20-09-2013, 03:09 PM
THE STS : AN ADVANCED DRIVER ASSISTANCE SYSTEM A SEMINAR REPORT
[attachment=58142]
ABSTRACT
Cooperative awareness in vehicular networks is probably the killer application
for vehicle-to-vehicle (V2V) communications that cannot be matched by
infrastructure-based alternatives even when disregarding communication costs. New
and improved driver assistance systems can be introduced by extending their reach to
sensors residing in neighboring vehicles, such as windshield-installed cameras. In
previous work, we defined theoretical foundations for a driver assistance system that
leverages on V2V communication and windshield-installed cameras to transform
vision-obstructing vehicles into transparent tubular objects. This paper presents an
implementation of the actual See-Through System (STS), where the communication
aspects are combined with the control and augmented reality components of the
system. This paper also presents a validation methodology and test result of the
system with multiple vehicles on a closed road segment. This evaluation shows that
the STS is able to increase the visibility of drivers intending to overtake, thus
increasing the safety of such critical maneuvers. It also shows that Dedicated Short
Range Communication (DSRC) provides the required latency for this delay-critical
inter-vehicle communication, which could hardly be guaranteed with infrastructure-
based communication technologies.
INTRODUCTION
Road traffic injuries are usually tolerated as an inherent risk of driving, even
though road traffic crashes caused over 1.27 million deaths in 2004. This problem is
not confined to developed countries but it rather has become a global health and
development problem of epidemic proportions. In the United States (US), the Fatal
Analysis Reporting System (FARS) provides a breakdown of accidents, where types
of crashes and fatalities can be analysed. There were 3,986 fatal head-on crashes in
2003, killing 5,063 people , which almost guarantees that 2 persons will die from
every headon crash. The FARS indicates that the vast majority of these crashes occurs
on rural, undivided, two-lane roads, which is to be expected since urban scenarios do
not usually provide scenarios for head-on crashes but rather side crashes. These head-
on crashes are the result of either deliberate an action such as executing a passing
maneuver or an inadvertent action causing a run-off-road. The latter is already
addressed by modern driver assistance technologies such as Lane-Keeping-System
(LKS), which is an efficient approach to mitigate the head-on crashes caused by the
inadvertent actions of drivers. However, regarding passing maneuvers there are no
available systems that help drivers on the decision of whether it is safe to engage on
such maneuvers.
STATE OF THE ART OF RELATED TECHNOLOGIES
Windshield Cameras and Computer Vision
The automotive industry is increasingly equipping vehicles with windshield
cameras that provide data to computer-vision based systems. Some examples of these
systems are automatic car following, object classification, and vehicle detection and
tracking. Image-based object detection methods have been proposed in various
research projects. In some cases, synchronized cameras were combined to extend the
number of parameters such as camera shot angle or camera focal length, combining
visual fields in the image-based object detection The visual processing of those
detection systems is based on vision algorithms that focus, for example, on neural
networks, temporal-spatial modeling, and fuzzy logic and are useful in the context of
the visual based vehicle guidance for applications such as road following or traffic
sign recognition. In this context, Gavrila and Philomin presented a shape-based object
detection method based on distance transforms, where object shapes are captured,
classified, and used for real-time vision onboard vehicles. Related to this, calculation
approaches for obtaining the distance between the driven car and the vehicle ahead
can vary, depending on the technology that is used to develop the algorithm.
Augmented Reality in Vehicular Environment
Modern cars are already converging to the concept of a virtualized windshield.
A basic approach is found in the replication of roadside traffic signs into in-vehicle
virtual traffic signs, either projected on the windshield or displayed on LCD screens
in the dashboard.
The earliest example of in-vehicle road signs only appeared in the 1990s with the
introduction of GPS-based navigation systems. Digital road maps that powered such
navigation devices included information about the speed limit of each road,
which was displayed as a digital, in-vehicle traffic sign on the screen of the
navigation device.
Related Assistance Systems
In the last decades, much effort has been put into developing innovative
ADASs to enhance driving safety and comfort. Examples of such systems are Global
Positioning System navigation, ACC, night vision or blind spot detection. Increasing
the driver’s awareness of the surrounding environment is what all these systems have
in common. In particular, cooperative ADASs that detect the existence of other
vehicles in the vicinity enable the driver to improve this awareness under low-
visibility conditions. The goal of the STS system is to increase the visual field of the
driver in overtaking maneuvers. Few works have been dedicated to this particular
issue: For example, the system that assists the driver in avoiding collision through a
Smart Dashboard system at a very low cost based on video-based analysis. Other
works address overtaking in the intelligent vehicle or different detection approaches
based on technologies such as optical flow. Most of them focus on the blind-spot
issue and lane-departure warning, alerting the driver of a potential hazard and, thus,
preventing accidents.
CONCLUSION
1. The STS, a cooperative Advanced Driver Assistance System performs as
expected and it meets tight safety requirements.
2. The latency introduced by the system is already quite low even when
considering that several components are software-based and were not
specifically designed to provide very low latency.
3. The augmented reality aspect of the STS is indeed representative of the
overtaking scenario.
4. STS and its implementation is a valid Advanced Driver Assistance System.
[attachment=58142]
ABSTRACT
Cooperative awareness in vehicular networks is probably the killer application
for vehicle-to-vehicle (V2V) communications that cannot be matched by
infrastructure-based alternatives even when disregarding communication costs. New
and improved driver assistance systems can be introduced by extending their reach to
sensors residing in neighboring vehicles, such as windshield-installed cameras. In
previous work, we defined theoretical foundations for a driver assistance system that
leverages on V2V communication and windshield-installed cameras to transform
vision-obstructing vehicles into transparent tubular objects. This paper presents an
implementation of the actual See-Through System (STS), where the communication
aspects are combined with the control and augmented reality components of the
system. This paper also presents a validation methodology and test result of the
system with multiple vehicles on a closed road segment. This evaluation shows that
the STS is able to increase the visibility of drivers intending to overtake, thus
increasing the safety of such critical maneuvers. It also shows that Dedicated Short
Range Communication (DSRC) provides the required latency for this delay-critical
inter-vehicle communication, which could hardly be guaranteed with infrastructure-
based communication technologies.
INTRODUCTION
Road traffic injuries are usually tolerated as an inherent risk of driving, even
though road traffic crashes caused over 1.27 million deaths in 2004. This problem is
not confined to developed countries but it rather has become a global health and
development problem of epidemic proportions. In the United States (US), the Fatal
Analysis Reporting System (FARS) provides a breakdown of accidents, where types
of crashes and fatalities can be analysed. There were 3,986 fatal head-on crashes in
2003, killing 5,063 people , which almost guarantees that 2 persons will die from
every headon crash. The FARS indicates that the vast majority of these crashes occurs
on rural, undivided, two-lane roads, which is to be expected since urban scenarios do
not usually provide scenarios for head-on crashes but rather side crashes. These head-
on crashes are the result of either deliberate an action such as executing a passing
maneuver or an inadvertent action causing a run-off-road. The latter is already
addressed by modern driver assistance technologies such as Lane-Keeping-System
(LKS), which is an efficient approach to mitigate the head-on crashes caused by the
inadvertent actions of drivers. However, regarding passing maneuvers there are no
available systems that help drivers on the decision of whether it is safe to engage on
such maneuvers.
STATE OF THE ART OF RELATED TECHNOLOGIES
Windshield Cameras and Computer Vision
The automotive industry is increasingly equipping vehicles with windshield
cameras that provide data to computer-vision based systems. Some examples of these
systems are automatic car following, object classification, and vehicle detection and
tracking. Image-based object detection methods have been proposed in various
research projects. In some cases, synchronized cameras were combined to extend the
number of parameters such as camera shot angle or camera focal length, combining
visual fields in the image-based object detection The visual processing of those
detection systems is based on vision algorithms that focus, for example, on neural
networks, temporal-spatial modeling, and fuzzy logic and are useful in the context of
the visual based vehicle guidance for applications such as road following or traffic
sign recognition. In this context, Gavrila and Philomin presented a shape-based object
detection method based on distance transforms, where object shapes are captured,
classified, and used for real-time vision onboard vehicles. Related to this, calculation
approaches for obtaining the distance between the driven car and the vehicle ahead
can vary, depending on the technology that is used to develop the algorithm.
Augmented Reality in Vehicular Environment
Modern cars are already converging to the concept of a virtualized windshield.
A basic approach is found in the replication of roadside traffic signs into in-vehicle
virtual traffic signs, either projected on the windshield or displayed on LCD screens
in the dashboard.
The earliest example of in-vehicle road signs only appeared in the 1990s with the
introduction of GPS-based navigation systems. Digital road maps that powered such
navigation devices included information about the speed limit of each road,
which was displayed as a digital, in-vehicle traffic sign on the screen of the
navigation device.
Related Assistance Systems
In the last decades, much effort has been put into developing innovative
ADASs to enhance driving safety and comfort. Examples of such systems are Global
Positioning System navigation, ACC, night vision or blind spot detection. Increasing
the driver’s awareness of the surrounding environment is what all these systems have
in common. In particular, cooperative ADASs that detect the existence of other
vehicles in the vicinity enable the driver to improve this awareness under low-
visibility conditions. The goal of the STS system is to increase the visual field of the
driver in overtaking maneuvers. Few works have been dedicated to this particular
issue: For example, the system that assists the driver in avoiding collision through a
Smart Dashboard system at a very low cost based on video-based analysis. Other
works address overtaking in the intelligent vehicle or different detection approaches
based on technologies such as optical flow. Most of them focus on the blind-spot
issue and lane-departure warning, alerting the driver of a potential hazard and, thus,
preventing accidents.
CONCLUSION
1. The STS, a cooperative Advanced Driver Assistance System performs as
expected and it meets tight safety requirements.
2. The latency introduced by the system is already quite low even when
considering that several components are software-based and were not
specifically designed to provide very low latency.
3. The augmented reality aspect of the STS is indeed representative of the
overtaking scenario.
4. STS and its implementation is a valid Advanced Driver Assistance System.