In this master's thesis, we studied non-normalized and normalized intensities of aerial laser scanning on selected forest stands and individual tree canopies in the test area of the Zelenica, Slovenia. The selected area was aerial laser scanned by four data sets recorded in two wavelengths (1064 nm and 1550 nm) and in different parts of the year (March, May, July and September). Intensity values were first analysed at randomly selected tree canopy peaks in the forest area, where we analysed values of reflected intensities according to species composition (coniferous, coniferous with deciduous, deciduous with coniferous or deciduous forest). Based on a field trip, orthophoto and digital canopy model, we selected 40 individual trees for six different tree species: spruce, pine, larch, beech, maple and ash. Only the points of the upper 3 m of tree canopies were considered in the analysis. First, a comparison was made of tree canopies based on geometry by comparing cross-sectional areas, floor areas and volumes. Secondary, a comparison of tree species based on normalized and non-normalized intensities was performed. Further, we have statistically significantly assessed the differences between the characteristics of intensities of individual tree species. We developed a model for predicting individual tree species from aerial laser data using a decision tree algorithm. Using the decision tree algorithm, we successfully identified trees as coniferous or deciduous (model accuracy 95%), and we also successfully identified selected tree species (model accuracy 60%).
|
