Automatic Stem Mapping in Three Dimensions by Template Matching from Aerial Photographs

Abstract

Recent developments in remote sensing have shown promising results for the detection and species classification of individual trees from high-resolution imagery of forests in leaf-on condition. Existing image-analysis algorithms produce a two-dimensional stem map in the observed imagery. This paper develops a new algorithm that generalizes existing template-matching algorithms to produce a stem map in the three dimensions of a forest in leaf-off condition. A self-contained algorithm to produce artificial template images is introduced. The image-analysis algorithm was applied to oak ( Quercus robur L.) grown under standard silvicultural treatment in Denmark. One analysis gave detection rates in the range 85-98% and root mean squared errors on the stem base coordinates below 8.1, 7.6 and 20.5 cm in the x -, y - and z -directions, respectively. It is concluded that it is indeed feasible to produce a quite accurate three-dimensional stem map. Keywords: 3dDenmarkDigital Image AnalysisOakPhotogrammetryQuercus Robur LRemote SensingStem DetectionTree DetectionThree-DIMENSIONAL