Smartphones save baby turtles

Field research in biology is notoriously difficult. In a famous 1982 study, Smithsonian entomologist Terry Irwin hand-counted the number of insects he encountered in a hectare of forest canopy in Panama. By doing the necessary calculations, he was then able to estimate the total insect biodiversity on Earth. Wouldn’t it be great if a cell phone could relieve him of some of this tedious work?

Harvard University biologist and computer scientist Walter Shearer has invented an optical-mechanical system that automatically identifies and counts specific animals while running on a Motorola Droid X2 mobile phone. This will help biologists make faster and more accurate assessments of the health of sensitive ecosystems.

Two years ago, Edwards Air Force Base in California’s Mojave Desert put out a call for a low-cost way to monitor the animals that live there. The area is one of the last refuges for the endangered desert tortoise and the endangered Mojave ground squirrel. Monitoring the health of the animal population in such a remote location is time-consuming and expensive. Mr. Shearer developed detection and classification algorithms that can tell the tortoises and squirrels apart using just a smartphone.

Automated camera traps have already been developed for this purpose, but they are not selective enough in what they photograph. “We have to manually review the photos one by one to identify the species and separate the interesting photos from the others. It’s an incredibly laborious job,” says Princeton biologist Siva Sundaresan, who works with Grevy’s zebras in Kenya. He says Mr. Shearer’s method could be particularly useful for biologists.

But how can a phone tell the difference between a squirrel and a rock or a bush? Mr. Shearer’s system starts by scanning the environment for objects that could be the animals it wants. It looks for sets of pixels that are new to the scene and then examines them to decide whether they represent one of the animals it has been trained to recognize. Instead of examining each individual pixel, Mr. Shearer’s algorithms analyze the content of a “frame” of video and look for patterns of pixels that identify the animal. The algorithms are not computationally intensive, so they can run just fine on a smartphone.

A paper presented last week at the Computer Vision Applications Laboratory in Clearwater, Florida, shows how well the algorithms work, with the system able to distinguish between three different species of ground squirrels 78 percent of the time, despite the animals being nearly identical. Mr. Shearer says the algorithms have already been improved, and the recognition rate is now about 85 percent.

The expert emphasizes that his goal is to develop a cheap, easy-to-use system that can automatically detect animals in any environment. More field tests are planned for next month and the team aspires to deliver a perfected system to the US Air Force by 2014.

Princeton population biologist Dan Rubinstein believes that machine vision systems will also help us understand fragile ecosystems in greater detail. “We won’t be generalizing from such a small scale to a massive scale,” he says. “We’ll be able to save ecosystems . ”

Another system being demonstrated, Hotspotter, identifies individual animals like zebras and giraffes by their stripes and spots. But it still requires some human guidance — something the Mojave Desert system doesn’t need. Mr. Rubinstein, who is working on developing Hotspotter, says such systems will allow biologists to examine animals and their actions on an individual basis. “We could start to create huge databases like who is who, tracking who each one is and how they move over time. We can use social networks to see if they are related to each other.