Selection of optimal tree top detection parameters in a context of effective forest management

Abstract

In the process of tree stand parameter estimation based on data from airborne laser scanning ALS, the detection of a single tree is an important starting point. The aim of this work is to develop optimal values of parameters in the process of detection of tops and segmentation of stands on the basis of ALS data analysis. The research was carried out on the basis of ALS data from raids carried out in 2007 and 2017 on a fragment of the Zajma forest district in the Zednia forest inspectorate (north-eastern Poland). Parameters analyzed included: Ground Sampling Distance [m], the level of smoothing of the Canopy Height Model (CHM) with the Gaussian filter (the size of the moving window, the value of standard deviation), the filtration of the output point cloud, as well as the application of the additional interpolation algorithm CHM based on the analysis of raster cells neighborhood. The research has shown that it is possible to indicate detection parameters that ensure a very high correlation between the number of automatically detected treetops and the number of trunks found during fieldwork. Importantly, the optimal detection parameters developed for remote-sensing materials from the years 2007 and 2017 differ slightly, which ensures generally high accuracy of ALS data and the possibility of implementing the values of these parameters in other research objects.