The effect of manual and automatic filtering methods on the estimation of DBH, tree height and crown spread for different time frames

By Luuk Dael

Abstract
LiDAR data is becoming more widely used in the field of forestry and forest inventories. Another one of these applications is the estimation of tree parameters. A key part in processing LiDAR point clouds for estimating tree parameters is the filtering of the point cloud. Many different methods are still being used for filtering and no clear framework exist to follow to achieve the most efficient results. One of the methods tom best filter point clouds remains manual segmentation. This thesis looks at the time it takes to manually segment trees and at what time frame it remains efficient. In addition, these trees will be put in an automatic filter to see if there is any additional improvements. A total of 13 trees were cleaned and filtered to obtain the results. A visual and numerical assessment showed that after 40 minutes the results for DBH, height and crown spread started improving a lot less than the earlier time frames. Using additional automatic filters after manual cleaning did not significantly improve or worsen the outcome, but did reduce the file sizes which makes it require less computational power.