Method for generating a monitoring image

1. A method for generating a monitoring image, comprising the following steps:

providing an image sequence of surroundings to be monitored using an imaging system;

determining at least one monitoring area and at least one periphery area of at least one image of the image sequence using a learning-based semantic segmentation method;

compressing the monitoring area of the at least one image of the image sequence with a first compression quality; and

compressing the periphery area of the at least one image of the image sequence with a second compression quality to generate the monitoring image, the second compression quality being lower than the first compression quality, wherein a mask for a particular image is generated using the monitoring area and/or periphery area of the particular image, and the monitoring image is generated based on an averaged multitude of masks of a corresponding multitude of respective images of the image sequence.

2. The method as recited in claim 1 , wherein the learning-based semantic segmentation method is carried out based on a trained neural convolutional network and/or an aggregate channel features method and/or an AdaBoost method.

3. The method as recited in claim 1 , wherein the surroundings to be monitored are a vehicle interior.

4. The method as recited in claim 1 , wherein the monitoring image is formed using the compressed monitoring area and the compressed periphery area.

5. The method as recited in claim 1 , wherein the periphery area of the at least one image of the image sequence is compressed with the second compression quality in that the at least one periphery area is softened and/or assigned predefined pixel values for image pixels of the periphery area before the periphery area is compressed with the first compression quality using a compression method.

6. The method as recited in claim 1 , wherein the periphery area of the at least one image of the image sequence is compressed with the second compression quality in that the at least one periphery area is replaced with a predefined image content.

7. The method as recited in claim 1 , wherein at least one monitoring area and at least one periphery area are determined from a plurality of images of the image sequence using the learning-based semantic segmentation method.

8. The method as recited in claim 1 , wherein image areas of the at least one image are segmented into a plurality of object classes using the learning-based semantic segmentation method, and the monitoring area and/or the periphery area is/are determined by assigning the image areas to the at least one monitoring area based on a first group of object classes, and by assigning image areas to the at least one periphery area based on a second group of object classes.

9. A method for training a learning-based semantic segmentation method for determining at least one monitoring area and at least one periphery area of at least one image of an image sequence, including a plurality of training cycles and a plurality of reference images of surroundings to be monitored using an imaging system, a particular reference image of the plurality of references images being assigned to a particular training cycle, and each training cycle including the following steps:

providing the particular reference image, at least one monitoring area and at least one periphery area being labeled in the particular reference image; and

adapting the learning-based semantic segmentation method to minimize a deviation from the particular labeled reference image during the determination of at least one monitoring area and at least one periphery area of the particular reference image using the learning-based semantic segmentation method, wherein a mask for a particular image is generated using the monitoring area and/or periphery area of the particular image, and the monitoring image is generated based on an averaged multitude of masks of a corresponding multitude of respective images of the image sequence.

10. The method as recited in claim 1 , wherein: (i) based on the generated monitoring image, a control signal for controlling an at least semi-automated vehicle is provided, and/or (ii) based on the generated monitoring image, a warning signal for warning a vehicle occupant is provided, and/or (iii) based on the generated monitoring image, an emergency signal for alerting a private and/or public security service and/or rescue service is provided.

11. A monitoring device configured to generate a monitoring image, the monitoring device configured to:

provide an image sequence of surroundings to be monitored using an imaging system;

determine at least one monitoring area and at least one periphery area of at least one image of the image sequence using a learning-based semantic segmentation method;

compress the monitoring area of the at least one image of the image sequence with a first compression quality; and

compress the periphery area of the at least one image of the image sequence with a second compression quality to generate the monitoring image, the second compression quality being lower than the first compression quality, wherein a mask for a particular image is generated using the monitoring area and/or periphery area of the particular image, and the monitoring image is generated based on an averaged multitude of masks of a corresponding multitude of respective images of the image sequence.

12. A non-transitory machine-readable memory medium on which is stored a computer program for generating a monitoring image, the computer program, when executed by a computer, causing the computer to perform the following steps:

providing an image sequence of surroundings to be monitored using an imaging system;

determining at least one monitoring area and at least one periphery area of at least one image of the image sequence using a learning-based semantic segmentation method;

compressing the monitoring area of the at least one image of the image sequence with a first compression quality; and

compressing the periphery area of the at least one image of the image sequence with a second compression quality to generate the monitoring image, the second compression quality being lower than the first compression quality, wherein a mask for a particular image is generated using the monitoring area and/or periphery area of the particular image, and the monitoring image is generated based on an averaged multitude of masks of a corresponding multitude of respective images of the image sequence.