| |
|
This page outlines the hardware and software requirements for using underwater stereo imagery to
make accurate three-dimensional measurements and particularly to estimate the volume and surface area of
complex objects.
Accurate three-dimensional measurement is based around the use of a calibrated stereo camera system.
The measured three-dimensional points and stereo camera location geometry can be combined to estimate the
volume and surface area of a geometrically complex object.
|
Capturing stereo images of a sponge for volume and surface area analysis
(note the optional artificial reference frame).
|
Camera System
Stereo cameras are the basis of the measurement system. The type of cameras and their geometric arrangement
and stability will determine the overall measurement accuracy of the system. To provide a stable and accurate
measurement system the cameras must be fired synchronously in time to provide a true stereo pair, and the
cameras must be mounted in a robust manner to maintain the camera calibration over time. Exact detail of the
physical camera separation and convergence angle will depend on the size of the objects to be measured, and
can be a compromise to cover a reasonable range of object sizes.
Key requirements:
-
Synchronised firing of cameras
-
Cameras robustly mounted in suitable housings, housings mounted on base bar
-
Ability to download imagery, charge and configure cameras without frequently opening housings -
Cameras must be capable of operation at a fixed focal length and aperture
Camera Calibration
In order to make accurate underwater three-dimensional measurements the stereo camera system must be calibrated.
The calibration procedure is critical to the overall accuracy of volume and surface area estimations, as these estimates are functions of the measured coordinates.
The camera calibration is performed by photographing a calibration fixture and processing the captured images.
The calibration fixture is photographed in-water to allow correct calibration of the camera system for underwater measurement.
Once the cameras are calibrated there is no need for re-calibration unless the camera housings are opened, or the cameras are physically knocked hard enough to alter their relative positions and orientations.
Key requirements:
Software
Measurement occurs as an office-based computer process after the cameras have been calibrated and the imagery captured.
In order to generate an accurate volume or surface area estimate, a series of stereo image will generally be required.
The images must capture all surface detail (if the detail cannot be seen in the images, it cannot be measured).
As a rule of thumb, around five images are usually required (see results in the JEMBE paper).
The software user must register all stereo images using a series of common points (minimum of three) visible between all stereo images.
This effectively allows three-dimensional coordinates derived from all stereo images to be transformed to a common datum.
The common points may be natural features on or around the object being measured, or can be artificially introduced targets that allow automation of the process (see artificial reference frame in the following figure).
The user must also identify points that define the surface of the object to be measured. Once the object surface is defined, generation of the volume and surface area estimate is automatic.
Key requirements:
|
|
|
|
