Geometric Accuracy Assessment of ADS40 Imagery under Various Network Configurations

Abstract

The ADS40 camera is a commercial example of the airborne Three-Line-Scanner (TLS). Its particular sensor geometry requires new approaches and a specialization of methods and algorithms to solve the triangulation problem. The goal of the paper is to investigate the geometric accuracy performance of an ADS40 camera under various network configurations and using different trajectory models and different sets of additional parameters for self-calibration. Investigations are performed by two independent research Groups, the Institute of Geodesy and Photogrammetry (IGP), ETH Zurich, and the Geomatics Laboratory of the University of Pavia. Identical image and point datasets are used by both parties to test different trajectory models and additional parameters for self-calibration. The test datasets contain three image blocks with 2000 m, 4000 m and 6000 m flying heights, which have been acquired over the PaviaTest Site (PTS), Italy using an ADS40 camera. The average ground resolutions are approximately 20 cm, 40 cm, and 60 cm for the three blocks. 50 signalized GCPs installed with white square markers of 60 cm side length are used in the tests. The tests wereperformed with different numbers and distributions of GCPs and with and without self-calibration procedures. In addition, two different trajectory models are tested at the IGP, ETH Zurich. Also, direct georeferencing without the use of GCPs has been performed. The RMSE values, given in pixels, for planimetry and height, obtained with triangulation and self-calibration for the best cases of each block are: 2000 m block: 0.20 and 0.25 4000 m block: 0.20 and 0.30 6000 m block: 0.13 and 0.24 The accuracy results of both Groups are consistent and show significant improvements when self-calibration is applied.