Vision based localization of a small UAV for generating a large mosaic image

This paper describes a way to generate mosaic image using a small lightweight UAV for effective remote sensing. Payload limits are the source of most error in small lightweight UAVs to generate accurate mosaic images. We have overcome inherent small lightweight UAV drawbacks by providing them with integration of a GPS receiver, inertial sensors, and an image sensor to generate a large mosaic image. Position and attitude data of a UAV is estimated using onboard sensors and image correlation from SIFT features between overlapped images. A nonlinear optimization algorithm is used to optimize the position and attitude of the UAV. We demonstrated that our proposed mosaic image effectively enables low-cost remote sensing using a small lightweight UAV.