Recent studies have shown that a third of the world’s coral reef-building species are facing extinction. According to researchers, climate change, coastal development, overfishing, and pollution are the major threats. It is strongly believed that threats to this particular ecosystem could also endanger the various species of fish which live their entire lives on reefs or those who use them as nurseries.
In August 2007, principal investigator (PI) Gareth J. Russell (Department of Biological Sciences, New Jersey Institute of Technology) and co-PIs Joseph Wilder (Center for Advanced Information Processing, Rutgers University) and Paul Boyle (New York Aquarium, Wildlife Conservation Society) joined forces to develop an underwater multi-camera system to observe, identify, count and, over time, track fish species found near coral reefs to understand the effects of environmental changes on those populations.
System set-up and configuration
The system consists of two high sensitivity 1.4 Megapixel Prosilica GC1380C color cameras fitted inside water-tight tubes equipped with housings for LED lighting. The cameras are located 3 feet away from each other and at a 90 degree angle to overlook an area of one cubic yard of water. Two side panels opposite each camera provide a uniform background for clearer image data. The cameras are connected to a PC via Cat5e cabling. Both GC1380C cameras are fitted with 8.5mm wide-angle lens and operate at full resolution (1360 x 1024) over the entire field. The cameras are triggered to capture images in pairs.
The Prosilica GC1380C is a very sensitive, high-resolution CCD camera with Gigabit Ethernet interface running at 20 fps at full resolution. It incorporates the incomparable Sony ICX285 CCD sensor with ExView technology providing high-sensitivity, low noise, excellent anti-blooming, and superb image quality, attributes necessary for underwater and low-light applications.
Implementation and tests
Currently nearing the end of year one, the system will be undergoing a series of tests over a three-year period to assess its performance and reliability.
The system is currently set-up in a small aquarium at the Rutgers Institute for Marine and Coastal Sciences (IMCS) in New Jersey. Built by the team at the IMCS, the 1,000 liter (265 gallon) aquarium is filled with sea water and simulated coral reef to recreate a nature-like habitat. During the course of year one, project members have been working on building the system hardware with optics, illumination and image sensing appropriate to the underwater reef environment, as well as developing the necessary software to operate both Prosilica cameras and capture and analyze images.
In order to obtain clear images for the study of the various species of coral reef fish, the team is working on several complex algorithms which will enable reliable segmentation of fish from their background (while distinguishing true fish images from floating debris), features extraction from pose-corrected images of fish (using size, shape, color and texture), and species classification.
In late 2008, the system will be moved to the Glover’s Reef exhibit at the New York aquarium. Opened in June 2005, the 623,000 liter (165,000 gallon) tank display recreates the ecosystem of a Caribbean coral reef and features over 25 different species of fish which can be found off the coast of Belize where the system will be moved to in year three. The team will focus on testing the system in this larger and more realistic environment with a wider array of species. The set-up will also be used to develop an educational image display for the aquarium.
In year three the system will be moved to the Glover’s Reef Marine Research Station in Belize for field trials in a natural environment where it will be tested for stability, robustness, power supply and data transmission. Extensive image data of free-swimming fish will be collected and analyzed to evaluate and adapt the algorithms if necessary, and assess the overall viability of the system.
In the long-term, and if successful, the team will be looking to duplicate the system to extend the research to other areas with similar ecosystems.