Method for determining body weight and / or seating position, control unit and motor vehicle

Abstract

Method for determining the body weight and / or seating position of a vehicle occupant (F) of a motor vehicle (10) comprising the following steps: - Obtaining an image (B) of the vehicle occupant (F) from a camera (20) of the motor vehicle (10), - Determining at least one physical characteristic (M) of the vehicle occupant (F) based on the image (B) by a feature module (24) of the motor vehicle (10), and - Determining body weight and / or sitting position by a classical machine learning module (26) based on the at least one determined body feature (M), wherein the classical machine learning module (26) uses a symbolic approach to machine learning, characterized in that the feature module (24) uses a classical machine learning method, in particular with a symbolic approach, when determining the at least one body feature (M).

Description

[0001] The invention relates to a method for determining the body weight and / or the seating position of a vehicle occupant of a motor vehicle, a control device for a motor vehicle and a motor vehicle.

[0002] Modern motor vehicles, such as cars and trucks, are equipped with safety systems, such as seat belt systems and airbag modules, whose protective effect can be optimized if they are appropriately tailored to the occupants of the respective seats. The weight and / or seating position of the occupants are particularly important in this regard.

[0003] However, control units of motor vehicles, especially for safety-relevant systems, should be designed to be as resource-efficient and robust, i.e., as low-complexity, as possible. Therefore, the object of the invention is to provide a method for determining the body weight and / or seating position of a vehicle occupant that can be implemented in a robust and resource-efficient manner.

[0004] The problem is solved by a method for determining the body weight and / or the seating position of a vehicle occupant of a motor vehicle, comprising the following steps: - Obtaining an image of the vehicle occupant from a vehicle camera, - Determining at least one body feature of the vehicle occupant based on the image by a feature module of the vehicle, and - Determining the body weight and / or sitting position using a classic machine learning module based on at least one determined body characteristic.

[0005] By using a classic machine learning module, a particularly robust, reliable and resource-saving determination of the body weight and / or seating position of the vehicle occupant is possible.

[0006] In the context of this invention, a classic machine learning module is understood to be a machine learning module that is not or does not comprise an artificial neural network.

[0007] For example, several different physical characteristics are identified and taken into account.

[0008] In one embodiment, a safety system, in particular a safety belt system and / or a restraint system, is controlled as a function of the determined body weight and / or the determined seating position in order to be able to optimally adapt the safety system to the vehicle occupant.

[0009] For example, the classic machine learning module uses a symbolic approach to machine learning, which determines body weight and / or sitting position in a resource-efficient but reliable manner.

[0010] In one embodiment, the machine learning module uses a classification method, in particular a random forest, to determine the body weight and / or sitting position. This allows the body weight and / or sitting position to be determined quickly.

[0011] Alternatively or additionally, the machine learning module can use a regression method, in particular a support vector machine, when determining the body weight and / or the sitting position, which enables a particularly reliable determination.

[0012] For example, the image is pre-processed, in particular by segmenting the body of the vehicle occupant, so that the determination of body weight and / or seating position can be further simplified.

[0013] For robust recognition of body parts, the feature module can determine at least one body feature using a machine vision method, in particular using the Canny algorithm. Camera parameters, such as calibration parameters, can be taken into account.

[0014] In order to provide a particularly resource-saving method, the feature module can use a classic machine learning method, in particular with a symbolic approach, when determining the at least one body feature.

[0015] For example, a random forest, i.e. a classification method consisting of several uncorrelated decision trees, can be used.

[0016] In one embodiment of the method, the feature module uses an artificial neural network when determining the at least one body feature, whereby a plurality of different body features can be distinguished.

[0017] For example, the feature module has an artificial neural network.

[0018] For example, the at least one body characteristic is the height of the body, in particular the upper body, the width of the body, in particular the upper body, the arm length of at least one arm, the abdominal circumference, the orientation of the upper body, the orientation of the head and / or the orientation of at least one arm, whereby the body weight and / or the sitting position can be precisely determined.

[0019] The orientation is determined, for example, relative to other parts of the body or to a part fixed to the vehicle, such as the seat or the B-pillar.

[0020] In order to further increase the effectiveness of the method, the at least one body characteristic used to determine the body weight and / or the sitting position can be selected by means of at least one data mining method.

[0021] In one embodiment of the invention, the image is a picture of the vehicle occupant sitting on a corresponding seat of the motor vehicle, whereby the body characteristics are determined in the relevant situation, in particular in the position relevant for the safety system.

[0022] Furthermore, the object is achieved by a control unit for a motor vehicle which is designed to carry out one of the methods described above.

[0023] The task is also solved by a motor vehicle with a camera and a previously described control unit.

[0024] The features and advantages explained for the method apply accordingly to the control unit and the motor vehicle and vice versa.

[0025] Further features and advantages of the invention will become apparent from the following description and the accompanying drawings, to which reference is made. In the drawings: - Fig. 1 schematically shows a motor vehicle according to the invention with a control unit according to the invention, - Fig. 2 shows a very schematic image of a vehicle occupant, and - Fig. 3 a flow diagram of a method according to the invention.

[0026] Fig. Figure 1 schematically shows a motor vehicle 10, in this case a passenger car. It is of course also conceivable that the motor vehicle 10 is a truck.

[0027] The motor vehicle 10 has a seat 12 on which a vehicle occupant F, here the driver of the motor vehicle 10, is seated.

[0028] Furthermore, the motor vehicle 10 has a steering wheel 14 in which a restraint system 16, e.g. a gas bag module, is arranged, a safety belt system 18 ( Fig. 3), a camera 20 directed towards the seat 12, and a control unit 22.

[0029] The control unit 22 is connected to the restraint system 16, the safety belt system 18 and the camera 20 for data exchange and controls in particular the restraint system 16, the safety belt system 18 and the camera 20.

[0030] The control unit 22 also has a feature module 24 and a classic machine learning module 26.

[0031] In Fig. 2 schematically shows an image B of the seat 12 together with the vehicle occupant F, which was taken by means of the camera 20 and which is used by the control unit 22 to determine the body weight and / or the seating position of the vehicle occupant F.

[0032] For this purpose, the control unit 22 carries out the Fig. 3 described procedures.

[0033] In a first step S1, the camera 20 takes an image B of the vehicle occupant F, as shown, for example, in Fig. 2 is shown.

[0034] In the next step S2, the camera 20 transmits the image B to the control unit 22, which accordingly receives the image B.

[0035] Image B is preprocessed by control unit 22 in a next step S3. This segments the body of vehicle occupant F. Other objects fixed to the vehicle can also be detected and segmented, such as seat 12 or the B-pillar 28 of motor vehicle 10.

[0036] Already in this step, a Canny algorithm (Canny edge detector) can be used for edge detection.

[0037] Parameters of camera 20, in particular calibration parameters, can also be taken into account.

[0038] In a next step S4, the preprocessed image is passed on to the feature module 24 and processed there.

[0039] The feature module 24 determines at least one body feature M of the vehicle occupant F based on the preprocessed image B.

[0040] A body characteristic M can be, for example, the following: the height of the upper body M1, the width of the upper body M2, the arm length M3, the abdominal circumference M4 and / or the orientations of the various body parts to one another or to a vehicle-fixed part, such as the seat 12 or the B-pillar 28.

[0041] For example, the orientation of the upper body, the orientation of the head and / or the orientation of the arms are determined as body feature M.

[0042] Either the length and / or orientation of one of the arms, both arms separately or both arms together can be determined.

[0043] The feature module 24 determines those body features M which have been predetermined and selected beforehand, for example at the factory.

[0044] The selection of which body characteristics M are suitable for determining body weight and / or sitting position and should be used was determined in a preceding step S0.

[0045] The determination is made, for example, using a data mining procedure, manually or in another suitable way.

[0046] For example, the feature module 24 uses a classic machine learning method with a symbolic approach to determine the body features M, rather than an artificial neural network. For example, a random forest is used.

[0047] In addition, the feature module 24 may perform a machine vision method, e.g., using a Canny algorithm and / or at least one camera parameter, such as a calibration parameter of the camera.

[0048] Alternatively or in addition to the classic machine learning method, it is also conceivable for the feature module 24 to use an artificial neural network to determine the at least one body feature M. For this purpose, the feature module 24 then comprises, for example, an artificial neural network 30.

[0049] The determined body characteristics M are then transmitted to the classic machine learning module 26 of the control unit 22 in a next step S5.

[0050] In step S6, the classical machine learning module 26 determines the weight and / or seating position of the vehicle occupant F based on the determined body characteristics M.

[0051] The classic machine learning module 26 does not use an artificial neural network, but rather a symbolic approach to machine learning.

[0052] For example, a classification method, in particular a random forest, is used to determine the body weight and / or seating position of the vehicle occupant F based on the body characteristics M.

[0053] Alternatively or additionally, the machine learning module 26 can use a regression method, for example a support vector machine, to determine the weight and / or position of the vehicle occupant F.

[0054] Regardless of whether one, both, or a combination of these methods is used, the classical machine learning module 26 receives the body weight and / or seating position of the vehicle occupant F.

[0055] In a next step S7, the control unit 22 controls the safety belt system 18, the restraint system 16 and / or other restraint systems of the motor vehicle 10 on the basis of the determined weight and / or the determined seating position.

[0056] For example, in the case of a heavy vehicle occupant, a belt tensioner of the safety belt system 18 is given a command to tighten the seat belt more.

[0057] The restraint system 16 can be controlled such that different chambers of the gas bag are inflated or not inflated in the restraint case, depending on whether the vehicle occupant is heavy or less heavy or whether the seating position is upright or more reclining.

[0058] Due to the use of a classic machine learning module, the resources required to determine the body weight and / or seat position of the control unit 22 are reduced. Furthermore, the effort required to train the classic machine learning module 26 is significantly reduced. Reference symbol 10 motor vehicle 12 seats 14 Steering wheel 16 Restraint system 18 Safety belt system 20 Camera 22 Control unit 24 Feature module 26 classic machine learning module 28 B-pillar 30 artificial neural network B Image F vehicle occupant M physical feature

Claims

[1] Method for determining the body weight and / or the seating position of a vehicle occupant (F) of a motor vehicle (10) with the following steps: - Obtaining an image (B) of the vehicle occupant (F) from a camera (20) of the motor vehicle (10), - Determining at least one body feature (M) of the vehicle occupant (F) based on the image (B) by a feature module (24) of the motor vehicle (10), and - determining the body weight and / or the sitting position by a classical machine learning module (26) based on the at least one determined body feature (M), wherein the classical machine learning module (26) uses a symbolic approach to machine learning, characterized by that the feature module (24) uses a classic machine learning method, in particular with a symbolic approach, when determining the at least one body feature (M). [2] Method according to claim 1, characterized bythat a safety system, in particular a safety belt system (18) and / or a restraint system (16) is controlled as a function of the determined body weight and / or the determined seating position. [3] Method according to claim 1 or 2, characterized by that the machine learning module (26) uses a classification method, in particular a random forest, when determining the body weight and / or the sitting position. [4] Method according to one of the preceding claims, characterized by that the machine learning module (26) uses a regression method, in particular a support vector machine, when determining the body weight and / or the sitting position. [5] Method according to one of the preceding claims, characterized by Preprocessing the image (B), in particular by segmenting the body of the vehicle occupant (F). [6] Method according to one of the preceding claims, characterized bythat the feature module (24) determines the at least one body feature (M) by means of a machine image recognition method, in particular using the Canny algorithm. [7] Method according to one of the preceding claims, characterized by that the feature module (24) uses an artificial neural network (30) when determining the at least one body feature (M). [8] Method according to one of the preceding claims, characterized by that the at least one body characteristic (M) is the height of the body (M1), in particular of the upper body, the width of the body (M2), in particular of the upper body, the arm length (M3) of at least one arm, the abdominal circumference (M4), the orientation of the upper body, the orientation of the head and / or the orientation of at least one arm. [9] Method according to one of the preceding claims, characterized bythat the at least one body characteristic (M) used to determine the body weight and / or the sitting position is selected by means of at least one data mining method. [10] Method according to one of the preceding claims, characterized by that the image (B) is a picture of the vehicle occupant (F) sitting on a corresponding seat (12) of the motor vehicle (10). [11] Control device for a motor vehicle (10) which is designed to carry out a method according to one of the preceding claims. [12] Motor vehicle with a camera (20) and a control unit (22) according to claim 11.

Images