Method for the classification of parking spaces in a surrounding region of a vehicle with a neural network
The invention relates to a method for the classification of parking spaces in a surrounding region of a vehicle with a driver support system, wherein the vehicle comprises at least one first surroundings sensor and a second surroundings sensor, comprising the steps of receiving first sensor data out of the surrounding region by the driver support system from the at least one first surroundings sensor, recognizing a parking-space-like partial region of the surrounding region in the first sensor data, requesting second sensor data acquired by the at least one second surroundings sensor out of the parking-space-like partial region by the driver support system as soon as the parking-space-like partial region is recognized in the first sensor data, transmitting the requested second sensor data to a vehicle-side computing unit comprising a deep neural network (DNN), and classifying the parking-space-like partial region into categories with the DNN, wherein the categories comprise legal, parkable parking spaces and illegal, non-parkable parking spaces. The invention also relates to a driver support system, in particular a parking assistance system, for a vehicle for the acquisition of parking spaces. The invention further relates to a vehicle with a driver support system.
1. A method for the classification of parking spaces in a surrounding region of a vehicle with a driver support system, wherein the vehicle comprises at least one first surroundings sensor and one second surroundings sensor, the method comprising:
receiving first sensor data out of the surrounding region by the driver support system from the at least one first surroundings sensor;
recognizing a parking-space-like partial region of the surrounding region in the first sensor data;
requesting second sensor data acquired by the at least one second surroundings sensor out of the parking-space-like partial region by the driver support system as soon as the parking-space-like partial region is recognized in the first sensor data;
transmitting the requested second sensor data to a vehicle-side computing unit comprising a deep neural network (DNN); and
classifying the parking-space-like partial region into categories with the DNN, wherein the categories comprise legal, parkable parking spaces and illegal, non-parkable parking spaces,
wherein the deep neural network (DNN) is pre-trained for the classification of parking spaces, and wherein the pre-training of the deep neural network (DNN) is done through vehicle-side recording of second sensor data comprising image data.
2. The method according to claim 1 , wherein receiving first sensor data with the at least one first surroundings sensor comprises a reception of distance sensor data of at least one vehicle-side distance sensor, wherein distance sensor data comprises a reception of ultrasonic sensor data, radar sensor data, or laser scanner data, by the driver support system.
3. The method according to claim 1 , wherein the reception of the second sensor data comprises a reception of image data of one or a plurality of images and/or video sequences with at least one camera system comprising one or a plurality of cameras by the driver support system.
4. The method according to claim 1 , wherein as soon as the parking-space-like partial region has been acquired by the driver support system on the basis of the first sensor data, a trigger signal is generated for the at least one second surroundings sensor, through which the second surroundings sensor is made to acquire the second sensor data from the surrounding region comprising the parking-space-like partial region.
5. The method according to claim 1 , wherein at least one geometrical dimension, determined on the basis of the first sensor data, of the parking-space-like partial region is transmitted to the vehicle-side computing unit comprising at least one item of information about the parking-space-like partial region, together with the requested second sensor data.
6. The method according to claim 1 , wherein the result of the classification of the parking space into categories by the deep neural network (DNN) is transmitted to the driver support system, wherein the legal, parkable parking spaces comprise parallel parking spaces and bay parking spaces.
7. The method according to claim 1 , wherein the classification of the parking-space-like partial region is performed with the deep neural network (DNN) that has a convolutional neural network (CNN).
8. The method according to claim 1 , wherein the classification of the parking-space-like partial region into parkable parking spaces and illegal, non-parkable parking spaces comprises a classification of the parkable parking spaces into parallel parking spaces and bay parking spaces.
9. A driver support system comprising a parking assistance system for a vehicle for the acquisition of parking spaces in a surrounding region of the vehicle, comprising at least one first surroundings sensor for the acquisition of first sensor data from the surrounding region, at least one second surroundings sensor different from the first surroundings sensor for the acquisition of second sensor data from the surrounding region, and a vehicle-side computing unit comprising a deep neural network (DNN) for the classification of a parking-space-like partial region into categories,
wherein the deep neural network (DNN) is pre-trained for the classification of parking spaces, while the pre-training of the deep neural network (DNN) is done through vehicle-side recording of second sensor data comprising image data.
10. The driver support system according to claim 9 , wherein the at least one first surroundings sensor comprises at least one distance sensor that is chosen from ultrasonic sensors, radar sensors and laser scanners, and the at least one second surroundings sensor comprises at least one camera system that preferably comprises one or a plurality of cameras.
11. The driver support system according to claim 9 , wherein the deep neural network (DNN) comprises a convolutional neural network (CNN) that is a binary convolutional neural network (CNN).
12. A vehicle with the driver support system comprising the parking assistance system, according to claim 9 .
13. A non-transitory machine-readable medium comprising a plurality of machine-readable instructions executed by one or more processors, the plurality of machine-readable instructions causing the one or more processors to perform operations for a classification of parking spaces in a surrounding region of a vehicle with a driver support system, wherein the vehicle comprises at least one first surroundings sensor and one second surroundings sensor, the operations comprising:
receiving first sensor data out of the surrounding region by the driver support system from the at least one first surroundings sensor;
recognizing a parking-space-like partial region of the surrounding region in the first sensor data;
requesting second sensor data acquired by the at least one second surroundings sensor out of the parking-space-like partial region by the driver support system as soon as the parking-space-like partial region is recognized in the first sensor data;
transmitting the requested second sensor data to a vehicle-side computing unit comprising a deep neural network (DNN); and
classifying the parking-space-like partial region into categories with the DNN, wherein the categories comprise legal, parkable parking spaces and illegal, non-parkable parking spaces,
wherein the deep neural network (DNN) is pre-trained for the classification of parking spaces, while the pre-training of the deep neural network (DNN) is done through vehicle-side recording of second sensor data comprising image data.