“Dimidium facti qui coepit habet”
(Orazio, Epistole, I, 2, 40)
This research is an integral part of a larger
project, AutoMERS , that CEII is pursuing with the collaboration
of two Departments of the University of Aberdeen, the Department of Zoology
and the Department of Engineering, alongside other research groups within
Scotland’s universities. One of the aims of the AutoMERS Project is to
investigate, using different types of probe, deep sea life; among these
probes are different kinds of cameras, still image and movie camera, with
different sizes and resolutions, providing a large amount of data to analyse.
The purpose of the research is to analyse this kind of data from the image
and video processing point of view, to improve them and to give some pointers
for future improvements and innovations of the cameras.
1.1. CEII
The Centre for Environmental and Industrial Imaging
brings together expertise in both fundamental and applied imaging from
across the Faculty of Science and Engineering in the University of Aberdeen.
The experience within the centre covers such areas as optics and instrumentation,
computational image analysis, novel architecture computing, pattern recognition
and artificial intelligence; the imaging modalities include optical, fluorescent,
holographic, electrical impedance, ultrasound, radar and infrared. The
centre provides a focus for this work, helping to provide a multi-disciplinary
team approach to solving key environmental and industrial problems; it
is currently tackling leading-edge problems such as the automatic recognition
and classification of biological objects, and the representation and interpretation
of 3-D image information.
There is a range of projects in progress, supported
by funding from Research Councils, the European Union, and industry; these
are the main research areas, interesting for our point of view:
Image acquisition,
Image enhancement and representation,
Pattern recognition,
Feature extraction and measurement,
Multi-resolution and compression techniques.
1.2. OCEANLAB AND AUTOMERS
The Aberdeen University Ocean Research Laboratory,
OceanLab, is based in the Zoology Department; the research group includes
biologists and engineers investigating the ecology and behaviour of marine
animals. The Autonomous Marine Environment Research Stations (AutoMERS)is
a joint initiative between the Universities of Aberdeen, Bristol and St.
Andrews and the Scottish Association for Marine Science (SAMS) with the
aim of creating a UK fleet of Unmanned Autonomous Sub-sea Platforms, capable
of undertaking research missions at depths down to 6000m for periods of
up to twelve months.
Figure 1.1 : Example of autonomous landers.
Current and recent research projects include:
Investigating the behaviour and ecology of deep-sea
fish,
Deep-sea fisheries: novel assessment methods,
Development of new technology for oceanographic research.
The research uses, as seen in Figure 1.1, with autonomous
"Landers", like the Aberdeen University Deep Ocean Submersible (AUDOS),
that are able to photograph and track deep-sea fish. They are capable of
operating at depths of 5000 m and consist of an open framework upon which
a scientific payload (time-lapse camera, current meter, fish-tracking sonar
…) is mounted together with buoyancy, ballast and ballast release system.
The lander descends from the ship to the working depth by free-fall under
its own ballast whereupon the scientific mission commences, controlled
by an on-board processor. At the end of the mission the ballast is shed
by acoustic command from a ship and the lander ascends by virtue of positive
buoyancy; radio beacons, strobe lights and visual markers activated on
the surface aid location and recovery of the lander. It is possible to
find more information about OceanLab and AutoMERS project on the web site
http://www.oceanlab.abdn.ac.uk/.
1.3. IMAGES
At the moment we are working with only two different
kinds of data: HDSI (High Definition Still Image) and VNDI (Video Normal
Definition Image).
High Definition Still Images: these images are captured
by scanning 35 mm video camera slides with a Polaroid “SprintScan 35 Plus
Film Scanner” . Every slide shows a still image of the target at intervals
of 1 minutes one to another, so there are plenty of different still images
showing different objects. It is possible to obtain a high definition picture
from this 23.8 ´ 33.8 mm slide by scanning at 2700 dpi. These are
the typical features of the picture: Windows or OS/2 bitmap image, size
approx. 3710 ´ 2440 pixels (for a total of about 9 million pixels),
RGB, 24 bit/pixel (8 for each matrix-colour), file size about 26 Mbytes,
Video Normal Definition Image: this video is a MPEG-2
video, 720 ´ 576 pixel size, RGB, 24 bit/pixel (8 for each matrix-colour),
25 frame/second recorded with a video camera on a tape. It shows the same
target of the HDSI images but with moving images; it usually takes images
for some fraction of time, every some interval of time, for example 5 seconds
of video every 10 minutes.
1.4. OBJECTS
Each image shows a typical view with some objects
inside. The objects represented in the image could be divided into one
of 3 kinds of object, shown in Figure 1.2:
The background: sand, stones, rocks, pebbles, usually
coloured grey/light brown,
The cross part formed by the dark brown/black centre
sinker and the light grey/blue cross with the red tips and some letters
and numbers usually written on to it. Usually this part is positioned on
the central area of the image,
None, one or more interesting objects like fish,
crabs, lobsters, starfish and others with different features. This is the
part of the image of interest to the researchers.
Figure 1.2 : Typical image from AutoMERS project
(coloured version in Appendix: Figure A.1).
One of the purposes of this project is to improve
the quality, the visual definition of these animated objects as regard
to the other kind of objects. The principal goal however is to have a trade
off between these two principals purposes, typical for the image and video
processing:
Improve the compression of the image or the video,
or even a part of these, as to the part regarding the first two kinds of
object, to attempt to reduce the amount of bits transmitted or saved,
Improve the quality of the image or the video, or
a part of them, as the part of the third kind of objects to attempt to
improve the understanding of the properties, the features and the characteristics
of the objects.
E-mail to: eug67@supereva.it
e.ballini@eng.abdn.ac.uk