GEOREFERENCING OF UAV IMAGES FOR SHIP DETECTION IN MARITIME SURVEILLANCE

摘要

Continuous acquisition of precisely georeferenced images of the marine area is necessary to prevent and handle accidents and illegal activities on the sea. Precisely georeferenced images can be obtained through postprocessing of high resolution images by using sensors' External Orientation (EO) parameters and Ground Control Point (GCP) coordinates. Surveillance of ocean areas from satellite and radar images has limitations such as low resolution, lengthy acquisition time and high expenditure. Using UAV systems, we can collect high-resolution images at low cost while have no confidence in terms of accuracy caused by a little number of GCP (e.g.: ships, land such as island) and the performance of mounted GPS/INS sensors. In this study, we apply georeferencing techniques on the UAV high-resolution images and verify the accuracy of the georeferencing results. Adopted georeferencing method is based on the bundle adjustment where initial EO parameters determined by GPS/INS data are adjusted with GCP coordinates and image points produced by an image matching process. The georeferencing results are verified by comparing them with ground truth which can be acquired by Check Point (CP). If the accuracy of the results is satisfying with demands for marine surveillance, we can employ the UAV systems and the image georeferencing technique to monitor marine accidents and illegal activities.