Provide an Automatic Method to Compute Approximate Exterior Orientation Parameters and Tie-Points Coordinate of Ultra-Light UAV Images in Order to Accurate Photogrammetry Block Formation

Today, various Unmanned Aerial Vehicles (UAV) Photogrammetry systems have been developed and utilized for close range aerial image acquisition and 3D topographic mapping. They have several advantages compared with classical manned aerial photogrammetry systems such as less cost, more accessibility, higher safety, shorter data acquisition process, and requires less skilled persons and specialized equipment. In this research a low-cost ultra-light UAV Photogrammetry system is used which equipped with a non-metric low-cost digital camera and a semi-automatic navigation system. Due to using these low cost components and general ultra-light UAV instability, some geometric and radiometric problems in thousands of small sized UAV aerial images are happen. The most important problems are (1) high tilt and rotation of vertical aerial images (2) Irregularity in image arrangement makes non-conventional standard overlap/side-lap coverage, (3) lower-quality images due to illumination changes, image motion and low signal-to-noise ratio, (4) high geometric and unstable image distortions come from non-metric off-the-shelf camera. In addition, since conventional photogrammetric software are designed and worked based on the characteristics of images taken from metric cameras mounted on manned aircraft, it is impossible to do automatic feature extraction and matching successfully on UAV images. These problems have caused to a big gap between imaging and aerial triangulation steps. The main goal of this research is dissolving the mentioned gap by automatic image processing of images were captured by ultra-light UAV and prepare input data to photogrammetric software for aerial triangulation. The results of this research show by eliminate this gap, relative high precision can be achieved from photogrammetric block bundle adjustment of these images. Also the results show that providing spatial products of these images such as Digital Terrain Model and Ortho-photo mosaic is possible.