Three-dimensional Visualization of Forest Landscapes by VRML

En-Mi Lim and Tsuyoshi Honjo

Graduate School of Science and Technology, Chiba University

1-33 Yayoi-cho, Inage-ku, Chiba, 263 Japan

(Correspondence

Name: En-Mi Lim

Phone and fax : +81-47-308-8896

E-mail: limeunmi@green.h.chiba-u.ac.jp

Address: Kankyoshokusai Lab., Faculty of Horticulture, Chiba University

648 Matsudo, Matsudo-shi, Chiba-ken, Japan, 271-8510)

Source Reference

Lim, E. and Honjo, T., 2003b. Three-dimensional visualization forest of landscapes by VRML. Landscape and Urban Planning, 63,175-186.

Abstract

Computer technology has been used to develop three-dimensional (3-D) forest landscape visualization systems that include the function of three-dimensional digital plant modeling. While the earlier systems accurately simulated forest landscapes, they lacked sufficient speed and could not adequately perform walk-through simulations.

The objective of this study is to describe the forest landscape visualization procedure capable of walk-through simulations and its application. We developed a forest landscape visualization system using Virtual Reality Modeling Language (VRML). This visualization system works with data of forest stands rather than individual trees. To confirm the feasibility of the system, we visualized an actual forest landscape with thousands to tens of thousands of trees. We also simulated a variety of forest landscapes and showed how this system can be used to simulate the changes of forest landscapes that occur as a result of natural processes or man-made disturbances such as planting, thinning, and harvesting.

This visualization system was capable of walk-through simulation and the three-dimensional images rendered by the system enabled us to effectively visualize the forest landscape resources. The visualization by computer graphics also made it possible to confirm the accuracy of forest data.

Keywords: Forest landscape visualization; Forest landscape simulation; Walk-through simulation; VRML; Computer graphics

1. Introduction

In the field of forest resource management, forest landscape visualization has been mainly used for accurately analyzing existing forest landscape resources and assessing the visual impact of proposed forest operation plans (Fridley et al., 1991; Lange, 1994; Orland, 1994; Bergen et al., 1998; McGaughey, 1998).

Over the past 30 years, the improved capabilities of computer hardware and software have allowed us to simulate and visualize natural complex forms and phenomena such as plant growth and the effects of changes in atmospheric conditions and light (Ervin and Hasbrouck, 1999). There are commercial visualization systems such as World Construction Set, Bryce, and VistaPro etc., which produce realistic terrain images. McGaughey (1998) and Muhar (2001) reviewed the features of these systems in landscape visualization.

Many researchers have developed algorithms for digital plant modeling (Oppenheimer, 1986; De Reffye et al., 1988, 1991; Prusinkiewicz et al., 1988), and three-dimensional (3-D) digital plant modeling systems have been used to develop forest landscape visualization systems such as the AMAP system (De Reffye et al., 1988; Perrin et al., 2001), the Vantage Point system (Fridley et al., 1991; Bergen et al., 1998), and the SmartForest (Orland et al., 1994; Orland, 1997). These 3-D visualization systems place individual trees on a digital terrain model (DTM) via a graphic user interface (GUI), and the images they render have nearly reached the level of photographic realism. Accordingly, the digital plant modeling techniques of these systems can be used to render forest landscapes accurately.

While these 3-D visualization systems are highly realistic, they have not yet achieved sufficient speed in modeling to allow users to use the visualization as a decision support tool (Orland et al., 1994). In many presentations using these systems, static images or animations generated using a series of static images are mainly used (Bergen et al., 1998; Bishop, 2001; Perrin et al., 2001).

Honjo and Lim (2001) developed a system for real-time rendering of landscapes using Virtual Reality Modeling Language (VRML). With their system, actual gardens with thousands of plants could be visualized in real-time in walk-through simulations on a personal computer. However, the system proved inadequate for presenting forest landscapes, as forest data are generally managed not in units of individual trees, but forest stands. A forest stand is a group of trees that have similar structures and are in the same growth stage. Forest stand tables describing the distributions of the species, sizes, and ages of the dominant trees in individual stands provide data for the visualization of a forest landscape.

In this study, we developed a forest landscape visualization system capable of walk-through simulation. By walk-through simulations, precise recognition is possible in selecting alternative plans. We incorporated several functions to the VRML system of Honjo and Lim (2001) to internalize and process data on forest stands for the simulation of large-scale forest landscapes. To confirm the feasibility of this forest landscape visualization system, we visualized an actual forest landscape with thousands to tens of thousands of trees. We also simulated a variety of forest landscapes and showed how this system can be used to simulate the changes of forest landscapes that occur as a result of natural processes or man-made disturbances such as planting, thinning, and harvesting.

2. Methods

2.1 About the VRML

VRML is one of Web3d technologies, which are used to deliver interactive 3-D objects and worlds across the Internet. Several Web3d technologies such as Pulse3D, Cult3D, Viewpoint and Shockwave3D etc. are developed or being developed now but only VRML can be practically used for walk-through simulation.

VRML is a high-performance language for 3-D visualization on the WWW (World Wide Web). As a programming language and library for 3-D computer graphics, VRML has many functions such as shading, setting objects, projection, and texture mapping. Virtual reality worlds can be easily built on the WWW with this technology.

VRML 1.0 was introduced in 1994 and VRML 2.0 (97) with more dynamic and interactive functions was made in 1996. GeoVRML and X3D, which are the successors of VRML, are currently being developed. In this study, VRML 97 was used in the present system.

Users working with a browser that supports VRML can easily download programs written in VRML from the WWW and view 3-D images on their personal computers. These VRML browsers are available for the Windows, Macintosh and Unix operating systems, as well as other platforms. In this study, Cosmo Player2.1.1 (Silicon Graphics Inc.) was used as a VRML browser with Internet Explorer6 (Microsoft Inc.) on Windows (Microsoft Inc.).

We tested several VRML browsers on Internet Explorer and on Netscape with Windows (98/Me/2000) and Macintosh (OS 9). The results are shown in Table 1. The rendering speeds of these VRML browsers were almost the same.

To write and run VRML code only a VRML browser and Internet browser are required. Cosmo Player and other VRML browsers can be downloaded as freeware, and the development environment can be built economically (Honjo and Lim, 2001).

Table 1

Test results of VRML browser

2.2 Visualization of forest landscapes by VRML

2.2.1 Visualization procedure

In this study, we intended to develop a data driven visualization of a forest landscape, which represents the underlying database. The visualization procedure using VRML was divided into the three steps (Fig.1). In the first step, the 3-D digital data of the terrain are obtained from a contour map and then data on the attributes and locations of the dominant trees of each stand are obtained from a forest stand table and stand map. In the second step, a conversion program is used to convert the data on the terrain and vegetation in the forest landscape into VRML format. In the final step, the 3-D image of the forest landscape is generated on a local computer.

Fig. 1 Forest landscape visualization procedure using VRML Fig. 2 Three different strategies to model plants as geometric objects in VRML (Ginkgo biloba L.)

2.2.2 Tree data and modeling

Information on forest resources is managed not in units of individual trees, but in forest stands and information on forest stands is presented in tables and maps. A forest stand table contains information on the area of a stand, together with the species, size, age, and density of the dominant trees in the stand (Fig.1). A forest stand map expresses the boundaries of each stand on a contour map. In many cases, forest stand maps are drawn from aerial photographs.

In this study, data on stand attributes, i.e., species, sizes, and ages of the dominant trees within each stand, were easily obtained from a forest stand table. The data on the locations of the trees were obtained from the forest stand map. The 3-D plant models were placed at constant intervals inside a stand boundary, according to the density of the trees within the stand.

VRML employs three different strategies to model plants as geometric objects. The 3-D plants can be easily modeled using simple objects such as cones, cylinders, and cubes. The size of the VRML program file used to generate the visualization shown in Fig. 2a was about 0.6KB. While this model provides recognizable quantitative information about a forest stand, the visual image it produced is far from realistic. Leaves, twigs and the trunk of a tree are described by sets of polygons (Fig. 2b). While the use of small numbers of polygons makes these models appear unrealistic, increasing the number of polygons generates an image that appears considerably more realistic. However, the number of polygons that have to be rendered can be immense, varying from thousands to millions. The 3-D plant shown in Fig. 2b consists of 7,705 polygons and the size of the file for this rendering is 6.0MB. Given the time required to generate such large numbers of polygons, walk-through simulations are difficult to achieve in the case of forest simulations. The third strategy is texture mapping of 2-D plant images on two planes, as shown in Fig. 2c. A plant texture recorded in a transparent GIF format is mapped on two planes that cross each other. In this case, the size of the file was about 0.7KB. Through the use of texture mapping, we obtained 3-D realism with a small file within a relatively short rendering time.

2.2.3 2-D Plant images

To express plants by texture mapping, we used computer graphic images of plants made by AMAP (Atelier de Modelisation de Architecture de Plants). A database of 2-D plant images was generated by AMAP to provide plant textures.

AMAP is a high-performance visualization system for landscape planning that was originally developed in the early 1980s by CIRAD (Center Internationale Recherche Agricultural Development) and later refined in work by De Reffye et al. (1988). Many researchers have simulated various landscapes using the AMAP system (Honjo et al., 1992; Morimoto, 1993; Saito et al., 1993; Honjo and Takeuchi, 1995; Perrin et al., 2001).

AMAP can model more than 300 types of plants, including flowers, bushes, and trees. The AMAP system can also generate 3-D models of a given plant at different ages (Fig. 3a), at different times of the year (Fig. 3b), and before and after pruning (Fig. 3c). AMAP also generates different shapes for the same plant type by changing the seed number of a plant.

2.2.4 Digital terrain data and modeling

When there are elevation data on a grid, the terrain is easily visualized in VRML using a node (command used in VRML) called ElevationGrid.

To obtain elevation data on a grid from data on control points, we developed a conversion program of terrain in Visual Basic (Microsoft Inc.). To convert data from the contour map, the map was scanned into a computer, the data were read, and then the conversion program was used to convert the control points along the contour lines into a lattice raster format. The 3-D terrain model was generated from the contour map.

2.2.5 Conversion program to VRML format

We developed a conversion program in Visual Basic to convert data on forest stands into the VRML format.

In this conversion program, individual trees were set on the terrain model at set intervals according to the tree density within the stands. To give the forest a more natural appearance, this program has a function to change the intervals between trees randomly.

Fig. 3 Example of 2-D plant images and effects obtained from AMAP

3. Visualization of real forest landscapes

In 1916, Tokyo University established a 5,821 ha forest in the Chichibu-Tama National Park of Saitama Prefecture for research and educational purposes. The forest is set in a cool temperate, boreal zone and has mountainous terrain. We simulated a part of this forest using the VRML system.

We simulated a forest stand (1150m X 740m) planted with hinoki (Chamaecyparis obtusa) (Fig. 3a). To perform real time rendering of landscape, the hinoki (Fig. 3a) image was converted to a low-resolution image (about 1KB). The VRML system populated the forest stand with about 2,630 trees, placed at intervals of 5m (Fig. 4b). The system took about one minute to generate the image whose screen resolution was 1024X768 pixels, running on a Windows platform with an AMD Athlon 900MHz processor and 640MB of RAM. Once the image was completely constructed, the dynamic images were rendered smoothly (about several frames per second) as the viewpoint was changed. The graphic quality of the simulated VRML image was close to that of a photographic image.

When we compared the image with the photograph from several viewpoints, the shape of one section of the simulated stand in Fig. 4b differed from that of the photograph in Fig. 4a due to an inaccuracy in the original forest information provided. It is very difficult to find such a mistake of data input in forest resource management without using visualization. Thus, we learned that the accuracy of forest data could be confirmed by rendering recorded forest information with computer graphics.

We then simulated more forest stands simultaneously using the VRML system (Fig. 4c). In total, about 15,000 trees were planted at intervals of 5m. It took the system about three minutes to process the data, but once the processing was complete, the dynamic images were rendered smoothly (about 50 frames per minute) and walk-through simulation by shifting viewpoints was possible. To render images smoothly, we had to use low-resolution plant images (about 1KB). Also, a certain amount of RAM is necessary. The image of Fig. 4c was made on a computer with at least 512 MB of RAM.

Fig. 4 Examples of forest stand simulations

4. Simulation of forest landscapes

Next, we generated a variety of forest landscapes and showed how the system could be used to simulate the changes of forest landscapes that occur as a result of natural processes such as seasonal changes or human induced disturbances such as planting, thinning, and harvesting.

Using the age and seasonal effects of AMAP shown in Fig. 3, we can easily simulate forest landscapes changed by the growth of the plants and simulate seasonal changes between summer and winter of a forest landscape, including colored and fallen leaves. In Fig. 5, we simulated seasonal changes in a mixed-species stand using the images of five species of trees, including cherry trees. The seasonal difference between spring and summer is illustrated by rendering the cherry trees with cherry blossoms (Fig. 5a) and without them (Fig. 5b). The VRML system can be used to simulate a variety of forest landscapes composed of various species of trees at various stages of growth.

Fig. 6 illustrates a simulation of a plantation landscape using the image of the 15-year-old hinoki shown in Fig. 4a, with simulated changes in the forest landscape resulting from different densities of plantation within the stands. Through 3-D images produced by the VRML system, we can intuitively understand how silvicultural activities such as cutting and thinning can be expected to change the structures of forests, and accordingly adjust proposed plans on the basis of their expected visual impact.

Through the use of various functions mentioned in this study, the VRML system enables forest managers to visualize alternative plans, therefore forest landscapes may be more efficiently managed. It is also a useful decision-support tool for policymakers and the general public.

Fig. 5 Simulation of seasons using the seasonal effects of AMAP in a mixed-species stand

Fig. 6 Stand planted with a single species of tree simulated with different densities of plantation

5. Conclusions

In this study, we developed a system for forest landscape visualization capable of walk-through simulations using VRML and 2-D plant images. To confirm the performance and feasibility of this visualization system, we simulated an actual forest landscape with thousands to tens of thousands of trees.

Through the real-time rendering of 3-D images with this visualization system and its capability for walk-through simulation, we could intuitively realize forest landscapes. In addition, by comparing photographic images with the computer graphic images generated by the system, we could confirm the accuracy of the recorded data.

Acknowledgements

We would like to thank Dr. Kaoru Saito, University of Tokyo for providing us with the data of the Tokyo University Forest in Chichibu.

References

Bergen S. D., McGaughey. R. J., Fridley. J. L., 1998. Data-driven simulation, dimensional accuracy and realism in a landscape visualization tool. Landscape and Urban Planning, 40, 283-293.

Bishop, I. D., 2001. Predicting movement choices in virtual environments. Landscape and Urban Planning, 56, 97- 106.

De Reffye, P., Edelin, C., Francon, J., Jaeger, M., Puech, C., 1988. Plant Models Faithful to Botanical Structure and Development. Computer Graphics, 22, 151-158.

De Reffye, P., Elguero, E., Costes, E., 1991. Growth units construction in trees: a stochastic approach. Acta Biotheoretica, 39, 325-342.

Ervin, S. M., Hasbrouck, H., 1999. Thirty Years of Computing in Landscape Architecture. Landscape Architecture，89(11), 54-56.

Fridley. J. L., McGaughey. R. J., Lee. F. E. 1991. Visualizing Engineering Design Alternatives on Forest Landscape. Presented at the 1991 ASAE Annual International Meeting. Paper No. 917523. ASAE, 2950 Niles Rd., St. Joseph, MI 49085-9659 USA.

Honjo, T., Lim, E., 2001. Visualization of landscape by VRML system. Landscape and Urban Planning, 55, 175-183.

Honjo, T., Saito, K., Kumagai Y., 1992. A Study of the Application of Plant Modeling and Its Visualization on Landscape Prediction. Journal of the Japanese Institute of Landscape Architecture, 55(5), 301-306. (In Japanese with an English summary)

Honjo, T., Takeuchi, S., 1995. Visualization of Local Environment. The Bulletin of School of High-Tech. Tokai Univ., 5, 177-181. (In Japanese with an English summary)

Lange, E., 1994. Integration of computerized visual simulation and visual assessment in environmental planning. Landscape and Urban Planning, 30, 99-112.

McGaughey. R. J. 1998. Techniques for visualizing the appearance of forestry operations. Journal of Forestry, 96(6), 9-14.

Morimoto, Y., 1993. Computer Visualization of the Vegetation of Katsura Imperial Garden by a Plant Modeling and Visualization System. Journal of the Japanese Institute of Landscape Architecture, 57(5), 113-120. (In Japanese with an English summary)

Muhar, A., 2001. Three-dimensional modeling and visualization of vegetation for landscape simulation. Landscape and Urban Planning, 54, 5-17.

Oppenheimer, P. E., 1986. Real Time Design and Animation of Fractal Plants and Trees. Computer Graphics, 20, 55-64.

Orland, B., 1994. Visualization techniques for incorporation in forest planning geographic information systmes. Landscape and Urban Planning, 30, 83-97.

Orland, B., 1997. SmartForest-II: Forest Visual Modeling for Forest Pest Management and Planning. USDA Forest Service, FPM-MAG, State and Private Forestry. 30 pp.

Orland, B., Radja, P., Su, W., 1994. SmartForest: an interactive forest data modeling and visualization tool. In: Greer, J.D.(Ed.), Remote Sensing and Ecosystem Management, Proceeding of the Fifth Forest Service Remote Sensing Applications Conference. America Society for Photogrammetry and Remote Sensing, Bethesda, MD, pp. 283-292.

Perrin, L., Beauvais, N., Puppo, M., 2001. Procedural landscape modeling with geographic information : the IMAGIS approach. Landscape and Urban Planning, 54, 33-47.

Prusinkiewicz, P., Lindenmayer, A., Hanan, J., 1988. Developmental Models of Herbaceous Plants for Computer Imagery Purposes. Computer Graphics, 22, 146-160.

Saito, K., Kumagai Y., Honjo, T., Ishida, Y., Lecoustre, R., De Reffye, P.，1993. Photo-realistic Forest Landscape Simulation - Application of GIS and plant modeling technique -. Nicograph Paper Contest, 9, 226-236. (In Japanese with an English summary)