Research On Visible Light Computational Spectral Imaging Technology Based On Dual Camera System

Spectrum is one of the effective information to characterize objects and scenes,and spectral imaging is an effective way to obtain the spectrum of a scene.In the human visual system,the RGB color sensing function of the visual cone cells can perceive the spectral information of nature,which means that the images captured based on the RGB color format are familiar to the human visual system,so most imaging devices use RGB sensors to capture images.However,describing the target with the values of R,G,and B color components will have the phenomenon of metamerism,and cannot accurately describe the response of the target at different wavelengths.In contrast,the spectral imaging technology obtains target information from spatial and spectral dimensions,which can overcome the problem of metamerism.Spectral imaging technology can record the reflection response of natural scene light at different wavelengths,and record the physicochemical properties of some substances that are difficult to be reflected by RGB images,such as the material,morphology,surface smoothness and other characteristics of the target.In this thesis,we focus on the visible light computational spectral imaging technology based on the dual camera system.In order to solve the problem of the low resolution in spectral dimension of the filter wheel spectral imaging system,we propose a high spectral resolution imaging method based on the filter wheel dual camera system in this thesis,and design a multi-spectral computational reconstruction algorithm based on interpolation compensation,so that the spectral imaging system has both high spectral resolution and high spatial resolution.First,the filter wheel dual-camera imaging system is used to collect multi-spectral images and RGB images.Then,the discrete spectral response curves are obtained from the multi-spectral images.Finally,the interpolation compensation of the spectral response curve is performed and the high spectral resolution imaging is achieved according to the mapping relationship between the RGB three-channel data and the spectral high-dimensional data and the energy conservation theorem.The experimental results show that our method can efficiently achieve imaging with a spectral resolution of 5nm or even higher while maintaining the spatial resolution,and has good accuracy and robustness compared with the comparison methods.In order to further research the visible light computational spectral imaging technology and improve the resolution and accuracy of the spectral imaging system,a prism-mask-based dual-camera spectral imaging system is researched in this thesis.The method is based on prism-mask multispectral video imaging system,and combines the image fusion technology to fuse the information of the spectral image and RGB image acquired by the dual-camera system,improving the problem of low spatial resolution of the traditional prism-mask multispectral video imaging system.It further extends the application of spectral computational imaging technology in practical scenarios.The experimental results show that this method can achieve high spectral resolution and high spatial resolution imaging with has good accuracy,which expends the application of spectral computational imaging technology in scenarios with high real-time requirements.