OPTIMIZED SYSTEMS

Facial-Recognition Software Identifies Apes in the Wild

Rebecca Kutzer-Rice | Date: 09-07-11 | Comments

New facial-recognition software is aiding research on endangered species by providing more accurate and detailed information about individual animals.

While facial-recognition technology is helping catch crooks and prevent terrorism attacks, scientists are now using it for a wilder purpose. A new software system uses facial recognition to identify individual animals within an endangered population. The tool could make protecting endangered species easier and more successful for scientists around the world.

Current methods of observing animals in the wild are largely a matter of guesswork. While scientists can record various animal behaviors, they are often unsure if they are observing one or several different animals. Analyzing hours of video footage also requires significant time and effort from scientists.

The software uses features like eyes and mouth to identify individuals within a wild population. (Source: Fraunhofer-Gesellschaft)

The new animal-recognition software, currently viable only for primates, is a part of a joint German project between the Fraunhofer Institutes for Integrated Circuits (IIS) and Digital Media Technology (IDMT) and the Max Planck Institute for Evolutionary Anthropology.

The software does the most tedious work for the scientists by searching through videos and photographs and assigning each image to a specific animal.

"The pictures are first filtered to find the ones on which the primates' faces can be seen," Alexander Loos from the IDMT in Ilmenau explained in a statement. Scientists at Fraunhofer�s IIS created a detection program to execute this task. The program is able to detect faces in both videos and photographs. A second software module then identifies visible faces with specific individuals. "Our software analyzes the primates' faces using special algorithms," said Loos.

In a trial run at the Leipzig Zoo in Germany, the facial-recognition software had an 83 percent success rate at identifying chimpanzees.

Although successful, scientists will need to further improve the software for use in the field, where photographs are often of low quality. "The algorithms are strongly affected by external influences," said Loos. "In poor light, or if the faces are partially occluded, the recognition rates quickly drop to below 60 percent."

In addition to facial recognition, the new software analyzes audio to assign certain sounds to specific individuals. One ape, for example, might drum its chest, while another might make grunting noises. These types of data collection are making further behavioral analysis easier and more accurate for the scientists.

This facial-recognition project is another example of how technological innovations can help protect endangered species and lead to a smarter planet.