Invited Speakers

Underwater Optical Imaging: Past, Present, and Prospects
Jules Jaffe
Scripps Oceanography, U. C. San Diego

Abstract: This lecture will discuss the current state of underwater optical imaging in the context of the physics, technology, biology, and history. This encompasses not only the history of human’s ability to see underwater, but also the adaptations that various organisms living in oceans or lakes have developed. The continued development of underwater imaging systems at military, commercial, and consumer levels portends well for both increased visibility and accessibility by these various segments. However, the fundamental limits imposed by the environment, as currently understood, set the ultimate constraints. The physics, biology, computer modeling, processing, and the development of technology that ranges from simple cameras and lights to more advanced gated and modulated illumination will be described. Finally, the future prospects for continuing advancements, as envisioned, will be discussed. Challenges to the Computer Vision community will be presented, as a natural aspect of future work.

Bio: Jules Jaffe is a research oceanographer with the Marine Physical Laboratory at Scripps Oceanography, U. C. San Diego. His research interests are broadly concerned with the invention of new technology for observing oceanic phenomena and the development of inverse techniques for their interpretation. Although his research has focused primarily on ocean ecology he has also worked in biomedical applications. In the realm of ocean ecology, several acoustic systems invented by Jaffe resulted in the first-ever behavioral observations of zooplankton in situ. In addition, the small-scale distribution of phytoplankton, as mapped via an autonomously deployed imaging fluorometer has provided unprecedented views of oceanic biota. Current projects under way in his lab concern the development of the, first ever, underwater, diver held microscopes that are being used to image many of the fundamental processes that occur at these small size scales. A new generation of miniature, sensor-equipped drifters have also been developed to gain insights into coastal circulation and larval transport. Additional work to characterize the camouflage displays of cuttle fish that were elicited with a cephalopod virtual reality fish tank have resulted in new insights into the behavior of these fascinating animals. An underwater plankton microscope has resulted in 10^7 images and efforts to classify them are currently underway. The National Science Foundation, the Office of Naval Research, California Sea Grant, The ARMY foundation for breast cancer research, The Seaver Institute of Los Angeles and the Keck Foundation have supported his research. Dr. Jaffe is a fellow the Acoustic Society of America and has been a visiting Miller Professor at UC Berkeley as well as a H. Burr Steinbuck Visiting Scholar at the Woods Hole Oceanographic Institution. He recently was awarded a "best paper" for his presentation at the International Ocean Optics meeting that took place in 2012. More information can be found via his web site: jaffeweb.ucsd.edu.

NOAA Fisheries Strategic Initiative on Automated Image Analysis
Benjamin L. Richards
NOAA Pacific Islands Fisheries Science Center

Abstract: The National Marine Fisheries Service (NMFS) of the United States National Oceanic and Atmospheric Administration (NOAA) is increasingly using optical data streams to augment traditional surveys of targeted marine resources. A report by the National Task Force for Improving Stock Assessment found the greatest impediment to producing accurate, precise, and credible stock assessments to be a lack of adequate input data. If properly employed, optical survey platforms can reduce sampling error, can increase sampling intensity and can increase the spatio-temporal area or number of species surveyed. However, as with other optical surveillance systems, the volume of data produced quickly exceeds the processing capability of human analysts. To date, this problem has been overcome by subsampling the available data or by using data streams and analysis techniques that many not yield the most accurate and precise representation of the population in question. To make the most effective and efficient use of optical data streams, automated classification algorithms—similar to those that have been developed for the human surveillance, biomedical and terrestrial wildlife communities—must be developed for the marine realm. While many of the baseline techniques will be similar, operating underwater produces some unique challenges. Variable light fields that change with respect to depth and moving water bodies, particulate matter, and the camera air-water interface are unique to the marine world. The mission of the NMFS Fisheries Strategic Initiative on Automated Image Analysis is to develop guidelines, set priorities, and fund projects to develop broad-scale, standardized, and efficient automated analysis of still and video imagery for use in marine resource assessment. The goal of this project is to develop a robust, open-source software toolkit allowing marine resource and stock assessment scientists to easily convert optical data streams into species-specific, size-structured abundance estimates.

Bio: Ben completed his PhD in Zoology at the University of Hawaii in 2011, where he studied the ecology and patterns in habitat preference of large-bodied reef fish. In 1998 Ben joined the NOAA Florida Keys National Marine Sanctuary (FKNMS) to assist with the development of the FKNMS Research and Monitoring Plan and Final Environmental Impact Statement. In 2004 Ben moved to the Coral Reef Ecosystem Division of the NOAA Pacific Islands Fisheries Science Center (PIFSC) in Honolulu, modeling the distribution of Pacific reef fish abundance and biomass associated with various natural and anthropogenic environmental factors. Input data for these studies predominantly resulted from diver-based visual surveys. Inter-observer variation and the paucity of data from below diver depths pushed his research toward the use of camera-based technologies for optical sampling. Since moving to the PIFSC Stock Assessment Division in 2010, his research has focused on the use of advanced fishery-independent sampling technologies, including optical camera systems, to assess species-specific, size-structured abundance for Hawaii bottomfish assemblages. The volume of data produced by these camera systems has quickly overwhelmed the capabilities of human analysts. His research focus has thus expanded to include the development of automated classification algorithms to produce species-specific, size-structured abundance estimates from underwater optical surveillance video of reef and bottomfish.

Ben currently serves as a member of the National Marine Fisheries Service Advanced Sampling Technology Working Group and is chair of the NMFS Strategic Initiative on Automated Image Analysis. His research currently focuses on the distribution of marine resources along gradients of natural and anthropogenic factors and how advanced sampling technologies and automated analysis tools can provide enhanced data for stock assessment and ecosystem-based management.

Data Acquisition and Analysis in the Fish4Knowledge Project
Robert B. Fisher
School of Informatics, Edinburgh University

Abstract: The talk will present an overview of the data acquisition and analysis from the Fish4Knowledge EU funded research project, which applied computer vision methods to complex underesea image data. Fish4Knowledge collected and analysed subsea video from up to 10 cameras off the coast of Taiwan, which were observing coral reef fish. Over 1 billion detections of more than 100 million fish were analysed from about 100K hours of video, with 23 species recognised to an accuracy of 80+%. Developing mechanisms to acquire, analyse and present results over such a large dataset was the major challenge of the project. Having such a large dataset allowed analysis of the abundance of particular species (here D. reticulatus) over time. It also led to interesting individual re-identification problems, for species that were 'resident' and might be re-observed many times, such as the clownfish (A. clarkii).

Bio: Prof. Robert B. Fisher FIAPR,FBMVA received a BS (Mathematics, California Institute of Technology, 1974), MS (Computer Science, Stanford, 1978) and a PhD (Edinburgh, 1987). Since then, Bob has been an academic at Edinburgh University. His research covers topics in high level and 3D computer vision, focussing on reconstructing geometric models from existing
examples, which contributed to a spin-off company, Dimensional Imaging. More recently, he has also been researching video sequence understanding, in particular attempting to understand observed animal behaviour. The research has led to 13 authored or edited books and about 250 peer-reviewed scientific articles. He has developed several on-line computer vision resources, with over 1 million hits. Most recently, he has been the coordinator of an EC STREP project acquiring and analysing video data of 1.4 billion fish from over about 20 camera-years of undersea video of tropical coral reefs. He is a Fellow of the Int. Association for Pattern Recognition (2008) and the British Machine Vision Association (2010).