Are Charles Jensen

Wide-beam or single-transmit acquisitions often reduce local spatial coherence, breaking the narrowband model assumed by high-resolution array spectral estimators such as IAA and Capon and thereby degrading performance. We propose a discrete prolate spheroidal sequence (DPSS) subspace projection of delay-focused aperture data. This projection suppresses incoherent off-angle energy and restores local spatial coherence, enabling scanline-wise adaptive spectral estimation under severe model mismatch. Each delay-focused aperture vector is projected onto a DPSS subspace spanned by the first K eigenvectors corresponding to a small angular bandwidth. The approach is lightweight, with precomputation and a per-point complexity of O(MK), and integrates naturally into standard delay-focused processing pipelines. Frequency–angle plots reveal how the projection reconstructs coherent ridge structures that are otherwise obscured by wide‑beam incoherence. Simulations in both plane-wave and diverging-wave ultrasound scenarios demonstrate improved resolution and contrast in single-transmit wide-beam imaging. Qualitative results on recorded channel data from the public PICMUS dataset provide an experimental sanity check and validation, indicating that the same coherence-restoration behavior is observed in real recordings. All experimental validation in this work is confined to ultrasound imaging; assessment of other array-processing applications is left for future work.

Seabirds are among the most threatened avian taxa globally, with over half of all species in decline. Accurate population estimates are essential for tracking trends and informing conservation, yet traditional survey methods are limited by logistical challenges, high costs, and the potential for wildlife disturbance, particularly in remote coastal areas. Unoccupied aerial vehicles (UAVs or drones) offer an efficient and low-disturbance alternative, but the vast volumes of imagery they produce are often labour-intensive to analyse. In this study, we combined drone imagery with deep learning techniques to estimate colony size and abundance of surface-nesting seabirds based on counts of visible individuals. High-resolution aerial imagery was collected from 163 colonies along the southern and central Norwegian coastline over three breeding seasons (2022–2024), covering a total of 7.67 km2. A convolutional neural network (Faster R-CNN with ResNet-101 backbone) was trained on 131 annotated orthomosaics and evaluated on 32 additional annotated orthomosaics from geographically distinct colonies. Across all data, 23,062 individual seabirds were annotated. Colonies hosted an average of 141.5 ± 193.9 individuals and 4.1 ± 2.3 focal species per site. At a confidence threshold of 0.7, the model achieved a detection rate of 87.5 % and a macro F1-score of 0.88. It performed well across multiple focal species, including terns, gulls, and cormorants, and remained robust in mixed-species colonies. Most errors involved false negatives or confusion among visually similar species. Our results demonstrate the potential for deep learning models to support efficient, scalable, and low-disturbance seabird monitoring across diverse habitats, reducing manual annotation effort and informing conservation practice.

This report details the data analysis products developed within the SeaBee project, a national infrastructure for drone-based services for use in coastal and aquatic research, mapping and monitoring of habitats, animal communities, and anthropogenic impacts. We present an advanced, automated data analysis pipeline that leverages deep learning for two primary tasks: pixel-wise thematic mapping of coastal habitats and object detection for counting wildlife. The pipeline utilizes models such as U-Net and Faster R-CNN to process high-resolution drone imagery (RGB, MSI, and HSI) and incorporates a novel hierarchical classification structure for habitat mapping and a robust method for detecting out-of-distribution (OOD) samples. We demonstrate the pipeline's pixel-wise mapping effectiveness through extensive experiments at three diverse Norwegian coastal sites—Remøy, Vega, and Ølbergholmen—achieving high accuracy in mapping complex habitats like kelp forests and various substrate types. Furthermore, the object detection framework shows strong performance in the automated counting and classification of 11 seabird species and coastal seals, offering a significant improvement in efficiency over traditional survey methods. The results confirm that the SeaBee pipeline is a powerful, scalable tool for environmental research and management, though we also discuss challenges such as data imbalance and model generalizability that will inform future work. This research is funded by the Research Council of Norway, project ID #296478, to the Norwegian Infrastructure for drone-based research, mapping, and monitoring in the coastal zone (SeaBee).

To leverage Earth observation (EO) data for large scale analysis, automatic methods is a prerequisite. Since 2012, deep learning (DL) models have brought about a revolutionary change in the analysis of image data and are currently considered state-of-the-art for a broad spectrum of EO tasks. However, a bottleneck with supervised DL models is that they often require a vast amount of labelled data to be trained, and the research community has therefore started to explore alternatives to supervised learning. During the last years, foundation models (FM) signify a change of thinking in computer vision. FMs are trained on a vast volume of unlabeled data and can identify complex patterns due to their large-scale learning capabilities. Typically, an additional head or decoder (small network) is added to the FM, which is trained and adapted to various use-cases by means of a small amount of labelled data. FM have also started to be explored for EO applications, however, current EO-based FMs are limited in terms of handling different modalities with large differences in resolution. Modern FMs are often based on transformers and are trained using self-supervised learning (SSL). There are several SSL schemes in place, including masked autoencoders (MAE) where we mask part of the input data and learn the model to predict the masked data. This is not useful by itself, but the model learns compressed representation of the data, which can be leveraged in downstream applications. This potentially makes the FM more useful than models trained on a limited set of labeled data. The Norwegian Computing Center and UiT – The Arctic University of Norway are in collaboration with user partners Romanian National Meteorological Administration, Danish Meteorological Institute, Polar View and Norwegian Water Resources and Energy Directorate developing a multi-modal FM. The FM is designed to process data from the satellites Sentinel-1 SAR, Sentinel-2 and Sentinel-3 OLCI and SLSTR. The FM is based on vison transformers (ViT) but utilizing the same principle as the USat approach to handle the different resolutions between the modalities. The training of the FM is based on the MAE approach, and to ensure that the SSL work efficiently, we have developed a smart sampling scheme during that provides relevant and diverse training data. In addition to SSL, we have also created a learning task in regressing climate variables from the ERA5 dataset. To train the FM, over 20 TB of Sentinel data was collected and processed using the LUMI supercomputer. The multi-modal FM is demonstrated on the following use-cases: mapping of snow, flood zone mapping, mapping and monitoring of sea ice, iceberg detection, early draught warning and mapping of wetlands. The resolution of the target use-case products are vastly different, e.g. for snow mapping we aim for a ground sampling distance (GSD) of 250m whereas for flood zone mapping we aim for a GSD of 10m. We have therefore trained two versions of the FM: one aiming for high-resolution products with GSD between 10 – 60m, and one aiming for low-resolutions products with GSD above 100m. The downstream tasks are implemented using the open-source framework TerraTorch, which is a flexible fine-tuning framework for geospatial FMs. TerraTorch supports common fine-tuning tasks such as image segmentation and pixel-wise regression along with a selection of task-specific decoder heads.

To leverage Earth observation (EO) data for large scale analysis, automatic methods is a prerequisite. Since 2012, deep learning (DL) models have brought about a revolutionary change in the analysis of image data and are currently considered state-of-the-art for a broad spectrum of EO tasks. However, a bottleneck with supervised DL models is that they often require a vast amount of labelled data to be trained, and the research community has therefore started to explore alternatives to supervised learning. Inspired by the progress in large language models, foundation models (FM) are now being applied extensively in computer vision. FMs are trained on a vast volume of unlabeled data and can identify complex patterns due to their large-scale learning capabilities. Typically, an additional head or decoder (small network) is added to the FM, which is trained and adapted to various use-cases by means of a small amount of labelled data. FM have also started to be explored for EO applications, however, current EO-based FMs are limited in terms of handling different modalities with large differences in resolution. Modern FMs are often based on transformers and are trained using self-supervised learning (SSL). There are several SSL schemes in place, including masked autoencoders (MAE) where we mask part of the input data and learn the model to predict the masked data. This is not useful by itself, but the model learns compressed representation of the data, which can be leveraged in downstream applications. This potentially makes the FM more useful than models trained on a limited set of labeled data. The Norwegian Computing Center and UiT – The Arctic University of Norway are in collaboration with user partners Romanian National Meteorological Administration, Danish Meteorological Institute, Polar View and Norwegian Water Resources and Energy Directorate developing a multi-modal FM. The FM is designed to process data from the satellites Sentinel-1 SAR, Sentinel-2 and Sentinel-3 OLCI and SLSTR. The FM is based on vison transformers (ViT) but utilizing the same principle as the USat approach to handle the different resolutions between the modalities. The training of the FM is based on the MAE approach, and to ensure that the SSL work efficiently, we have developed a smart sampling scheme during that provides relevant and diverse training data. In addition to SSL, we have also created a learning task in regressing climate variables from the ERA5 dataset. To train the FM, over 20 TB of Sentinel data was collected and processed using the LUMI supercomputer. The multi-modal FM is demonstrated on the following use-cases: mapping of snow, flood zone mapping, mapping and monitoring of sea ice, iceberg detection, early draught warning and mapping of wetlands. The downstream tasks are implemented using the open-source framework TerraTorch, which is a flexible fine-tuning framework for geospatial FMs. The FM4CS model is one of the first to apply Sentinel-3 data, which makes it attractive for climate applications.

Updating urban-area maps is crucial for urban planning and development. Traditional methods of updating urban-area maps based on aerial photography are labor-intensive and struggle to keep pace with rapid urban development. Automated algorithms for detecting new and removed buildings based on bi-temporal images typically either rely on comparing mono-temporal building detection outputs or requiring examples of new and removed buildings for training. This study presents a novel method using self-supervised learning principles to train a distinct object-change scoring network. It repurposes segments of the (potentially imperfect) delineations used in single-temporal detector training, harnesses bi-temporal data attributes, and leverages the assumption that most buildings remain unchanged over time. This eliminates the need for explicit examples of new or removed buildings, while still overcome usual constraints of post-detection output-mask comparison methods. We provide precision-recall curves and examples demonstrating the improved performance of the suggested approach. Furthermore, we discuss several immediate algorithmic variations that hold the potential for even further enhancements in performance.

A typical approach to reduce speckle in coherent imaging systems, is to average same-target images with different speckle realizations. We study settings where such realizations come from applying different transducer-array element-weights at reception, referred to here as receive compounding. An effect of such compounding is reduced spatial resolution, causing smearing of point-like image structures, filling of cysts and expansion of hyperechoic regions. In this paper we study how these unwanted effects can be mitigated by combining the compounding with a small, phase-based, adaptive steering of the array at reception. The adaptivity is based on a criterion akin to that of the Capon beamformer; a minimum-output distortionless response. Here, the distortionless part ensures that however we steer, we have a uniform at-focus response. We have applied this adaptive steering in combination with several receive compounding techniques on simulated Field II, phantom and in vivo data. The results show that all the studied compounding techniques respond to this positively in light of the mentioned unwanted effects. The technique based on Thomson’s multitaper method even surpassed the non-compounded equivalent in reproducing the geometry of structures. The speckle reduction, as measured by the change in the pixel mean to standard deviation ratio, is indeed lower, and there are subtle changes in the spatial speckle patterns when applying steering; however, we believe that in most cases the negative effects are tolerable in light of the benefits gained. The suggested approach is intuitive and easily implemented.