Sun, May 10, 2009 page12

• Program can help naturalists turn over a new leaf

  Researchers have developed a field guide device now in prototype for iPhones. Users can snap a photo of a leaf from a tree and the device will scan its library to identify the species

  • By Anne Eisenberg / NY TIMES NEWS SERVICE
  • • 
    • 
    • 

  The traditional way to identify an unfamiliar tree is to pull out a field guide and search its pages for a matching description. One day people may pull out a smartphone instead, photograph a leaf from the mystery tree and then have the phone search for matching images in a database.

  A team of researchers financed by the National Science Foundation has created just such a device — a hand-held electronic field guide that identifies tree species based on the shape of their leaves, said Peter Belhumeur, a professor of computer science at Columbia and a member of the team.

  The field guide, now in prototype for iPhones and other portable devices, has been tested at three sites in the northeastern US, including Plummers Island in Maryland and Central Park in New York, said John Kress, a research botanist and curator at the Smithsonian Institution in Washington, who is also on the research team. The computer program compares the leaf snapshot to a library of leaf images.

  

  A researcher uses a newly developed program that can identify leaves by searching a field guide on a PC or a phone. Researchers say it could be useful not only to hikers but also to scientists compiling data.

  PHOTO: NY TIMES NEWS SERVICE

  “We believe there is enough information in a single leaf to identify a species,” he said. “Our brains can’t remember all of these characteristics, but the computer can.”

  The tree guide, and other electronic guides to nature being developed, may be used one day not only by backyard botanists, hikers and children on field trips, but also by scientists and volunteers to compile data for environmental inventories, or as part of species discovery.

  Computer applications that automatically identify, for example, a particular bird call or tree species or even an individual dolphin may prove popular with the public, said Bryan Heidorn, a program manager at the National Science Foundation who tried a prototype of the tree program at the Smithsonian. (Heidorn is not connected with the financing of the project.)

  The computer tree guide is good at narrowing down and finding the right species near the top of the list of possibilities, he said.

  “Instead of flipping through a field guide with 1,000 images, you are given five or 10 choices,” he said.

  The right choice may be second instead of first sometimes, “but that doesn’t really matter,” he said. “You can always use the English language — a description of the bark, for instance — to do the final identification.”

  The tree guide will be specific to trees and shrubs of the northeastern US, Kress said.

  “To apply this system to a forest in Costa Rica, you would have a different library of images,” he said, because there is a different set of species.

  Sean White, who created the tree-identifying prototypes for mobile devices as part of his graduate work in computer science at Columbia, hopes to develop mobile applications for the Google, Nokia and Microsoft platforms as well.

  The public cannot yet download the software for automatic tree ID, Belhumeur said, although that is planned for the future. The program will probably show up first in educational kiosks at, for example, the Smithsonian, where people will be able to bring in a leaf to have it identified ­automatically, Kress said.

  The computer matching is done using images of recently acquired leaves, Belhumeur said, typically 20 leaves from each species, but sometimes more, that are in the image library. For Plummers Island, the team collected 5,013 leaves representing 157 species; for the trees of Central Park, it collected 4,320 leaves representing 144 species, said White of Columbia. The matching algorithms were developed by David Jacobs, an associate professor of computer science at the University of Maryland, College Park.

  Among other automatic identification programs for flora and fauna is one developed by a team of researchers at Eckerd College in St Petersburg, Florida, which includes Kelly Debure, an associate professor of computer science. The program identifies individual dolphins based on photographs of their dorsal fins, which have characteristic nicks and notches.

  At one time, the matching was done by hand, “but the task becomes daunting when the catalog grows enormous,” Debure said.

  The group primarily studies dolphins in Tampa Bay. The Florida Aquarium there may use the program at a kiosk, she said, so that people can identify individual dolphins in photographs, perhaps leading to a greater concern with water quality and boating regulations affecting dolphins. The software is free to download at code.google.com/p/darwin-ec and is being used by groups around the world, Debure said.

  Matthew Brown, supervisor of the soil, water and ecology lab at the Central Park Conservancy, thinks the programs can be useful.

  “If people are walking through the park and they come upon a tree that’s not in their field guide they can snap a photo, send it in, and get the name back and find out more information,” he said.

  But he believes in traditional education, too.

  “People don’t have to take botany for four years, but we shouldn’t lose sight of the value of learning information. If a computer can figure it all out, we can get lazy,” he said.
• Most Popular

  • 1

    Bologna advance to face AC Milan in Coppa final

  • 2

    Taiwan ranked as the cheapest travel destination

  • 3

    Wolves hold off Lakers 116-104 to take series lead

  • 4

    Taiwan in Time: Taiwan’s first high school for Taiwanese students

  • 5

    Taiwan would benefit from integrated theater: expert