非可控条件下基于深度神经网络的彩色图像光谱估计OACSTPCD

Spectral reflectance can be estimated from color images,and currently,spectral estimation based on color images is generally carried out under controlled conditions,that is,the light source and the camera settings of color cameras are strictly fixed,so as to ensure that the main factor causing the change of camera response values is only the spectral reflectance,so that the mapping relationship between camera response values and the spectral reflectance can be established accurately.However,there are many factors causing the variation of camera response values under uncontrolled conditions,which increases the difficulty of spectral estimation.In this paper,a deep neural network framework is designed based on the residual neural network with the help of its powerful nonlinear fitting ability to estimate the spectral reflectance from the RGB values of color images.A large number of color images were collected by smart phones under uncontrolled conditions,and the proposed network was verified by five-fold cross validation.The results show that the proposed deep neural network framework can achieve higher accuracy than two existing spectral estimation algorithms,which enables more accurate spectral estimation based on color images under uncontrolled condi-tions.