A4 Vertaisarvioitu artikkeli konferenssijulkaisussa
Utilising Simulated Tree Data to Train Supervised Classifiers
Tekijät: Rönnholm P, Wittke S, Ingman M, Putkiranta P, Kauhanen H, Kaartinen H, Vaaja MT
Toimittaja: A. Yilmaz, J. D. Wegner, R. Qin, F. Remondino, T. Fuse, I. Toschi
Konferenssin vakiintunut nimi: International Society for Photogrammetry and Remote Sensing
Julkaisuvuosi: 2022
Journal: International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Kokoomateoksen nimi: XXIV ISPRS Congress “Imaging today, foreseeing tomorrow”, Commission II
Tietokannassa oleva lehden nimi: XXIV ISPRS CONGRESS IMAGING TODAY, FORESEEING TOMORROW, COMMISSION II
Lehden akronyymi: INT ARCH PHOTOGRAMM
Sarjan nimi: International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Vuosikerta: 43-B2
Aloitussivu: 633
Lopetussivu: 639
Sivujen määrä: 7
ISSN: 1682-1750
eISSN: 2194-9034
DOI: https://doi.org/10.5194/isprs-archives-XLIII-B2-2022-633-2022
Verkko-osoite: https://doi.org/10.5194/isprs-archives-XLIII-B2-2022-633-2022
Rinnakkaistallenteen osoite: https://research.utu.fi/converis/portal/detail/Publication/177175882
The aim of our research was to examine whether simulated forest data can be utilized for training supervised classifiers. We included two classifiers namely the random forest classifier and the novel convolutional neural network classifier that utilizes feature images. We simulated tree parameters and created a feature vector for each tree. The original feature vector was utilised with random forest classifier. However, these feature vectors were also converted into feature images suitable for input into a YOLO (You Only Look Once) convolutional neural network classifier. The selected features were red colour, green colour, near-infrared colour, tree height divided by canopy diameter, and NDVI. The random forest classifier and convolutional neural network classifier performed similarly both with simulated data and field-measured reference data. As a result, both methods were able to identify correctly 97.5 % of the field-measured reference trees. Simulated data allows much larger training data than what could be feasible from field measurements.
Ladattava julkaisu This is an electronic reprint of the original article. |  |
