UMass Amherst Researchers Collaborate on High-Tech Conservation Project
AMHERST, Mass. - Conservationists may someday be able to monitor environmentally sensitive regions with a few clicks of a computer mouse, tracking land use and wildlife populations, and making decisions based on the very latest information, thanks to a University of Massachusetts project funded with a $1.8 million grant from the National Science Foundation. The project combines the expertise of Dana Slaymaker, of the forestry and wildlife management department, and Edward Riseman, leading a large team from the computer science department. The team has done extensive work in re-creating 3-D environments from aerial images. Other team members include Howard Schultz, Allen Hanson, Paul Utgoff, Eliot Moss, and Kathryn McKinley, all of the computer science department, and Jack Finn, of forestry and wildlife management.
The project is aimed at helping conservationists "who truly understand vegetation and environmental problems of their region, but who may not have extensive computer experience and accurate environmental data," says Slaymaker, who specializes in updating conservation maps through the use of aerial photography and videography. Slaymaker has flown over environmentally threatened areas in countries such as Peru, Botswana, Madagascar, and Rwanda, and created detailed computer maps based on the images captured during his flights. But computerizing the images is extremely labor-intensive. That’s where Riseman of the Computer Vision Lab and the computer scientists come in.
Usually, conservationists must make do with outdated topographic maps at very coarse elevation levels, in which critical detail of the three-dimensional vegetation is missing, Riseman says. In order to provide both fine resolution and a visually accurate look, the UMass researchers turn to "stereoscopic reconstruction" ? creating a digital elevation map by combining two or more images of the same site, taken from different vantage points, and using those combined images to create a 3-D surface, Riseman says. This results in depth information attained in a manner "similar to the way human beings experience depth perception, by focusing our two eyes, simultaneously, on an object." In this instance, the two "eyes," or cameras, can be hundreds or even thousands of feet apart. The plan is to combine close-up images, taken from low-flying aircraft, with larger-scale images, taken either by satellites or high-flying aircraft. Computer scientists will then piece together the images and create a 3-D terrain with a remarkably fine resolution of one meter or less, a major improvement over the standard 30 or 90 meter maps.
The combination of images offers the best of both worlds, according to Riseman: the enormous amounts of information offered in high-altitude pictures, and the fine detail available from close-up views. All of these images will correspond with geographic points as mapped by the Global Positioning System, the technology associated with the satellite system that enables people with receivers to locate their position on the globe within approximately 100 feet. Mostly used by sailors in the past, this technology is now available as an option in some new cars. Combining the various factors was an idea Slaymaker came up with several years ago; Riseman aims to automate the system, significantly improve it to use 3D information, and make it widely available. The fine resolution will help in identifying year-to-year changes, such as encroaching development, or the growth or dwindling of particular species of plants or animals, Riseman said.
It is this recreated landscape that conservationists, sitting at computers, can virtually "fly" over. While the fly-through feels like a game, it has a very serious purpose, Riseman notes. Relying on his or her knowledge of the area, a user can update the map. Riseman offers this example: a conservationist might want to examine an area’s vegetation, including pine, birch, and grass, or a combination of the species. The person can click on an area that he or she knows is red pine, creating a computer label which will cause all the other areas of red pine to show up. The maps can be updated by selecting areas amounting to much less than one percent of a visible area, Riseman said. "You end up with computer labels that show what the environment is at a specific moment in time." In addition, the program may enable conservationists to locate and classify actual animals, tallying them for wildlife surveys.
Groups that have expressed interest in high-resolution terrain reconstruction and classification include Conservation International, the World Wildlife Foundation, and the Nature Conservancy. The aerial terrain mapping project could enable very close monitoring of forests that are under pressure from development. "Rather than relying on estimates, conservationists will have facts, and will know where to concentrate their efforts," Slaymaker said. Corporations are interested in the idea as well, to prove that they haven’t harmed the landscape while extracting oil or harvesting other natural resources.