Instrumentation & Measurement Magazine 23-8 - 18

Polarimetric Dynamic Vision
Sensor p(DVS) Principles
Martin Nowak, Anthony Beninati, Nicolas Douard, and George C. Giakos

T

his contribution relates to the activities of the TC-19
on "Imaging Signals and Systems." In a rapidly
changing global economy, experiencing an unparalleled integration of science and technology, the development
of efficient imaging systems is of extreme significance. The
scope of TC-19 is to exchange and disseminate knowledge as
well as bridge multidisciplinary areas leading to the creation
of new knowledge and establish global collaborative multidisciplinary opportunities by tightening collaborations among
government, industry, academia, and healthcare industry. The
Committee explores a variety of computer vision and visualization systems spanning from medical imaging to defense,
industry, and consumer electronics. These imaging systems
use statistical or sophisticated artificial algorithms and require
the integration of several technological and computational
disciplines, such as detector physics and phenomenology,
calibration, metrology, data acquisition, data analysis, image
processing, feature extraction, and classification. Often, coexisting competing design requirements and trade-off introduce
significant challenges to the design of imaging technologies.
For instance, Autonomous Vehicles (AV)s, Unmanned Aerial
Vehicles (UAV)s , Unmanned Underwater Vehicle (UUV)s, microsatellites, nanosatellites, and picosatellites for defense and
civilian applications are characteristic examples with trade-off
between design requirements for high performance imaging
and maintaining low-power consumption, high-data transmission, low storage and memory, and reduced payload. To
address these competing design challenges, TC-19 introduced
the polarimetric neuromorphic vision principles which blend
bioinspired principles such as human cognition with the vision of certain arthropods, and aquatic biological species.
This article articulates the broad principles of bioinspired
vision. These principles integrate human cognitive vision
with biological vision properties, such as polarization of light.
Second, the neuromorphic or retina vision principles are introduced. Special emphasis is paid on the unique properties of the
event-based Dynamic Vision Sensors (DVS)s which operate
asynchronously based on differential light intensity changes.
Third, the principles of polarimetric Dynamic Vision Sensors

18	

p(DVS), accompanied by experimental study on the detection of rotating targets at different speeds, are introduced. At
the end, feature extraction and classification of acquired images using a deep learning network are exposed. The purpose
of this study is to introduce bioinspired vision architectures,
integrating human cognition capabilities, such as computation and memory emulating neurons and synapses, together
with the polarization vision capabilities of certain animal species that would ultimately revolutionize and give rise to the
next-generation highly efficient augmented intelligence vision systems. The presented p(DVS) system would be able
to exhibit high processing speed and high spatial-temporal
contrast, while operating at low bandwidth, with low power
consumption, low memory, and low storage capabilities.

Bioinspired Vision
Bioinspired computer vision systems are a promising way
to build more adaptive, fast sensing, low-power, robust, and
agile robotic vision solutions [1]-[4]. The term "bioinspired vision" encompasses vision structures that span from human
cognitive vision architectures, to arthropods, and marine species vision. Arthropods, certain marine invertebrate species,
and human cognitive vision demonstrate unparalleled vision
capabilities that are complementary to each other. As matter
fact, the morphological and biochemical parameters of the biological retina, the decentralized vison-sensor architectures of
the arthropods coupled with polarization sensitivity, and the
cognition-assisted human vision combined with neural network processing capabilities are distinct features of these eye
architectures.
Polarimetric vision, attributed mainly to the vision characteristics of certain animal species, offers unparalleled
detection and classification capabilities for a wide range of scenarios. Due to its intrinsic potential high-background scatter,
indeed, polarized imaging can yield high-contrast images under low-light conditions, and cluttered environments, while it
provides information related to the object geometrical, material, molecular and chemical composition. Bioinspired vision
architectures, combining both human cognition capabilities,

IEEE Instrumentation & Measurement Magazine	
1094-6969/20/$25.00©2020IEEE

November 2020
Instrumentation & Measurement Magazine 23-8

Table of Contents for the Digital Edition of Instrumentation & Measurement Magazine 23-8
