Searching the World’s Herbaria: A System for Visual Identification of Plant Species
1
Searching the World’s Herbaria: A System for Visual Identification of Plant Species Peter N. Belhumeur 1 , Daozheng Chen 2 , Steven Feiner 1 , David W. Jacobs 2 , W. John Kress 3 , Haibin Ling 4 , Ida Lopez 3 , Ravi Ramamoorthi 1 , Sameer Sheorey 2 , Sean White 1 , and Ling Zhang 3 1 Columbia University, 2 University of Maryland, 3 National Museum of Natural History, Smithsonian Institution, 4 Temple University http://herbarium.cs.columbia.edu Overview Accelerate specimen identification by Type Specimen Digital Collection • 90,000 Type Specimens (1/4 of all plant species) • Digitized and available online and on DVD Funded in part by NSF Grant IIS-03- 25867 1. Photograph Leaf • Leaf photographed on plain background • GPS coordinates, date/time, etc. saved 2. Segmentation • E-M using saturation and value • Domain heuristics • Random sampling for real- time 3. Recognition • Inner Distance Shape Context • Fast Nearest Neighbor search • Histogram smoothing Collection of Isolated Leaves • Flora of Plummers Island, near Washington, DC 5,013 leaves of 157 species • Woody Plants of Baltimore-Washington, DC 7,481 leaves of 245 species • Trees of Central Park, NYC 4,320 leaves of 144 species • Photograph and digitize herbaria • Make data available in the field • Automate species search • Build and test prototype field guide Demonstrations • US Congressional Office Building • Smithsonian Institution Congressional Night • National Geographic Society’s BioBlitz • Microsoft Faculty Summit • Science Pavilion at NMNH (Smithsonian) • Field Studies: Plummers Island and Forest of VA 4. Visualization and Matching • Specimen and ranked results displayed • Zoomable interface for inspection • Geocoded, time-stamped results recorded Recognition Results Future Goals • Woody plants of United States • Mobile phone system • Tropical flora
description
Searching the World’s Herbaria: A System for Visual Identification of Plant Species. Peter N. Belhumeur 1 , Daozheng Chen 2 , Steven Feiner 1 , David W. Jacobs 2 , W. John Kress 3 , Haibin Ling 4 , Ida Lopez 3 , Ravi Ramamoorthi 1 , Sameer Sheorey 2 , Sean White 1 , and Ling Zhang 3 - PowerPoint PPT Presentation
Transcript of Searching the World’s Herbaria: A System for Visual Identification of Plant Species
Searching the World’s Herbaria: A System for Visual Identification of Plant SpeciesPeter N. Belhumeur1, Daozheng Chen2, Steven Feiner1, David W. Jacobs2, W. John Kress3, Haibin Ling4, Ida Lopez3, Ravi Ramamoorthi1, Sameer Sheorey2, Sean White1, and Ling Zhang3
1Columbia University,2 University of Maryland,3 National Museum of Natural History, Smithsonian Institution, 4Temple Universityhttp://herbarium.cs.columbia.edu
OverviewAccelerate specimen identification by
Type Specimen Digital Collection• 90,000 Type Specimens (1/4 of all plant species)
• Digitized and available online and on DVD
Funded in part by NSF Grant IIS-03-25867
1. Photograph Leaf
• Leaf photographed on plain background
• GPS coordinates, date/time, etc. saved
2. Segmentation
• E-M using saturation and value
• Domain heuristics
• Random sampling for real-time
3. Recognition
• Inner Distance Shape Context
• Fast Nearest Neighbor search
• Histogram smoothing
Collection of Isolated Leaves• Flora of Plummers Island, near Washington, DC
5,013 leaves of 157 species
• Woody Plants of Baltimore-Washington, DC
7,481 leaves of 245 species
• Trees of Central Park, NYC
4,320 leaves of 144 species
• Photograph and digitize herbaria
• Make data available in the field
• Automate species search
• Build and test prototype field guide
Demonstrations• US Congressional Office Building
• Smithsonian Institution Congressional Night
• National Geographic Society’s BioBlitz
• Microsoft Faculty Summit
• Science Pavilion at NMNH (Smithsonian)
• Field Studies: Plummers Island and Forest of VA
4. Visualization and Matching
• Specimen and ranked results displayed
• Zoomable interface for inspection
• Geocoded, time-stamped results recorded
Recognition Results
Future Goals• Woody plants of United States
• Mobile phone system
• Tropical flora
