Carolin Fischer
Avdelingsleder/forskningssjef
Lenker
Linked inBiografi
Carolin has a Bachelor and Master in Forest Sciences and Wood technology from the Technical University (TU) in Munich.
She gained her PhD with the topic “Density and bending properties of Norway spruce structural timber – Inherent variability, site effects in machine strength grading and possibilities for presorting” in 2016 from the Norwegian University of Life Sciences (NMBU).
Carolin is the Head of Department for Forest Operations and Digitalisation at NIBIO. Her research work focuses on traceability along the forest value chain and wood quality evaluation early in the wood production chain. Her work includes also the coordination of SmartForest, a senter for research driven innovation (SFI), led by NIBIO.
Sammendrag
The European Union Deforestation Regulation (EUDR) mandates traceability of timber that makes up wood products from its harvest site to the end product to ensure sustainable wood sourcing. This study proposes a cost-effective, image-based method for tracing logs using alphabetic codes printed onto logs at the harvest site. These codes are detected and interpreted through a two-stage system leveraging deep learning models. The detection stage employs YOLOv8 to locate tracking codes in images of log piles. It is trained and evaluated on a dataset of 125 images, achieving an F1-score of 0.811 on unseen images. The recognition stage, trained on 1,020 images, uses YOLOv8 models to detect individual characters and their positions within each code. On a set of unseen images, the interpretation stage is able to identify 92.8% of the individual logs despite the limited quality of the printer and degradation of the codes due to stem wetness. Analysis indicates that errors predominantly arise in the character detection step. Compared to existing traceability approaches, this method is more cost-effective than RFID tags and attains higher accuracy than image-based biomarker tracking methods.
Forfattere
Henrik Persson Stefano Puliti Tuomas Yrttimaa Vikash Ghildyal Rasmus Astrup Johan Holmgren Carolin Fischer Nicolas Cattaneo Mikko VastarantaSammendrag
Tracing wood along the value chain is needed to preserve information about wood origin and understand associations between standing tree characteristics and the resulting wood quality. Linking timber products with standing trees without artificial marking remains a formidable challenge where detailed reconstruction of trees’ architectural characteristics could provide a solution. The objective is to develop an automated method for single-tree tracing using dense laser scanning from standing trees, leveraging branch patterns as unique fingerprints. A secondary objective is to explore how these fingerprints can be derived from computer tomography (CT) scans at sawmills, enabling the reconstruction of individual branch patterns. We use the AI algorithm BranchPoseNet to detect tree whorls and individual branch vectors from a terrestrial laser scanner-derived point cloud. A tree's unique fingerprint is derived by presenting the branch origins as a function of height and azimuth around the stem. This fingerprint is then reconstructed from a mobile laser scanner-derived point cloud collected from the same trees as well as from CT scans of knots and their positions in processed logs. By minimizing residuals between corresponding branch locations between the initial and reconstructed fingerprints, individual trees can be accurately linked, enabling full traceability from living trees to sawlogs. Preliminary results indicate that this approach is feasible for pine trees and that a limited number of unique branch connections may be sufficient for tree traceability in managed forest stands. More testing is needed to assess the performance of other species. We conclude that this method can be integrated into industry practices, being viable for automatically tracing trees from the harvested forest stands to the sawmill, thereby closing critical gaps in the value chain and enabling the attribution of additional information (e.g., origin, carbon sequestration potential) to wood products and other forest-based applications without artificial marking of logs.
Forfattere
Tuomas Yrttimaa Aapo Erkkilä Teemu Kamula Carolin Fischer Nicolas Cattaneo Mostafa Hoseini Juha Hyyppä Mikko VastarantaSammendrag
Efficient use of forest resources requires identifying the origin of wood to determine its intended purpose before harvesting. This purpose depends on the wood’s quality, which is influenced by the tree’s growth process and only fully revealed during processing at the sawmill. Identifying which attributes of a standing tree align with the quality requirements of sawn timber necessitates linking forest-collected data to information obtained at the sawmill. However, a nondestructive approach for establishing this connection without artificial marking of logs is currently unavailable. We propose a potential solution employing “tree fingerprints”—biometric patterns that capture trees’ unique branching arrangements along the stem. The tree architecture reflects a hierarchical growth pattern shaped by the interplay between genetics and the environment. Environmental variation leads to unique resource availability between individuals, and thus we assume that each tree develops distinct architectural characteristics, akin to the uniqueness of a human fingerprint. To investigate whether this uniqueness can be captured using terrestrial laser scanning (TLS), we conducted an experiment with 65 Scots pine (Pinus sylvestris L.) trees in a managed boreal forest stand. We derived tree fingerprints from two independent TLS data acquisitions (September 2021, November 2022) and matched corresponding fingerprints. In total, 52 trees (80.0%) were identified based on their architectural characteristics. The results showed that identifying ≥10 branch origins from independent reconstructions was sufficient to establish architectural uniqueness, resulting in 100% identification accuracy (n = 20 trees). These findings suggest that tree fingerprints can be used to condense the complex three-dimensional tree architecture into a two-dimensional pattern of points representing unique branch arrangement. Further, we demonstrate how this tree fingerprinting concept could be expanded across laser scanning methods to enable operational-scale wood traceability if point cloud data of standing trees is collected during forest operations and corresponding sawlogs are scanned at sawmills using X-ray computed tomography. Existing incentives support this kind of development: laser scanners on harvesters can assist operators, and sawlog scanning is essential for optimising timber yield. Seamlessly integrating wood traceability into industry practices would enable automated recording of data that can be further used for linking architectural characteristics of standing trees, grown within specific site conditions, to sawlog properties. This integration would enhance understanding of how tree architecture, environmental factors, and forest management influence desired properties of processed wood, enabling more informed decision-making for the wood procurement process.
