Estimating the population of species within an ecosystem is a task of critical importance. Be it for the conservation of endangered species or effective farming of a particular species, having an accurate account of their conditions and numbers is vital. The National Oceanic and Atmospheric Association (NOAA) recognizes this, and is pioneering a novel approach to tracking fish populations, leveraging Artificial Intelligence (AI) and Computer Vision technologies.
The Traditional Approach and Its Limitations
Traditionally, counting fish involves dragging a net along a seafloor, capturing a certain quantity of fish, and extrapolating these measurements across a larger area. This method, while effective in certain regions, has significant limitations. In particular, it fails in regions where the ocean floor is unreachable or where bottom trawling is unfeasible.
Automated Image Analysis: A Strategic Initiative by NOAA
NOAA is tackling this issue head-on with its Automated Image Analysis Strategic Initiative. The aim is to develop software that can automatically analyze video footage to count and size the fish present in the imagery. This technology, known as CamTrawl, is a giant leap forward in marine biology data collection and analysis. By accelerating the data analysis process, researchers can process a summer’s worth of work in just a day, aiming to eventually reach an accuracy rate of 80%.
Computer Vision: The Future of Marine Ecosystem Analysis
The integration of computer vision into NOAA’s analysis procedures represents a shift in how marine ecosystem data is collected and analyzed. By automating the counting and sizing process, researchers can redirect their efforts to solve other problems within the marine ecosystem.
The Power of AI in Marine Biology
Artificial Intelligence has revolutionized countless fields, and marine biology is no exception. NOAA’s exploration into AI usage for fish counting has the potential to drastically improve the efficiency, accuracy, and scale of marine ecosystem monitoring.
Update
In their ongoing commitment to developing advanced methods of fish counting, NOAA has been testing new technologies like autonomous submarines and drones to generate more accurate fish population counts and richer marine ecosystem understanding[1]. Additionally, the NOAA Pacific Islands Fisheries Science Center launched the OceanEYEs Citizen Science project in September 2020. Volunteers are asked to identify fish species in underwater images from the survey. The work done by OceanEYEs volunteers is being used to develop artificial intelligence solutions for counting fish[2].
No longer just an exploratory concept, NOAA has started utilizing machine learning and artificial intelligence to count fish with increased accuracy and efficiency. The technology is capable of detecting when fishing gear is used or a catch is hauled in, and storing these catch events on a hard drive or wirelessly transmitting the data[3]. An integral part of this project is the citizen science project OceanEYEs, where volunteers help review images from the annual bottomfish survey. Their input is used to train advanced AI tools for fish counting and stock assessments[4].
NOAA has also begun testing the use of an underwater camera system called the MOUSS for its annual bottomfish survey. A new software package named VIAME, which uses AI to detect, identify, and count bottomfish, is being tested to streamline the survey process. According to predictions, VIAME will be fully operational within two to five years[5].
References
[1] The Atlantic
[2] Citizenscience.gov
[3] Fondriest.com
[4] Biology.washington.edu
[5] Hawaiinewsnow.com
