In order to improve the localization accuracy for Unmanned Aerial Vehicle（UAV）in flighta localization algorithm is proposed based on machine vision.This algorithm can make UAV fly in the environment where Global Positioning System (GPS) signal is disturbed or is in failure.First of alltwo modified coordinate transformation algorithmsTrf1 and Trf2were obtained based on ScaleInvariant Feature Transform（SIFT）KDTree search image matching algorithm and the transformation between coordinate systems.With the help of Trf1the first coordinate transformation algorithmground coordinates of the target matched points were calculated out to build up a map of the flight environment.The second algorithmTrf2could calculate the realtime locations of UAV.Subsequentlythe localization algorithm in UAV flight was obtained based on the two modified coordinate transformation algorithmsand experiment was carried out based on the algorithms.Experiment results show that errors of calculated UAV positions are within 12 cmwhich demonstrates the accuracy of this method.