ROV Video Transects

MBARI’s remotely operated vehicles (ROVs) frequently perform benthic video transects, where the ROV flies in a straight line just above the seafloor over kilometer distances. The main video camera records all of the benthic animals seen, providing valuable information to benthic ecologists on what species are present and in what numbers.

While identifying and counting the animals in the video is relatively straightforward, determining their size is much more difficult. Although the zoom angle of the camera can be calculated, the distance between the camera and the animals is not known and accurate size measurements can’t be made.

Parallel Lasers

The first approach to solving this problem was to mount two parallel lasers on the camera, aimed in the direction that the camera is looking. Because the distance between the lasers is known (typically 29 cm), the two laser dots on the seafloor provide a size reference to which seafloor animals can be compared.

There are, however, significant limitations to this technique. Because the seafloor is tilted at an unknown angle with respect to the camera, only animals lying directly between the laser dots can be accurately sized. In addition, many animals are much longer than they are wide and are more difficult to size if they aren’t oriented on the line between the two laser dots. Due to these issues, many animals seen in the video can’t be measured.

Helical Lasers

We developed Laser Measure to address these limitations. Instead of just two parallel lasers, Laser Measure uses three lasers aimed in a helical pattern to create three laser dots on the seafloor. If we assume that the seafloor is flat, as it is for most of the areas we study with video transects, then the locations of the three laser dots in the video images can tell us the distance and the orientation (i.e. tilt) of the seafloor. Given these parameters, we can measure the size of any object in the image regardless of its location or orientation in the image.

The system must be calibrated each time the three lasers are mounted on the ROV’s main camera. Once the ROV is on the seafloor, a calibration target with a checkerboard pattern is held in front of the camera in various positions by the ROV’s robotic manipulator. This calibration allows us to determine two sets of parameters: first, the intrinsic parameters such as the camera’s focal length and field of view, both of which change with the camera’s zoom; and second, the extrinsic parameters, which are the locations in 3D space of the target and the laser dots falling upon it. The intrinsic and extrinsic parameters in turn allow us to determine the precise locations and pointing directions of the three lasers. The calibration procedure takes about 15 minutes and is performed just before the video transect is started.

After the ROV dive, the calibration video is analyzed to generate a calibration file, which is then used in the Laser Measure software to make measurements of still images from the video transect. The Laser Measure system has been tested in Monterey Bay and by measuring artificial animals of known size, we are seeing measurement errors of 10–15%. We are working to improve the calibration and analysis algorithms to further reduce this error.