Automobile safety air bag control method and system

Abstract

The invention provides an automobile safety air bag control method and system.The method comprises the steps that collision twinning is carried out according to driving data of a target vehicle, personnel data of personnel in the vehicle and air bag data of the target vehicle, and the personnel risk degree of air bag control of different pop-up logics is determined; generating a pre-control instruction of air bag control of the corresponding pop-up logic when the personnel risk degree is minimum; performing collision verification according to the posterior data of the collision twinning, and if the verification is passed, executing the pre-control instruction; and if the verification is not passed, determining a target airbag control instruction according to the actual collision data and a preset airbag control instruction library, and performing corresponding control. According to the automobile safety air bag control method and system, the final air bag control strategy is determined according to the time windows suitable for the database matching mode and the risk degree evaluation mode of the person colliding with the twin and the collision verification result, and the suitability of air bag control is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile airbag control, and in particular to an automobile airbag control method and system. Background Art

[0002] Typically, when a vehicle is struck by a collision, its airbag deployment logic proceeds as follows: first, sensors collect data; then, the vehicle's computer system performs collision detection and classification based on the data collected; finally, based on the collision detection and classification results, it determines whether the collision has reached the vehicle's airbag triggering threshold. If so, the airbag deploys. While this airbag deployment window is relatively short, it relies on preset threshold rules and simple classification output instructions, which can easily lead to inappropriate deployment. For example, if the collision detection results indicate that the steering wheel airbag should deploy with the greatest degree of force, but the driver's sitting posture is inappropriate, the driver's injuries may be more severe under the airbag control decision with the greatest degree of force than under the airbag control decision with a less severe degree of force.

[0003] In view of this, there is an urgent need for a vehicle airbag control method and system to at least solve the above-mentioned deficiencies. Summary of the Invention

[0004] One of the purposes of the present invention is to provide a method and system for controlling automobile airbags, which performs collision twinning of the collision situation of people in the vehicle within the fuzzy perception time, and pre-calculates and generates pre-control instructions; introduces a posteriori data to determine whether the subsequent corresponding collision situation actually occurs. If so, the pre-control instructions are directly triggered; otherwise, the target airbag control instructions matched in a preset database are triggered according to the actual collision data, and the final airbag control strategy is determined according to the appropriate time window and collision verification results of the two methods of database matching and personnel risk evaluation of the collision twin, thereby improving the suitability of the airbag control.

[0005] An embodiment of the present invention provides a vehicle airbag control method, comprising:

[0006] Based on the target vehicle's driving data, the occupant's data, and the target vehicle's airbag data, collision twinning is performed to determine the occupant's risk level under different airbag deployment logics.

[0007] Generate pre-control instructions for airbag control corresponding to the ejection logic when the risk level of personnel is minimum;

[0008] Perform collision verification based on the collision twin's a posteriori data, and execute pre-control instructions if the verification passes;

[0009] If the verification fails, the target airbag control instruction is determined and corresponding control is performed based on the actual collision data and the preset airbag control instruction library.

[0010] Preferably, the driving data of the target vehicle includes the surrounding environment information of the target vehicle and the state of the target vehicle itself.

[0011] Preferably, the personal data of the occupants of the vehicle include: the person's body shape, the person's seat position, the person's sitting posture and the seat belt wearing condition.

[0012] Preferably, the airbag data includes airbag distribution and achievable deployment states of the airbag.

[0013] Preferably, collision twinning is performed based on the driving data of the target vehicle, the data of the occupants in the vehicle, and the airbag data of the target vehicle to determine the risk level of the occupants under different airbag deployment logics, including:

[0014] Based on the target vehicle's driving data, determine whether a collision will occur within the fuzzy perception time;

[0015] Based on the collision scenario and the current occupant data, the trajectory of the occupants after the collision is predicted.

[0016] Preview the post-collision state of occupants based on their trajectory and airbag control with different deployment logics;

[0017] Determine the personnel risk level based on the preset personnel risk assessment template and post-collision status.

[0018] Preferably, the steps for obtaining the fuzzy perception duration are as follows:

[0019] Extract collision deduction factor values ​​based on driving data;

[0020] Determining the derivation degree corresponding to the collision factor value set according to the collision derivation factor value and the input value of the collision model;

[0021] Obtaining a preselected fuzzy perception duration preset for a derivation target corresponding to a collision factor value set having a derivation degree greater than or equal to a preset derivation degree threshold;

[0022] The preselected fuzzy perception time is adjusted upward according to the fatigue monitoring result of the driver to obtain the fuzzy perception time.

[0023] Preferably, the steps for obtaining fatigue monitoring results are as follows:

[0024] Intervene and collect in-car conversations when the recommended route deviates from the actual route;

[0025] The fatigue monitoring video stream, driving time, and fatigue values ​​corresponding to the semantic representation of in-car conversation are normalized and summed to obtain the fatigue monitoring results.

[0026] Preferably, the steps for obtaining the fatigue value represented by the fatigue monitoring video stream are as follows:

[0027] Based on the fatigue monitoring video stream, the changes in the driver's blink frequency, eye openness and the time interval of the head following swing when turning are extracted;

[0028] The fatigue value represented by the fatigue monitoring video stream is quantified based on the changes in blink frequency, eye openness and head swing time interval when turning.

[0029] Preferably, the steps for obtaining the fatigue value corresponding to the semantic representation of the in-vehicle conversation are as follows:

[0030] Expand the in-car conversation semantics on a preset timeline according to the corresponding extraction time;

[0031] Match the route change query semantics on the timeline. If a match is successful, use the in-car conversation semantics of the other party after the route change query semantics as the query response;

[0032] Input the query response into the preset reasonable route change judgment model and output a reasonable value;

[0033] According to the reasonable value and the preset fatigue value reading table, read the fatigue value corresponding to the reasonable value.

[0034] An embodiment of the present invention provides an automobile airbag control system, comprising:

[0035] The collision twinning module is used to perform collision twinning based on the target vehicle's driving data, the occupant data, and the target vehicle's airbag data to determine the occupant danger level of airbags controlled by different deployment logics;

[0036] A pre-control instruction generation module is used to generate pre-control instructions for airbag control corresponding to the ejection logic when the risk level of the personnel is minimized;

[0037] A first control module is configured to perform collision verification based on the a posteriori data of the collision twin and execute pre-control instructions if the verification passes;

[0038] The second control module is used to determine the target airbag control instruction and perform corresponding control based on the actual collision data and the preset airbag control instruction library if the verification fails.

[0039] The beneficial effects of the present invention are:

[0040] The present invention performs collision twinning of the collision situation of the occupants of the vehicle within the fuzzy perception time, and pre-calculates and generates pre-control instructions; introduces a posteriori data to determine whether the subsequent corresponding collision situation actually occurs. If so, the pre-control instructions are directly triggered; otherwise, the target airbag control instructions matched in a preset database are triggered according to the actual collision data. The final airbag control strategy is determined according to the appropriate time window and collision verification results of the two methods of database matching and occupant risk evaluation of the collision twin, thereby improving the suitability of the airbag control.

[0041] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be achieved and obtained through the structures specifically pointed out in this application document.

[0042] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0044] Figure 1 Schematic diagram of a vehicle airbag control method according to an embodiment of the present invention;

[0045] Figure 2 Schematic diagram of an automobile airbag control system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0047] The embodiment of the present invention provides a method for controlling an automobile airbag. Figure 1 As shown, including:

[0048] Step 1: Perform collision twinning based on the target vehicle's driving data, the occupant data, and the target vehicle's airbag data to determine the occupant risk level of airbags controlled by different deployment logics;

[0049] Among them, the target vehicle is the vehicle that needs to be controlled by the automobile airbag; the driving data includes: the surrounding environment information of the target vehicle and the target vehicle's own state. The surrounding refers to the range of 50 meters from the target vehicle. The surrounding environment information includes: the relative position relationship between static objects and the target vehicle, the three-dimensional data of static objects, the relative position relationship between dynamic objects and the target vehicle, the three-dimensional data of dynamic objects, the movement speed and direction of dynamic objects, etc. The own state includes: the three-dimensional data of the target vehicle, the movement speed and direction of the target vehicle; the personnel data includes: the personnel's body shape, the personnel's seat position, the personnel's sitting posture and the wearing of the seat belt; the airbag data includes: the airbag distribution and the achievable airbag deployment state;

[0050] The method of performing collision twinning based on the target vehicle's driving data, the occupant's data, and the target vehicle's airbag data to determine the occupant's risk level under airbag control with different ejection logics includes:

[0051] Based on the target vehicle's driving data, determine whether a collision will occur within the fuzzy perception time;

[0052] When determining whether a collision is about to occur, the existing collision model is used to determine the surrounding environment information of the target vehicle and the target vehicle's own state. The fuzzy perception duration is the time it takes for the driver to perceive the danger and react, for example, 2 seconds.

[0053] Based on the collision scenario and the current occupant data, the trajectory of the occupants after the collision is predicted.

[0054] The collision scenario is the collision scenario derived from the collision model, including: the collision position and collision force of different collisions, where different collisions refer to all collisions from the first collision in which a collision is derived to the complete stationary state of the target vehicle; the occupant data at the current moment refers to the occupant's body shape, seat position, sitting posture, and seatbelt wearing status at the current moment; the occupant's trajectory is predicted based on a trajectory prediction model, which is obtained by using a machine learning model to learn historical occupant collision data (with historical collision scenarios, historical pre-collision occupant data, and historical post-collision occupant trajectories), where the airbag in the historical occupant collision data is not triggered;

[0055] Preview the post-collision state of occupants based on their trajectory and airbag control with different deployment logics;

[0056] Airbag control with different deployment logics refers to the deployment timing and degree of deployment at different airbag distribution locations. Previewing the post-collision state of occupants refers to creating a digital twin of the corresponding in-vehicle situation based on the predicted trajectory of the occupants after the collision and the airbag control with different deployment logics, previewing the state in which the trajectory of the occupants after the collision is blocked by the deployed airbags (post-collision state).

[0057] Determine the risk level of personnel based on the preset risk assessment template and post-collision status;

[0058] The preset personnel risk assessment template is a standardized assessment system for quantifying the risk of injury to vehicle occupants. For example, the corresponding risk assessment parameters (head acceleration, chest compression, and neck torque) are calculated based on the motion trajectory of the various parts of the vehicle occupant's body after the force is applied. The risk assessment parameters are multiplied and summed with their corresponding preset normalized weights to obtain the vehicle occupant's risk value. The average of the risk values ​​is used as the risk level.

[0059] Step 2: Generate pre-control instructions for the airbag control corresponding to the ejection logic when the risk level of the personnel is minimum;

[0060] Among them, the pre-control instruction is the airbag control instruction that may be executed based on the collision pre-exploration;

[0061] Step 3: Perform collision verification based on the posterior data of the collision twin. If the verification passes, execute the pre-control instruction;

[0062] The a posteriori data is relevant data to verify whether the collision deduced based on the collision model actually occurred, such as the collision sensor data of the target vehicle within twice the fuzzy perception time. A successful verification means that the collision scenario obtained based on the a posteriori data analysis is consistent with the collision scenario deduced by the collision model.

[0063] Step 4: If the verification fails, determine the target airbag control command and perform corresponding control based on the actual collision data and the preset airbag control command library.

[0064] Among them, verification failure means: the collision scenario obtained based on the analysis of the a posteriori data is inconsistent with the collision scenario derived from the collision model, or it is determined based on the analysis of the a posteriori data that no collision occurred; the actual collision data is: the collision sensor data when the driver perceives the danger and operates the vehicle but still a collision occurs; the preset airbag control instruction library stores one-to-one corresponding collision sensor data and airbag control instructions. When the actual collision data is consistent with a group of collision sensor data in the airbag control instruction library, the corresponding associated airbag control instruction will be used as the target airbag control instruction.

[0065] The working principle and beneficial effects of the above technical solution are:

[0066] The present invention performs collision twinning of the collision situation of the occupants within the fuzzy perception period based on the driving data of the target vehicle, the data of the occupants and the airbag data of the target vehicle, and pre-calculates and generates pre-control instructions.

[0067] In actual car collision accidents, the driver may not be able to perceive the danger and make decisions in time within the fuzzy perception time. In this case, when the collision verification is passed, the pre-control instruction is directly triggered. This overcomes the problem that the time window for determining the airbag control instruction by the collision twin is long after the actual collision, which leads to the inability to determine and trigger the instruction in time. In addition, compared with directly matching the target airbag control instruction based on the actual collision data, the pre-control instruction does not rely too much on the preset threshold rules and simple classification output instructions, and the airbag control instruction is more appropriate; if the driver can perceive the danger and make decisions in time within the fuzzy perception time, the decision result may be successful avoidance, or a collision may still occur. If the danger is successfully avoided, the airbag will not be triggered to pop out. Otherwise, the target airbag control instruction is determined based on the airbag control instruction library.

[0068] The present invention performs collision twinning of the collision situation of the occupants of the vehicle within the fuzzy perception time, and pre-calculates and generates pre-control instructions; introduces a posteriori data to determine whether the subsequent corresponding collision situation actually occurs. If so, the pre-control instructions are directly triggered; otherwise, the target airbag control instructions matched in a preset database are triggered according to the actual collision data. The final airbag control strategy is determined according to the appropriate time window and collision verification results of the two methods of database matching and occupant risk evaluation of the collision twin, thereby improving the suitability of the airbag control.

[0069] In one embodiment, the steps for obtaining the blur perception duration are as follows:

[0070] Extract collision deduction factor values ​​based on driving data;

[0071] The collision derivation factor values ​​are: quantitative indicators related to potential collision risks extracted from driving data, such as following distance, lane departure standard deviation, etc.

[0072] Determining the derivation degree corresponding to the collision factor value set according to the collision derivation factor value and the input value of the collision model;

[0073] The input values ​​of the collision model are: standardized parameters of the input parameters required for the collision model to perform collision deduction, such as: when the following distance and lane deviation standard deviation must be reached before the collision model can be input for collision deduction; the deduction degree is the triggering degree value for the collision deduction of the collision model, which is the reciprocal of the average deviation between the collision deduction factor value and the input value of the corresponding factor type;

[0074] Obtaining a preselected fuzzy perception duration preset for a derivation target corresponding to a collision factor value set having a derivation degree greater than or equal to a preset derivation degree threshold;

[0075] Among them, the preset derivation degree threshold is manually pre-set; the derivation target corresponding to the collision factor value set refers to the collision risk scenario jointly described by the collision factor value set, such as: the risk scenario of rear-end collision with the preceding vehicle, the risk scenario of collision with a roadside flower bed; the pre-selected fuzzy perception duration preset for the derivation target is the time for the driver to perceive and react to the collision accident type preset for the derivation target, such as: the pre-selected fuzzy perception duration for the collision accident type of collision with a flower bed is: 1 second, and the pre-selected fuzzy perception duration for the collision accident type of rear-end collision is: 2 seconds;

[0076] The preselected fuzzy perception time is adjusted upward according to the fatigue monitoring result of the driver to obtain the fuzzy perception time.

[0077] Among them, the fatigue monitoring result is the perception result of the driving status of the driver in the car based on the multimodal sensing device (for example: the camera preset on the vehicle's A-pillar). When the driver is tired, his reaction time will become longer. Therefore, the pre-selected fuzzy perception time is extended to obtain the fuzzy perception time.

[0078] The working principle and beneficial effects of the above technical solution are:

[0079] The timeliness of a driver's response to collision risk is affected by their driving status. Long-distance driving can cause drivers to react slowly and fail to avoid risks in a timely manner. Therefore, quantitative indicators related to potential collision risks are extracted from driving data, compared with the input values ​​of the collision model, and the derivation degree corresponding to the collision factor value set is calculated. The fuzzy perception duration is determined based on the pre-selected fuzzy perception duration of the derivation target of the corresponding collision factor value set with a derivation degree greater than or equal to a preset derivation degree threshold and the driver's fatigue monitoring results, thereby improving the suitability of the derivation duration setting for subsequent collision derivations.

[0080] In one embodiment, the steps for obtaining fatigue monitoring results are as follows:

[0081] Intervene and collect in-car conversations when the recommended route deviates from the actual route;

[0082] The recommended route is the navigation route of the vehicle computer; the actual route is the vehicle's real-time route. When the recommended route and the actual route deviate, such as due to the driver being distracted and missing an intersection, the moment of deviation is used as the opportunity to collect in-vehicle conversations.

[0083] The fatigue monitoring video stream, driving time, and fatigue values ​​corresponding to the semantic representation of in-car conversation are normalized and summed to obtain the fatigue monitoring results.

[0084] Among them, the more obvious the driver's fatigue behavior in the image analysis of the fatigue monitoring video stream, the greater the corresponding fatigue value; the longer the driving time, the greater the corresponding fatigue value; the greater the degree of confirmation of the driver's fatigue by the discussion content corresponding to the semantics of the in-car conversation, the greater the corresponding fatigue value; the normalization rules are pre-set manually.

[0085] The working principle and beneficial effects of the above technical solution are:

[0086] The present invention introduces the fatigue monitoring video stream, driving time and in-vehicle conversation semantics of the route deviation time to determine the fatigue monitoring result, which is more comprehensive.

[0087] In one embodiment, the steps for obtaining the fatigue value represented by the fatigue monitoring video stream are as follows:

[0088] Based on the fatigue monitoring video stream, the changes in the driver's blink frequency, eye openness and the time interval of the head following swing when turning are extracted;

[0089] Among them, the blink frequency change is: the change in the number of blinks per minute; the eye openness change is: the change in the exposed eye area per minute; the follow-up time interval of the head swing when turning is: the time interval between the head turning to check the rearview mirror and following the turn when the car turns;

[0090] The fatigue value represented by the fatigue monitoring video stream is quantified based on the changes in blink frequency, eye openness and head swing time interval when turning.

[0091] Among them, the larger the increment of blink frequency, the smaller the increment of eye openness, and the larger the increment of follow-up time interval, the larger the fatigue value represented by the corresponding fatigue monitoring video stream. The specific quantitative proportional relationship is manually preset.

[0092] The working principle and beneficial effects of the above technical solution are:

[0093] When determining the fatigue value corresponding to the representation of the fatigue monitoring video stream, the present invention extracts the driver's blinking frequency changes, eye opening changes and head following swing changes when turning from the fatigue monitoring video stream to quantify the fatigue value represented by the fatigue monitoring video stream, thereby improving the comprehensiveness and representation accuracy of the representation basis.

[0094] In one embodiment, the steps for obtaining the fatigue value corresponding to the semantic representation of the in-vehicle conversation are as follows:

[0095] Expand the in-car conversation semantics on a preset timeline according to the corresponding extraction time;

[0096] The in-car conversation semantics are expanded on the time axis in the order of the extraction time corresponding to the in-car conversation semantics; each point on the preset time axis corresponds to a specific moment;

[0097] Match the route change query semantics on the timeline. If a match is successful, use the in-car conversation semantics of the other party after the route change query semantics as the query response;

[0098] The route change query semantics include: querying the reason for the route change, such as: "Did I take the wrong turn?", "Shouldn't I get off at that intersection?";

[0099] Input the query response into the preset reasonable route change judgment model and output a reasonable value;

[0100] The preset reasonable route change judgment model is an AI model that determines whether a route change is reasonable. It will combine real-time traffic conditions and responses to inquiries to determine the reasonableness of the route change (reasonable value).

[0101] Based on the reasonable value and the preset fatigue value reading table, the fatigue value corresponding to the reasonable value is read. The preset fatigue value reading table stores the conversion relationship between the reasonable value and the fatigue value corresponding to the semantic representation of the in-vehicle conversation. The conversion relationship satisfies the relationship that the smaller the reasonable value, the larger the fatigue value. The specific conversion ratio is pre-set manually.

[0102] The working principle and beneficial effects of the above technical solution are:

[0103] When the recommended route deviates from the actual route, it is not necessarily because the driver is tired and drives in the wrong direction. It may also be because the driver has chosen a more optimal route. Therefore, a detailed analysis is required.

[0104] During the analysis, the semantics of the in-car conversation are expanded on a timeline and matched against the semantics of route change queries. If a match is found, the semantics of the in-car conversation following the route change query on the timeline are used as the response. This response is then fed into a reasonable route change determination model that combines real-time road conditions with the response to determine the reasonableness of the route change. The model then outputs a reasonable value. The fatigue value corresponding to the reasonable value is read from a pre-set fatigue value table, improving the accuracy of driver fatigue determination in the event of deviations between the recommended route and the actual route.

[0105] The embodiment of the present invention provides a vehicle airbag control system, such as Figure 2 As shown, including:

[0106] Collision twinning module 1, used to perform collision twinning based on the target vehicle's driving data, the occupant data, and the target vehicle's airbag data, and determine the occupant danger level of airbags controlled by different deployment logics;

[0107] Pre-control instruction generation module 2, used to generate pre-control instructions for airbag control corresponding to the ejection logic when the risk level of the personnel is minimum;

[0108] The first control module 3 is used to perform collision verification based on the a posteriori data of the collision twin and execute the pre-control instruction if the verification passes;

[0109] The second control module 4 is configured to determine a target airbag control instruction and perform corresponding control based on actual collision data and a preset airbag control instruction library if the verification fails.

[0110] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A vehicle airbag control method, characterized in that: include: Based on the target vehicle's driving data, the occupant's data, and the target vehicle's airbag data, collision twinning is performed to determine the occupant's risk level under different airbag deployment logics. Generate pre-control instructions for airbag control corresponding to the ejection logic when the risk level of personnel is minimum; Perform collision verification based on the collision twin's a posteriori data, and execute pre-control instructions if the verification passes; If the verification fails, the target airbag control instruction is determined and corresponding control is performed based on the actual collision data and the preset airbag control instruction library.

2. The automobile airbag control method according to claim 1, wherein: The driving data of the target vehicle includes the surrounding environment information of the target vehicle and the state of the target vehicle itself.

3. The automobile airbag control method according to claim 1, wherein: The personal data of the occupants of the vehicle include: the person's body shape, the person's seat position, the person's sitting posture and the seat belt wearing status.

4. The automobile airbag control method according to claim 1, wherein: The airbag data includes the airbag distribution and the achievable deployment states of the airbag.

5. The automobile airbag control method according to claim 1, wherein: Based on the target vehicle's driving data, the occupant data, and the target vehicle's airbag data, collision twinning is performed to determine the hazard level of the airbags controlled by different deployment logics, including: Based on the target vehicle's driving data, determine whether a collision will occur within the fuzzy perception time; Based on the collision scenario and the current occupant data, the trajectory of the occupants after the collision is predicted. Preview the post-collision state of occupants based on their trajectory and airbag control with different deployment logics; Determine the personnel risk level based on the preset personnel risk assessment template and post-collision status.

6. The automobile airbag control method according to claim 5, characterized in that: The steps for obtaining the fuzzy perception duration are as follows: Extract collision deduction factor values ​​based on driving data; Determining the derivation degree corresponding to the collision factor value set according to the collision derivation factor value and the input value of the collision model; Obtaining a preselected fuzzy perception duration preset for a derivation target corresponding to a collision factor value set having a derivation degree greater than or equal to a preset derivation degree threshold; The preselected fuzzy perception time is adjusted upward according to the fatigue monitoring result of the driver to obtain the fuzzy perception time.

7. The automobile airbag control method according to claim 6, characterized in that: The steps to obtain fatigue monitoring results are as follows: Intervene and collect in-car conversations when the recommended route deviates from the actual route; The fatigue monitoring video stream, driving time, and fatigue values ​​corresponding to the semantic representation of in-car conversation are normalized and summed to obtain the fatigue monitoring results.

8. The automobile airbag control method according to claim 7, characterized in that: The steps for obtaining the fatigue value corresponding to the fatigue monitoring video stream are as follows: Based on the fatigue monitoring video stream, the changes in the driver's blink frequency, eye openness and the time interval of the head following swing when turning are extracted; The fatigue value represented by the fatigue monitoring video stream is quantified based on the changes in blink frequency, eye openness and head swing time interval when turning.

9. The automobile airbag control method according to claim 7, characterized in that: The steps for obtaining the fatigue value corresponding to the semantic representation of the in-car conversation are as follows: Expand the in-car conversation semantics on a preset timeline according to the corresponding extraction time; Match the route change query semantics on the timeline. If a match is successful, use the in-car conversation semantics of the other party after the route change query semantics as the query response; Input the query response into the preset reasonable route change judgment model and output a reasonable value; According to the reasonable value and the preset fatigue value reading table, read the fatigue value corresponding to the reasonable value.

10. An automobile airbag control system, characterized in that: include: The collision twinning module is used to perform collision twinning based on the target vehicle's driving data, the occupant data, and the target vehicle's airbag data to determine the occupant danger level of airbags controlled by different deployment logics; A pre-control instruction generation module is used to generate pre-control instructions for airbag control corresponding to the ejection logic when the risk level of the personnel is minimized; A first control module is configured to perform collision verification based on the a posteriori data of the collision twin and execute pre-control instructions if the verification passes; The second control module is used to determine the target airbag control instruction and perform corresponding control based on the actual collision data and the preset airbag control instruction library if the verification fails.

Images