Automatic calibration of laser stripe bathymetric mapping systems

Abstract

This thesis explores the problems associated with mapping large, regular-shaped areas of the seafloor at great depth, such as by sparse or dense grid survey. Specifically, it focuses on reducing the errors in reconstructions made by laser stripe mapping systems, which are especially well-suited to this task. These mapping errors arise from errors in pose estimates and in the modelling and calibration of sensors. While the former generalises across many problems and is the focus of much research, the latter are specific to the mapping system and application, and so are under explored for laser stripe mapping systems, particularly the calibration of their extrinsic parameters. Existing calibration methods for these have deficiencies. This work develops an automatic calibration method that avoids those deficiencies by not relying on the presence of distinct features in the terrain nor requiring multiple cameras. Firstly, terrain models are explicitly conceptualised and a featureless correspondence problem is solved, with the side effect of highlighting an important design consideration for laser stripe systems. Then, error metrics are developed to facilitate optimisation of laser parameters. Substitute parameters and observation models are developed to facilitate automatic calibration, and which also allow swift detection of infeasible laser parameters for given sets of observations. End-to-end mathematical models, that predict observations for varying errors in parameters, are then constructed and used to validate a proposed method of automatic calibration for single laser systems, and a method for improving the calibration of important parameters for dual laser systems. Finally, complex simulations are used to validate the practical application of a method of automatic calibration for single laser systems passing over inclined and largely smooth terrain. This method is ultimately applicable to any laser stripe system used to map in such a way as to cause overlapping transects of measurements, such as when using a grid survey approach, and at any depth.

More information

Identifiers

ORCID for David Stanley: orcid.org/0000-0003-3699-3203

ORCID for Miguel Massot Campos: orcid.org/0000-0002-1202-0362

ORCID for Adrian Bodenmann: orcid.org/0000-0002-3195-0602

Catalogue record

Export record