The invention provides an unmanned aerial vehicle (UAV) multispectral image fast splicing method, comprising the steps of determining an optimal wave band and a best matching algorithm; performing feature matching on remaining images with the presence of potentially overlapping regions by the determined optimal wave band and the best matching algorithm; taking a camera projection center and a heading angle corresponding to each image as optimization parameters, and iteratively solving an objective function to obtain values of optimal camera projection center coordinates and an optimal heading angle corresponding to each image; and further realizing multispectral image splicing. The method of the invention is advantageous in that an optimal wave band descriptor is used to select a representative wave band for feature point matching, out-of-band elements obtained by single wave band matching calculation are assigned to the remaining wave bands so as to ensure the independence and consistency between the wave bands after multi-band images are spliced, and the problems of the redundant calculation generated by separate processing of each wave band or the inconsistency of splicing results generated by the wave band grouping and splicing processing are avoided.