Electric automobile safety belt self-adaptive system based on intelligent identification and dynamic adjustment

Abstract

The invention discloses an electric automobile safety belt self-adaptive system based on intelligent identification and dynamic adjustment, which belongs to the field of automobile safety belts and comprises a visual identification unit, a pressure sensing unit, a control host, an electric adjusting unit, a collision protection unit and a mechanical memory unit. The body type, physiological state and sitting posture information of the passengers are comprehensively captured, differential adjustment strategies are made for different groups, the limitation that a traditional safety belt is single in adaptation is solved, the protection requirements of various passengers are fully met, potential safety hazards caused by improper adaptation are avoided, the riding safety guarantee level of the different groups is improved, and the safety of the passengers is improved. Meanwhile, the system dynamically adjusts key parameters of the safety belt and corrects bad sitting postures in time, the defect that static adjustment cannot respond to sitting posture changes is overcome, the contradiction that a traditional safety belt is tight and uncomfortable or loose and unstable is solved, passengers can obtain reliable protection and comfortable experience in the whole riding process, and dual optimization of protection and comfort is achieved.

Description

Technical Field

[0001] This application belongs to the field of automotive seat belts, and specifically relates to an adaptive seat belt system for electric vehicles based on intelligent recognition and dynamic adjustment. Background Technology

[0002] With the rapid development of the new energy vehicle industry, the adaptability and personalization of automotive safety systems have become important directions for industry upgrades. Seat belts, as a core component of passive safety in automobiles, directly affect passenger safety. Currently, mainstream automotive seat belt designs are mostly based on standard-sized dummies, focusing on the physical characteristics of typical adults to meet basic collision protection needs in common scenarios. However, with the increasing prevalence of automobiles, the passenger population is showing significant diversity, with the needs of various passengers, including children, the elderly, individuals with special physiological conditions, and adults of different body types, becoming increasingly prominent.

[0003] Existing car seat belt systems have significant technical limitations and cannot meet the personalized protection needs of diverse passenger groups. Specific problems include: 1. Single basis for adaptation, lack of targeted consideration for passenger differences: Traditional seat belts are designed based on a uniform standard body shape, which does not fully take into account the differences in body shape, physiological state and sitting posture of different passengers. As a result, they are extremely unsuitable for non-standard body shape passengers such as children, the elderly and people with special physiological conditions, and it is difficult to form an effective protective fit. 2. Limited sensing methods, unable to fully capture key passenger information: Existing technologies mostly rely on manual adjustment or simple mechanical adjustment structures, lacking the ability to fully perceive passenger body characteristics, physiological state and real-time sitting posture, and are unable to obtain the core data required for adaptive adjustment, thus making it difficult to formulate targeted adjustment strategies. 3. Protection needs are difficult to meet, posing significant safety hazards: Due to insufficient adaptability, some passengers may experience discomfort or protection failure when using traditional seat belts. For example, small passengers may face the risk of the seat belt getting stuck in their necks, while large passengers may experience the problem of the seat belt slipping off. Improper adaptation directly causes safety hazards and seriously affects the level of passenger safety protection for different groups. Summary of the Invention

[0004] To address the aforementioned issues, this application provides an adaptive seatbelt system for electric vehicles based on intelligent recognition and dynamic adjustment. This system offers the advantages of enabling personalized adaptive adjustment of the seatbelt and providing passengers with reliable protection and a comfortable experience throughout their journey.

[0005] This application provides an adaptive seatbelt system for electric vehicles based on intelligent recognition and dynamic adjustment, including a visual recognition unit, a pressure sensing unit, a control host, an electric adjustment unit, a collision protection unit, and a mechanical memory unit, wherein: The visual recognition unit and the pressure sensing unit work together to collect passenger feature information. The control host analyzes the feature information to classify human body types and calculates seat belt adaptation parameters by combining safety and comfort dual-objective constraints. The electric adjustment unit dynamically adjusts the structural shape, contact position, width, and tension of the seat belt according to the seat belt adaptation parameters; The collision protection unit predicts collision risks and initiates protective actions in advance based on real-time monitoring data from the visual recognition unit, while dynamically matching the protection intensity according to human body type. When the visual recognition unit fails, the feature information collected by the pressure sensing unit is used to complete the recognition and adaptation calculation; when the pressure sensing unit fails, the feature information collected by the visual recognition unit is used to complete the recognition and adaptation calculation; if both the visual recognition unit and the pressure sensing unit fail, the mechanical memory unit locks the last confirmed seat belt safety position.

[0006] In some embodiments, the visual recognition unit collects passenger feature information, including: The dual cameras, which combine infrared and visible light, capture data from multiple key areas of the passenger's body, such as the head, shoulders, elbows, chest, and neck, through skeletal key point detection. The system reconstructs the three-dimensional body contour of passengers based on data from key areas, while simultaneously capturing the characteristic information of passengers' sitting posture, forming a comprehensive and continuous set of feature information.

[0007] In some embodiments, the pressure sensing unit collects passenger characteristic information, including: A flexible sensor array is evenly distributed in key stress areas such as the waist and buttocks of the seat cushion and backrest to capture the pressure distribution data of the passenger and seat contact surface in real time, generate the corresponding pressure heat map and extract the pressure distribution features. When the visual recognition unit is obstructed by external objects such as pillows or clothing, resulting in incomplete feature information collection, the control host calls up the pressure distribution features in the pressure heat map to supplement passenger body shape-related feature information and improve the feature information set.

[0008] In some embodiments, the control host classifies human body types, including: The system performs a multi-dimensional comprehensive analysis of passengers' gender, age, body type, and physiological characteristics, and completes the classification by matching the characteristic information with preset classification standards. Five body types were identified: children, lean adults, obese adults, groups with special physiological conditions, and the elderly. Each body type has its own independent feature information database.

[0009] In some embodiments, the dual-objective constraint calculation of seat belt adaptation parameters includes: The optimization objectives are to minimize passenger neck pressure and pelvic displacement, while setting safety distance constraints between the seat belt and key skeletal parts such as the clavicle and effective coverage constraints of the lap belt on the hip. By balancing safety and comfort goals and combining feature data of corresponding human body types in the feature information database, the combination of seat belt structure, position, width and tension parameters suitable for the current passenger is calculated.

[0010] In some embodiments, the electric adjustment unit performs dynamic adjustment, including: Differentiated adjustment strategies are developed for different body types: children are fitted with four-point harnesses, slender adults have their retractor tension parameters optimized, obese adults have their shoulder strap width adjusted, groups with special physiological conditions have their lap belt contact position lowered, and the elderly have their pretension response threshold lowered.

[0011] In some embodiments, the control host analyzes the feature information of the passenger's sitting posture captured by the visual recognition unit, including: The control host continuously receives passenger posture feature information captured by the visual recognition unit to determine whether the posture is within the preset comfortable and safe range; If a passenger's sitting posture deviates from the preset range, the system will immediately issue an audible alert and display a warning light signal on the instrument panel; If the passenger does not correct the situation within a reasonable time after being warned, the control unit will drive the electric adjustment unit to automatically tighten the seat belt to the preset safe contact position. At the same time, it will continuously receive passenger posture information and automatically fine-tune the tension to a comfortable state after the passenger's posture returns to the preset comfortable and safe range.

[0012] In some embodiments, the collision protection unit performs protective actions, including: The collision protection unit receives passenger head position feature information continuously collected by the visual recognition unit and comprehensively predicts the collision risk level by combining it with the vehicle's driving status; Based on the collision risk level, the graded pretensioning action is initiated in advance. At the same time, the corresponding adaptation data in the human body type and characteristic information database output by the control host is received, and the seat belt force limiting parameters that match the passenger's physical tolerance are dynamically set, forming a linkage protection with the vehicle collision detection signal.

[0013] In some embodiments, when the visual recognition unit fails and the system detects that the integrity of the feature information it collects is lower than the set standard, it automatically switches to the recognition mode dominated by the pressure sensing unit; the control host simulates the passenger's body shape outline through the pressure distribution data collected by the pressure sensing unit, calls the historical adaptation parameters stored in the system for correction, and completes the matching calculation in combination with the preset body shape template library to generate seat belt adaptation parameters that meet the basic safety requirements.

[0014] In some embodiments, the mechanical memory unit locks the last confirmed seatbelt safety position, including: Real-time storage of seat belt safety position data that has been verified and adapted after each adjustment; When the system is powered off, the mechanical memory unit automatically locks the last confirmed safe position through the built-in mechanical locking structure; When a collision occurs, the collision protection unit prioritizes receiving the vehicle collision detection signal and coordinates with the electric adjustment unit to simultaneously perform pretensioning and force limiting actions.

[0015] Compared with the prior art, this application has the following advantages: 1. By integrating multimodal perception and multidimensional human body classification, personalized adaptive adjustment of seat belts is achieved: The system comprehensively captures passenger body shape, physiological state and sitting posture information, formulates differentiated adjustment strategies for different groups, solves the limitation of traditional seat belts with single adaptation, fully meets the protection needs of various passengers, avoids safety hazards caused by improper adaptation, and improves the level of passenger safety protection for different groups. 2. Construct a dynamic adjustment mechanism with dual objectives of safety and comfort, combined with real-time posture monitoring and abnormal correction functions, to achieve dual optimization of protection and comfort: The system dynamically adjusts key parameters of the seat belt and corrects poor posture in a timely manner. This avoids the drawbacks of static adjustment being unable to respond to changes in posture, and solves the contradiction between the traditional seat belt being too tight and uncomfortable or too loose and unstable, allowing passengers to obtain reliable protection and a comfortable experience throughout the entire journey. 3. Design a cross-validation system for vision and pressure sensing and a full-scenario failure backup mechanism, combined with the safe position locking function of the mechanical memory unit, to form a comprehensive safety redundancy guarantee: the system compensates for the defects of a single module through dual perception verification, the other module quickly takes over when any module fails, and the effective safe position is locked when both modules fail, covering various extreme scenarios, reducing the risk of failure, and improving the reliability and stability of system operation. 4. Adopting a linkage protection design of early collision risk prediction and graded pre-tensioning force limiting, the system optimizes the pertinence and effectiveness of collision protection: The system detects collision risks in advance and initiates protection, while dynamically matching the protection intensity according to the physical tolerance of different groups, avoiding secondary injuries caused by traditional uniform protection, making collision protection more scientific and adaptable, and minimizing the physical injury to passengers from collisions.

[0016] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart of the invention is shown; Figure 2 A schematic diagram of the hardware module connection of the present invention is shown; Figure 3 A flowchart illustrating the implementation of the posture monitoring and correction method of the present invention is shown. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] It should be noted that the electric vehicle seat belt adaptive system based on intelligent recognition and dynamic adjustment is an in-vehicle safety protection system that integrates multimodal perception, intelligent decision-making and dynamic execution. It collects passenger feature information through visual recognition and pressure sensing, and after analysis and classification by the control host, it drives the electric adjustment unit to achieve adaptive matching of seat belt structure, position, width and tension. At the same time, it combines a collision protection unit and a mechanical memory unit to build a full-link safety protection system of recognition-adjustment-protection-backup. It is suitable for all types of new energy vehicles and can solve problems such as poor seat belt adaptability for passengers with special body shapes, static adjustment failure and insufficient collision protection.

[0021] The specific implementation methods of this application are illustrated below through examples: See Figure 1 and Figure 2 ,in Figure 1 A flowchart illustrating an embodiment of this application is shown. Figure 2 The diagram shows a hardware module connection diagram of an embodiment of this application. The electric vehicle seat belt adaptive system based on intelligent recognition and dynamic adjustment provided in this embodiment includes a vision recognition unit, a pressure sensing unit, a control host, an electric adjustment unit, a collision protection unit, and a mechanical memory unit. Each unit realizes data interaction and command transmission through the vehicle CAN bus. Among them, the vision recognition unit includes a fusion camera module of DMS (Driver Monitoring System) and OMS (Occupant Monitoring System), the pressure sensing unit is a seat-integrated flexible pressure sensor array, the control host adopts the SOC chip integrated in the vehicle instrument host, the electric adjustment unit includes a seat belt electric height adjuster, a pneumatic shoulder strap unfolding mechanism, and a variable torque retractor, the collision protection unit consists of a side impact sensor and a graded pretensioning force limiter, and the mechanical memory unit is a storage module with a built-in mechanical locking structure.

[0022] I. Feature Information Acquisition and Collaborative Calibration Mechanism In this embodiment, the visual recognition unit and the pressure sensing unit work together to complete the comprehensive collection and calibration of passenger feature information, providing data support for subsequent human body classification and adaptation calculation.

[0023] The visual recognition unit collects passenger feature information by employing a dual-camera module that fuses infrared and RGB visible light spectra. This module is integrated into the top of the vehicle interior and the B-pillar. Using the OpenPose skeletal keypoint detection algorithm, and with the passenger's consent, it captures real-time 3D coordinate data of 18 to 25 key points on the passenger's body, including the head, shoulders, elbows, chest, and neck. Based on this keypoint data, the system generates a complete body contour model of the passenger through 3D reconstruction technology. Simultaneously, it continuously captures data on changes in the passenger's posture, including torso tilt angle, shoulder position offset, and head posture, forming a continuous set of dynamic feature information. The infrared camera addresses blind spots in complex lighting conditions such as low light and backlight, while the RGB camera ensures the clarity of feature details. Dual-spectrum fusion achieves a keypoint detection accuracy of ≥98%, meeting the recognition needs of different driving scenarios.

[0024] The pressure sensing unit collects passenger feature information in the following way: a flexible pressure sensor array is used, evenly distributed in the buttock pressure area of ​​the seat cushion and the lumbar and back pressure areas of the backrest. The sensor array density is one sensing node per 10cm², which can capture the pressure distribution data of the passenger-seat contact surface in real time and generate a corresponding pressure heat map. The system extracts feature parameters such as pressure peak position, pressure distribution range, and pressure gradient from the pressure heat map as auxiliary judgment basis for passenger body shape (such as the wide pressure distribution of obese people and the small concentrated pressure of children) and sitting posture. When the visual recognition unit is obstructed by external objects such as pillows or heavy clothing, resulting in the integrity of human body key point acquisition being less than 80%, the control host automatically calls the pressure distribution features of the pressure sensing unit to supplement key information such as the thickness and width of the passenger body, correct the error of the three-dimensional contour model, and ensure the integrity of the feature information set.

[0025] It is important to note that both the visual recognition unit and the pressure sensing unit are set to a sampling frequency of 30Hz to achieve real-time synchronous updates of feature information. The control host aligns the data collected by the two units using timestamps to avoid calibration deviations caused by data delays. When the sampling frequency of either unit drops below 20Hz, the system determines that the unit is in an unstable state, automatically increases the sampling frequency of the other unit to 40Hz, and extends the data buffer time to ensure the continuity of feature information.

[0026] II. Human Body Type Classification and Adaptation Parameter Calculation After receiving the feature information set synchronously uploaded by the visual recognition unit and the pressure sensing unit, the control host completes the human body type classification through multi-dimensional analysis, and calculates the seat belt adaptation parameters based on the dual-objective constraints of safety and comfort.

[0027] The specific implementation of human body type classification is as follows: The control host has a built-in trained CNN (Convolutional Neural Network) classification model, which is generated through feature data training of five types of human samples (children, lean adults, obese adults, pregnant women, and the elderly over 65 years old, corresponding to males and females respectively). During the classification process, the system first extracts gender features (based on shoulder width, torso proportions, etc.), age features (based on skeletal contour tightness, body proportions, etc.), body shape dimension features (based on contour circumference, pressure distribution range, etc.), and physiological state features (based on abdominal pressure distribution, torso posture, etc.) from the feature information set, and then matches the above features with preset classification criteria: Children: Age characteristics match ≤12 years, body circumference characteristics below 60% of adult standards; Lean adults: Body circumference characteristics are between 60% and 85% of the adult standard, and the peak pressure distribution is concentrated; Obese adults: Body circumference characteristics exceeding the adult standard by 115%, pressure distribution range covering ≥80% of the seat cushion area; Special physiological condition group (pregnant women): Abdominal pressure distribution shows a uniform diffusion characteristic, and the trunk tilt angle is ≥15°; Older adults: age characteristics matched ≥65 years, bone contour tightness is less than 70% of the adult standard.

[0028] Each human body type corresponds to an independent feature information database, which stores data such as typical body shape parameters, stress tolerance thresholds, and comfortable contact ranges for that group, providing a basis for calculating adaptation parameters.

[0029] The specific implementation method for calculating the adaptation parameters is as follows: The control host adopts a dual-objective optimization model, with the optimization objectives being to minimize neck pressure and minimize the predicted pelvic displacement. The objective function expression is: The objective function is: {Objective function} = min(α_neck pressure + β_predicted pelvic displacement). α and β are weighting coefficients (α=0.6, β=0.4, set based on the principle that safety takes precedence over comfort). The model also sets two constraints: the distance between the shoulder strap and the collarbone must be ≥3cm (to avoid choking), and the lumbar belt must cover at least 80% of the hip bone (to ensure effective lumbar protection).

[0030] The control unit combines typical data of the current human body type from the feature information database, inputs it into a dual-objective optimization model, and calculates the appropriate seat belt structure (e.g., four-point, conventional), contact position (e.g., shoulder strap sinking distance, lap belt offset position), shoulder strap width, and retractor torque, forming the final seat belt fitting parameters. For example, the fitting parameters for children are: four-point seat belt structure, shoulder strap sinking 5-8cm, and retractor torque ≤3N·m; the fitting parameters for obese adults are: conventional seat belt structure, shoulder strap pneumatically extended to 8cm width, and retractor torque ≥6N·m.

[0031] III. Dynamic Adjustment Implementation of the Electric Adjustment Unit After receiving the seat belt adaptation parameters output by the control host, the electric adjustment unit completes the dynamic adjustment of the seat belt through the coordinated action of each actuator, ensuring that the adaptation parameters are implemented.

[0032] The core actuators of the electric regulating unit include: Electric seat belt height adjuster: Installed on the upper part of the B-pillar of the vehicle, driven by a stepper motor, with an adjustment accuracy of 0.1cm, it can adjust the height of the shoulder belt up and down according to the shoulder belt contact position requirements in the fitting parameters, with an adjustment range of 0 to 15cm; Pneumatic shoulder strap deployment mechanism: Integrated inside the shoulder strap, driven by a miniature air pump, it can achieve stepless adjustment of the shoulder strap width from the standard 4cm to a maximum of 8cm, with a response time of ≤0.5s; Variable torque retractor: Built-in torque sensor and electromagnetic adjustment module, torque adjustment range is 2~8N·m, and seat belt tension can be adjusted in real time according to the adaptation parameters; Waist belt displacement adjustment mechanism: Installed at the fixed end of the waist of the safety belt, it is driven by a linear motor to realize the displacement adjustment of the waist belt in the front-back and up-down directions, with an adjustment range of 0-10cm.

[0033] The specific implementation of dynamic regulation strategies for different human body types is as follows: For children: The electric adjustment unit first switches the seat belt structure to a four-point type (the shoulder strap and waist belt are linked and fixed through the built-in electromagnetic buckle). Then the electric height adjuster drives the shoulder strap to sink 5-8cm to ensure that the shoulder strap avoids the neck area. At the same time, the retractor torque is set to ≤3N·m to avoid excessive tension that may compress the child's body. For slender adults: The electric adjustment unit maintains the standard shoulder strap width (4cm), the retractor torque is set to 4-5 N·m, and the electric height adjuster adjusts the shoulder strap to 3-5cm below the collarbone according to the shoulder position in the 3D contour model to ensure a close fit to the shoulder and prevent slippage; For obese adults: The electric adjustment unit activates the pneumatic shoulder strap deployment mechanism to extend the shoulder strap width to 6-8cm (adaptively selected according to body circumference), and the retractor torque is set to 6-8N·m to increase the contact area between the seat belt and the body. At the same time, the lap belt displacement adjustment mechanism slightly adjusts the lap belt upward by 2-3cm to ensure hip bone coverage ≥80%. For special physiological groups (pregnant women): The waist belt displacement adjustment mechanism shifts the waist belt downward by 5-7cm to form a pressure-distributing ring on the abdomen, avoiding direct pressure on the abdomen from the waist belt. The shoulder strap width remains at the normal 4cm, and the retractor torque is set to 3-4N·m to balance protection and comfort. For the elderly: The pretension response threshold of the retractor is reduced by 20% (the normal threshold is 5m / s², and it is set to 4m / s² for the elderly), while the shoulder strap tension is slightly adjusted to 80% of that of a slender adult to reduce the pressure on the elderly.

[0034] It is important to note that during the adjustment process, the electric adjustment unit will send the execution result back to the control host after each action is completed, such as the actual position and torque value after adjustment. The control host compares the deviation of the actual value with the adaptation parameter. If the deviation exceeds 0.5cm (position deviation) or 0.3N·m (torque deviation), the actuator will be driven to perform a secondary fine adjustment to ensure that the adjustment accuracy meets the requirements.

[0035] IV. Posture Monitoring and Abnormal Correction Mechanism The control host continuously receives passenger sitting posture feature information uploaded by the visual recognition unit and determines in real time whether the current sitting posture is within the preset comfortable and safe range. This range is set based on ergonomic data, including torso tilt angle ≤30°, shoulder offset ≤5cm, head deviation from the frontal direction ≤45°, etc.

[0036] The specific implementation process of sitting posture monitoring and correction is as follows: Figure 3 As shown: After DMS starts scanning, it simultaneously completes body type classification and sitting posture modeling. The control unit analyzes the sitting posture model. If it determines that the sitting posture is normal (all parameters are within the preset range), it drives the electric adjustment unit to dynamically fine-tune the tension according to the comfort curve to maintain a comfortable state. If the system determines that the sitting posture is abnormal (any parameter exceeds the preset range), it will immediately perform two warning actions: 1. The in-car audio will emit a continuous 3-second beeping warning tone (frequency 800Hz); 2. The instrument display screen will show a red light warning signal and pop up a text prompt "Please adjust your sitting posture".

[0037] If, within 10 seconds of the warning being issued, the visual recognition unit detects that the passenger's posture has not been corrected (the abnormal parameters have not yet returned to the preset range), the control unit drives the electric adjustment unit to automatically tighten the seat belt to the preset safe contact position. This position is calculated based on the minimum safe constraint of the current body type, and the tightening force is 1.2 times the adaptive tension, using physical constraint to remind the passenger to adjust their posture. In the tightened state, the visual recognition unit continuously monitors changes in posture. Once it detects that the posture has returned to the preset range, the control unit immediately drives the electric adjustment unit to restore the seat belt tension to the original comfort value, ending the correction process.

[0038] It should be noted that during the abnormal correction process, if the passenger manually triggers the seat belt release button, the system will pause the tightening action but will still maintain the audible and visual warnings until the posture correction is completed; if the posture is not corrected after 3 consecutive warnings, the system will prompt "Continued abnormal posture will affect the collision protection effect" through the instrument panel and keep the seat belt tightened until the vehicle is turned off.

[0039] V. Collision Protection Unit's Linked Protection Implementation The collision protection unit works in conjunction with real-time monitoring data from the visual recognition unit and vehicle driving status data (vehicle speed, acceleration, steering angle, etc.) to predict collision risks and initiate coordinated protective actions.

[0040] The specific implementation of collision risk prediction is as follows: the visual recognition unit continuously collects passenger head position feature information (collection frequency 30Hz), and combines it with driving data transmitted by the vehicle's CAN bus to control the host to calculate the collision risk level (low, medium, and high). When the predicted collision risk level is "high" (i.e., the estimated collision time is ≤300ms), the collision protection unit initiates graded pretensioning action in advance. This pretensioning timing is 200ms earlier than the traditional seat belt system, allowing sufficient time for the passenger to adjust their body posture.

[0041] The specific process for calculating the collision risk level is as follows: First, the system receives dynamic data such as passenger head position / displacement rate and torso tilt angle change rate collected in real time by the visual recognition unit, as well as driving data such as vehicle speed, longitudinal / lateral acceleration, and relative distance / relative speed to obstacles transmitted by the vehicle CAN bus. After time stamp alignment and sliding window filtering preprocessing, the data is converted into five core risk factors: collision time prediction value, absolute value of vehicle acceleration, passenger head displacement rate, passenger torso tilt angle deviation, and obstacle type weight. These factors are then standardized according to a score of "0-100". The final risk value is obtained by weighting and summing the risk values ​​according to the following weights: TTC (40%), acceleration (25%), head displacement rate (15%), torso tilt deviation (10%), and obstacle weight (10%). Finally, the risk level is divided according to the risk value: <30 points are low risk (no immediate collision risk, only continuous monitoring), 30-59 points are medium risk (potential collision risk, the collision protection unit enters standby state and triggers a yellow light warning), and ≥60 points are high risk (immediate collision risk, estimated collision time ≤300ms).

[0042] The specific implementation method of graded pretensioning and force limiting adjustment is as follows: The collision protection unit receives the human body type information and risk level output by the control host, and drives the graded pretensioning force limiter to perform the corresponding action: Children: The preload force should be set to 60% of the normal force, with a force limit threshold of ≤4kN, to avoid excessive preload force causing damage to children's bones; For lean adults: Set the pre-tightening force to 80% of the normal force, with a force limit of 6-7 kN, to balance the pre-tightening fixation effect with the body's tolerance; For obese adults: the pretension force is set to 100% of the normal force, with a force limiting threshold of 7-8 kN, to ensure effective restraint of body displacement during a collision; For special physiological groups (pregnant women): the pre-tightening force should be set to 50% of the normal force, with a force limit threshold of 5-6 kN, to reduce the impact on the abdomen; For the elderly: the pre-tightening force is set to 70% of the normal force, with a force limit threshold of 5-6 kN, which is suitable for the stress tolerance range of the elderly.

[0043] When the vehicle's side impact sensor detects an actual collision signal, the collision protection unit fuses the collision signal with the prediction result. If the prediction action has been executed, the force limiting parameters are finely adjusted according to the actual collision intensity. If the prediction has not been made in advance, the pretensioning and force limiting actions corresponding to the human body type are immediately activated to ensure that the seat belt is always in the best protective state when a collision occurs.

[0044] VI. Failure Protection and Mechanical Memory Function Implementation In this embodiment, the system is equipped with multiple failure protection mechanisms to ensure that basic safety functions can still be guaranteed when a single unit fails or the system is powered off.

[0045] Protection mechanism in case of visual recognition unit failure: When the system detects that the integrity of the feature information collected by the visual recognition unit is less than 60% for 3 consecutive seconds, it is determined to have failed. At this time, the system automatically switches to the recognition mode dominated by the pressure sensor unit. The control host uses the pressure distribution data collected by the pressure sensor unit and combines it with a preset body type template library (storing typical pressure distribution models of 5 body types) to complete the matching calculation and generate seat belt adaptation parameters that meet basic safety requirements. For example, if the pressure distribution range determines that the body type is obese, the default adaptation parameters for that type are directly called (shoulder strap width 8cm, retractor torque 7N·m) to ensure that the seat belt has basic protection capabilities; at the same time, the instrument display shows the prompt "Visual recognition failed, switched to pressure sensor mode" to inform the passenger of the current system status.

[0046] Protection mechanism for pressure sensing unit failure: When the system detects that the pressure sensor array's collected data has not changed for 3 consecutive seconds or the pressure value is abnormal, such as the pressure in the entire area being 0 or exceeding the sensor's range, it is determined to have failed. At this time, the system automatically switches to the recognition mode dominated by the vision recognition unit. The control host corrects the error under no-pressure calibration conditions through dynamic updates of the 3D contour model, and optimizes it by combining historical adaptation parameters to generate adaptation parameters. If the vision recognition unit is not obstructed, the accuracy of its generated adaptation parameters can meet the core safety requirements, with only the comfort-oriented fine-tuning function being limited.

[0047] Protection mechanism in case both the visual recognition unit and the pressure sensing unit fail: When both units fail, the mechanical memory unit immediately activates, automatically locking the seatbelt's safe position—the last one to pass adaptation verification—through a built-in mechanical locking structure. The mechanical memory unit stores data from each adjustment in real time, including seatbelt position, width, and torque, and performs adaptation verification after each adjustment. Adaptation verification is determined by factors such as the passenger's stable seating posture and the absence of abnormal warning feedback, ensuring the validity of the stored safe position data. After locking, the seatbelt remains in that position until the system returns to normal or the vehicle is turned off and restarted.

[0048] System power failure protection mechanism: When the vehicle loses power, the mechanical memory unit triggers the mechanical locking structure through the built-in backup power supply to lock the last confirmed safe position, preventing the seat belt position from shifting due to power failure, and ensuring that passengers can still receive basic safety protection before power is restored.

[0049] In summary, the electric vehicle seatbelt adaptive system based on intelligent recognition and dynamic adjustment provided in this application embodiment achieves feature acquisition through visual-pressure collaborative perception, addresses the protection pain points of passengers with special body types based on differentiated adaptation strategies for multiple body types, and constructs a full-scenario seatbelt protection system by combining posture monitoring, early collision prediction, and multiple failure protection mechanisms. This effectively improves the adaptability, comfort, and safety of electric vehicle seatbelts and is suitable for safety configuration upgrades of various new energy vehicles.

[0050] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A smart identification and dynamic adjustment based electric vehicle safety belt adaptive system, characterized in that, The system comprises a visual recognition unit, a pressure sensing unit, a control host, an electric adjustment unit, a collision protection unit and a mechanical memory unit, wherein: The visual recognition unit and the pressure sensing unit cooperate to collect the feature information of the passenger, and the control host analyzes the feature information to divide the human body type and calculate the seat belt fitting parameters in combination with the safety and comfort double target constraints; The electric adjustment unit dynamically adjusts the structure, contact position, width and tension of the seat belt according to the seat belt fitting parameters; The collision protection unit predicts the collision risk based on the real-time monitoring data of the visual recognition unit and starts the protection action in advance, and dynamically matches the protection intensity according to the human body type; When the visual recognition unit fails, the feature information collected by the pressure sensing unit is used to complete the identification and fitting calculation; when the pressure sensing unit fails, the feature information collected by the visual recognition unit is used to complete the identification and fitting calculation; if both the visual recognition unit and the pressure sensing unit fail, the mechanical memory unit locks the last confirmed safe position of the seat belt.

2. The smart identification and dynamic adjustment based electric vehicle safety belt adaptive system according to claim 1, characterized in that, The visual recognition unit collects the feature information of the passenger, including: A dual-camera that fuses infrared and visible light spectrum captures multiple key position data, including the passenger's head, shoulders, elbows, chest and neck, through bone key point detection; Based on the key position data, the passenger's three-dimensional body profile is reconstructed, and the feature information of the passenger's sitting posture is captured to form a comprehensive and continuous feature information set. 3.The smart identification and dynamic adjustment based safety belt adaptive system for electric vehicles according to claim 2, characterized in that, The pressure sensing unit collects the feature information of the passenger, including: A flexible sensor array is uniformly arranged in the key stress area, including the waist and hips of the seat cushion and backrest, to capture the pressure distribution data of the passenger and the seat contact surface in real time, generate the corresponding pressure heat map and extract the pressure distribution features; When the visual recognition unit is blocked by external objects, resulting in incomplete feature information collection, the control host calls the pressure distribution features in the pressure heat map to supplement the passenger body type related feature information and improve the feature information set.

4. The smart identification and dynamic adjustment based electric vehicle safety belt adaptive system according to claim 3, characterized in that, The control host divides the human body type, including: Multi-dimensional comprehensive analysis is performed on the gender, age, body type dimension and physiological state characteristics of the passenger, and classification is completed through matching of the feature information with the preset classification standard; Five human body types are divided, including children, thin adults, obese adults, special physiological state groups and the elderly, and each human body type corresponds to an independent feature information library.

5. The smart identification and dynamic adjustment based electric vehicle seat belt adaptive system of claim 4, wherein, The double-target constraint calculation of the seat belt fitting parameters includes: The optimization targets are to minimize the neck pressure and minimize the pelvic displacement, and the safety distance constraints between the seat belt and the human clavicle and the effective coverage constraints of the waist belt on the hip are set; By balancing the safety and comfort targets, the feature data of the corresponding human body type in the feature information library is combined to calculate the seat belt structure, position, width and tension parameter combination that fits the current passenger. 6.The smart identification and dynamic adjustment based safety belt adaptive system for electric vehicles according to claim 4, characterized in that, The dynamic adjustment of the electric adjustment unit includes: Differentiation of customized adjustment strategies for different body types, children corresponding to switch four-point structure of seat belt, thin body type adults corresponding to optimize the tension parameters of retractor, obese adults corresponding to adjust the shoulder belt width, special physiological state group corresponding to lower the waist belt contact position, the old group corresponding to lower the pre-tightening response threshold. 7.The smart identification and dynamic adjustment based safety belt adaptive system for electric vehicles according to claim 2, characterized in that, The control host analyzes the feature information of the passenger's sitting posture captured by the visual recognition unit, including: The control host continuously receives the passenger's sitting posture feature information captured by the visual recognition unit, and judges whether the sitting posture is in the preset comfortable and safe range; If the passenger's sitting posture deviates from the preset range, the system immediately issues a sound prompt and displays a light warning signal through the instrument; If the passenger does not correct within a reasonable time after the warning, the control host drives the electric adjustment unit to automatically tighten the seat belt to the preset safe contact position, and continuously receives the passenger's sitting posture feature information, and automatically adjusts the tension to the comfortable state when the passenger's sitting posture returns to the preset comfortable and safe range. 8.The smart identification and dynamic adjustment based electric vehicle safety belt adaptive system according to claim 4, characterized in that, The collision protection unit executes protection actions, including: The collision protection unit receives the passenger's head position feature information continuously collected by the visual recognition unit, and comprehensively judges the collision risk level combined with the vehicle driving state; According to the collision risk level, start the grading pre-tightening action in advance, and receive the corresponding adaptive data in the human body type and feature information library output by the control host, dynamically set the safety belt force limiting parameter matched with the passenger's body tolerance, and form a linkage protection with the vehicle collision detection signal. 9.The smart identification and dynamic adjustment based electric vehicle safety belt adaptive system according to claim 3, characterized in that, When the visual recognition unit fails and the system detects that the integrity of the feature information collected by it is lower than the set standard, automatically switch to the recognition mode dominated by the pressure sensing unit; The control host simulates the passenger's body outline through the pressure distribution data collected by the pressure sensing unit, calls the historical adaptive parameters stored by the system for correction, and completes the matching calculation combined with the preset body type template library to generate safety belt adaptive parameters that meet the basic safety requirements. 10.The smart identification and dynamic adjustment based electric vehicle safety belt adaptive system according to claim 1, wherein, The mechanical memory unit locks the last confirmed safety belt safety position, including: Real-time storage of safety belt safety position data that has passed adaptive verification after each adjustment; When the system is powered off, the mechanical memory unit automatically locks the last confirmed safety position through the built-in mechanical locking structure; When a collision occurs, the collision protection unit preferentially receives the vehicle collision detection signal and synchronously executes the pre-tightening and force limiting action with the electric adjustment unit.

Images