[0001] The invention relates to a driver assistance system for a motor vehicle, comprising
an imaging apparatus adapted to capture images of the surrounding of a motor vehicle,
and a processing device adapted to perform image processing on the captured images
and to control a driver assistance device based on the result of said image processing,
wherein said processing device comprises a risk assessment estimator adapted to estimate
a traffic risk level.
[0002] In driving situations the risk level varies substantially. For example, an empty
highway has much lower risk than a highly trafficked city.
[0003] Assessing the risk level can be beneficial for informing the driver so that the driver
can adapt to an adequate alertness level. The functionality is beneficial in different
ways for the range of manually controlled to semi-autonomous to fully autonomous vehicles.
[0004] US 7,672,764 B2 discloses a driver assistance system including various sensors for recording driver's
conditions and vehicle operations by a driver. The driver's conditions in combination
with the vehicle operations are used to determine a risky situation of the vehicle
based inter alia on an averaged driver's condition.
[0005] The problem underlying the present invention is to provide a driver assistance system
with improved traffic risk level estimation.
[0006] The invention solves this problem with the features of the independent claims.
[0007] The invention describes a driver assistance system where the risk assessment estimator
estimates the traffic risk level on the basis of input source data from at least one
input source including data captured by the imaging apparatus. The image data captured
by the imaging apparatus contains valuable information in addition to sensor data
from other vehicle sensors used in the prior art. By using this image data in the
risk assessment estimator for the estimation of the traffic risk level, the accuracy
and reliability of the traffic risk level estimation can be enhanced, contributing
to a better performance of the driver assistance system.
[0008] In the present application, traffic risk level refers to the amount of alertness
required by the driver. For example the traffic risk level would be low when driving
on an empty highway where no threats are near. On the other hand, a risky situation
(high risk level) can for example be a situation where it is common for pedestrians
to be near the road and/or where the speed of the ego vehicle is relatively high given
the regulations of the current road.
[0009] Preferably the appearance of the scene surrounding the motor vehicle extracted from
captured images is used by said risk assessment estimator. Herein, scene refers to
the landscape or background scenery, as opposed to detected objects like other vehicles,
pedestrians, cyclists, animals etc. By analysing the appearance of the scene surrounding
the motor vehicle, it can be determined in what type of environment the vehicle is
located in, which closely correlates with the traffic risk level. For example, the
environment of the vehicle may preferably be classifiable by the risk assessment estimator
into one of city environment, highway environment, rural road environment.
[0010] Preferably the risk assessment estimator considers the whole image in the determination
of said traffic risk level. This holistic approach is advantageous over prior arts
systems detecting and analyzing objects in the images, only, which form only small
parts of the whole image. The considering of the whole image makes use of the complete
image information contained in the image, where valuable information can also be contained
in image parts not forming discrete objects like pedestrians, other vehicles, pavement,
cross walk signs, bus stops etc.
[0011] In a preferred embodiment of the invention, the risk assessment estimator considers
each pixel of a captured image as a separate input source in the determination of
said traffic risk level. Through the analysis of the captured images on an individual
pixel level, maximum information can be used in the determination of the traffic risk
level, avoiding any inaccuracies originating from averaging of the image over multipixel
areas.
[0012] Preferably the risk assessment estimator is trained by using a learning system, in
particular a deep learning system, like a deep neural network. The network is preferably
directly connected to each pixel of the captured images, and to other sensors used
as input sources for said traffic risk level estimation. The learning system is preferably
trained using training data rated by a plurality of persons with respect to the estimated
traffic risk level. That is, each person rates the estimated traffic risk level of
each training image according to his best evaluation, and for example the average
of all ratings is taken as the true traffic risk level. The learning system could
automatically learn to recognize aggressive driving, pedestrians (zebra) crossings,
school areas, etc. Alternatively or in addition, the classifier could be trained by
using publically available databases of accidents, containing corresponding GPS positions,
and recording data at those locations. The classifier can then learn to recognize
environments that look similar to actual environments where real accidents happen.
[0013] In order to improve the accuracy of the traffic risk level estimation, the use of
a plurality of input sources for the risk assessment estimator, generally as many
input sources as possible correlating with the traffic risk level, is desirable.
[0014] A first category of preferred input sources provide ego vehicle data, in particular
ego vehicle dynamics data, as input source data. Possible examples of input sources
falling under this category are acceleration sensors, yaw sensor, roll sensor, pitch
sensor, speed sensor, braking sensor, steering wheel angle sensor.
[0015] A second category of preferred input sources provide data of other objects in the
environment of the motor vehicle, in particular other object's dynamics data. Other
(discrete) objects may for example be moving objects like other vehicles, pedestrians,
bicyclists, large animals. Immovable objects like traffic signs, poles, buildings,
trees etc. are also possible. Possible examples of input sources falling under this
category are the imaging apparatus, a radar apparatus, a LIDAR apparatus, a backward
looking camera. Also, it can be preferable to use a data memory or storing information
about detected other objects as an input source.
[0016] Other possible input sources provide input source data comprising: speed limits obtained
by sign recognition or a satellite navigation receiver; biometric information; road
conditions; ambient conditions, like ambient temperature or ambient humidity; stored
data of previous hazard areas or conditions; location of crosswalks obtained by image
recognition or a satellite navigation receiver.
[0017] The assessed or estimated traffic risk level can be used in different manners. In
one embodiment, the driver assistance system comprises a risk level indicator adapted
to indicate an estimated risk level to the driver. For example, one or more diodes
could be used to indicate the risk level to the driver via color coding, e.g. covering
the color spectrum between green (low risk level) to yellow (middle risk level) to
red (high risk level). Alternatively or in addition, the estimated risk level may
preferably be used for requesting safety-relevant actions by the driver depending
on said estimated risk level. For example, at a first (lower) risk level, the driver
could be requested to have his hands on the driving wheel, while at a second (higher)
risk level, the driver could be requested to have his eyes on the road. Also in autonomous
driving applications, the estimated risk levels may be used by a corresponding driver
assistance device.
[0018] Also, information could be downloaded from the cloud during driving, e.g. traffic
jam data, using driving dynamics of other vehicles (similar to electronic maps available
in the internet), or average driver aggressiveness which may vary from region to region
and from time to time of the day.
[0019] In the following the invention shall be illustrated on the basis of preferred embodiments
with reference to the accompanying drawings, wherein:
- Fig. 1
- shows a schematic drawing of a driver assistance system in a motor vehicle; and
- Fig.2
- shows a schematic drawing of a risk assessment classifier in such a driver assistance
system.
[0020] The driver assistance system 10 is mounted in a motor vehicle and comprises an imaging
apparatus 11 for capturing images of a region surrounding the motor vehicle, for example
a region in front of the motor vehicle. Preferably the imaging apparatus 11 comprises
one or more optical imaging devices 12, in particular cameras, preferably operating
in the visible and/or infrared wavelength range, where infrared covers near IR with
wavelengths below 5 microns and/or far IR with wavelengths beyond 5 microns. In some
embodiments the imaging apparatus 11 comprises a plurality imaging devices 12 in particular
forming a stereo imaging apparatus 11. In other embodiments only one imaging device
12 forming a mono imaging apparatus 11 can be used.
[0021] The imaging apparatus 11 is coupled to a data processing device 14 adapted to process
the image data received from the imaging apparatus 11. The data processing device
14 is preferably a digital device which is programmed or programmable and preferably
comprises a microprocessor, microcontroller a digital signal processor (DSP), and/or
a microprocessor part in a System-On-Chip (SoC) device, and preferably has access
to, or comprises, a data memory 25. The data processing device 14 may comprise a dedicated
hardware device, like a Field Programmable Gate Array (FPGA) or an Application Specific
Integrated Circuit (ASIC), or an FPGA and/or ASIC part in a System-On-Chip (SoC) device,
for performing certain functions, for example controlling the capture of images by
the imaging apparatus 11, receiving the electrical signal containing the image information
from the imaging apparatus 11, rectifying or warping pairs of left/right images into
alignment and/or creating disparity or depth images. The data processing device 14,
or part of its functions, can be realized by a System-On-Chip (SoC) device comprising,
for example, FPGA, DSP, ARM and/or microprocessor functionality. The data processing
device 14 and the memory device 25 are preferably realised in an on-board electronic
control unit (ECU) and may be connected to the imaging apparatus 11 via a separate
cable or a vehicle data bus. In another embodiment the ECU and one or more of the
imaging devices 12 can be integrated into a single unit, where a one box solution
including the ECU and all imaging devices 12 can be preferred. All steps from imaging,
image processing to possible activation or control of driver assistance device 18
are performed automatically and continuously during driving in real time.
[0022] Image and data processing carried out in the processing device 14 advantageously
comprises identifying and preferably also classifying possible objects (object candidates)
in front of the motor vehicle, such as pedestrians, other vehicles, bicyclists and/or
large animals, tracking over time the position of objects or object candidates identified
in the captured images, and activating or controlling at least one driver assistance
device 18 depending on an estimation performed with respect to a tracked object, for
example on an estimated collision probability. The driver assistance device 18 may
in particular comprise a display device to display information relating to a detected
object. However, the invention is not limited to a display device. The driver assistance
device 18 may in addition or alternatively comprise a warning device adapted to provide
a collision warning to the driver by suitable optical, acoustical and/or haptic warning
signals; one or more restraint systems such as occupant airbags or safety belt tensioners,
pedestrian airbags, hood lifters and the like; and/or dynamic vehicle control systems
such as braking or steering control devices. Information about detected objects may
be stored in said data memory 25.
[0023] The processing device 14 has access, for example via a vehicle data bus, to data
from vehicle sensors 20 (21, 22, 23, ...) other than the imaging apparatus 12, which
are called other sensors 20 in the following for simplicity. The other sensors 20
comprise for example one or more acceleration sensors, a yaw sensor, a roll sensor,
a pitch sensor, a speed sensor, a braking sensor, a steering wheel angle sensor, a
radar apparatus, a LIDAR apparatus, a backward looking camera, a satellite navigation
receiver, a biometric sensor adapted to obtain biometric information of the driver,
ambient temperature sensor, ambient humidity sensor.
[0024] In the processing device 14 a risk assessment classifier 30 is realized, for example
by software, which is adapted to classify the traffic risk level, as defined above,
into one of a plurality of at least three, preferably at least five, more preferably
at least ten, for example 256 possible traffic risk level values ranging from (very)
low risk to (very) high risk. The risk assessment classifier 30 is shown in more detail
in Figure 2. The risk assessment classifier 30 comprises an artificial neural network
31 (deep neural network) depicted only schematically in Figure 2. The network 31 uses
as separate input sources every pixel of images 13 captured by the imaging apparatus
12 as well as other sensor data 36, i.e., sensor data 32, 33, 34, 35 originating from
the other sensors 20 in Figure 1 (vehicle sensors 21, 22, 23, ...).
[0025] In an initial procedure, which may be executed in the development of the driver assistance
system 10, the network 31 learns how to assess the traffic risk level of a traffic
situation. This is done by feeding reference data into the network 31, composed of
reference images 13 taken for example by driving a test car, and reference sensor
data 36 measured at the time of the corresponding reference images 13. All reference
images 13 have been assessed by a group of test persons in advance, who rate the traffic
risk level of the traffic situation shown in each reference image 13. The true traffic
risk level may be calculated as an average of all traffic risk levels estimated by
the test persons, and is fed to the network 31 as set data together with the corresponding
reference data.
[0026] Similarly to reference input fed to the network 31 during training, reference output
13 could be obtained by risk level annotations (fed into the network 31 as signal
40) performed by the marking team at the developer of the vision system 10.
[0027] After the network 31 has been trained how to assess the traffic risk level of numerous
traffic situations, representative of essentially all traffic situations occurring
in practice, the expert network 31 is implemented into cars for everyday usage. In
the car, the images 13 captured by the imaging apparatus 12 together with the corresponding
sensor data 36 measured at the time of capturing the images 13 are fed into the network
31. Based on all input sources, the network calculates online and outputs the traffic
risk level 40 belonging to the specific traffic situation shown in the image 13 under
consideration and corresponding sensor data 36.
[0028] The calculated traffic risk level can be used for different applications. In Figure
1, for example, a schematic traffic risk indicator 19 is shown with three indicator
LEDs, where a green LED indicates low risk, a yellow LED indicates middle risk and
a red LED indicates high risk. Of course, one multicolor LED can be used instead of
multiple single color LEDs. Also, more than three colors, for example up to 256 colors
can be used for indicating the traffic risk level in a more differentiated manner.
1. A driver assistance system (10) for a motor vehicle, comprising an imaging apparatus
(12) adapted to capture images (13) of the surrounding of a motor vehicle, and a processing
device 14) adapted to perform image processing on the captured images and to control
a driver assistance device (18, 19) based on the result of said image processing,
wherein said processing device (14) comprises a risk assessment estimator (30) adapted
to estimate a traffic risk level (40), characterized in that said traffic risk level (40) is estimated by said risk assessment estimator (30)
on the basis of input source data (13, 36) from at least one input source (12, 20),
wherein said input source data comprises image data (13) captured by said imaging
apparatus (12).
2. The driver assistance system as claimed in claim 1, characterized in that the appearance of the scene surrounding the motor vehicle extracted from captured
images (13) is used by said risk assessment estimator (30).
3. The driver assistance system as claimed in any one of the preceding claims, characterized in that said risk assessment estimator (30) considers the whole image (13) in the determination
of said traffic risk level (40).
4. The driver assistance system as claimed in any one of the preceding claims, characterized in that said risk assessment estimator (30) considers each pixel of a captured image (13)
as a separate input source in the determination of said traffic risk level (40).
5. The driver assistance system as claimed in any one of the preceding claims, characterized in that said risk assessment estimator (30) comprises a learning network (31), in particular
an artificial neural network.
6. The driver assistance system as claimed in any one of the preceding claims, characterized in that said learning network (31) is trained using training data rated by a plurality of
persons with respect the estimated traffic risk level.
7. The driver assistance system as claimed in any one of the preceding claims, characterized in that input source data (36) comprises ego vehicle data, in particular ego vehicle dynamics
data.
8. The driver assistance system as claimed in any one of the preceding claims, characterized in that said at least one input source (20) comprises one or more of an acceleration sensor,
yaw sensor, roll sensor, pitch sensor, speed sensor, braking sensor, steering wheel
angle sensor.
9. The driver assistance system as claimed in any one of the preceding claims, characterized in that said input source data (36) comprises data of other objects in the environment of
the motor vehicle, in particular other objects dynamics data.
10. The driver assistance system as claimed in any one of the preceding claims, characterized in that said driver assistance system (10) comprises a data memory (25) for storing information
about detected other objects.
11. The driver assistance system as claimed any one of the preceding claims, characterized in that said at least one input source (20) comprises one or more of said imaging apparatus,
a radar apparatus, a LIDAR apparatus, a backward looking camera.
12. The driver assistance system as claimed in any one of the preceding claims,
characterized in that input source data (13, 36) comprises one or more of:
- speed limits obtained by sign recognition or a satellite navigation receiver;
- biometric information;
- road conditions;
- ambient conditions, like ambient temperature or ambient humidity;
- stored data of previous hazard areas or conditions;
- location of crosswalks obtained by image recognition or a satellite navigation receiver.
13. The driver assistance system as claimed in any one of the preceding claims, characterized in that the driver assistance system (10) comprises a risk level indicator (19) adapted to
indicate an estimated risk level (40) to the driver.
14. The driver assistance system as claimed in any one of the preceding claims, characterized in that the estimated risk level (40) is used for requesting safety-relevant actions by the
driver depending on said estimated risk level.
15. A driver assistance method for a motor vehicle, comprising an imaging apparatus (12)
adapted to capture images (13) of the surrounding of a motor vehicle, and a processing
device (14) adapted to perform image processing on the captured images (13) and to
control a driver assistance device (18, 19) based on the result of said image processing,
wherein said processing device (14) comprises a risk assessment estimator (30) adapted
to estimate a traffic risk level (40), characterized in that said traffic risk level (40) is estimated by said risk assessment estimator (30)
on the basis of input source data (13, 36) from at least one input source (12, 20),
wherein said input source data comprises image data (13) captured by said imaging
apparatus (12).