Publikasjoner
NIBIOs ansatte publiserer flere hundre vitenskapelige artikler og forskningsrapporter hvert år. Her finner du referanser og lenker til publikasjoner og andre forsknings- og formidlingsaktiviteter. Samlingen oppdateres løpende med både nytt og historisk materiale. For mer informasjon om NIBIOs publikasjoner, besøk NIBIOs bibliotek.
2026
Sammendrag
Longitudinal analysis of a patient’s screening history is fundamental to mammography interpretation. Yet existing deep learning models struggle with the irregular, continuous-time nature of screening data, where patient histories involve unevenly timed multi-view exams paired with evolving textual reports. To address this gap, we introduce Dynamo, a novel multi-modal pretraining dynamic framework for longitudinal mammography with two key innovations to model exam timelines as latent trajectories, i.e., a coarse-fine grained exam-level temporal encoder based on Neural Controlled Differential Equations (Neural CDEs) and a Temporal Visual Question Answering (TVQA) loss for query-driven masked token prediction conditioned on temporal context. We perform comprehensive evaluations on downstream tasks including risk prediction using large-scale datasets (EMBED and CSAW-CC), BI-RADS assessment, and breast density classification. Across all benchmarks, Dynamo achieves overall gains over state-of-the-art vision-only and CLIP-style models, improving both calibration and temporal reasoning. On the EMBED dataset, representative gains include 11.81% relative reduction in Risk Prediction’s Brier Score, 3.59% relative improvement in BI-RADS ϰ, and 2.39% relative improvement in BI-RADS AUC in zero-shot settings. Our code is available at this URL.
Sammendrag
Collection, processing and provision of comprehensive geometric information of forest roads is decisive for its technical classification to facilitate sustainable timber supply chains. An automized classification system based on the mobile proximal sensor platform RoadSens was developed, applied and validated through a case study approach in Eastern Norway. Six sample roads of various vegetation stages were surveyed through RoadSens and complemented through sampled total station measurements for validation purposes. The determined geometric parameters road slope, curvature and width were used for technical classification following the national forest road standard. Road width was identified as the main constraint in meeting the standard, resulting in a general downgrading of the sampled roads according to its technical class. The results showed a root mean square error (RMSE) ranging from ±0.53 to 1.50 m (12–33%) depending on the road and vegetation stage compared to the validation data. Despite these accuracy constraints, the application case study already indicates a general need for improvement of road data acquisition and updating of associated databases. The study underscores that, despite the challenges and limitations, there is a clear need for an automated sensing and classification system, which offers a cost-effective alternative to manual surveying and requires less specialized expertise.
Forfattere
Stephan Hoffmann Mostafa Hoseini Moritz Wingartz Mahmoud Rajabi Helle Ross Gobakken Rasmus AstrupSammendrag
A functional and low-impact forest road network is essential for sustainable forest management, yet maintaining such infrastructure is costly and requires monitoring tools that are reliable and simple enough for operational use. We present an automated approach to detect, map, and evaluate forest road surface deterioration, designed to support end-users, including those with limited road expertise, to indicate required maintenance actions. The system relies on data collected by the vehicle-mounted near-field sensor platform RoadSens, which integrates stereo camera imagery with GNSS-based geo-referencing to capture detailed road surface information. Collected data are processed within a monitoring and scheduling environment using a YOLOv8 object detection model trained on nearly 14,000 annotated images. The model identifies six key deterioration features: potholes, wheel ruts, gullies, washboards, stones, and vegetation. These detections are used to locate maintenance-relevant features and classify road segments into three deterioration levels based on coverage thresholds, which are then visualized through a traffic-light system. A case study on a forest road in southern Norway demonstrated the system’s ability to detect and classify maintenance needs. While performance was strong for more uniform features such as vegetation, irregular structures like wheel ruts proved more challenging, occasionally leading to misclassification of actual maintenance requirements. Nevertheless, the findings confirm the technical feasibility of integrating object detection models into data-driven forest road maintenance scheduling. Future improvements will require larger and more diverse training datasets, as well as classification frameworks tailored to local conditions and specific road-user needs.309671 -
Sammendrag
Time and motion studies in forest operations benefit from video-based analysis, but manual annotation is time consuming. This pilot study aims to reduce analysis time by developing a deep-learning framework that classifies dashcam video into four work elements: crane out, cutting and processing, driving, and processing. Using a 3D ResNet-50 (PyTorchVideo) trained on manually annotated clips, the model achieved validation F1 = 0.88 and precision = 0.90, showing that spatiotemporal CNNs can capture rele-vant motion and appearance cues in forest environments. Overfitting indicates that more diverse data and better class balance are needed, but the approach shows clear potential to scale automated work-element monitoring and efficiency analysis.
Forfattere
Nicolas CattaneoSammendrag
Det er ikke registrert sammendrag
Sammendrag
Rapporten presenterer resultater fra prosjektet «Demonstrasjon av omstilling til selektive hogster (DEMO)». Hovedmålet har vært å øke kunnskapen blant skogeiere, forvaltning og kunnskapsleverandører om omstilling av ensaldret skog til flersjiktet skog egnet for bledning. Metoden som har blitt testet er inspirert av forsøk i Nord-Amerika med omstilling av douglasgranplantasjer til skog med naturlig dynamikk. Der har man brukt ‘variable-density thinning’ (VDT) for å skape varierende tetthet som skal sette i gang foryngelsesprosesser på forskjellige tidspunkter og romlig spredt i bestandet. Metoden er her utviklet videre for å tilpasses omstilling til bledning i norsk skog med maskinell tynning. Det gis en beskrivelse av metoden, inkludert krav til egnede bestand. Det er gjennomført 15 praktiske drifter, samt etablert 4 feltforsøk hvor VDT testes ut. Erfaringene fra disse driftene utgjør grunnlaget for våre resultater og konklusjoner. Hovedfokuset har vært å teste metoden i grandominert skog. Det har imidlertid også vært interesse blant skogforvaltere for å teste metoden i furubestand. En tilpasning for å legge til rette for framtidig bledning i furudominert skog er derfor også beskrevet. De praktiske driftene og forsøksfeltene muliggjør undersøkelser av den videre utviklingen i bestandene de neste årtiene. Stikkord er etablering av foryngelse, vekstreaksjoner, stabilitet mot vindfelling, trærnes reaksjoner i tørkeperioder, og reaksjoner på skogbildet blant rekreasjonsbrukere.
Sammendrag
Det er ikke registrert sammendrag
Forfattere
Ishita Ahuja Arne Steffenrem Irena Fundova Helmer Belbo Torstein Myhre Inger Sundheim FløistadSammendrag
When regenerating clearcut areas in Norway and several other countries, tree seedlings are planted adjacent to stumps of harvested trees to reduce snow load and provide shading, despite limited scientific evidence supporting this practice. This study investigated the role of tree-stumps as planting microsites in the establishment of Norway spruce (Picea abies (L.) Karst.) seedlings. We assessed the growth and survival of seedlings from two provenances: Undesløs seed orchard (60.7°, 140 m), consisting of tested parents from the lowland around 63–65°N, and seed collected from forests in the M4 provenance (64–65°N, 350–449 m), at two microsite types in Trøndelag County, Norway: beside stumps (Microsites-B) and at a distance from stumps (Microsites-D). Undesløs seedlings exhibited 100% survival at Microsites-B, whereas M4 seedlings showed higher survival at Microsites-D. Provenance had a significant effect on seedling height and diameter, while microsite type had no significant effect on these parameters. In 2022, significant differences in height and diameter were observed between provenances at Microsites-B. Phenotypic variations, including chlorotic, green, and brown needles, occurred in seedlings of both provenances across both microsite types. Overall, this single-site experiment provided no evidence that planting beside stumps improves growth/survival compared with planting at a distance away.
Forfattere
Janine Schweier Raffaele Spinelli Francesco Latterini Natascia Magagnotti Rodolfo Picchio Stelian A. Borz Csongor Horvath Leo G. Bont Stephan HoffmannSammendrag
Purpose of review This review traces the development of mini forestry crawlers (MFCs) from earlier small-scale skidding machines to modern remote-controlled tool carriers, and evaluates their current applications, technological characteristics, operational performance, safety, soil impact, ergonomics, and automation potential. Recent findings Recent studies show that MFCs have evolved from simple extraction-oriented machines into multifunctional platforms whose suitability depends on machine class, task–machine matching, site conditions, and work organisation. Field and bench studies report productivity, soil impacts, operator workload, remote-controlled felling performance, non-harvesting applications, and early automation functions. Summary Mini forestry crawlers are most effective in constrained-access settings and in tasks that align with their limited payload while benefitting from high manoeuvrability, remote operation, and multifunctionality. Their advantages are therefore conditional on careful deployment, particularly with respect to soil moisture, turning intensity, traffic frequency, and operator workload. Future evaluations should adopt integrated performance metrics that jointly assess productivity, soil response, and human workload under realistic operating conditions.
Forfattere
Binbin Xiang Maciej Wielgosz Stefano Puliti Kamil Král Martin Krůček Azim Missarov Rasmus AstrupSammendrag
The segmentation of forest LiDAR 3D point clouds, including both individual tree and semantic segmentation, is fundamental for advancing forest management and ecological research. However, current approaches often struggle with the complexity and variability of natural forest environments. We present ForestFormer3D, a new unified and end-to-end framework designed for precise individual tree and semantic segmentation. ForestFormer3D incorporates ISA-guided query point selection, a score-based block merging strategy during inference, and a one-to-many association mechanism for effective training. By combining these new components, our model achieves state-of-the-art performance for individual tree segmentation on the newly introduced FOR-instanceV2 dataset, which spans diverse forest types and regions. Additionally, ForestFormer3D generalizes well to unseen test sets (Wytham woods and LAUTx), showcasing its robustness across different forest conditions and sensor modalities. The FOR-instanceV2 dataset and the ForestFormer3D code are publicly available at https://bxiang233.github.io/FF3D/.