Image and Vision Computing '01 |
Keynote Speakers:
| John Fryer | Andre Gagalowicz | Ralf Reulke |John Fryer
Professor of PhotogrammetryDepartment of Civil Engineering
& Surveying, University of Newcastle, Australia.
E-mail: cejgf@cc.newcastle.edu.au
Video imagery and forensic photogrammetry
Video recording is widely used to capture images in locations prone to criminal activity, for example banking institutions. A vital piece of forensic evidence can often be the height determination of a bank robber imaged during a crime scene incident. Some justice systems are often claimed to weigh heavily in favour of the accused, so the presentation of forensic evidence in the form of photogrammetric height determinations must be concise and explicit. A jury of non-experts must be certain in their own minds that they understand the physical processes involved in the height determination of, for example, a masked bank robber.
This paper presents the results of real investigations into some video imagery taken during a series of bank robberies. The methods devised to extract height information and the component lengths of sawn-off shotguns and rifles is believed to be unique in that all measurements were made in object space and are totally independent of lens distortion. Object space is a concept which a non-expert jury can understand, rather than the mathematical modeling associated with the electronic imaging and camera/lens calibration required for an image space solution.
Other forensic investigations with which the author has been involved
will also be discussed.
Biography
John Fryer is the Professor of Photogrammetry and Head of the Department of Civil, Surveying and Environmental Engineering at the University of Newcastle.Photogrammetry is the science of determining the size and shape of an object by making precise measurements on photographs or images. John Fryer's research has involved the application of close-range photogrammetric techniques and the analysis of lens distortions in both digital and film cameras. For the period 1992-96, John was President of Technical Commission V of the International Society for Photogrammetry and Remote Sensing (94 member countries).
His research with small format cameras has involved him with several interesting projects. Some have involved underwater photography with subjects as diverse as coral growth rate studies, corrosion evaluation on oil platforms and the monitoring of underwater projectiles (Trident II missiles leaving submarines). Current research involves the use of video- and image-processing technology for a variety of measurement tasks such as architectural facades, dam monitoring studies, Aboriginal rock art and forensic applications.
John has always had a love of sport and played cricket and tennis in
Newcastle competitions for nearly two decades. His interest in sport
led to his election as President of the University of Newcastle Sports
Union (1993-1998) and a Director of the Hunter Academy of Sport 1993-98.
He has published over 170 technical reports and scientific
papers, has contributed chapters to books, and is co-author of a text book
on surveying.
| back to the top |
Andre Gagalowicz
Research Director, MIRAGES
INRIA Rocquencourt, France.
E-mail: Andre.Gagalowicz@inria.fr
Computer vision and graphics collaboration. Application to Post-production
We present a new technique, based upon image analysis/synthesis collaboration, which allows to track a chosen 3D object automatically in an image sequence. The application we have in mind is rotoscopy, for postproduction applications. For this application, tracking has to be performed at a pixel precision which explains that, in most postproduction companies, such operation is still performed manually. A technician proceeds image after image and draws the contours of the object that they want to suppress (rotoscopy is the name of this operation).We propose to use vision tasks to speed up the procedure.
The solution we provide is restricted to rigid objects
and supposes that we dispose of a model of the object to track:
it is a model-based approach. The model is first positioned manually on
the first image of the sequence. This operation is equivalent to calibrating
the camera using the obect to track. Afterwards, an animation model is used
to predict the position of the object on the next frame and this position
is tuned by a pseudo-annealing technique minimising the distance between
the real object image and the predicted one (using the object model). Examples
showing the stability of the technique will be shown as well as remaining
problems and possible extensions.
Biography
Dr. Andre Gagalowicz is a research director at INRIA, FRANCE. He was the creator of the first laboratory involved in image analysis/synthesis collaboration techniques. He graduated from Ecole Superieure d'Electricite in 1971 (engineer in Electrical Engineering), obtained his PHD in Automatic Control from the University of Paris XI, Orsay, in 1973, and his state doctorate in Mathematics (doctoctorat d'Etat es Sciences) from the University of Paris VI (1983). He is fluent in English, German, Russian and Polish and obtained a bachelor degree in Chinese from the University of Paris IX, INALOCO in 1983.His research interests
are in 3D approaches for computer vision, computer graphics, and their
cooperation and also in digital image processing and pattern
recognition. In 1985, he received prizes for best scientific communication
and the best technical slide at the Eurographics'85 conference. In 1991, he was
awarded the second price of the Seymour Cray competition and one papers was
selected by Computers and Graphics journal as one of
the three best publications of this journal for the last ten years. He
took part to the redaction of eight books and has written more than one
hundred publications.
| John Fryer | Andre Gagalowicz | Ralf Reulke | back to the top |
Ralf Reulke
Head of Theory and Signal ProcessingGerman Aerospace Centre (DLR),
Institute of Space Sensor Technology and Planetary Exploration, Germany.
E-mail: Ralf.Reulke@dlr.de
Design and application of high resolution imaging systems
Recent camera technology provides new solutions for high-resolution image acquisition. These cameras are based on CCD - line sensors and fulfil both: large field of view and high spatial resolution. Single- or multi - line cameras have been designed for spaceborne, airborne and terrestrial scanners to provide high-resolution imagery. Such cameras allow new approaches of 3D scene visualization and new directions in airborne and close range photogrammetry. This has consequences on camera design and development, calibration, image acquisition and data processing.
During the past two years the company LH Systems and the German Aerospace Center (DLR) have developed the commercial airborne digital sensor ADS40. This high-resolution imaging system is able to fulfil both photogrammetric and remote sensing requirements. The new sensor was introduced in mid 2000 and will complete the digital chain for airborne photogrammetric data processing.
The design principles for this high-resolution sensor are based on a single optics / single focal plate solution. To meet the requirements, customized CCD-line structures were used. A staggered line arrangement allows a significant spatial resolution enhancement. CCD-line sensors have a linear and reproducible intensity response and are therefore measurement devices in contrast to traditional photo film systems. This has an impact on the instrument design: Digital photogrammetric sensors have also multispectral analysis capabilities. Multispectral CCD-lines can incorporate in the focal plate. The hardware is on the today's technological edge and can handle 120MPixel/s with 14 bits radiometric dynamic and an SNR better than 8 bits. The data transfer between camera head and camera computer is based on fibre channel connection. The hardware includes all aspects of real time data normalization, correction and processing. The on-board data compression is based on the standard JPEG algorithm.
In addition to the technical design, the performances of ADS40 will be
shown on examples to demonstrate the main features of digital sensors:
high accuracy, wide field of view, high radiometric dynamics, high
signal-to-noise-ratio, in-track stereo capability, and multispectral
capability.
Biography
Dr. Ralf Reulke received his M.Sc and Ph.D. degrees in Physics from the Humboldt-University of Berlin in 1980 and 1984, respectively. He is presently Head of Theory and Signal Processing at DLR (German Aerospace Centre). He has an expertise of 18 years in simulation and optimization of remote sensing systems, data and stereo processing, as well as remote sensing with multisensor systems, which has been documented through 1 book, 18 papers in international journals, more than 50 conference papers and 1 report. He is active in sensor design and development for space missions, and also for airborne and terrestrial instruments.| John Fryer | Andre Gagalowicz | Ralf Reulke | back to the top |
Last updated: Tuesday 2nd October 2001