Abstract:
When using conventional illumination in the underwater environment, the
greatest problem faced has always been that of back-scatter. This reduces
visual quality both to the naked eye and through the use of underwater
cameras. This project continues research into a hybrid underwater laser
viewing system, combining a laser scanning technique with conventional
underwater viewing system technology.
The use of carefully positioned lighting can greatly reduce the problem of
back-scattered light, improving image definition and contrast. With the use of
a laser light source this can be improved even further within a more flexible
system. Utilising two scanning devices a narrow stripe of light can be formed
which is scanned over a target area. Small successive regions of any object
lying within that area can then be illuminated, isolated and extracted using
image processing to compile a complete image of the object.
For the purposes of surf zone and shallow water mines counter measures, the
system requires reliable machine algorithm based image processing
techniques, to compile a constructed image of target objects within highly
turbid regions. Projected stripes take on various characteristics dependent on
the geometry and aspect of the object which they fall upon and the conditions
experienced. High levels of turbidity leading to significant noise, creates
considerable difficulties in the isolation of the stripe region and the extraction
of meaningful visual information. The success of any technique would rely
upon the overcoming of these difficulties, allowing the system to offer the
operator the most useful visual information permissible in the final compilation
of the image.
Through various image processing techniques these factors have been
tackled and their effects on the final image, greatly reduced. The result is the
creation of highly stable and effective processing techniques allowing
improved object definition and recognition, in levels of high turbidity.