Computer vision applications using multispectral UAS imagery: comparing pixel and object-based methods for automatic classification of river landscapes

The use of unmanned aerial system (UAS) imagery in environmental sciences has rapidly increased due to the ease of use and low cost [1]. In general, the ecological informatics community widely uses spatial data analyses, especially via GIS. A growing number of applications now include UAS imagery, which can provide centimetre scale multispectral data [2]. Processing this information requires the application of image-based remote sensing techniques. Specifically, the differences between unsupervised and supervised image classification methods are discussed, with a focus on pixel and object-based computational methods [3]. Examples relevant to ecological studies are presented using multispectral imagery collected of river landscapes to illustrate how UAS data can be used to classify complex spatial features such as vegetation and submerged regions of different depths, including turbulent flows and complex lighting and shade conditions. REFERENCES: 1. Arif, M.S.M., Gülch, E., Tuhtan, J.A., Thumser, P., Haas, C., 2016. An investigation of image processing techniques for substrate classification based on dominant grain size using RGB images from UAV. Int. J. Remote Sens. 0, 1–23. https://doi.org/10.1080/01431161.2016.1249309 2. Hugenholtz, C.H., Whitehead, K., Brown, O.W., Barchyn, T.E., Moorman, B.J., LeClair, A., Riddell, K., Hamilton, T., 2013. Geomorphological mapping with a small unmanned aircraft system (sUAS): Feature detection and accuracy assessment of a photogrammetrically-derived digital terrain model. Geomorphology 194, 16–24. https://doi.org/10.1016/j.geomorph.2013.03.023 3. Black, M., Carbonneau, P., Church, M., Warburton, J., 2014. Mapping sub-pixel fluvial grain sizes with hyperspatial imagery. Sedimentology 61, 691–711. https://doi.org/10.1111/sed.12072