1
Title: DELINEATION AND IDENTIFICATION OF INDIVIDUAL TREES IN THE EASTERN DECIDUOUS FOREST
Authors: Timothy A. Warner, Jong Yeol Lee and James B. McGraw
Download: treesintheforest.pdf
Abstract:
The Ecological Evaluation using Remote Sensing (EERS) group at West Virginia University is studying the health and status of West Virginia's forests using high spatial resolution imagery. Central to our work is a focus on classification and mapping of trees. This paper reports on our initial findings regarding the delineation of individual trees, and discusses future directions we hope to pursue. In a separate paper (Key et al, in this volume) we discuss tree species classification using multi-temporal imagery. Delineation of individual trees in the Eastern Deciduous Forest is challenging due to the variety of scales of tree canopy size, the relatively flat topography of the canopy, and the complex mosaic of the individual crowns. Nevertheless, the shadows between crowns provide a good first cut for identifying tree boundaries. A rank normalization is required to reduce problems due to variable illumination and vignetting. The size of the moving window used in this normalization is crucial in determining the scale of shadows that are enhanced. A window approximately the size of the average tree tends to enhance branching within the crown, whereas a window approximately three times the size of the average tree enhances individual tree crowns. The shadows are, however, in short, separate segments that do not isolate the trees. These segments can be connected by orientation information obtained from a direction of minimum texture algorithm. For each pixel in the image, texture is calculated over narrow groups of pixels (1 pixel wide by 11 long) centered on the pixel of interest.
The orientation of these groups is incremented by a small angle until all directions have been tested. The direction with the lowest texture is written out to a new file. A rule-based algorithm is currently being developed to use this information to join shadow segments.
Comments:
[Howard] Segmentation based on shadows. No 3D information is used. The data appear to be very similar to ours.
2
Title: Multi-source object-oriented classification of landcover using very high resolution imagery and digital elevation model
Authors: Akiko Harayama and Jean-Michel Jaquet
Download: article_ecognition.pdf
Abstract:
With high-resolution imagery such as ortho-photos and object -oriented software, semi-automatic mapping of urban land cover is possible. This approach allows low cost, rapid and standardized results, as shown by a pilot study carried out in Geneva
area (Switzerland).
3
Title: Measuring individual tree crown diameter with lidar and assessing its influence on estimating forest volume and biomass
Authors: Sorin C. Popescu, Randolph H. Wynne, and Ross F. Nelson
Source: Can. J. Remote Sensing, Vol. 29, No. 5, pp. 564–577, 2003
Download: measuringtreecrowndiameter.pdf
Abstract:
The main objective of this study was to develop reliable processing and analysis techniques to facilitate the use of
small-footprint lidar data for estimating tree crown diameter by measuring individual trees identifiable on the threedimensional
lidar surface. In addition, the study explored the importance of the lidar-derived crown diameter for estimating
tree volume and biomass. The lidar dataset was acquired over deciduous, coniferous, and mixed stands of varying age
classes and settings typical of the southeastern United States. For identifying individual trees, lidar processing techniques
used data fusion with multispectral optical data and local filtering with both square and circular windows of variable size.
The crown diameter was calculated as the average of two values measured along two perpendicular directions from the
location of each tree top by fitting a fourth-degree polynomial on both profiles. The lidar-derived tree measurements were
used with regression models and cross-validation to estimate plot level field-measured crown diameter. Linear regression
was also used to compare plot level tree volume and biomass estimation with and without lidar-derived crown diameter
measures from individual trees. Results for estimating crown diameter were similar for both pines and deciduous trees, with
R2 values of 0.62–0.63 for the dominant trees (root mean square error (RMSE) 1.36 to 1.41 m). Lidar-measured crown
diameter improved R2 values for volume and biomass estimation by up to 0.25 for both pines and deciduous plots (RMSE
improved by up to 8 m^3/ha for volume and 7 Mg/ha for biomass). For the pine plots, average crown diameter alone
explained 78% of the variance associated with biomass (RMSE 31.28 Mg/ha) and 83% of the variance for volume (RMSE
47.90 m^3/ha).
Comments:
[Howard] Discusses the USGS Field Inventory Analysis (FIA), which is relavant to Digital Government. Uses lidar for high-resolution DEMs. The algorithms are written in ENVI. Maybe we could setup a direct test of their algorithm?
4
Title: A Segmentation-Based Method to Retrieve Stem Volume Estimates from 3-D Tree Height Models Produced by Laser Scanners
Authors: Juha Hyyppä, Olavi Kelle, Mikko Lehikoinen, and Mikko Inkinen
Source: IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. 39, NO. 5, PP. 969-975, MAY 2001
Download: lidarsegmentation.pdf
Abstract:
In the boreal forest zone and in many forest areas, there exist gaps between the forest crowns. For example, in Finland, more than 30% of the first pulse data of laser scanning reflect directly from the ground without any interaction with the canopy. By increasing the number of pulses, it is possible to have samples from each individual tree and also from the gaps between the trees. Basically, this means that several laser pulses can be recorded per m^2. This allows detailed investigation of forest areas and the creation of a three-dimensional (3-D) tree height model. Tree height model can be calculated from the digital terrain and crown models both obtained with the laser scanner data. By analyzing the 3-D tree height model by using image vision methods, e.g., segmentation, it is possible to locate individual trees, estimate individual tree heights, crown area, and, by using that data, to derive the stem
diameter, number of stems, basal area, and stem volume. The advantage of the method is the capability to measure directly physical dimensions from the trees and use that information to calculate the needed stand attributes.
This paper demonstrates for the first time that it is possible to accurately estimate standwise forest attributes, especially stem volume (biomass), using high-pulse-rate laser scanners to provide data, from which individual trees can be detected and characteristics of trees such as height, location, and crown dimensions can be determined. That information can be applied to provide estimates for larger areas (stands). Using the new method, the following
standard errors were demonstrated for mean height, basal area and stem volume: 1.8 m (9.9%), 2.0 m^2/ha (10.2%), and 18.5 m^3/ha (10.5%), respectively. The precision obtained is better than that in conventional standwise forest inventories.
Comments:
[Howard] The lidar DEMs are similar to our high-res DEMs, which makes the peaking finding algorithm relavant to our work.