Method of generating a training image for an automated vehicle object recognition system
In a method of generating a training image for teaching of a camera-based object recognition system suitable for use on an automated vehicle which shows an object to be recognized in a natural object environment, the training image is generated as a synthetic image by a combination of a base image taken by a camera and of a template image in that a structural feature is obtained from the base image and is replaced with a structural feature obtained from the template image by means of a shift-map algorithm.
1. A method of teaching a camera-based object recognition system suitable for use on an automated vehicle, said method comprising:
using a camera to provide a base image of an object in a natural object environment, wherein the base image includes an initial structural feature;
providing a template image that corresponds to the object, wherein the template image defines a substitute structural feature comparable to the initial structural feature; and
generating a training image for teaching the system, wherein the training image is generated by replacing the initial structural feature with the substitute structural feature by using a shift-map algorithm, wherein the replacement of the initial structural feature is only carried out by rearrangement of pixels of the base image, wherein the base image and the template image are compared with one another to determine a distance dimension indicating the similarity of the images, the distance dimension is used as an optimization criterion for the shift-map algorithm, and both the base image and the template image are transformed into a canonical-reference frame before the base image and the template image are compared using the distance dimension.
2. The method in accordance with claim 1 , wherein the rearrangement of pixels of the base image is restricted to a portion of the base image.
3. The method in accordance with claim 1 , wherein the rearrangement of pixels of the base image is restricted to a central-portion of the base image.
4. The method in accordance with claim 1 , wherein optimization criteria are made use of for the shift-map algorithm which comprise one of a maintenance of proximity relationships of pixels of the base image, an avoidance of tonal value discontinuities in the generated training image, and a maintenance of a similarity between the base image and the template image.
5. The method in accordance with claim 1 , wherein the substitute structural feature comprises a texture and/or a pattern.
6. The method in accordance with claim 1 , wherein the template image is an image taken by a camera.
7. The method in accordance with claim 1 , wherein the template image is a graphically generated image.
8. The method in accordance with claim 1 , wherein the base image and the training image each show a road sign in a road environment.
9. A method of generating a training image for teaching of a camera-based object recognition system that is suitable for use on an automated vehicle and which shows an object to be recognized in a natural object environment, wherein the training image is generated as a synthetic image by combination of a base image taken by a camera and of a template image in that an initial structural feature is removed from the base image and is replaced with a substitute structural feature obtained from the template image by means of a shift-map algorithm, wherein the replacement of the initial structural feature is only carried out by rearrangement of pixels of the base image, wherein the base image and the template image are compared with one another to determine a distance dimension indicating the similarity of the images, the distance dimension is used as an optimization criterion for the shift-map algorithm, and both the base image and the template image are transformed into a canonical-reference frame before the base image and the template image are compared using the distance dimension.