Converting input image data from a plurality of vehicle cameras of a surround-view system into optimised output image data
View Patent ↗The present disclosure relates to a machine learning method, to a method and to a device for converting input image data from a plurality of vehicle cameras of a panoramic-view system into optimized or enhanced output image data. The method for converting input image data from a plurality of vehicle cameras of a panoramic-view system into optimized or enhanced output image data includes input image data acquired by the vehicle cameras and having a current brightness or color distribution provided to a trained artificial neural network. the trained artificial neural network is configured to convert the input image data having the current brightness or color distribution into optimized or enhanced output image data having different output brightness or color distribution; and the trained artificial neural network is configured to output the output image data.
1. A method for machine learning of a conversion of input image data from a plurality of vehicle cameras of a panoramic view system into enhanced output image data by at least one artificial neural network, wherein the machine learning is effected with a multiplicity of training image pairs in such a manner that a first image having a first brightness or color distribution is provided at the input of the at least one artificial neural network and a second image of the same scene having a different second brightness or color distribution, respectively, is provided as a target output image, wherein the first brightness or color distribution comprises the first brightness distribution and the different second brightness or color distribution comprises the different second brightness distribution, and at least one factor is established as a measure of a difference between the t second brightness distribution and the first brightness distribution of a training image pair and is provided to the at least one artificial neural network.
2. The method according to claim 1 , wherein the training image pairs are produced for each of the plurality of vehicle cameras in that the first image having a first brightness and the second image having a second brightness are acquired simultaneously or immediately one after another with different exposure times.
3. The method according to claim 1 , wherein the at least one artificial neural network is trained jointly for all of the plurality of vehicle cameras.
4. The method according to claim 1 , wherein the at least one artificial neural network has a common input interface and two separate output interfaces, wherein the common input interface has shared feature representation layers, wherein brightness-converted image data are output at the first output interface, wherein ADAS-relevant detections of at least one ADAS detection function are output at the second output interface and wherein the outputs of both output interfaces are enhanced as part of the training.
5. A method for converting input image data from a plurality of vehicle cameras of a panoramic-view system into enhanced output image data, the method comprising:
a) input image data acquired by the plurality of the vehicle cameras and having a current brightness or color distribution are provided to a trained artificial neural network;
b) the trained artificial neural network is configured to convert the input image data having the current brightness or color distribution into enhanced output image data having different output brightness or color distribution, respectively; and
c) the trained artificial neural network is configured to output the enhanced output image data
wherein a factor is additionally provided to the trained artificial neural network during the input image data acquisition in block a) and the conversion is controlled as a function of the factor in the conversion of the input image data in block b).
6. The method according to claim 5 , wherein the factor is estimated and, during the estimation, the current brightness or color distribution of the input image data, previously captured image data and/or a history of the factor is/are taken into account.
7. The method according to claim 5 , wherein a separate factor is estimated or determined for each of the plurality of the vehicle cameras.
8. A non-transitory memory storing a computer program, the computer program, when executed by a data processing unit having at least one processor, configures the data processing unit to carry out the method according to claim 5 .
9. The method according to claim 5 , wherein the trained artificial neural network has a common input interface and two separate output interfaces, wherein the common input interface has shared feature representation layers, wherein brightness-converted image data are output at the first output interface, and wherein ADAS-relevant detections of at least one ADAS detection function are output at the second output interface.
10. The method according to claim 5 , wherein the trained artificial neural network is part of an on-board ADAS detection neural network having a shared input interface and two separate output interfaces, wherein a first output interface of the two separate output interfaces is configured to output the enhanced output image data and a second output interface of the two separate output interfaces is configured to output ADAS-relevant detections.
11. A device having at least one data processing unit comprising at least one processor, the data processing unit configured to convert input image data from a plurality of vehicle cameras of a panoramic-view system into enhanced output image data, comprising:
an input interface which is configured to receive the input image data having a current brightness or color distribution from the plurality of the vehicle cameras;
a trained artificial neural network which is configured to convert the input image data having the current brightness or color distribution into output image data having a different output brightness or color distribution;
wherein the trained artificial neural network is part of an on-board ADAS detection neural network having a shared input interface and two separate output interfaces, wherein a first output interface of the two separate output interfaces is configured to output the enhanced output image data, and wherein a second output interface of the two separate output interfaces is configured to output ADAS-relevant detections.
12. The device according to claim 11 , wherein the data processing unit is implemented in a hardware-based image pre-processing stage.
13. The device according to claim 11 , wherein the shared input interface has shared feature representation layers, wherein the brightness-converted or color-converted image data are output at the first output interface, and wherein the ADAS-relevant detections are of at least one ADAS detection function and are output at the second output interface.
14. The device according to claim 11 , wherein a factor is additionally provided to the trained artificial neural network and the conversion is controlled as a function of the factor in the conversion of the input image data.
15. The device according to claim 14 , wherein the factor is estimated and, during the estimation, the brightness or color distribution of current image data, previously captured image data and/or a history of the factor is/are taken into account.
16. The device according to claim 14 , wherein a separate factor is estimated or determined for each of the plurality of the vehicle cameras.