The accuracy of traditional single image based on geo-location method using Unmanned Aerial Vehicle (UAV)is too low to meet the needs of precise strike. To solve the problem, a new UAV geo-location method is presented in this article. The mathematical models are constructed by linearizing the colinearity equations and iteratively computing the pose angles and focal length of the camera. At least three images of the target, along with at least three identifiable common points among the images, are needed for reckoning camera pose angles and focal length. The three dimensional (3D) coordinates of ground target are calculated using forward intersection. The new method can get the target coordinates without dependence on Digital Elevation Model (DEM). Measured values of camera pose angles are not used directly for geo-location. Two of the three primary error sources in the traditional UAV target location approach are eliminated. Simulation and real image experiment results show that the accuracy of the estimated target location is very high if the UAV is navigated by Differential Global Positioning Systems (DGPS).