UP team develops coral monitoring system

A team of researchers from the National Institute of Physics of the University of the Philippines is featured in the January 2006 issue of Laser Focus World, an international magazine for the photonics and optoelectronics industry, for customizing an image processing system to monitor corals.

The magazine article is based on the work of Ma. Sheila Angeli Marcos, Maricor Soriano, and Caesar Saloma published at Optics Express 13 (22) (Oct. 31, 2005), an international peer-reviewed journal on optics.

Titled "Classification of Coral Reef Images from Underwater Video Using Neural Networks," the work details a system through which corals can be classified automatically as living, dead, or just rubble.

According to the magazine article, this is the first time a "close-up automatic image recognition of coral reefs" has been done.

This is due to the difficulty of classifying natural objects that come in a variety of colors and textures and change appearances from different perspectives.

The NIP team used a "neural network-based vision system" that concentrated on the color and texture of the corals – parameters used by marine scientists themselves to classify corals.

It achieved as high as 79.7 percent recognition rate during tests using pre-classified images taken from underwater video of a small portion of Australia’s Great Barrier Reef. The images consisted of 98 of live coral, 43 of dead, and 44 of sand.

According to the magazine article, humans can classify corals by species using video images and determine if they are living or dead with higher accuracy, but "the process is slow."

"At the moment, we can classify living and dead coral and sand," Dr. Caesar Saloma, one of the researchers and NIP director, was quoted as saying.

"However, the applicability of the current performance of our system would depend on the problem of the marine scientist. If a need for rapid assessment of living and non-living areas is desired, then our system would suffice," Saloma added.

Their work involved designing the network architecture, training rates, and other parameters for the vision system.

"Increasing the number of categories analyzed by the neural net should boost the classification accuracy," Saloma said.

The NIP is a leading research center for physics in Asia. Its prolific research output regularly comes out in international peer-reviewed publications. — Jo Florendo Lontoc