MODELLING SINGLE TREES FROM TERRESTRIAL LASER SCANNING DATA IN A FOREST RESERVE

摘要

A terrestrial laser scanning survey has been performed on a forest reserve site. This paper represents the processing of the dataset. The main objective of this study was to extract single tree parameters automatically from this dataset. Three general, metric parameters were estimated, the tree position, the stem diameter at breast height, and tree height. These single tree data are equally important for the monitoring of reserves and management planning of primarily economic forest sites. Two basically different processing approaches were proved for each object; a geometrical and a statistical. The main difference between them occurs principally at tree positioning, one detects trees with tracing of laser beams (it's called crescent moon method), the other works with statistical clustering of measurements. The recognized trunks were modelled with fitted circles in different height layers, which make the instant diameter determination in any height possible. These slices and the truncated cones between form the model of the trunk. Two methods were tested for height estimation. One was a direct method based upon the Digital Surface Model; however a new method was developed to predict the tree-top based upon the modelled trunk and then the nearest local maximum was selected to that predicted point in the DSM. The other was an indirect method, where the modelled stems were described as bodies of revolutions, and the heights were extrapolated. The accuracies of the methods were investigated by means of field survey and photogrammetric measurements based on very high-resolution aerial images. Both the clustering and crescent moon methods were appropriate in tree positioning and stem modelling. The height estimation based upon the DSM yielded good results. The extrapolation method was tested for European Larch only. The height, in one-third of the cases, was estimated with this method within tolerance. This method can be taken into consideration in the case of leaves on scanning in very dense, multi-storey forests. The stem detection was successful in four fifth of the 213 trees on a 30 m radius sample circle. The estimations of the diameter at breast height were within ±2 cm in half of the samples. The height estimation based on this leaves-off scanning was ±1-4 m accurate depending on the methods.