Field research in biology is renowned for its difficulty. In a famous study of 1982 the entomologist of the Smithsonian Foundation Terry Erwin measured by hand the number of insects encountered in one hectare of forest cover in Panama. By making the necessary calculations he then estimated the approximate total biodiversity of Earth's insects. Would not it be nice if a cell phone could relieve him of part of this tedious job?
Biologist and computer scientist at Harvard University Walter Siere devised an optical mechanical system that automatically recognizes 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.
Squirrels and turtles stand out
Two years ago, the Edwards Air Base in the Mojave Desert, California, issued a call for a cheap way to track the animals that live there. The area is one of the last shelters of the desert turtle, as well as the endangered Squirrel Territory of Moheb. Monitoring the health of the animal population in such a remote location is time consuming and costly. Mr. Syrer developed detection and classification algorithms that can distinguish turtles and squirrels with the help of only one smart mobile.
For this purpose, automated trap cameras have already been developed, but these are not quite selective as to what they are photographing. "We have to manually review the photos one by one to locate the species and separate the interesting photos from the others. It's an extremely tedious job " says Princeton biologist Siva Sudanthesan, who is engaged in Grevery's zebras in Kenya. He says Mr. Sirer's method might be very useful for biologists.
But how can a phone capture the difference between a squirrel and a stone or a bush? Mr. Syrer's system starts by scrutinizing the environment to search for objects, which could be the animals he wants. It searches for sets of pixels that are new to the scene and then examines them to decide if they represent any of the animals trained to recognize. Instead of examining each separate pixel, Mr. Siere's algorithms analyze the content of a video frame and search for pixel motifs identified with the animal. Algorithms do not require intensive processing, so they can "run" fine on a smartphone.
An article presented last week at the Visual Applications Laboratory in Florida, Florida, shows how well the algorithms work, as the system can distinguish between three different species of squirrels in the 78% of cases, despite the fact that the animals are almost identical . Mr. Sirer says the algorithms have already been improved and the recognition rate is now about 85%.
How to save animals
The expert points out that his goal is to develop a cheap, handy system that can automatically detect animals in any environment. More on-site testing is scheduled for next month and the team aspires to deliver a sophisticated system to US Air Force as 2014.
The Princeton biologist Dan Rubinstein believes that mechanical optical systems will also help us to understand sensitive ecosystems in greater detail. "We will not generalize from such a small scale on a mass scale" he stresses. "We can save ecosystems".
Another system introduced, the Hotspotter, recognizes individual animals such as zebras and giraffes by their stripes and spots. But it still requires some human guidance - something the Mojave Desert system does not need. Mr. Rubinstein, who is working on the development of the Hotspotter, argues that the species' systems will allow biologists to examine animals and their actions on an individual basis. "We could start creating huge databases like who is who, watching what everyone is and how it moves over time. We can use social networks to see if they are related ".