Plankton Image Analysis

Date/Time:

15 Mar 2012
10:25-10:50

Location:

South Galleries

Seminar Details
Current technologies available for the study of marine organisms remain limited in comparison to the technologies available to physical oceanographers (high resolution, high-speed data acquisition rate and rapid analysis). Even though nets technologies have become quite sophisticated and the use of Multiple Opening Closing Net systems enable discrete vertical sampling coupled with detailed environmental data, such samples still require the task of being processed manually â?? a time-consuming and costly effort. To address these issues, Bellamare, LLC has designed a towed, high-resolution digital imaging system, capable of sampling large volume of water at once, sufficient for accurate quantification of meso-zooplankton in their environment. The instrument captures images of organisms in situ and informs about the precise position and time of each organism as well as their spatial and vertical distribution. It also simultaneously captures environmental data of the organisms surroundings. Images acquired are high resolution, enabling clear identification of organisms (e.g. larvaceans, gelatinous zooplankters, chaetognaths, larval fish), often to family or generic level. Moreover when towing the instruments at 5 knots, 162 liters of water are imaged every seconds, which is greater than an order of magnitude improvement over other imaging systems. Bellamare, LLC has partnered with Traklogik Corporation to automate plankton recognition, enabling fast data analysis of collected images. Such a tool allows for increase sampling frequencies, leading to better taxa monitoring and eventually leading to a greater capacity for improved scientific inquiries. For classification, the newly developed software utilizes a unique combination of semi-analytical methods for generalization of image patterns, with the diversity-oriented recognition technology based on Radial-Based Neural Network, with a dedicated super-compact and energy efficient processor allowing for real-time applications. In this presentation, we provide examples of images captured, show the first plankton-recognition results and discuss methods of mathematical and computational approach. Charles Cousin, President, Bellamare; Marina Murzina, , President, Traklogik.