Self-adaptive adjusting system and method for transverse position of automobile safety belt

Abstract

The invention discloses a self-adaptive adjusting system and method for the transverse position of an automobile safety belt, and belongs to the technical field of automobiles. The system comprises a transverse position adjusting mechanism which comprises a sliding rail arranged on a seat backrest and an adjusting executing mechanism moving along the sliding rail; the safety belt braid is fixed through an open type guide ring arranged on the adjusting executing mechanism. The multi-modal sensing module is used for collecting multi-modal shape data and dynamic posture data of the passenger; and the controller is electrically connected with the multi-mode sensing module, the transverse position adjusting mechanism and the vehicle control unit and used for determining a static optimal transverse position according to the multi-mode body shape data of the passenger, predicting the posture of the passenger according to the dynamic posture data and the vehicle driving state, generating a dynamic adjusting instruction and driving the adjusting executing mechanism to transversely move along the sliding rail. The body shape of the passenger is predicted through multi-modal feature fusion, the posture change of the passenger is actively pre-judged, the transverse position of the safety belt is adaptively adjusted in time, and the condition of shoulder sliding or neck tightening is effectively avoided.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of automobiles, and particularly relates to a horizontal position self-adaptive adjusting system and adjusting method for automobile safety belts. BACKGROUND

[0002] The statements in this section merely provide background information related to the present application and do not necessarily constitute prior art.

[0003] As one of the most important passive safety devices on an automobile, the rationality of the wearing position of a safety belt directly affects the safety protection effect and riding comfort of a driver and passengers. At present, the wearing position of a safety belt is affected by the position of a retractor, a buckle and a lower fixing point. Usually, after the positions of these arrangement points are determined, the position of the safety belt relative to the shoulder of a human body is fixed. This will result in the fact that different body types of personnel may have the situation of shoulder slipping or neck squeezing when they ride, which seriously affects the riding comfort and even the safety protection performance.

[0004] In view of the above problems, an adaptive adjusting type safety belt is proposed in the prior art. The adaptive adjusting type safety belt collects information through a position sensor, a weight sensor, a camera and the like in a passenger identification module, analyzes the body type of a passenger according to the collected information, drives an adjusting execution module to adjust the position of a safety belt guide ring and a buckle, so that the wearing of the safety belt can be matched with the passenger, and the riding comfort is ensured. However, there are certain technical defects in the prior art: (1) In the prior art, the identification of the body type mainly depends on two-dimensional images and weight data, which lacks analysis of shoulder contour information, and the shoulder feature recognition error of special body types (such as pregnant women, short and fat passengers, etc.) is large, which is easy to cause deviation of the adjusting position; (2) The prior art is mostly passive adjustment according to real-time posture data. The adjustment is usually made according to the changed posture after the posture of the passenger changes. The adjustment response is lagged, and the situation of shoulder slipping or neck squeezing may occur before the posture of the passenger changes, which affects the comfort and safety. SUMMARY

[0005] To solve the above problems of the prior art, the present application provides a horizontal position self-adaptive adjusting system and adjusting method for an automobile safety belt. The body type of a passenger is predicted through multi-modal feature fusion, the horizontal position of a safety belt on the shoulder of a passenger is automatically adjusted according to the body type of the passenger, and the horizontal position of the safety belt is timely self-adaptively adjusted through active prediction of the posture change of the passenger, so as to avoid the situation of shoulder slipping or neck squeezing due to the posture change, and improve the comfort while ensuring the safety.

[0006] In a first aspect, the present application provides a horizontal position self-adaptive adjusting system for an automobile safety belt.

[0007] A kind of automobile safety belt transverse position self-adapting adjustment system, comprising: Transverse position adjustment mechanism, including slide rail arranged on seat back and adjustment execution mechanism moving along slide rail, safety belt webbing is fixed through the open guide ring arranged on adjustment execution mechanism; Multi-modal sensing module, for collecting passenger multi-modal body shape data and dynamic posture data; Controller, electrically connected with multi-modal sensing module, transverse position adjustment mechanism, vehicle controller, for determining the static optimal transverse position of safety belt according to passenger multi-modal body shape data, and predicting passenger posture to optimize transverse position according to dynamic posture data and vehicle driving state, and generating dynamic adjustment instruction based on transverse position, drive adjustment execution mechanism moves along slide rail transversely.

[0008] Further technical solutions, the multi-modal sensing module includes: Image acquisition unit, using high-frequency RGB camera, is installed in the area of vehicle interior rearview mirror, for collecting the video frame time sequence image of seated passenger; Pressure sensing array, including several pressure sensors, is distributed in the key stress area of seat cushion and seat back, for collecting multi-dimensional pressure data of seated passenger; Millimeter wave radar, installed in the area of vehicle interior rearview mirror, for obtaining three-dimensional contour point cloud data of seated passenger shoulder.

[0009] Further technical solutions, according to passenger multi-modal body shape data, determine static optimal transverse position, comprising: Preprocessing is carried out to passenger multi-modal body shape data; The preprocessed multi-modal body shape data is input into body shape classification model, and image features, pressure features and point cloud features are extracted respectively, and fusion features are obtained through feature fusion and attention mechanism, and then the classification results of passenger body shape are output based on fusion features; With fusion features as input, the static optimal transverse position of safety belt is output through the gradient boosting tree model trained corresponding to the body shape type identified.

[0010] Further technical solutions, the multi-modal body shape data includes video frame time sequence image, multi-dimensional pressure data and shoulder three-dimensional contour point cloud data; Preprocessing is carried out to passenger multi-modal body shape data, comprising: For video frame time sequence image, Gaussian filter is used to remove noise, and Canny edge detection operator is used to extract shoulder contour image; For multi-dimensional pressure data, normalization processing is carried out; For shoulder three-dimensional contour point cloud data, environmental noise points are removed through RANSAC algorithm, and effective points of shoulder are retained.

[0011] Further technical solutions, the passenger body type is divided into children, small, standard, obese and pregnant women, and an independent gradient boosting tree model is trained for each type of body shape. The extracted fusion features are input, and the static optimal transverse position of the safety belt is output. The calculation formula is: ; In the above formula, is the weight coefficient, is the mth decision tree, and M is the total number of inputs, is the fusion feature vector.

[0012] Further technical solutions, according to the dynamic posture data and the vehicle driving state, the passenger posture is predicted, and then the transverse position of the safety belt is optimized through PID closed loop control, including: According to the set acquisition frequency, the video frame time sequence image of the passenger is obtained, and the real-time coordinates of the passenger shoulder key points in each image are extracted. At the same time, the tension of the safety belt fabric and the acceleration and steering angle of the vehicle driving are collected in real time; According to the real-time coordinates of the passenger shoulder key points, combined with the optimal position of the shoulder key points of the body type adaptation, the shoulder offset is obtained, and the shoulder offset sequence is constructed; The shoulder offset sequence and the vehicle driving state sequence of a set time length before the current time are input, and the shoulder offset prediction value of the next time period is output through the prediction model based on bidirectional LSTM; According to the shoulder offset prediction value and the real-time tension, PID closed loop control is adopted to calculate the movement amount of the adjusting actuator, and the transverse position of the safety belt is adjusted.

[0013] Further technical solutions, the calculation of the shoulder offset is:

[0014] In the above formula, represents the shoulder offset, , represents the shoulder key point coordinates, , represents the optimal position of the shoulder key points of the body type adaptation.

[0015] Further technical solutions, according to the shoulder offset prediction value and the real-time tension, PID closed loop control is adopted to calculate the movement amount of the adjusting actuator, which is: Based on the shoulder offset prediction value and the real-time tension , the comprehensive error is calculated as: ; Based on the comprehensive error, the movement amount of the adjusting actuator is calculated by PID closed-loop control, and the adjusting actuator is driven to move along the slide rail in the horizontal direction according to the calculated movement amount. ; In the above formula, 、 represents the weight, 、 represents the minimum value and the maximum value of the pulling force, 、 、 is a PID coefficient.

[0016] In a second aspect, the present application provides a method for self-adaptive adjustment of horizontal position of a vehicle safety belt.

[0017] A method for self-adaptive adjustment of horizontal position of a vehicle safety belt is realized based on the vehicle safety belt horizontal position self-adaptive adjustment system proposed in the first aspect, and comprises the following steps: After the occupant is seated and the safety belt is buckled, the multi-modal body shape data and dynamic posture data of the occupant are collected in real time; According to the multi-modal body shape data of the occupant, the static optimal horizontal position is determined, the dynamic adjustment instruction is generated based on the horizontal position, and the adjusting actuator is driven to move in the horizontal direction along the slide rail; During the starting and driving of the vehicle, the occupant's posture is predicted according to the real-time collected dynamic posture data and the driving state of the vehicle to optimize the horizontal position, and the dynamic adjustment instruction is generated based on the horizontal position to drive the adjusting actuator to move in the horizontal direction along the slide rail.

[0018] In a third aspect, the present application further provides a vehicle comprising the vehicle safety belt horizontal position self-adaptive adjustment system proposed in the first aspect, or performing the method for self-adaptive adjustment of horizontal position of a vehicle safety belt proposed in the second aspect.

[0019] The above one or more technical solutions have the following beneficial effects: 1. The present application proposes a vehicle safety belt horizontal position self-adaptive adjustment system and method, which collects video frame time sequence images by means of high-frequency RGB camera, collects multi-dimensional pressure data of key areas of the seat by means of pressure sensing array, and collects shoulder three-dimensional point cloud data of the rearview mirror area in the vehicle by means of millimeter wave radar, constructs a multi-modal data system of image+pressure+three-dimensional point cloud, completely covers key body shape characteristics such as shoulder width, shoulder thickness, shoulder curvature, and weight distribution, and makes up for the defect of single data dimension of traditional schemes; the body shape of the occupant is predicted through multi-modal feature fusion, and the static optimal horizontal position is accurately calculated by using the gradient boosting decision tree (GBDT) model independently trained for this type of body shape according to the body shape of the occupant, so that the safety belt position of different body type occupants (especially pregnant women and obese people) is matched with the physiological characteristics of their shoulders, and the horizontal position of the safety belt on the occupant's shoulders is automatically adjusted to improve comfort.

[0020] 2、Considering that the vehicle acceleration and the steering angle and other states can reflect the influence of the driving condition on the passenger posture, therefore, the application actively judges the posture change of the passenger according to the posture change of the passenger and the driving state of the vehicle, so that the horizontal position of the safety belt can be adjusted before the passenger deviates due to the driving condition, the effective restraint of the safety belt on the shoulder is always maintained, the shoulder slipping or the neck pressing situation due to the posture change is avoided, and especially in the emergency condition, the protection performance caused by the improper safety belt position can be reduced.

[0021] The advantages of the additional aspects of the application will be partially given in the following description, partially will become obvious from the following description, or will be known by the practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0022] The drawings accompanying the specification of this application form a part thereof, serve to provide further understanding of the application, and together with the description of the exemplary embodiments of the application, serve to explain the application, and do not constitute improper limitations on the application.

[0023] Figure 1 The isometric view of the automobile safety belt horizontal position self-adapting adjustment system proposed by the embodiment of the application; Figure 2 The exploded schematic view of the automobile safety belt horizontal position self-adapting adjustment system proposed by the embodiment of the application; Figure 3 The principle schematic view of the automobile safety belt horizontal position self-adapting adjustment system proposed by the embodiment of the application; Figure 4 The working schematic view of the automobile safety belt horizontal position self-adapting adjustment system proposed by the embodiment of the application; wherein, a ) is the safety belt position schematic view of the obese passenger, b ) is the safety belt position schematic view of the small passenger.

[0024] 1, safety belt; 2, adjustment execution mechanism; 3, sliding rail; 4, seat; 5, pressure sensor; 6, camera. DETAILED DESCRIPTION

[0025] It should be noted that the following detailed description is exemplary only, is only for describing the specific embodiments, and is intended to provide further description of the application, and is not intended to limit the exemplary embodiments according to the application. Unless otherwise indicated, all technical and scientific terms used herein have the same meaning as generally understood by those skilled in the art to which the application belongs. In addition, it should be understood that when the terms "comprise" and / or "include" are used in the specification, they indicate the presence of a feature, step, operation, device, component and / or their combination.

[0026] Embodiment One To solve the problems of existing adaptive adjustment type seat belt system, such as body type recognition deviation and adjustment response lag, the embodiment provides a car seat belt transverse position adaptive adjustment system, which can accurately identify the body type of the occupant, determine the shoulder position of the occupant, automatically adjust the position of the seat belt on the shoulder, and predict the posture of the occupant at the next moment according to the posture of the occupant and the driving state of the vehicle, so as to respond to the adjustment of the position of the seat belt in advance, and ensure the comfort and safety of the occupant.

[0027] The car seat belt transverse position adaptive adjustment system provided by the embodiment, as shown in Figure 1 and Figure 2 , specifically includes a transverse position adjustment mechanism, a multi-modal sensing module and a controller.

[0028] The transverse position adjustment mechanism includes a slide rail 3 arranged on the seat back and an adjustment execution mechanism 2 moving along the slide rail, and the seat belt webbing is fixed on the open guide ring 7 arranged on the adjustment execution mechanism. Under normal circumstances, the seat belt 1 is installed on the set position in the body by bolts, the seat belt 1 webbing is fixed on the seat back 4 through the open guide ring 7 on the adjustment execution mechanism 2, the slide rail 3 is installed on the seat back 4 by bolts, the adjustment execution mechanism is connected with the slide rail 3 through a screw / turbine, the adjustment execution mechanism 2 is internally provided with a controller and a driving motor, the operation of the driving motor is controlled by the controller, the adjustment execution mechanism is driven by the driving motor to move transversely along the slide rail 3, and the transverse movement of the occupant's shoulder seat belt position is driven.

[0029] Through the design of the above transverse position adjustment mechanism, a stable and controllable hardware foundation can be provided for the transverse movement of the seat belt, which breaks through the limitation of the fixed position of the traditional seat belt and realizes the active adjustment of the transverse position of the seat belt.

[0030] Further, the multi-modal sensing module is used to collect multi-modal body shape data and dynamic posture data of the occupant, and the multi-modal sensing module includes an image acquisition unit, a pressure sensing array, and a millimeter wave radar. The image acquisition unit adopts a high-frequency RGB camera 6, which is mounted in the rearview mirror area of the vehicle through bolts, and can collect video frame time sequence images of the seated occupant on all seats, so as to ensure real-time capture of dynamic posture data (such as slight shoulder deviation) of the occupant, and avoid missing of posture information caused by low-frequency acquisition. The pressure sensing array includes a plurality of pressure sensors 5, which are distributed and mounted on the key stress areas of the seat cushion and the seat back through adhesion, and can collect multi-dimensional pressure data of the seated occupant, so as to obtain data such as body weight distribution and sitting posture (such as whether to lean forward), and make up for the defects of only focusing on the appearance shape, especially for more accurate adaptation to special sitting postures of the occupant such as pregnant women. The millimeter wave radar is mounted in the rearview mirror area of the vehicle through bolts, and can obtain three-dimensional contour point cloud data of the shoulder of the seated occupant. The data can fully reflect three-dimensional features such as shoulder thickness and shoulder curvature, solve the problem that the two-dimensional image cannot reflect the three-dimensional shape of the shoulder, and significantly reduce the recognition error of special body types such as obese and narrow-shouldered types, thereby providing key three-dimensional feature support for subsequent static optimal position calculation.

[0031] As shown in Figure 3 , the pressure sensor 5, the in-vehicle camera 6, and the millimeter wave radar are signal acquisition elements, the adjustment actuator 2 is a controller and an actuator, the slide rail 3 is an actuator, and the safety belt 1 and the seat 4 are carriers. The controller built in the adjustment actuator 2 is electrically connected with the multi-modal sensing module, the transverse position adjustment mechanism, and the vehicle controller. The controller determines the static optimal transverse position of the safety belt according to the multi-modal body shape data of the occupant, predicts the posture of the occupant according to the dynamic posture data and the driving state of the vehicle to optimize the transverse position, generates a dynamic adjustment instruction based on the transverse position, and drives the adjustment actuator to move transversely along the slide rail.

[0032] Specifically, after the occupant is seated on the seat 4 and buckles up the safety belt 1, the controller collects multi-modal body shape data and dynamic posture data of the occupant through the multi-modal sensing module, first analyzes the body shape of the occupant to determine the static optimal transverse position.

[0033] Considering that the traditional method of roughly judging two types of body shapes through pressure and two-dimensional images has the problems of large differences in shoulder features of special body types, deviation of unified adjustment logic, and nonlinear mapping relationship between body shape parameters and optimal positions of the safety belt, which is difficult to be covered by traditional rules, the embodiment further refines the body shape classification, accurately identifies different body shapes through multi-dimensional data, and determines accurate shoulder positions for different body shapes to solve the above problems. Specifically, the body shape analysis and transverse position determination step includes: S1.1, preprocessing the collected occupant multi-modal body shape data. Among them, the multi-modal body shape data includes video frame time sequence image, multi-dimensional pressure data and shoulder three-dimensional contour point cloud data: for the video frame time sequence image collected by the in-vehicle camera, Gaussian filtering is used to remove noise, and Canny edge detection operator is used to extract shoulder contour image; for multi-dimensional pressure data, normalization processing is performed, that is, the pressure value detected by the i-th pressure sensor is mapped to the interval [0, 1], and the formula is: ; for the shoulder three-dimensional contour point cloud data, the RANSAC algorithm is used to remove environmental noise points and retain the effective points of the shoulder. Through the above preprocessing steps, the noise in the multi-modal data can be filtered, the accuracy of subsequent feature extraction can be ensured, and the body shape recognition deviation caused by noise can be avoided.

[0034] S1.2, input the preprocessed multi-modal body shape data into the body shape classification model, the body shape classification model includes a feature extraction module, a feature fusion module and a body shape classification module, the image feature, the pressure feature and the point cloud feature are extracted through the feature extraction module, the multiple features are input into the feature fusion module, the fusion feature is obtained through the feature fusion and attention mechanism, and finally the classification result of the occupant body shape is output through the body shape classification module according to the fusion feature.

[0035] Among them, based on the feature extraction module, the 256-dimensional feature vector of the shoulder contour image is extracted through the ResNet-18 network , and the image feature contains shoulder width, shoulder peak distance and other information; the normalized pressure data is processed through the fully connected network, and the 64-dimensional feature vector is output, which reflects the weight distribution and sitting posture; the shoulder effective point cloud data is processed through the PointNet++ network, and the 128-dimensional feature vector is output, which contains shoulder thickness, shoulder curvature and other three-dimensional feature information.

[0036] Further, based on the feature fusion module, the above three types of features are spliced into a 448-dimensional vector , and then input into the attention mechanism module to automatically assign weights to obtain the fusion feature , wherein, , k represents the category.

[0037] Finally, in the body shape classification module, the body shape is divided into 5 categories: children, slim type, standard type, obese type and pregnant women by using softmax regression, and the classification probability is output.

[0038] S1.3, taking the fusion feature as input, outputting the static optimal transverse position of the safety belt through the gradient boosting tree model trained corresponding to the recognized body shape type.

[0039] ​Specifically, for each type of body shape, a separate gradient boosting tree (GBDT) model is trained to extract the static optimal lateral position of the seat belt. The formula is: ; In the above formula, is the weight coefficient, is the mth decision tree, and M is the total number of inputs, is the fusion feature vector.

[0040] By matching the corresponding trained gradient boosting tree model for the identified body type, rather than a single model covering all body types, the physiological characteristics of different body types (such as narrow shoulders in children and shoulder force deviation in pregnant women) can be fully adapted, ensuring the calculation accuracy of the static optimal lateral position, and avoiding shoulder slipping and neck compression when initially wearing.

[0041] Finally, after determining the static optimal lateral position, a dynamic adjustment instruction is generated based on the difference between the current position of the seat belt and the optimal lateral position, driving the adjustment actuator to move laterally along the slide rail to the optimal position to improve comfort. As shown in Figure 4 , for passengers of different body types, the embodiment can adjust the lateral position of the seat belt. Compared with traditional methods, the embodiment can further improve the accuracy of body type classification and effectively reduce the deviation of the seat belt position for special body types (such as pregnant women), achieving more accurate adjustment of the seat belt lateral position according to the passenger's body type.

[0042] Further, considering the real-time changes in passenger posture, existing methods only adjust passively after the body posture changes, which will cause the body to quickly deviate due to inertia during sudden acceleration / turning, and the adjustment lag will easily cause shoulder slipping or neck compression. Moreover, the dynamic interaction between the seat belt and the body is not considered during the adjustment process, which may result in excessive or insufficient adjustment and cause shoulder slipping or neck compression. To address this issue, in the embodiment, the passenger's posture can be predicted based on dynamic posture data and vehicle driving state, and then a PID closed-loop control is performed to optimize the lateral position of the seat belt, thereby avoiding adjustment lag and over-adjustment. The adjustment method specifically includes: S2.1, acquire the video frame time sequence image of the passenger according to the set acquisition frequency, and extract the real-time coordinates of the passenger's shoulder key points in each image , ); simultaneously, the tension of the seat belt fabric , the acceleration , and the steering angle of the vehicle are collected in real time.

[0043] In this embodiment, the posture image of the upper body of the occupant is collected at 60 frames per second, and the real-time coordinates of the shoulder key points are extracted by MediaPipe. At the same time, a tension sensor is installed in the safety belt, and the real-time tension is collected by the tension sensor to reflect the binding force between the safety belt and the human body.

[0044] S2.2, according to the real-time coordinates of the shoulder key points of the occupant, and combining the optimal position of the shoulder key points adapted to the body type, the shoulder offset is obtained, and the shoulder offset sequence is constructed.

[0045] Wherein, after determining the body type by the above-mentioned manner, the corresponding optimal position of the shoulder key points can be determined according to the body type; or according to the determined optimal static transverse position of the safety belt, moving a set distance along the safety belt, the corresponding optimal position of the shoulder key points can be determined. Further, the shoulder offset at different times can be calculated, which is: In the above formula, represents the shoulder offset, , represents the shoulder key point coordinates, , and the optimal position of the shoulder key points adapted to the body type.

[0046] S2.3, taking the shoulder offset sequence of a set time length before the current time and the vehicle driving state sequence as input, the shoulder offset prediction value of the next time period is output by the prediction model based on bidirectional LSTM.

[0047] Specifically, the offset sequence of a set time length (such as 1 second before) before the current time , and the acceleration and the steering angle of the vehicle driving are input into the prediction model based on bidirectional LSTM, the historical trend is captured by the forward LSTM, the future dependence is captured by the backward LSTM, and finally the offset sequence in the future 0.5 seconds is output by the linear activation. , which can be represented as: ， ; Wherein, 、 is a coefficient.

[0048] S2.4, according to the shoulder offset prediction value and the real-time tension, the movement of the adjusting actuator is calculated by PID closed-loop control, and the transverse position of the safety belt is adjusted.

[0049] ​In this embodiment, the movement amount of the adjustment mechanism is calculated according to the predicted offset and the real-time pulling force, so as to ensure that the offset is within the set range, to avoid shoulder slipping or neck tightening, and to ensure that the pulling force is within the comfortable range, to ensure the comfort of the passengers. The calculation process is as follows: Firstly, based on the shoulder offset prediction value and the real-time pulling force , the comprehensive error is calculated as follows: ; Secondly, based on the comprehensive error, the movement amount of the adjustment actuator is calculated through PID closed-loop control as follows: ; In the above formula, , represent the weights, , represent the minimum and maximum values of the pulling force, , , are the PID coefficients.

[0050] By using the above LSTM time series prediction and force feedback PID control method, the passive response is improved to active prediction control, which can realize the closed-loop control of "prediction-adjustment-feedback", and solve the problems of lag and over-regulation. Moreover, the PID closed-loop control combines the offset prediction value and the real-time pulling force, so as to dynamically adjust the movement amount of the adjustment actuator according to the error, to avoid the sudden tightening or small shoulder slipping caused by the excessive adjustment range, and to ensure that the seat belt always fits the shoulder during the dynamic process, and to balance safety and comfort.

[0051] The above system proposed in this embodiment can not only determine the static optimal transverse position through multi-modal body shape data, to solve the position deviation problem of the initial wearing of passengers of different body shapes, such as small and thin type neck tightening and obese type shoulder slipping, but also can predict the passenger posture by combining the dynamic posture and vehicle state, to generate dynamic adjustment instructions, to avoid the hysteresis of the existing passive adjustment, and to fundamentally balance safety and comfort.

[0052] Embodiment Two The embodiment provides an automobile seat belt transverse position self-adaptive adjustment method, which is realized based on the automobile seat belt transverse position self-adaptive adjustment system proposed in embodiment one, and comprises the following steps: After the passenger is seated and the seat belt is buckled, the multi-modal body shape data and dynamic posture data of the passenger are collected in real time; According to the multi-modal body shape data of the passenger, the static optimal transverse position is determined, the dynamic adjustment instruction is generated based on the transverse position, and the adjustment actuator is driven to move transversely along the slide rail; During the starting and driving of the vehicle, the occupant posture is predicted to optimize the lateral position according to the real-time collected dynamic posture data and the driving state of the vehicle, and then the dynamic adjustment instruction is generated based on the lateral position to drive the adjustment execution mechanism to move along the slide rail in the lateral direction.

[0053] Embodiment three The embodiment also provides a vehicle comprising the lateral position adaptive adjustment system of the vehicle safety belt proposed in the embodiment one or performing the adaptive adjustment method of the lateral position of the vehicle safety belt proposed in the embodiment two.

[0054] The steps and methods involved in the above embodiments two to three correspond to the embodiment one, and the specific implementation can refer to the related description part of the embodiment one.

[0055] Those skilled in the art should understand that the above-mentioned modules or steps of the present application can be realized by a general computer device, alternatively, they can be realized by program codes executable by a computing device, so that they can be stored in a storage device and executed by a computing device, or they can be respectively manufactured into integrated circuit modules, or a plurality of modules or steps among them can be manufactured into a single integrated circuit module. The present application is not limited to any specific combination of hardware and software.

[0056] The above description is only the preferred embodiments of the present application, and the specific embodiments of the present application are described in combination with the drawings, but it is not a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications or changes made on the basis of the technical solutions of the present application without creative labor are still within the protection scope of the present application.

Claims

1. A vehicle seatbelt lateral position adaptive adjustment system, characterized in that, include: The lateral position adjustment mechanism includes a slide rail mounted on the seat back and an adjustment actuator that moves along the slide rail. The seat belt webbing is fixed by an open guide ring mounted on the adjustment actuator. The multimodal sensing module is used to collect occupant multimodal body shape data and dynamic posture data; The controller is electrically connected to the multimodal sensing module, the lateral position adjustment mechanism, and the vehicle controller. It is used to determine the static optimal lateral position of the seat belt based on the occupant's multimodal body shape data, and to predict the occupant's posture based on dynamic posture data and vehicle driving status to optimize the lateral position. Then, it generates dynamic adjustment commands based on the lateral position to drive the adjustment actuator to move laterally along the slide rail.

2. The adaptive adjustment system for the lateral position of a car seatbelt as described in claim 1, characterized in that, The multimodal sensing module includes: The image acquisition unit uses a high-frequency RGB camera, which is installed in the rearview mirror area inside the vehicle to capture video frame time sequence images of seated occupants. The pressure sensor array, which includes several pressure sensors, is distributed in key stress areas of the seat cushion and seat back to collect multi-dimensional pressure data of seated occupants. Millimeter-wave radar is installed in the rearview mirror area inside the vehicle to acquire three-dimensional contour point cloud data of the shoulders of seated occupants.

3. The adaptive adjustment system for the lateral position of a car seatbelt as described in claim 1, characterized in that, Based on occupant multimodal body shape data, the optimal static lateral position is determined, including: Preprocess the multimodal body shape data of the occupants; The preprocessed multimodal body shape data is input into the body shape classification model, and image features, pressure features and point cloud features are extracted respectively. After feature fusion and attention mechanism, fused features are obtained, and the classification results of occupant body shape are output based on the fused features. Using the fused features as input, the static optimal lateral position of the seat belt is output through the gradient boosting tree model trained according to the identified body type.

4. The adaptive adjustment system for the lateral position of a car seatbelt as described in claim 3, characterized in that, The multimodal body shape data includes video frame time-series images, multidimensional stress data, and shoulder three-dimensional contour point cloud data; Preprocessing of occupant multimodal body shape data includes: For the time-series images of video frames, Gaussian filtering is used to remove noise, and the Canny edge detection operator is used to extract the shoulder contour image; For multidimensional stress data, normalization processing is performed; For the 3D contour point cloud data of the shoulder, the RANSAC algorithm is used to remove environmental noise points and retain the valid points of the shoulder.

5. The adaptive adjustment system for the lateral position of a car seatbelt as described in claim 3, characterized in that, The occupant body types are categorized into children, slim / small, standard, obese, and pregnant women. An independent gradient boosting tree model is trained for each body type. Using the extracted fusion features as input, the model outputs the static optimal lateral position of the seatbelt. The calculation formula is as follows: ； In the above formula, These are the weighting coefficients. Let m be the m-th decision tree. M For the total number of losses, To fuse feature vectors.

6. The adaptive adjustment system for the lateral position of a car seatbelt as described in claim 1, characterized in that, Based on dynamic attitude data and vehicle driving status, the occupant posture is predicted, and then optimized through PID closed-loop control to optimize the lateral position of the seat belt, including: The system acquires video frame time-series images of the occupants based on a set acquisition frequency, and extracts the real-time coordinates of key points on the occupants' shoulders in each image; at the same time, it also acquires the tension of the seat belt webbing, as well as the vehicle's acceleration and steering angle in real time. Based on the real-time coordinates of the occupant's shoulder key points and the optimal position of the shoulder key points for body shape adaptation, the shoulder offset is obtained, and a shoulder offset sequence is constructed. Using the shoulder offset sequence and vehicle driving state sequence of a set duration before the current moment as input, the predicted shoulder offset value for the next time period is output through a prediction model based on bidirectional LSTM. Based on the predicted shoulder offset and real-time tension, PID closed-loop control is used to calculate and adjust the movement of the actuator to adjust the lateral position of the seat belt.

7. The adaptive adjustment system for the lateral position of a car seatbelt as described in claim 6, characterized in that, The shoulder offset is calculated as follows: ； In the above formula, Indicates the shoulder offset, ( , ) represents the coordinates of the key points on the shoulder. , () indicates the optimal position of the shoulder key points for body type adaptation.

8. The adaptive adjustment system for the lateral position of a car seatbelt as described in claim 6, characterized in that, Based on the predicted shoulder offset and real-time tension, PID closed-loop control is used to calculate the movement of the regulating actuator, as follows: Based on predicted shoulder offset value and real-time tension The overall error is calculated as follows: ； Based on the comprehensive error, the movement of the regulating actuator is calculated using PID closed-loop control, as follows: ； In the above formula, 、 Indicates weight, 、 Indicates the minimum and maximum values ​​of the tension. 、 、 These are the PID coefficients.

9. A method for adaptive adjustment of the lateral position of a car seat belt, characterized in that, Based on the adaptive adjustment system for the lateral position of a car seat belt according to any one of claims 1-8, it includes: After the occupants are seated and fastened their seat belts, multimodal body shape data and dynamic posture data of the occupants are collected in real time. Based on the occupant's multimodal body shape data, the optimal static lateral position is determined, and a dynamic adjustment command is generated based on the lateral position to drive the adjustment actuator to move laterally along the slide rail; During vehicle startup and driving, based on real-time collected dynamic attitude data and vehicle driving status, the occupant posture is predicted to optimize the lateral position. Then, based on the lateral position, a dynamic adjustment command is generated to drive the adjustment actuator to move laterally along the slide rail.

10. A vehicle, characterized in that, Includes the vehicle seat belt lateral position adaptive adjustment system according to any one of claims 1-8, or performs the vehicle seat belt lateral position adaptive adjustment method according to claim 9.

Images