Behavior-based learning of visual characteristics from real-world traffic scenes for driver assistance systems
A computer-implemented method for setting-up and operating a driver assistance system, the method comprising the steps of: visually sensing the environment of a vehicle, sensing at least one rate of change of a parameter representing a status of the car, which parameter can be influenced by a driver manipulating a man/machine interface of the vehicle, structuring information gathered by the visual sensing by classifying the at least one sensed rate of change in different categories, in order to find visual features associated to a certain category of sensed rate of change and thus associated with a certain behavior of the driver.
1. A computer-implemented method for setting-up and operating a driver assistance system,
the method comprising the steps of:
visually sensing an environment of a vehicle,
sensing at least one rate of change of a parameter representing a status of the vehicle, which the parameter can be influenced by a driver manipulating a man/machine interface of the vehicle,
structuring information gathered by the visual sensing by classifying the at least one sensed rate of change in different categories, in order to find visual features associated to a certain category of sensed rate of change and thus associated with a certain behavior of the driver,
characterized in that
the step of structuring comprises splitting the visually sensed data stream in at least two categories of the at least one sensed rate of change,
preparing a histogram indicating how frequent a visual feature out of a set of different visual features occurred while the sensed rate of change indicated an occurrence of a category of a driver's behavior, wherein
the set of different visual features is gained by comparing the split visually sensed data streams for the categories exceeding a threshold value with a data stream portion of time periods during which the threshold values were not reached or with another data stream or data stream portion.
2. The method according to claim 1 , wherein, as a function of the structured information, a visual or acoustical signal is generated, or an effector of the vehicle is operated.
3. The method according to claim 1 , wherein the
at least one sensed rate of change comprises, “turning left”, “turning right” or “accelerating”.
4. The method according to claim 1 , wherein one additional category is provided for portions of the visually sensed data stream in time period during which no sensed rate of change was present or did not exceeded a predefined threshold value.
5. A computing unit, designed to carry out a method according to claim 1 .
6. A driver assistance system, equipped with a computing unit according to claim 5 , furthermore comprising visual or acoustical indicating means, and/or an interface for supplying control commands to at least one effector of the vehicle.
7. A computer software program product, performing a method according to claim 1 when run on a computing device.
8. A land, sea or air vehicle, comprising a driver assistance system according to claim 6 .
9. An autonomous robot, comprising a driver assistance system according to claim 6 .
10. The method according to claim 2 , wherein the effector of the vehicle comprises at least one of a steering wheel, an accelerator, or a brake.
11. The method according to claim 3 , wherein the sensed rate of change exceeds a predefined threshold value.