UAV photogrammetry for very-shallow-water mapping

Using drone platforms and RGB images acquired with passive optical cameras, image-based UAV bathymetry is a cost-effective and accessible method of inspecting and surveying the seabed in shallow coastal waters.

Various measurement techniques can be employed to inspect and survey a seabed in shallow coastal waters, including oceans, seas, lakes, rivers and artificial reservoirs. One such method is image-based UAV bathymetry, which uses drone platforms and RGB images acquired with passive optical cameras. This approach offers a cost-effective and accessible alternative to expensive technologies such as active Lidar or sonar sensors. The potential of this method is discussed in this article, as well as the key limitations and conditions that need to be considered.

According to Cosby et al. (2024), 30% of the world’s population lives within 50 kilometres of the coast. Therefore, it is crucial to understand these areas, their changes and the processes occurring within them. The spatial data obtained can be used by coastal zone managers for maritime navigation, fauna and flora research, erosion monitoring, archaeological mapping and much more. Consequently, new technologies that enable the rapid acquisition of spatial information about these areas are in demand.

One important condition

Mapping coastal areas using drones and RGB images is an easily accessible and relatively inexpensive technology. However, it can only be used effectively if the water is transparent and the seabed is fully visible. Water turbidity, surface caustics effects and suspended particles in the water can cause reconstruction problems and lack of data in the 3D seabed model. Figure 1 shows two images and their corresponding 3D models as examples. Gaps in the reconstructed surface models are clearly visible and are caused by reduced water transparency. Complete reconstruction of the model is only possible when the bottom surface is clearly visible.

Planning the measurement

Two aspects are crucial to plan the measurements: the purpose of the study, including the required image resolution and overall accuracy, and the available equipment. The first step is to set up and measure the ground control points (GCPs). While a minimum of three points is theoretically required, more are typically used in practice to improve accuracy and provide some quality control in the object space. The points should be distributed throughout the area and measured with an accuracy better than that obtained with geodetic instruments. As the measurement of points on the ground (water) could be difficult and time consuming, onboard GNSS RTK/PPK trajectories could also be used. These are more practical as no GCP would be needed, although the accuracy of the final 3D reconstruction would be lower.

The next step is to plan the UAV flight mission. At this stage, the required ground sample distance (GSD) – the size of a single pixel on the ground – defines the appropriate flight altitude. It is also important to ensure adequate image coverage for a complete 3D reconstruction of the surveyed area. 

In addition to the instrumental aspects, weather conditions must also be taken into consideration. To ensure stable drone flight, measurements should be taken on windless days. This also minimizes the formation of waves and ripples on the water’s surface. Lighting is also important; the best conditions occur when the sun is at an angle relatively low above the horizon, as this prevents strong reflections from being captured in the images.

Photogrammetric processing

Photogrammetry is a technique that enables the creation of 3D data of a surveyed area using a set of overlapping images. This process can be divided into four stages: image orientation, sparse point cloud generation, dense point cloud reconstruction and orthomosaic creation. Image orientation involves detecting a sufficient number of well-distributed tie points and applying a least squares adjustment process to derive camera poses and a sparse point cloud. For proper scaling, georeferencing and deformation minimization, some GCPs are used. Next, a dense point cloud is created and, at most, a digital surface model and an orthomosaic are produced. Many commercial and open-source tools are available to automatically carry out the entire photogrammetric process with very similar performance, although there is a general lack of transparency regarding quality control purposes.

The true depth-refraction correction

Photogrammetric processing alone is insufficient to create an accurate surface model of the seabed’s shape. This is due to the refraction effect, whereby light is refracted at the air-water interface, artificially reducing the surface depth, as illustrated in Figure 2. Well-known solutions based on Snell’s law can be employed to correct this effect (Dietrich, 2016), while newer approaches based on machine-learning algorithms are still under development (Agrafiotis & Demir, 2025).

Final products... and what next?

The final products of the photogrammetric processing are typically a point cloud, a digital surface model (DSM) and/or an orthomosaic. Figure 3 shows the results obtained from a study of one of the Polish lakes.

These geospatial products can be used for further environmental research, for example to detect, measure and monitor vegetation such as seagrass beds, to analyse morphological changes, to monitor coastal habitats or to study the impact of rapid climate change and pollution.

Conclusions

Image-based UAV bathymetry is a fast, effective and low-cost method for reconstructing shallow-water areas. It uses images collected with a UAV/drone platform and an RGB camera to create 3D data. Clear water and a visible bottom are prerequisites for its application. There are many ways to acquire and process data, and the most suitable method depends on the purpose of the measurement, the expected accuracy and the available equipment. Despite these constraints, this technology provides valuable spatial information and improves community understanding of changes in coastal areas.

References

Agrafiotis, P., & Demir, B. (2025). Deep learning-based bathymetry retrieval without in-situ depths using remote sensing imagery and SfM-MVS DSMs with data gaps. ISPRS Journal of Photogrammetry and Remote Sensing, 225, 341–361. https://doi.org/10.1016/j.isprsjprs.2025.04.020

Cosby, A. G., Lebakula, V., Smith, C. N., Wanik, D. W., Bergene, K., Rose, A. N., Swanson, D., & Bloom, D. E. (2024). Accelerating growth of human coastal populations at the global and continent levels: 2000–2018. Scientific Reports, 14(1). https://doi.org/10.1038/s41598-024-73287-x

Dietrich, J. T. (2016). Bathymetric Structure‐from‐Motion: extracting shallow stream bathymetry from multi‐view stereo photogrammetry. Earth Surface Processes and Landforms, 42(2), 355–364. Portico. https://doi.org/10.1002/esp.4060

Kujawa, P., & Remondino, F. (2025). A Review of Image- and LiDAR-Based Mapping of Shallow Water Scenarios. Remote Sensing, 17(12), 2086. https://doi.org/10.3390/rs17122086

Kujawa, P., Wajs, J., & Pleśniak, K. (2025). The approach to UAV image acquisition and processing for very shallow water mapping. International Journal of Applied Earth Observation and Geoinformation, 141, 104604. https://doi.org/10.1016/j.jag.2025.104604