The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Publications Copernicus
Download
Citation
Articles | Volume XLIII-B2-2022
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B2-2022, 633–639, 2022
https://doi.org/10.5194/isprs-archives-XLIII-B2-2022-633-2022
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIII-B2-2022, 633–639, 2022
https://doi.org/10.5194/isprs-archives-XLIII-B2-2022-633-2022
 
30 May 2022
30 May 2022

UTILISING SIMULATED TREE DATA TO TRAIN SUPERVISED CLASSIFIERS

P. Rönnholm1, S. Wittke1,2, M. Ingman1, P. Putkiranta1, H. Kauhanen1, H. Kaartinen2,3, and M. T. Vaaja1 P. Rönnholm et al.
  • 1Department of Built Environment, Aalto University, P.O. BOX 14100, 00076 AALTO, Finland
  • 2Department of Remote Sensing and Photogrammetry, Finnish Geospatial Research Institute, 02430 Masala, Finland
  • 3Department of Geography and Geology, University of Turku, 20500 Turku, Finland

Keywords: convolutional neural network, YOLO, simulation, feature image, classification

Abstract. 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.