Comparison of active and passive optical methods for mapping river bathymetry

Abstract

Optical methods for high resolution shallow water depth estimation are currently in a phase of rapid development driven by technical progress and growing interest in several scientific fields. Throughout the last years a diversity of potential applications has driven the creation of evolved methods and sensors. However, this trend primarily concerns Airborne Laser Bathymetry (ALB) while passive methods, particularly two-media photogrammetry, lack comparable advancements. A possible reason for the lesser amount of interest could be the limited spatial resolution when working with feature-based image matching. On account of this restraint, two-media photogrammetric methods concentrated on the reconstruction of individual points so far. Adopting the relatively new technique of dense image matching, this thesis takes a different approach and aims at a high-resolution description of the river bed as seen from aerial images through the water column. The effect of light refraction at the boundary between the two media, air and water, is theoretically investigated under varying conditions. Based on these findings, a practical refraction correction procedure is derived, implemented and embedded into the photogrammetric workflow. For evaluation under realistic conditions, the enhanced photogrammetric processing chain is applied to a set of aerial images of the pre-Alpine Pielach River in Lower Austria. Reference data are provided by a simultaneous ALB campaign. With both methods, digital terrain models (DTMs) including the river course and water depth models are derived, enabling direct comparison of the respective characteristics and quality assessment of the introduced photogrammetric procedure. Under favorable conditions, differences between the two methods rarely exceed 10-20 cm, but obstacles like overhanging trees or effects like sun glint are responsible for larger deviations. In the study area, a mean absolute difference between photogrammetric and ALB depth models of about 15 cm is achieved. Despite revealing some shortcomings in the demanding study area, the general capability of two-media photogrammetry to map shallow water bodies can be approved. Running through the procedure provides important insights into the possibilities and necessities of adapting the processing pipeline of commercial photogrammetric software packages to the peculiarities of the two-media problem. Besides, flights for data acquisition have to be planned carefully as well in order to restrict effects like sun glint or overexposure of bright objects while at the same time ideally resolving river bed texture. One of the most critical steps for active as well as passive methods is the extraction of the water surface which is needed for refraction correction. In addition to qualitative considerations, mainly the achievable degree of automatization is a question of particular importance since it decisively influences the practical applicability of a certain method for more extensive surveys.