Super-resolution textured digital surface model formation using aerial texel images taken from a low-cost, small unmanned aerial system

Abstract

Methods for the creation of Textured Digital Surface Models (TDSM), which are a 3D representation of a terrain surface, are of increasing interest. This paper investigates the prospects for the generation of higher resolution TDSMs using texel images taken from a low cost unmanned aerial system (UAS). Coarse measurements for attitude and position are obtained from a low cost GPS/IMU fitted into the UAS. Using these coarse measurements and the data from the texel image, the error in the camera position and attitude is reduced which helps in producing an accurate TDSM. The main reason for using texel images is that these images contain both 2D image data and 3D lidar data. Despite having many texel images as input, the out- put TDSM reported previously only has approximately same resolution as the input texel image. In this pa- per, we propose a method to combine these low resolution texel images to produce a TDSM with high resolution texture using multi-frame Wiener filter super- resolution. The final image has the resolution at the sub-pixel level instead of pixel level and because of this, the TDSM looks visually better. The improvement pro- vided by implementing super-resolution on the texture is reported and the final registered TDSM results are shown.