Model training method and method for predicting zero-gravity attitude diving trend of driver and passenger

Abstract

The invention relates to a model training method and a driver and passenger zero-gravity attitude diving trend prediction method, and the model training method comprises the steps: determining a pulley collision trend parameter; constructing a training data set based on the tackle collision trend parameters; using the training data set to train a submergence trend prediction model; wherein the diving trend prediction model is used for predicting the diving trend of a driver and passengers based on the tackle collision trend parameters; and obtaining the trained submergence trend prediction model. Compared with the prior art, the method has the advantages that the diving protection capability is improved, driver and passenger protection is optimized, the prediction precision is high, and the generalization capability is high.

Description

Technical Field

[0001] The present invention relates to the field of vehicle safety technology, and in particular to a model training method and a method for predicting a driver's and passenger's zero-gravity diving trend. Background Art

[0002] More and more vehicles are equipped with zero-gravity seats, but in the event of a collision, the driver and passengers in the zero-gravity seats are at risk of diving, and the seat belts will cause great damage to the abdomen and neck. In a head-on collision, the phenomenon of the seat belt's abdominal strap sliding off the human pelvis and directly invading the human abdomen is defined as diving. When diving occurs, the abdominal strap of the seat belt will slide off the pelvis and directly invading the human abdomen, squeezing the internal organs, while the shoulder straps will directly act on the human neck and head to cause damage, and the thighs will hit the dashboard to cause damage to the lower limbs. At this time, the seat belt not only fails to protect the personnel, but also causes more serious injuries.

[0003] Although the manufacturer requires the use of the zero-gravity function in a non-driving state in the instructions, it is inevitable that the driver and passengers will use it incorrectly during intelligent driving. At present, some manufacturers use seat rail collapse or back airbags to mitigate the risk of diving injuries during the collision of zero-gravity seats, but the cost is high and the mitigation effect is general. It is structurally destructive to the seat and increases maintenance costs. If the current diving trend and time of the driver and passengers can be predicted in advance before the dive occurs, the safety system will have more reaction time, thereby providing better protection for the driver and passengers. Real-time prediction of the current driver and passenger's zero-gravity posture diving trend can achieve better active and passive safety protection for the driver and passengers.

[0004] The prediction accuracy of the zero-gravity diving trend prediction scheme for passengers in the current related technologies is low, and it is impossible to achieve active and passive safety protection for the passengers. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a model training method and a method for predicting the diving trend of passengers in zero-gravity posture, which can improve the diving protection capability, optimize the protection of passengers and occupants, have high prediction accuracy and strong generalization ability.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] According to a first aspect of the present invention, a model training method is provided, the method comprising: determining a pulley collision trend parameter; constructing a training data set based on the pulley collision trend parameter; using the training data set to train a diving trend prediction model; wherein the diving trend prediction model is used to predict the diving trend of the driver and passenger based on the pulley collision trend parameter; and obtaining the trained diving trend prediction model.

[0008] In an implementation of the first aspect, the sled collision tendency parameter includes at least one of a driver and passenger state parameter, a zero-gravity seat state parameter, an auxiliary mechanism state parameter, a seat belt state parameter, and a vehicle state parameter.

[0009] In an implementation of the first aspect, determining the trolley collision trend parameters includes: based on at least one consideration factor among parameter acquisition complexity, diving trend correlation and parameter differentiation, determining the trolley collision trend parameters including at least one of driver and passenger state parameters, zero gravity seat state parameters, auxiliary mechanism state parameters, seat belt state parameters and vehicle state parameters.

[0010] In an implementation of the first aspect, the pulley state collision tendency parameter includes:

[0011] The driver and passenger status parameters include: the body size of the driver and passenger;

[0012] The zero-gravity seat state parameters include: seat cushion inclination angle and backrest inclination angle;

[0013] The vehicle state parameters include: vehicle speed.

[0014] In an implementation of the first aspect, constructing a training data set includes: obtaining a training data set based on a sled collision model; wherein the sled collision model is used to obtain vehicle collision simulation results based on different input conditions; the training data set includes the sled collision trend parameters and diving trend results, and the diving trend results include a diving state and a diving time.

[0015] In an implementation of the first aspect, the acquiring of a training data set based on the pulley collision model includes: acquiring an initial training data set based on the pulley collision model; and performing augmentation processing on the initial training data set to acquire an augmented training data set.

[0016] In an implementation of the first aspect, the augmenting the initial training data set to obtain the augmented training data set includes: augmenting the initial training data set using an interpolation method to obtain the augmented training data set.

[0017] In an implementation of the first aspect, the diving trend prediction model is a model constructed based on a deep neural network DNN.

[0018] According to a second aspect of the present invention, a method for predicting a diving trend of a driver and passenger in a zero-gravity posture is provided, the method comprising: obtaining a pulley collision trend parameter; inputting the pulley collision trend parameter into a diving trend prediction model to obtain a diving trend prediction result; wherein the diving trend prediction model is a trained diving trend prediction model obtained by using the method provided by the first aspect or any possible implementation method of the first aspect.

[0019] In an implementation of the second aspect, the obtaining of the pulley collision tendency parameters includes: obtaining part or all of the pulley collision tendency parameters based on computer vision technology.

[0020] Compared with the prior art, the present invention has the following beneficial effects:

[0021] 1. Improving diving protection capabilities: The model training method and the method for predicting the diving trend of passengers in zero-gravity posture provided by the present invention can greatly improve the diving protection capabilities for passengers, especially passengers in zero-gravity posture.

[0022] 2. Optimizing driver and passenger protection: The model training method and the driver and passenger zero-gravity diving trend prediction method provided by the present invention can better protect the driver and passengers during a vehicle collision and effectively reduce potential injuries.

[0023] 3. High prediction accuracy and strong generalization ability: The model training method and the method for predicting the diving trend of drivers and passengers in zero-gravity posture provided by the present invention are based on the diving trend prediction model constructed based on computer vision and deep learning, which has excellent prediction accuracy and generalization ability, and can adapt to different drivers and passengers and collision conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of a flow chart of a model training method provided in an embodiment of the present invention;

[0025] Figure 2 A schematic diagram of the seat cushion inclination angle and the backrest inclination angle provided by an embodiment of the present invention;

[0026] Figure 3 A schematic diagram of a training method for a diving trend prediction model in a certain scenario provided by an embodiment of the present invention;

[0027] Figure 4 A schematic diagram of the architecture of a dive trend prediction model in a certain scenario provided by an embodiment of the present invention;

[0028] Figure 5 A schematic diagram of the training process of a diving trend prediction model in a certain scenario provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0030] "One embodiment" or "embodiment" referred to herein refers to a particular feature, structure, or characteristic that may be included in at least one implementation of the present application.

[0031] Figure 1 It is a flowchart of the model training method provided in the embodiment of the present application. The present application provides method operation steps as described in the embodiment or flowchart, but may include more or fewer operation steps based on conventional or non-creative labor. The order of steps listed in the embodiment is only one of the many step execution orders of the model training method and the method for predicting the trend of the zero-gravity attitude dive of the driver and passenger, and does not represent the only execution order. The method should be implemented by software and / or hardware. Please refer to Figure 1 , model training methods can include:

[0032] Step S110: determining the pulley collision tendency parameter;

[0033] Step S120: constructing a training data set based on the pulley collision trend parameter;

[0034] Step S130: training a diving trend prediction model using a training data set; wherein the diving trend prediction model is used to predict the diving trend of the driver and passenger based on the pulley collision trend parameter;

[0035] Step S140: Obtaining a trained diving trend prediction model

[0036] The following is an introduction to the possible implementation methods of determining the pulley collision tendency parameter in step S110:

[0037] Optionally, the above-mentioned sled collision tendency parameter may include: at least one of: a driver and passenger state parameter, a zero-gravity seat state parameter, an auxiliary mechanism state parameter, a seat belt state parameter and a vehicle state parameter.

[0038] It is understandable that the mechanism of submergence and its influencing factors have been studied and applied for a long time. The occurrence of submergence is affected by many factors, mainly related to the following aspects:

[0039] (1) Driver and passenger status parameters:

[0040] Exemplarily, the driver and passenger status parameters may include: the driver and passenger's body size and posture, the passenger's foot position, etc.;

[0041] (2) Zero gravity seat status parameters:

[0042] Exemplarily, the zero-gravity seat state parameters may include: a seat cushion inclination angle and a backrest inclination angle;

[0043] The schematic diagram of the seat cushion inclination and backrest inclination is as follows Figure 2 As shown, angle A represents the seat cushion inclination angle of the current human body posture, and angle B represents the backrest inclination angle of the current human body posture.

[0044] (3) Auxiliary mechanism status parameters:

[0045] It is understandable that the design of the seat and seat belt auxiliary mechanisms is conducive to alleviating the risk of the driver and passengers diving, and the auxiliary mechanisms include diving stoppers, knee and back airbags, seat belt pre-tensioning mechanisms, seat belt force limiting mechanisms, and seat belt locking mechanisms;

[0046] Exemplarily, the auxiliary mechanism status parameters may include: whether the auxiliary mechanism is installed, the auxiliary mechanism installation position, etc.

[0047] (4) Seat belt status parameters:

[0048] It is understandable that the dive is related to the arrangement of the upper and lower anchor points of the safety belt anchor points;

[0049] Exemplarily, the seat belt status parameters may include: the position of the lower fixing point of the seat belt, the position of the D-ring of the seat belt, and the gap between the seat belt and the person.

[0050] (5) Vehicle status parameters:

[0051] It is understandable that the speed of collision is closely related to diving. Under the same posture, the faster the collision speed, the higher the diving risk;

[0052] Exemplarily, the vehicle state parameters may include: vehicle speed.

[0053] Optionally, the above step S110 may include: determining the sled collision trend parameters including at least one of the driver and passenger state parameters, zero-gravity seat state parameters, auxiliary mechanism state parameters, seat belt state parameters and vehicle state parameters based on at least one consideration factor among parameter acquisition complexity, diving trend correlation and parameter differentiation.

[0054] Optionally, in a certain scenario, based on at least one of the considerations of parameter acquisition complexity, diving trend correlation, and parameter differentiation, the above-mentioned sled state collision trend parameter may include:

[0055] Driver and passenger status parameters, including: driver and passenger body dimensions;

[0056] Zero gravity seat status parameters, including: seat cushion inclination and backrest inclination;

[0057] Vehicle status parameters, including: vehicle speed.

[0058] The scheme for training the diving trend prediction model using the above four parameters of the driver and passenger body size, seat cushion inclination, backrest inclination and vehicle speed is as follows: Figure 3 As shown, the input of the diving trend prediction model is four parameters: human body size, seat cushion inclination, backrest inclination and vehicle speed, and the output is the diving state and diving time.

[0059] The following describes an optional implementation of step S120 for constructing a training data set:

[0060] Optionally, the above-mentioned step S120 of constructing a training data set includes: obtaining a training data set based on a sled collision model; wherein the sled collision model is used to obtain vehicle collision simulation results based on different input conditions; the training data set includes sled collision trend parameters and diving trend results, and the diving trend results include diving state and diving time.

[0061] It is understandable that the above-mentioned sled collision model can be constructed based on a vehicle simulation model.

[0062] Optionally, the obtaining of a training data set based on the pulley collision model includes: obtaining an initial training data set based on the pulley collision model; and augmenting the initial training data set to obtain an augmented training data set.

[0063] Optionally, performing augmentation processing on the initial training data set to obtain an augmented training data set includes: performing augmentation processing on the initial training data set using an interpolation method to obtain an augmented training data set.

[0064] It can be understood that the initial training data set can be augmented by the interpolation method, thereby obtaining more comprehensive training data.

[0065] The following is an introduction to the training scheme of the diving trend prediction model in step S130:

[0066] Optionally, the subsurface trend prediction model is a model built based on a deep neural network DNN, and its structure is as follows: Figure 4As shown, it may include an input layer, a hidden layer and an output layer. The input of the input layer is the slide collision trend parameter, such as human body size, vehicle speed, seat cushion inclination angle and backrest inclination angle. The output of the output layer includes the diving state (that is, whether to dive) and the diving time. The diving trend prediction model constructed by using the deep neural network DNN has good prediction accuracy and generalization ability.

[0067] In addition, it is understandable that before the pulley collision trend parameters are input into the diving trend prediction model, the parameters may be normalized and regularized so that the model can process the input data more quickly and efficiently.

[0068] In addition, it is understandable that when training the diving trend prediction model, the momentum gradient method can be used for training, which is conducive to improving the training efficiency and accuracy of the model and optimizing the problems that the model may fall into local optimality.

[0069] In addition, it can be understood that the above-mentioned training data set can be divided into a training set and a test set, wherein the training set is used to train the diving trend prediction model, and the test set is used to test the diving trend prediction model until the diving trend prediction model reaches the expected accuracy or reaches a preset number of iterations.

[0070] See also Figure 5 ,The following provides a complete training scheme for the above-mentioned diving trend prediction model in a certain scenario, which mainly includes:

[0071] Step 1: Build a pulley collision model;

[0072] Step 2, obtaining the pulley collision trend parameters;

[0073] Step 3: interpolation is performed to increase the data sample size and obtain a collision diving trend data set;

[0074] Step 4: Divide the collision diving trend data set into a training set and a test set;

[0075] Step 5: Use the training set to train the diving trend prediction model, which mainly includes:

[0076] (1) Normalize the data and process the input and output functions;

[0077] (2) Set model training parameters;

[0078] (3) Training the latent trend prediction model (DNN model);

[0079] (4) Use the test set to test the model;

[0080] (5) If the model accuracy reaches the expected accuracy, the trained model is obtained and the training process ends;

[0081] If the model accuracy does not reach the expected accuracy, adjust the training and model parameters and return to the above step (3) to retrain the model.

[0082] Based on the same inventive concept, an embodiment of the present invention further provides a method for predicting a diving trend of a driver and passenger in a zero-gravity posture, the method comprising:

[0083] Step S210: obtaining the pulley collision tendency parameter;

[0084] Step S220: inputting the pulley collision trend parameter into the diving trend prediction model to obtain the diving trend prediction result; wherein the diving trend prediction model is a trained diving trend prediction model obtained by using any of the above-mentioned model training methods.

[0085] Optionally, the above step S210 may include: obtaining part or all of the pulley collision trend parameters based on computer vision technology.

[0086] It is understandable that when an image acquisition device is installed inside the vehicle, some or all of the slide collision trend parameters can be obtained based on the images collected by the image acquisition device. Taking the diving trend prediction model using human body size, seat cushion inclination, backrest inclination and vehicle speed as input as an example, the image acquisition device installed inside the vehicle can at least obtain parameters such as human body size, seat cushion inclination, backrest inclination, etc., thereby realizing the zero-gravity posture diving trend prediction of the driver and passenger based on vision and deep neural learning.

[0087] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A model training method, characterized in that: The method comprises: Determine the pulley collision tendency parameters; Based on the pulley collision trend parameter, construct a training data set; The diving trend prediction model is trained using the training data set; wherein the diving trend prediction model is used to predict the diving trend of the driver and passenger based on the pulley collision trend parameter; Obtain the trained dive trend prediction model.

2. The model training method according to claim 1, characterized in that: The sled collision tendency parameter includes at least one of a driver and passenger state parameter, a zero-gravity seat state parameter, an auxiliary mechanism state parameter, a seat belt state parameter, and a vehicle state parameter.

3. The model training method according to claim 2, characterized in that: Determining the pulley collision tendency parameter comprises: Based on at least one of the considerations of parameter acquisition complexity, diving trend correlation and parameter differentiation, a sled collision trend parameter is determined, which includes at least one of the driver and occupant state parameters, zero-gravity seat state parameters, auxiliary mechanism state parameters, seat belt state parameters and vehicle state parameters.

4. The model training method according to claim 2, characterized in that: The pulley state collision trend parameters include: The driver and passenger status parameters include: the body size of the driver and passenger; The zero-gravity seat state parameters include: seat cushion inclination angle and backrest inclination angle; The vehicle state parameters include: vehicle speed.

5. The model training method according to claim 1, characterized in that: The constructing of the training data set comprises: Based on the sled collision model, a training data set is obtained; wherein the sled collision model is used to obtain vehicle collision simulation results based on different input conditions; the training data set includes the sled collision trend parameters and diving trend results, and the diving trend results include diving state and diving time.

6. The model training method according to claim 5, characterized in that: The step of obtaining a training data set based on the pulley collision model includes: Based on the pulley collision model, an initial training data set is obtained; The initial training data set is augmented to obtain an augmented training data set.

7. The model training method according to claim 6, characterized in that: The augmenting the initial training data set to obtain the augmented training data set includes: The initial training data set is augmented by using an interpolation method to obtain an augmented training data set.

8. The model training method according to any one of claims 1 to 7, characterized in that: The diving trend prediction model is a model built based on a deep neural network DNN.

9. A method for predicting the diving trend of a passenger in zero gravity posture, characterized in that: The method comprises: Obtaining the pulley collision trend parameters; The pulley collision trend parameter is input into a diving trend prediction model to obtain a diving trend prediction result; wherein the diving trend prediction model is a trained diving trend prediction model obtained by the method as described in any one of claims 1 to 8.

10. The method for predicting the trend of a passenger diving in zero gravity posture according to claim 9, characterized in that: The step of obtaining the pulley collision trend parameter includes: Based on computer vision technology, some or all of the pulley collision trend parameters are obtained.

Images