Improvement of Direct Georeferencing using Unmanned Aerial Vehicle Images Acquired at the Maritime

Abstract

Recently, UAV images have been used to monitor various marine accidents and environmental pollution, such as ship accidents, oil spill and red tide detection. In case of acquired images on the marine, it is difficult to apply georeferencing method since it is hard to obtain tie points between images and ground control points (GCPs). To overcome this problem, we propose an improved direct georeferencing method to determine the exterior orientation parameters (EOPs) of the UAVoptical images. Improved georeferencing method is determining to EOPs, which is based on aerial triangulation using intermittently observed GCPs. In other words, Assuming that there is a bias in GPS/INS data, the relative relationship constructed as the result of direct georeferencing. That relationship is absolutely corrected using the ground control points based on aerial triangulation. We verified the accuracy of the georeferenced EO. The georeferenced images consist of 81 images. These images were classified into the ground or the ocean images according to the main target areas acquired altitude and place from the ground or the ocean. As the result of the study, The Root Mean Square Error (RMSE) of the CPsfor the EOPs of the images from the ground was determined to be 2.69 m based on the direct georeferencing and 0.52 m based on the proposed method. The RMSE of the CPs for the images from the marine was determined to be 3.44 m based on the d