A method to test differences between additional parameter sets with a case study in terrestrial laser scanner self-calibration stability analysis

This paper presents a new method for quantitatively assessing the impact and therefore the significance of differences between sets of additional parameters used to model systematic sensor errors. Focusing on bundle reconstruction at the sensor, simulation-based techniques have recently been proposed as superior methods to standard parameter-space hypothesis testing. This paper experimentally demonstrates the shortcoming of this approach and proposes an improved method that tests the effect of additional parameter differences on object reconstruction, which is generally of primary interest in photogrammetry. Additionally, the arbitrariness in selecting only one or two object space configurations is overcome with the new method by simulating a large number of randomly-generated but realistic control point networks for sensor orientation and a dense grid of points for object reconstruction. The experimental subject of the paper is a Faro 880 terrestrial laser scanner for which 10 sets of calibration parameters have been captured over a 13-month period. While standard parameter-space hypothesis testing indicates the instrument is not stable over this time, the new procedure shows that this is not true in all instances.