Low-cost high-solution multispectral imagery is attractive to various applications. This paper presents our ongoing research on developing Band-reconfigurable Unmanned Airship on Board Imaging System and focuses on the automatic registration of multispectral imageries at its each exposure. First, the overall architecture of developed Multispectral Imaging system as well as its main functions and core components are described. After that, motion compensation among collected multispectral imageries is discussed and based on Least Square technique, multi-homography model is developed to undo the motions deduced in multispectral imageries as well. Next, to reliably obtain control points for the estimation of motion parameters, one FFT-induced Multispectral Imagery Matching framework is proposed. Finally, experiments of registering multispectral imageries with different motion models are implemented, compared and analyzed as well as conclusions are conducted.