RiverDetectWood
RiverDetectWood: Automatic Detection and Quantification of In-Stream River Wood from Very High-Resolution Aerial Imagery.
Quantifying in-stream wood is central to river geomorphology and habitat studies, yet manual photo-interpretation of aerial imagery is slow and hard to reproduce. RiverDetectWood automates this workflow: it derives NDVI, brightness and Haralick texture indices from a very high-resolution orthophotograph (e.g. IGN BD ORTHO IRC), classifies pixels with a pre-trained Random Forest model, vectorises the wood class and computes per-element size and volume metrics (area, length, width, volume). It is exposed to QGIS as a Processing algorithm ("RiverDetectWood" provider), so it can be run from the Processing toolbox, from the Python console, or chained into Processing models. Requires Orfeo Toolbox (OTB) for the texture computation step, and the joblib/scikit-learn Python packages to load and run the classification model; the plugin can install these automatically from the Plugins menu. Method described in Grimmer et al. 2025 (Environmental Modelling & Software, https://doi.org/10.1016/j.envsoft.2025.106460) and Grimmer et al. 2025 (SoftwareX, https://doi.org/10.1016/j.softx.2025.102042). This QGIS plugin was developed by Romain Wenger (https://romainwenger.fr); the original Python classification tool it packages was developed by Gauthier Grimmer (http://g-grimmer.github.io/). Source code is GPL-3.0-or-later (see LICENSE).