> Research

Lidar Based Forest Visualization:

Modeling Forest Stands and User Sutdies

Ikuko Fujisaki, Research Assistant
David L. Evans, Associate Professor
Robert J. Moorhead, Professor
Derek Irby, Research Associate
Mahnas Jean Mohammadi-Aragh, Research Assistant
Scott D. Roberts, Associate Professor

Mississippi State University,
MS 39762

Abstract

This is an ongoing cooperative research project between the Department of Forestry and the Visualization, Analysis, and Imaging Laboratory  (VAIL) at Mississippi State University.  Descriptive forest data comprised of large, multi-dimensional datasets, are difficult to comprehend. LIDAR (Light Detection And Ranging)-derived forest measurement data have been incorporated into the fully interactive forest visualization system which has been developed at the VAIL.  Using the system, a three-dimensional (3-D) graphic forest image is displayed in the CAVE Automatic Virtual Environment, a room size-virtual environment facility.  The system allows a group of a small number of people to freely interact with the virtual forest.  The ultimate goal of the system is to provide a tool for forest management in resources definition and operational decision-making.  To be the valid data presentation tool, it the viewer should see proper forest characteristics by observing the Based-based virtual forests.  The potential users’ (i.e. people in the forestry domain) level of understanding of the information depicted will be examined by comparing their perception of graphic forests to the real forest characteristics. 

Background

In scientific research, visualization is a way to display data in an easily understandable form. Visualized data provides viewers new insights into a dataset.  Current display technologies are able to build fully interactive immersive virtual environments in which the viewers have a sense of presence in the graphic images.  Such immersive visualization systems have the potential to aid in the examination of forest inventory data which contain various measurement variables in a complex spatial context.

Among existing immersive virtual environment systems, the CAVE is superior with respect to field of view and visual acuity.  It is a room-sized stereoscopic virtual environment projection system which was developed by the Electronic Visualization Laboratory, University of Illinois, Chicago (Cruz-Neira et al., 1992).  In the CAVE, high-resolution 3-D images are projected on the walls, floor and ceiling.  Multiple viewers can interact with graphic images with a hand held controller. Applications of the CAVE are quite diverse and include vehicle simulation, training, medicine, and natural resources.  An expanded field of view, and provisions for visual acuity and interactivity, make forest examination a potential new application area of the CAVE virtual environment. 

A limitation in utilizing a forest visualization system is the requirement of reliable tree measurement data over a large area.  Since traditional ground-based forest inventory is time consuming and labor intensive, it is difficult to obtain measurement data in a timely manner.  Previous studies confirmed that airborne LIDAR data are useful to characterize the structural condition of pine stands (McCombs et al., In press).  Various measurement variables such as tree location, canopy height, and basal area can be estimated using LIDAR data either directly or indirectly.  Methods of LIDAR forest measurement have the potential to solve the problem of data requirements for utilization of forest visualization systems in practice.

Cognitive science is not a separable issue in a visualization system development.  To have a valid visualization system, understandable levels of graphic representation of the modeled environment should be confirmed.  A successful visualization system is able to display meaningful information for viewers in a usable manner. Graphic image representation and desired photo reality depend on the intended purpose of the system.  The purpose of this specific forest visualization system is to provide a silvicultural management aid.  Therefore, it is necessary to examine the extent that foresters can extract silviculturally meaningful information by viewing the virtual forest using the system.  Early integration of human factors is important for further system development and to gain better understanding regarding the utility of the system.

Forest Measurments From Lidar

In our forest visualization model, the form of each tree is based on: location, total height, diameter at breast height (DBH), crown radii, and height to the base of the live crown (BLC).  The study area includes different types (e.g. age and density) of pine stands in which ground-based measurements at sample plots were obtained as reference data.  Two of the measurement variables, location and total height, were directly measured from LIDAR data.  The other three variables, DBH, crown radii, height to BLC of each tree were estimated from total height of identified trees using field-measured data in the study area.  

Identifying Tree Location and Estimating Total Height

LIDAR data obtained by a four return system with a small footprint of the LIDAR pulse (0.11 m) were used.  Two layers, a digital terrain model (DTM) and a canopy surface model (CSM), were created using raw LIDAR return data.    To create the DTM, the last return data were interpolated and filtered using a focal minimum function to keep pixels with relatively low z values within a specified window (kernel) (Figure 1).  The CSM was created by interpolating the first return data (Figure 2).  By subtracting the DTM from the CSM, a vegetative height surface from the ground was created.  To identify the location of peaks of trees and their heights, a model defined the pixels that have a local maximum value within a specified kernel passed over the vegetation height surface (assumed tree tops) (Figure 3).  Stem density of the stands was determined by using Identified-identified counts adjusted by the field sample data.

 
 Figure 1. Filtered DTM 

    
Figure 2. Canopy surface layer 


Figure 3. Canopy height measurement

Establishing a Relationship between Identified Total Tree Height and Other Variables Stand specific regression coefficients were estimated from the field-measured data of sample plots.  Figures 4, 5, and 6 show positive relationships between the total height and other measurement variables of one of the stands in the study area.  The general equation for a height-diameter curve (log ht = a + b DBH-1/2) was used to estimate DBH. Height to BLC and crown radii were estimated using simple linear regression since both had a significant linear relationship with height (p<0.05).

 
Figure 4.  Tree height vs. DBH  

 
Figure 5.  Tree height vs. height to BLC     

 
Figure 6. Tree height vs. crown radii 

User Tests

The goal of the forest visualization system is to provide a management tool to monitor forest stand conditions.  A central question for system use is “how close is the virtual forests to the real forests?”.  To answer this question, a series of user studies with people in the forestry domain is planned to examine the viewer’s degree of information extraction by observing the graphic forest in an immersive virtual environment.

Representation of Graphic Forests Is a graphic forest representative of the real forest? Before using Measurement-measurement data, field-measured forest stand data will be incorporated with the system.  The first test is to examine graphic representation of the system.  Human subjects will view a video recorded in the field and graphic forest images for the same pre-determined paths and their perceptions of specific forest characteristics will be compared.


Figure 7. Example of graphic forest stand 


 Figure 8. Example from field-based video of forest stand

Viewers’ Perceptions of Graphic Forests in the Virtual Environment

In general, higher fidelity is preferred in visualization systems because simple images may under-represent the real environment.  However, since complex models are computationally intensive, speed of interaction is limited by enhancing photo-reality.  The influence of photo-reality on viewers’ perceptions will be tested using different graphic forest models to identify an optimal level of graphical detail for use (Figure 9).  Viewers’ perceptions of optimized graphic forest environments will then be compared with the real forest characteristics. The subjects will explore Measurement-measurement based graphic forest stands in an immersive virtual environment (Figure 10). 


Figure 9. Example of simple tree model


Figure 10. Example of graphic forest image in the CAVE

Current Status

LIDAR measured tree data of different types of forest stands were incorporated with the forest visualization system.  The utility of the system is being examined for pine forests; however, understory hardwood species in different graphic form will be introduced to improve representation of the true ground conditions.  Currently, the research is at the user studies with the human subjects stage. A series of user tests will be conducted by following cognitive science approaches.  The result of the experiments will be an integration of human factors to ongoing visualization system development.    

Acknowledgements

This research is sponsored by the NASA Earth Sciences Application Directorate as part of the Remote Sensing Technologies Center monitored by NASA/SSC. This paper has been approved for publication as Journal Article No. FO223 of the Forest and Wildlife Research Center, Mississippi State University.

References

Cruz-Neira, C., D.J. Sandin, T.A. Defantl, R.V. Kenyon, and J.C. Hart. (1992). The Cave: Audio visual experience automatic virtual environment. Communications of the ACM. 35(6): 64-72. McCombs, J.W., S.D. Roberts, and D.L. Evans. (In press). Influence of fusing LIDAR and multi-spectral imagery on remotely sensed estimates of stand density and mean tree height in a managed loblolly pine plantation. Forest Science.