Coral reefs are rapidly degrading under the pressure of increasing human activities, and at a faster pace than our current actions to prevent further the degradation. Comprehensive and fully comparable information about the state of coral reefs is urgently needed to provide timely advice to management authorities. Under these premises, rapid and standardised survey technologies at a global scale are needed to provide such information and complement current regional efforts. The XL Catlin Seaview Survey combines existing technologies to merge efforts for a high quality and standardised approach.
The XL Catlin Seaview Survey approach is centred on high definition imagery, to document and study coral reefs worldwide. Based on the concept that ‘a picture is worth a thousand words’, this scientific survey is looking at extracting valuable information from coral reef imagery while archiving an unbiased proof of the condition of surveyed reefs. This approach involves rapid acquisition of high-resolution imagery over large extensions of reefs. Using computer vision and machine learning technologies, the imagery is rapidly analysed for coral reef structure across multiple spatial scales, from metres to kilometres.
The XL Catlin Seaview SVII camera was specially built for this project. It is designed to be a highly robust underwater vehicle capable of taking 360° high-resolution images in almost all weather conditions whilst on lengthy remote expeditions. A timer triggers the camera allowing it to take pictures at three-second intervals, which is controlled by an Android tablet, encased in an underwater housing. The head of the camera is designed to remain sealed whilst on expedition to avoid flooding issues – data and power are transferred through a wet link docking station.
Shallow Reef Survey Methodology
Lead scientist – Dr Manuel Gonzalez-Rivero
Using diver-propelled vehicles and customised HD cameras (SVII), divers survey unprecedented areas of coral reef over multiple reef sites. A depth sensor logger carried by the diver monitors the survey depth every second. Surveys are set to a standard depth of 10 metres (+/- 2 metres) following the contour of the reef at this depth. By standardising the survey depth, we avoid the inclusion of confounding factors (i.e. light, wave energy) when comparing the structure of coral reef communities amongst multiple spatial scales.
During the survey, the SVII camera captures three high-definition images every three seconds during the fifty minute dive. These images allow us to continuously record the full 360° environment along a 2 kilometres stretch of reef.
The transect location is selected by looking at capturing a vast variety of coral reef configurations or habitats to best represent the fore-reefs of a region.
This variability is captured by looking at multiple drivers of coral reef structure, such as environmental regimes (i.e. hurricane incidence, temperature regimes, ocean colour, etc.), local information, geomorphology (atoll, fringing reef, barrier reefs) as well as anthropogenic influence (i.e. coastal development, fishing pressure).
Every image captured for the XL Catlin Global Reef Record has a geographical tag that allows us to accurately describe a specific section of reef that can be revisited in time and compared against previous available information. To achieve this, a GPS unit is synchronised in time to the SVII cameras and tethered from the surface by the divers.
In order to accurately and consistently locate the benthic community structure, image scale is required to standardise the area of the reef to be analysed. A customised altitude logger has been fitted to the SVII camera, which is also synchronised to the time of the SVII cameras. This altimeter, connected to a doppler transducer, logs the altitude of the camera with respect to the seabed and depth. Using this information and the optical properties of the lens of the SVII camera, the footprint area of each downward facing camera is estimated.
After finishing the surveys, images are post-processed, scaled and stitched to recreate a 360° panorama of the reef. Each image along the transect allows us to explore the complete environment of a reef. At present, each transect feeds our database with up to 1000 panoramas from a particular reef section. In total, every field campaign provides about 30-40 thousand images, which challenge the capacity of traditional analysis methods, to rapidly extract the scientific information from each image. Using computer vision and machine learning algorithms developed by SCRIPPS Institute of Oceanography and the University of California, we rapidly analyse our imagery at a pace, which is 50 times faster than traditional approaches.
CoralNet, the platform developed by SCRIPPS to analyse coral reef imagery, uses a “point-count approach” where points are randomly laid over the image to identify the organisms under each point and calculate the percentage of cover per image. The software learns from continuous expert annotations that repeatedly identify the categories under each point on the image. Combined with computer vision algorithms, the optical properties of a section within the image under each point are archived to identify each coral category.
As experts annotate the imagery, the machine also learns which attributes of a particular section in the image (i.e. colour, texture) best describe different groups of species (i.e., brain corals, tubular sponges, fleshy macroalgae). Following the training sessions the algorithms are ready to crunch through the large set of images and extract coverage for each group of species present in the image. This information allows us to rapidly understand the composition of the reef from metres to kilometres across large reef systems.