Estimating Forest Volume using Lidar CHM Image with Local-based filtering methods
2011
Hochschulschrift
Zugriff:
99
The three-dimensional spatial distribution information of forest stand can be obtained through the penetrability and multiple reflections of LIDAR, so we can get the accurate forest height coverage model, which can be a great aid to the calculation of the height of forest stand. Usually, we get the canopy height model (CHM) by performing interpolation to the LIDAR point cloud data. But the canopy is not smooth and there are irregular protuberances and pits on its surface, or gap may be formed because the density of the point cloud data of LIDAR is not enough, errors are produced when we estimate the number of plant of forest stand and the estimation of volume of timber of forest stand is also affected. This research aims at discussing whether segmentation results of different smoothing degrees have differences in the number of plant surveyed when the image segmentation is performed to CHM images and discussing whether they can be combined using the classification of terrain features. In the end, we estimate the total timber volume of forest stand of Cryptomeria japonica, Chamaecyparis formosensis and broad-leaves mixed stands around the recreation area of Alisan based on the number of superior trees estimated by LIDAR and the height distribution curve of Weibull function. The research results show the overall accuracy for using local repair smoothing method to detect the number and location of umber of stems is 82%, 74% for median method and 69% for mean value method. If the terrain feature classification is used, the overall accuracy will drop to 76%; so it is not recommended. While, the height distribution curve of Weibull function built with the survey data of sample trees can represent the distribution of each height level in sample areas of the 3 kind of sample trees. Then, estimate the number of overtopped trees with Weibull function, based on distribution frequency of height of superior trees surveyed, and calculate the timber volume of overtopped trees. Finally, the sum of the timber volumes of dominant trees and overtopped trees will represent the total timber volume of forest stand estimated by LIDAR; results for each sample area: the average estimated timber volume, the real volume of timber, RMSE and the error rate of the sample area of Cryptomeria japonica fortune are 68.72 m3/plot, 71.35 m3/plot, 9.45 m3/plot and 13% respectively; the average estimated timber volume, the real timber volume, RMSE and the error rate of the sample area of Chamaecyparis formosensis area are 32.29 m3/plot, 38.60 m3/plot, 6.31m3/plot and 17% respectively; the average estimated timber volume, the real timber volume, RMSE and the error rate of the sample area of broad-leaves mixed stands area are 18.11 m3/plot, 22.01 m3/plot, 3.90 m3/plot and 28% respectively; all of the above are underestimated and the error rate of broad-leaved trees is the broad-leaves mixed stands area.
Titel: |
Estimating Forest Volume using Lidar CHM Image with Local-based filtering methods
|
---|---|
Autor/in / Beteiligte Person: | Ji, Xiang-Yu ; 紀祥鈺 |
Link: | |
Veröffentlichung: | 2011 |
Medientyp: | Hochschulschrift |
Sonstiges: |
|