Method and device for classifying an object for a vehicle

1. A method for classifying an object for a vehicle, the method comprising:

reading in, using a camera situated on the vehicle, a camera image of an object external to the vehicle;

detecting the object including classifying, using a device including hardware situated in the vehicle, the object from the camera image using a generic classifier, the generic classifier being a trained classifier including a learned object model which was learned in off-line training;

checking plausibility of a result of the detecting by comparing parameterized models and/or templates with image data;

extracting, using the device, at least one specific object feature assigned to the object from the camera image;

creating, using the device, a specific classifier using the specific object feature; and

on-line training the specific classifier.

2. The method of claim 1 , wherein in the classifying, an additional object from an additional camera image, read in using the camera, is classified, using the device and using the generic classifier and the specific classifier, in the extracting, at least one specific object feature assigned to the object is extracted, using the device, from the additional camera image read in using the camera, and, in the creating, the specific classifier is adapted, by the device, using the additional specific object feature.

3. The method of claim 1 , wherein in the creating, the specific classifier is stored in a ring memory unit.

4. The method of claim 1 , further comprising:

ascertaining the camera image using a surroundings detector.

5. The method of claim 1 , further comprising:

storing the camera image if, in the classifying, a classification inaccuracy with respect to the classification of the object is below a threshold value.

6. The method of claim 5 , wherein in the storing, the camera image is sent to a training unit situated externally to the vehicle.

7. The method as recited in claim 1 , wherein the camera is a camera of a surroundings detector of the vehicle.

8. The method as recited in claim 1 , wherein the plausibility checking includes creating a 3D-wireframe model of the object.

9. A device for classifying an object for a vehicle, comprising:

a processor device including hardware and situated in the vehicle, the processor device configured to perform the following:

reading in, using a camera situated on the vehicle, a camera image of an object external to the vehicle;

detecting the object including classifying the object from the camera image using a generic classifier, the generic classifier being a trained classifier including a learned object model which was learned in off-line training;

checking plausibility of a result of the detecting by comparing parameterized models and/or templates with image data;

extracting at least one specific object feature assigned to the object from the camera image;

creating a specific classifier using the specific object feature; and

online-training the specific classifier wherein the at least one specific object feature is a color of the object.

10. The device as recited in claim 9 , wherein the learned object model is a learned vehicle model, and the object external to the vehicle is another vehicle.

11. The device as recited in claim 9 , wherein in the classifying, an additional object from an additional camera image, read in using the camera, is classified by the device using the generic classifier and the specific classifier, in the extracting, at least one specific object feature assigned to the object is extracted, by the device, from the additional camera image read in using the camera and, in the creating, the specific classifier is adapted, by the device, using the additional specific object feature.

12. The device as recited in claim 9 , wherein the plausibility checking includes creating a 3D-wireframe model of the object.

13. A non-transitory computer readable medium having a computer program, which is executable by a processor, comprising:

a program code arrangement having program code for classifying an object for a vehicle, by performing, when executed by the processor, the following:

reading in, using a camera situated on a vehicle, a camera image of an object external to the vehicle;

detecting the object including classifying, using a device including hardware situated in the vehicle, the object from the camera image using a generic classifier, the generic classifier being a trained classifier including a learned object model which was learned in off-line training;

checking plausibility of a result of the detecting by comparing parameterized models and/or templates with image data;

extracting, using the device, at least one specific object feature assigned to the object from the camera image; and

creating, using the device, a specific classifier using the specific object feature; and

online-training the specific classifier.

14. The non-transitory computer readable medium of claim 13 , wherein in the classifying, an additional object from an additional camera image, read in using the camera, is classified using the device using the generic classifier and the specific classifier, in the extracting, at least one specific object feature assigned to the object is extracted, using the device, from the additional camera image read in using the camera and, in the creating, the specific classifier is adapted, by the device, using the additional specific object feature.

15. The method as recited in claim 1 , wherein the learned object model is a learned vehicle model, and the object external to the vehicle is another vehicle.

16. The non-transitory computer readable medium as recited in claim 13 , wherein the learned object model is a learned vehicle model, and the object external to the vehicle is another vehicle.

17. The non-transitory computer readable medium as recited in claim 13 , wherein the plausibility checking includes creating a 3D-wireframe model of the object.