Visual Perception Microsystems based on Structural Computational Paradigms
Funding Agency: MURST (Italian Ministry of Scientific and
Technological Research)
Funding period: 1998-2000
URL:
Contact Person
Giacomo.M. BISIO
Department of Biophysical and Electronic Engineering
University of Genoa
Via Opera Pia, 11a
I-16145 Genova, ITALY
tel: (+39) 010 353 2756
fax: (+39) 010 353 2777
e-mail: bisio@@dibe.unige.it
Participants
Coordinator:
PSPC-lab, DIBE-University of Genoa I
Partners:
IMS-lab, DIE-University of Pavia I
EOLAB, DIEE-University of Cagliari I
Abstract
The project is aimed at the design and implementation of integrated
microsystems for machine vision. These systems will consist of an array of
optical sensors together with analog interfaces and processing modules
whose characteristics suit a specific application domain. The results of
this research will impact on the development of sensorial communication
interfaces, in surveillance systems, in robotics and as vision aids for
elderly and handicapped people.
The research conducted at DIBE focusses on signal processing based on the
dynamic behavior of distributed computational structures (e.g., lattice
networks), and on algorithmic solutions to visual perception tasks based on
local operators. Specifically, the research program is devoted to the
definition, implementation and testing of the algorithmic solutions to
perceptual problems (stereo depth map estimation, or discrimination of
forward and backward movements from the observer) that well suit a
computational interpretation in terms of lattice networks.
Synopsis
Aims:
Under an applicative perspective, the project aims to develop a novel
microsystem prototype for visual applications that is characterized by
stand-alone perceptual capabilities such as stereo depth map estimation and
discrimination of forward and backward movements of objects relative to the
observer. Techical and theoretical contributions of the project to
key-themes in the field of perceptual microsystems with complex
functionalities are related to (1) computational
circuits for spatiotemporal signals, (2) algorithmic solutions to vision
tasks constrained to the available computational resourses, (3) advanced
CMOS image sensors, (4) the management, at system level, of computational
flow for analog and digital signals. Specifically, the research program
focusses on the first two key-themes: signal processing based of the
dynamic behaviour of distributed computational structures (e.g., lattice
networks); algorithmic solutions to visual perception tasks based on local
operators. To evaluate the impact of this approach on system performance it
will be necessary to consider the second two key-thems at the level of the
state of the art.
Approaches and Methodologies:
To achieve these goals it is necessary to follow a novel design apporach at
architectural level upon which the algorithmic solutions oriented to analog
processing and the system's computational requirements hinge. In
particular, at circuital level, one should follow a mixed approach in order
to take full advantage of the potentialities of analog processing together
with the flexibility provided by digital hardware.
The research activities are organized as follows: (1) functional
specification of the microsystem; (2) algorithmic solutions and
simulations; (3) VLSI design and implementation of functional primitives;
(4) specification of the demonstrator and its implementation. Operatively,
the research program is organized in three worparts:
The expected results from the three workparts converge into the final
demonstrator devised to achieve early-vision perceptual functionalities in
depth perception and motion analysis from dynamic stereo images. The main
components of the demonstrator are: (1) 2D imager integrated with analog
processing modules performing Gabor filtering; (2) application units (i.e.,
specific processing modules for the target application);(3) board for
system integration and communication toward the host PC.
Impact and Exploitation plans:
Implementation of stand-alone visual perception microsystems to be used as
sensorial communication interfaces in surveillance systems, in robotics and
as vision aids for elderly and handicapped people.
These issues could be related to the development of CMOS image sensors (to
be used in modern multimedia systems, computer peripherals, portable
consumer products, telemedicine, etc.) based on the integration of sensing
devices and read-out circuitry and processing. However, it is worthy to
note that our microsystem do have functionalities of different nature,
oriented not yo signal transmission but to signal
processing/interpretation, in order to gain perceptual capabilities that
can guide the behaviour of an artificial system, such as autonomous robots,
and surveillance devices.