Feature ArticlePrototype Imaging Devices For Jelly Zooplankton
By Simone Marini • Annalisa Griffa • Anne Molcard
Imaging systems play an important role in the investigation of the ocean interior. Cameras used in fixed-point, ocean observatories in the coastal and deep ocean or mounted on ROVs from ships have contributed greatly to our knowledge of the ocean ecosystem, both at the sea bottom and in the water column. Also, systems operated from ships, such as a video plankton recorder (VPR) and underwater vision profiler (UVP), have provided great insights on zooplankton distribution in the world ocean, complementing information from bioacoustics and nonimaging optics on biomass abundance.
Architecture of the image acquisition device (a), the two prototypes (b, c), and Prototype 1 during an experiment at the Aquarium of Genoa (d).
Most of the imaging systems in use today are cabled, therefore avoiding power and communication problems due to management and transfer of massive visual data. In the last two decades, there has been an increasing interest in unmanned autonomous platforms communicating in real time via satellite, for both fixed-point and moving instruments. Examples are vertical profilers, gliders and Lagrangian instruments like drifters and floats. The possibility of using imaging sensors mounted on such platforms is very appealing, but at the moment still challenging. Imaging systems have the same constraints as other physical and biogeochemical sensors presently under development, including size, consumption and low cost, plus the additional challenge due to the size of the data. Image data cannot be realistically transferred in real time via satellite unless they are first analyzed online from the sensor, recognizing the image content and encoding it in a suitable format for remote communication. Systems of this type are not available yet, to the authors' knowledge.
Here, we present a low-power, stand-alone (i.e., uncabled) imaging system, Guard1, consisting of two separate modules for acquisition and analysis, which can be used for applications such as regional experiments where instruments are released for a finite time and then recovered without need for real-time communication.
The image-acquisition device is relatively small and low-power and can be mounted on fixed-point or moving platforms, acquiring data from days to months. The analysis component, also tailored for low consumption, works offline once the data are recovered and uses pattern recognition techniques to identify relevant contents, encoding them in alphanumeric strings.
Further developments are underway to unify the two components as a single device. The present system can be seen as a first step toward a complete and integrated imaging sensor suitable for remote communication.
The specific application we are presently targeting is the detection of gelatinous zooplankton (jellies), whose distribution in the world's oceans is still unknown. Jellies play an important role in the trophic mesopelagic communities, with implications for the carbon cycle and for fisheries. However, they cannot be monitored using traditional net-based techniques because of their fragile structure. Imaging systems are particularly suited for monitoring jellies, and they can also be envisioned as a useful tool for early warning of jelly invasions, which can be very disruptive to human activities.
Even though the specific application of Guard1 is jelly-oriented, the methodology is general and can be used for other applications, such as fish or bottom-feature detection.
Image Acquisition Component
Two prototypes have been built and tested that both rely on commercial off-the-shelf cameras, but differ in terms of the type of acquisition devices.
Prototype 1 is based on a trail camera, modified for the project and endowed with a fish-eye lens. Prototype 2 has a more flexible and sophisticated architecture, consisting of a programmable controller based on the Arduino open-source electronic prototyping platform able to command a consumer-grade digital camera and a lighting system consisting of two high-performing LEDs. A sensor acquires the intensity of the natural light and delivers this information to the controller, which synchronizes the camera with the LEDs.
The programmable controller allows for the use of any programmable camera for image acquisition. Moreover, a CPU for pattern recognition can be installed and triggered by the controller to process the acquired images, while the communication of the relevant information can be custom managed.
Prototype 2 absorbs 6 milliwatts in standby and 25 milliwatts during shooting. It can operate for approximately eight months with a sampling rate of one picture every 10 minutes by using three battery packs consisting of three DD-size, 1.2-volt, 36 ampere-hour batteries (10-minute sampling is appropriate for Lagrangian platforms that move with the currents). Considering an acquisition rate of one picture every minute, the system can work for 33 days continuously using the same battery pack.
Both prototypes were tested in stand-alone mode by positioning them externally to floating or fixed-point platforms. The instruments were first checked in a controlled environment in May to July 2011, using jellyfish tanks provided by the Aquarium of Genoa (Genoa, Italy). Then several tests were performed in situ in the Ligurian Sea in spring and summer 2011, 2012 and 2013. The system was mounted on surface drifters in the Gulf of La Spezia and on drifters and Argo profilers in front of Cannes and Toulon, France, in collaboration with the JellyWatch monitoring French regional program in spring and summer 2012. To continue this article please click here.
Simone Marini has a master's degree in computer science and a doctoral degree in electronic and computer engineering. His current research activity deals with pattern analysis, recognition and classification of marine ecological data, with special interest in pattern recognition and classification of visual data acquired by AUVs.
Annalisa Griffa is a senior scientist at Consiglio Nazionale delle Ricerche, ISMAR-CNR in La Spezia, Italy, and adjunct professor at Rosenstiel School of Marine and Atmospheric Science at the University of Miami. She carries out research on transport processes in the ocean, Lagrangian techniques, data analysis and assimilation. She holds a Ph.D. in oceanography from Scripps Institution of Oceanography.
Anne Molcard is a professor at the University of Toulon, France, and is a member of the Mediterranean Institute of Oceanography. She is a physical oceanographer, and her research activity involves modeling and observations based on the physical oceanography theory for the development of Lagrangian methodologies, and exploiting the complementarity between Eulerian (coastal radar and traditional in-situ measurements) and Lagrangian data, observed and simulated.