-Sean Hoyt

I have been busy constructing an FSK reader to interface TI glass tags and for future moisture, density, and diameter research. I also have continued my work towards the UW Brockman Tree Tour. Much work has gone into the construction of UW map navigation, database lookup, and media presentation via a Pocket PC device for the tree tour.

Analysis of three-dimensional forest structure from airborne laser scanner data

- Hans-Erik Andersen

Objective: Extraction of detailed forest structure information, including stem density, tree heights, crown dimensions, crown form characteristics, and foliage density, is important for a variety of forest management applications, including resource inventory, harvest planning, and fire fuels monitoring. The emergence of high-resolution remotely sensed data, such as LIDAR, allows for many forest measurement tasks to be (at least partially) automated. The automated extraction of forest measurement information from remotely sensed data can potentially decrease inventory costs and support more precise and intensive approaches to resource monitoring and management.

Methods: Algorithms are currently being developed to extract crown dimensions and locations directly from "raw" (unfiltered) LIDAR data. These algorithms build upon promising results of a mathematical morphology-based approach developed and presented last year. In addition, algorithms are being developed to infer canopy coverage and the distribution of crown bulk density (important for canopy fuels monitoring) across the landscape. Field data has been acquired at study areas within the Blue Ridge installation at Capitol State Forest and Fort Lewis Military Reservation.

Results: Output of one algorithm applied to a small (0.2 ha) area is shown graphically in Figure 1. The algorithm is currently being applied to selected areas within the Capitol Forest study area to compare automated measurements to field-based measurements within numerous geo-referenced growth plots.

Timeframe: Results of this research will be presented in May and June, 2002.

Elastic Surface Draping - Modeling Terrain and Canopy Over 3D Spatial Dataset

- Adnan Hyder Yusuf

Introduction: LIDAR technology has been used to map terrains for some time now. However its application to densely foliaged terrains has been limited by the lack of proper methods to analyze the data. The present project aims to develop a method of visualizing the data graphically with the application of methods used in cloth animation. The working principle is to drape a piece of fabric over the data points in 3D space. As the fabric settles over the dataset, a shape can be made out which would otherwise be difficult to visualize from the data points themselves. The method must be able to render a shape accurate to a certain extent with minimal computation on a microcomputer. The current project aims at developing an approach that attains all these objectives.

Methodology:

The microcomputer modeling and simulation of cloth draped over an object1,2 has been a subject for research in the computer graphics and textile engineering community for well over a decade. In the current research, we plan to use this approach to solve the visualizing problem of LIDAR dataset. The required results are different for the two cases: in the former case, the objective is to model the cloth as realistically as possible, whereas in the latter, the emphasis is tracing out the shape of the dataset. However, the similarity in the two scenarios is also apparent.

Figure: the fabric

In our case, we envision the fabric to be a mesh of threads that intersect at right angles. At each intersection, a particle of finite mass¹ is positioned. The threads themselves are of negligible weight, and are assumed to have a predefined spring constant so that they function as springs. The mass of the fabric is a total of the particle masses. Thus the fabric is a mesh of particles that can be deformed within certain limits.

The 3D spatial dataset is assumed to be a set of one-sided constraints in that it allows the fabric to move away from it only in the positive vertical axis. Initially, the fabric is assumed to be positioned over the dataset so that their horizontal projections coincide. Once the fabric is allowed to fall freely over the dataset under the effect of gravity, each point in the dataset comes in contact with a particle in the mesh, thus becoming a point of constraint. Since the mesh has a much higher resolution than the dataset, we can safely assume that the dataset points come in contact with particles only and not threads.

Figure: Fabric draped over one-sided constraints

In addition to spring forces and gravity, we also assume a normal force of predefined magnitude acting on each particle. Therefore, a particle that coincides with a data point will have both forces acting vertically. On the other hand, on a suspended particle, the normal force will act perpendicularly to its orientation, and planar spring forces will act on all four thread pieces connected to it. In either case, these two forces will be adjustable, thus allowing control over how closely the fabric clings to the dataset surface.

Since we are only interested in the equilibrium state of the fabric, we do not attempt to compute the transient states of the fabric as it settles. However, once it does, we apply Finite Element Analysis techniques to determine the final state of each particle in the mesh. Currently, we envision the fabric to be a rectangular mesh³ for Finite Element Analysis purposes: the values of the nodes inside the rectangular area can be interpolated. Once the force equilibrium for each node in the fabric has been reached, their respective positions in three-dimensional coordinates can be determined.

This fabric surface can then be used in a graphical algorithm to be rendered or texture-mapped. Lighting models can be applied to produce a realistic terrain image. An image such as an aerial photograph can be mapped onto the surface to produce a realistic terrain model.

There are other algorithms that exist to interpret the data to result in similar visual representations. In most cases, these algorithms tend to filter or smooth out the data to produce realistic results. This however has the inevitable consequence of losing useful data in the process. In the process that we plan to develop, we do not filter the data in any way. The results obtained should therefore be more faithful in representing the dataset.

The challenge in this approach is to develop an algorithm that determines the states of each particle of the fabric with minimal computation.

References:

1. Breen, D.E., House, D.H. and Wozny, M.J.,"Predicting the Drape of Woven Cloth Using Interacting Particles", SIGGRAPH '94 Conference Proceedings, pp. 365-372
2. Weil, J., "The Synthesis of Cloth Objects", Computer Graphics (Proc. SIGGRAPH), Vol. 20, N0. 4, pp. 359-376, 1986.
3. Segerlind, L.J., "Applied Finite Element Analysis", 1976, John Wiley & Sons, Inc.

- Matthew Walsh

This quarter was spent primarily learning the details of using ArcMap. More time than was originally anticipated went into getting familiar with the data editing and manipulation tools within ArcMap. I also spent time learning what was available in Arc Spatial Analysis and also using the data in ArcScene.

Time was spent cleaning and creating data for use within Envision. Using sites at Tiger Mountain, I have completed a number of sample runs while varying distances and also varying the tree representations to see what limitations may be encountered when representing stands in the area based on the DNR's data.

	a long distance view at Tiger Mountain

Based on the sample images run through this last quarter it is noticeable that there are a number of problems within the DNR inventory data that need to be modified. This next quarter will be filled with cleaning the data and then running harvest options for the area based on pre-determined harvest areas.

- Jianyang Zheng

During the past winter quarter, I studied the fundamentals of GPS and INS under the guide of Dr. Ahmed, based on the reference "Global Positioning Systems, Inertial Navigation, and Integration". Furthermore, with the help of Dr. Wang, I specified the objectives, methodology, and work plan of the research. I have also had several discussions with Doug St. John which were very fruitful for my research. With GeoExplorer3, I surveyed in the campus and analyzed the data with GIS. The following is part of the surveyed points and trail (in red) in ArcMap.

The next step is to analyze the data with statistics tools. GIS can provide the coordinate data of the GPS-surveyed features and the 'true' positions surveyed with traditional method. I also plan to go to Capital Forest and specify the field procedures. I should get the DGPS data before April14th. According to the work plan, I now also need to be trained by ApplAnix. The site work is planned to start on April 14th.

- Sam Pittman

Over the past quarter we have formulated a forest corporation decision support model which allows the decision making process to be analyzed hierarchically. The model we are pursuing is general enough to consider as many levels as necessary, however forest planning problems will likely only use two or three. The decision making process can best be described using a bilevel model. An upper level authority must make a strategic decision which allocates the company's (forests) resources under strategic level constraints and a set of equilibrium constraints. These equilibrium constraints describe the optimal reaction of lower level agents to the upper level authority's decision. The equilibrium constraints need not represent an actual agent, they may also be thought of as an agent acting in the best interest of a forest which he/ she has been given the responsibility of stewardship. They may be separate land trusts, large landscapes that are part of a national forest or simply the different regional divisions of a company. Over the next few quarters we will be developing an algorithm which solves this type of model with imperfect information among levels and applying these developments to a reasonable size test problem.