square method and the ordinary partial least square method are used to calculate the distance of the different distance color blocks, and the reflectivity prediction results are obtained. The experimental results show that the composite partial least squares method is better than the traditional partial least squares method. The number of filters varies with the number of channels. However, there is no uniform regulation on the number of filters in a multispectral image system. And the effect of the number of filters on the accuracy of spectral reflectivity reconstruction is unknown. In this paper, we use different number of filters combined with CCD camera to compose different multispectral image systems, and use partial least square method and basis vector method combined with neural network to reconstruct spectral reflectivity. For the partial least square method and the basis vector combined neural network method, the higher the precision of spectral reflectivity reconstruction is with the increase of the number of filters.
【学位授予单位】：华东交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP183

【参考文献】