Georreferenciación directa en RPAS: ¿mito o realidad?

Abstract

Nowadays we are witnessing real boiling in the development of RPAs (Remotely Piloted Aircrafts) aiming to many different purposes in the civil sphere. One of these fields of application is the generation of large scale mapping products derived from images taken from these platforms. As it is well known this task require a high level of accuracy that is currently based in surveying reference points by means of well established techniques. But as has happened with the capture of imagery that has become autonomous, there is a demand for solutions that can offer independence of all topographic field works required for GCP's (Ground Control Points) measurement. The level of accuracy required by methodologies of direct or semi-direct images orientation only seems achievable for high precission airborne inertial units commonly seen as the only way to obtain accurate position and attitude of the image sensor systems. These high end systems costs can reach thousands of euros, so that accuracy seems exclusive just for certain sectors or professionals. Thanks to the kindness of the companies Geobit Consulting and Droning Sistemas Aéreos, it has been possible to have an ow developed RPA system with built in GNSS L1 / L2 to perform various tests. This has made possible to verify the accuracy that can be achieved aiming to direct georeferencing or a priori external orientation given by onboard sensors for the production of orthophotos and other products. Tests were carried out in different scenarios: mining, corridor or areas with poor texture in order to determine the repeatability in the offered precision. Many factors have been found to influence the final accuracy, such as satellite coverage, flight scenario, data redundancy, overlapping percentages, or stability in calibration sensor. All these parameters have been considered in different projects conducted for the analysis of the datasets. The results are strongly influenced by the type of sensor used as well. This paper exposes the results obtained in the different scenarios, leaving open the possibility of further testing in order to compare the results obtained.