Research Of Multi-spectral Imaging System

The color of object is determined by the spectrum reflected from the surface of the object, which depends on the object's reflectivity within the electromagnetic spectrum of visible light. The traditional imaging devices such as digital camera and scanners capture color images by RGB sensors. The approach of obtaining object's spectral reflectance in this way may have metamerism, but in a multi-spectral imaging system, such problem does not exist. Multi-spectral imaging system is characterized by optical channels composed of a filter wheel which is composed of multi-band pass filters. The object colors are obtained by acquiring multi-spectral images and recovering spectral reflectance.Multi-spectral image contains more spectral information compared to the RGB color image, so the reflectivity of the image reconstruction is more accurate. This thesis presents a spectral reflectance reconstruction algorithm based on the partial least squares regression. Mathematical model of the algorithm is created by extended imaging equipment response using higher order polynomial and selecting the appropriate dimension for the partial least squares regression. Experimental results show that the proposed method is superior to the conventional Wiener estimation and Pinv method both in the spectral error and colorimetric errors.All multi-spectral imaging systems which use multiple filters have a common shortcoming caused by the changes in optical path. Because in the actual system it's impossible to arrange all filters in the same plane, each filter has a tilt angle relative to the principal axis. Even a small angle will cause significant distortion in the image. In addition, the difference of the filter's thickness and refractive index cause deviation on the optics. Thus the image in each channel is blurred. This thesis explores auto-focus way to solve problem of focusing image in multi-spectral imaging system. Besides the transmission of image data plays a very important role both in the exchange of mutual information and in the analysis of the actual system. This thesis explores the use of FTP file transfer technology in the transmission of multi-spectral image and data.