The NYT ran a quick “novelties” story on smartphone applications for identifying species of trees by matching cameraphone images of leaves to a database of leaf shapes — Foliage Field Guides for Cellphones. I hope the EOL project is taking notes. Lightweight iPhone-type apps may be the quickest development path towards some exciting new citizen science initiatives.
The promise of automated species identification has been a taxonomic dream for some time. Species identification mobile applications could be a platform for more widespread and persistent biodiversity surveys, e.g., bringing more users to collaborative birding surveys now coordinated online. Given enough uptake and use of such applications, one could imagine an ambient level of biodiversity awareness emerging from this data, a Wikipedia-like record of species sightings. Even a simple map showing the geographic range of sightings of a particular species over time would be hugely interesting.
Science and science fiction has been awaiting the device that would enable this type of information since before Captain Kirk first held a tricorder. Ecologist Dan Janzen articulated the vision of a handheld DNA analyzer just a few years ago:
Imagine a world where every child’s backpack, every farmer’s pocket, every doctor’s office and every biologist’s belt has a gadget the size of a cell-phone. A free gadget. Pop off a leg, pluck a tuft of hair, pinch a piece of leaf, swat a mosquito, and stick it in on a tuft of toilet tissue. One minute later the screen says Periplaneta americana, Canis familiaris, Quercus virginiana, or West Nile virus in Culex pipiens.
However, we may not have to wait for science to catch up with fiction. I would wager that image-based species identification — next-generation versions of the iPhone app covered by the New York Times — will be enough to correctly identify 90% of the species images that are thrown at them. Most of us would use these in our back yards and beyond, not in the Amazonian rainforest. Facial-recognition algorithms are getting increasingly sophisticated, and there will be no shortage of amateur taxonomists to help systems learn when they mis-identify recognizable species. Given a march towards better cameras and faster mobile data networks, it’s not unreasonable to assume that one could do real-time matching on an uploaded image to narrow the range down to a given species by querying a number of EOL-type taxonomic databases. It’s ultimately a tractable software problem, not an incredibly difficult hardware problem.
I’ve been waiting for my jetpack since we entered the 21st century, and will probably have to wait a lot longer. But a handheld species identification device — sounds like I could be hiking with a beta version on my iPhone pretty soon!