Library Item
Computer Vision Applications for Assessing Coral Reef Health
Description:
Anthropogenic stresses to coral reefs ecosystems coupled with climate change impacts are causing unprecedented declines in coral reefs globally, yet appropriate technology to monitor the health of coral reefs at large temporal and spatial scales does not exist. I present here the development of a computer vision enabled photographic coral ecology monitoring tool which will provide a method for rapid, large scale, objective monitoring of coral reef health, on spatial and temporal scales appropriate to the scale of such as global warming and ocean acidification.
Using corrected, high resolution digital images of tagged coral reef targets acquired over a 5 year period in Panama, traditional labor-intensive hand annotation and interpretation results are compared to computer vision enabled automated results, in terms of bleaching and recovery of individual corals across ten species. This NSF-supported project will use this initial photographic dataset to train the computer vision system for interpretation of additional repeat photographic transect datasets, including over 5 years at the NSF-sponsored Moorea Coral Reef Long Term Ecological Research (LTER) site and a 4 year time series for a site on the Great Barrier Reef (GBR).
During this comparison period, this automated system will be coordinated with additional enabling technologies including geo-referenced survey methods using enhanced Red/Green/Blue (RGB) image cameras, a unique multi-spectral fluorescence camera, a towed imaging system that can rapidly survey many kilometers of reef, and an autonomous vehicle for surveying deep reefs.
