Håkon Næss Sandum, Hans Ole Ørka, Oliver Tomic, Erik Næsset, Terje Gobakken. (2026). Semantic segmentation of forest stands using deep learning. Silva Fennica vol. 60 no. 1 article id 25010. https://doi.org/10.14214/sf.25010

Keywords: forest management; image segmentation; remote sensing; stand delineation; U-Net

Highlights: Deep learning enables automated stand delineation that closely replicates expert human interpretation; The proposed approach has the potential to reduce time and cost required for operational stand delineation; Performance declines in highly complex forest environments, highlighting the need for further refinement.

Abstract | Full text in HTML | Full text in PDF | Author Info

Forest stands are the fundamental units in forest management inventories, silviculture, and financial analysis within operational forestry. Over the past two decades, stand borders have typically been delineated through manual interpretation of stereographic aerial images. This is a time-consuming and subjective process, which limits operational efficiency and introduces inconsistencies. Substantial effort has been devoted to automating the process, using various algorithms together with aerial images and canopy height models constructed from airborne laser scanning (ALS) data, but the manual interpretation remains the preferred method. Deep learning (DL) methods have demonstrated great potential in computer vision, yet their application to forest stand delineation remains unexplored in published research. This study presents a novel approach, framing stand delineation as a multiclass segmentation problem and applying U-Net-based DL-framework. The model was trained and evaluated using multispectral images, ALS data, and an existing stand map created by an expert interpreter. Performance was assessed on independent data using overall accuracy, a standard metric for classification tasks that measures the proportions of correctly classified pixels. The model achieved a pixel-level overall accuracy of 0.72. These results demonstrate the strong potential for DL-based stand delineation to be faster and more objective than manual methods. However, a few key challenges were noted, especially for complex forest environments. In these environments, model predictions showed over-segmentation and complex, irregular stand boundaries.

Sandum, Faculty of Environmental Sciences and Natural Resource Management, Norwegian University of Life Sciences, NMBU, P.O. Box 5003, NO-1432 Ås, Norway https://orcid.org/0009-0001-5764-2544 E-mail: hakon.nass.sandum@nmbu.no
Ørka, Faculty of Environmental Sciences and Natural Resource Management, Norwegian University of Life Sciences, NMBU, P.O. Box 5003, NO-1432 Ås, Norway https://orcid.org/0000-0002-7492-8608 E-mail: hans.ole.orka@nmbu.no
Tomic, Faculty of Science and Technology, Norwegian University of Life Sciences, NMBU, P.O. Box 5003, NO-1432 Ås, Norway https://orcid.org/0000-0003-1595-9962 E-mail: oliver.tomic@nmbu.no
Næsset, Faculty of Environmental Sciences and Natural Resource Management, Norwegian University of Life Sciences, NMBU, P.O. Box 5003, NO-1432 Ås, Norway E-mail: erik.naesset@nmbu.no
Gobakken, Faculty of Environmental Sciences and Natural Resource Management, Norwegian University of Life Sciences, NMBU, P.O. Box 5003, NO-1432 Ås, Norway https://orcid.org/0000-0001-5534-049X E-mail: terje.gobakken@nmbu.no

article id 155, category Research article

Minna Räty, Annika Kangas. (2010). Segmentation of model localization sub-areas by Getis statistics. Silva Fennica vol. 44 no. 2 article id 155. https://doi.org/10.14214/sf.155

Keywords: eCognition; form height; Getis statistics; image segmentation; local indicators of spatial association

Abstract | View details | Full text in PDF | Author Info

Models for large areas (global models) are often biased in smaller sub-areas, even when the model is unbiased for the whole area. Localization of the global model removes the local bias, but the problem is to find homogenous sub-areas in which to localize the function. In this study, we used the eCognition Professional 4.0 (later versions called Definies Pro) segmentation process to segment the study area into homogeneous sub-areas with respect to residuals of the global model of the form height and/or local Getis statistics calculated for the residuals, i.e., G_i^*-indices. The segmentation resulted in four different rasters: 1) residuals of the global model, 2) the local G_i^*-index, and 3) residuals and the local G_i^*-index weighted by the inverse of the variance, and 4) without weighting. The global model was then localized (re-fitted) for these sub-areas. The number of resulting sub-areas varied from 4 to 366. On average, the root mean squared errors (RMSEs) were 3.6% lower after localization than the global model RMSEs in sub-areas before localization. However, the localization actually increased the RMSE in some sub-areas, indicating the sub-area were not appropriate for local fitting. For 56% of the sub-areas, coordinates and distance from coastline were not statistically significant variables, in other words these areas were spatially homogenous. To compare the segmentations, we calculated an aggregate standard error of the RMSEs of the single sub-areas in the segmentation. The segmentations in which the local index was present had slightly lower standard errors than segmentations based on residuals.

Räty, University of Helsinki, Department of Forest Sciences, P.O. Box 27 (Latokartanonkaari 7), FI-00014 University of Helsinki, Finland E-mail: minna.s.raty@helsinki.fi
Kangas, University of Helsinki, Department of Forest Sciences, P.O. Box 27 (Latokartanonkaari 7), FI-00014 University of Helsinki, Finland E-mail: ak@nn.fi

Click this link to register to Silva Fennica.

If you are a registered user, log in to save your selected articles for later access.

Contents alert

Your selected articles

Follow @SilvaFennica