At present, satellite and aerial remote sensing are common ways to collect data for territorial resources monitoring in most countries, but they are not effective or rapid enough. Compared with traditional ways of obtaining images, the UAV based platform for photogrammetry and remote sensing is a more flexible and easy way to provide high-resolution images with lower cost. So building UAV based platforms is becoming a hot field throughout the whole world. However, there are also some problems with UAV images, e.g. the views of UAV images from UAV are smaller than those of traditional aerial images, so these images with small views should be pasted together in order to increase the visual field. Therefore, mosaicking UAV images is a critical task. The homographies between sequence images will be affected by the accumulated errors, which will lead to drifts of the position of each image in the mosaic. In this paper, we introduce a two-step optimization method for mosaicking UAV sequence images which can correct the homographies and improve the position of each image in the mosaic. Experimental results will also be presented.