A pose prediction method and device for driving a digital human model construction

Abstract

The application provides a posture prediction method and device for driving digital human model construction, and relates to the technical field of vehicle ergonomics. The method is based on a prediction sample data set, and uses a posture prediction model to solve occupant posture data set. The posture prediction model comprises an anchor point positioning sub-model, a spine reconstruction sub-model and a limb analysis sub-model. The anchor point positioning sub-model is based on occupant physical parameters and vehicle interior parameters to solve hip joint characteristic parameters and eye point characteristic parameters. The spine reconstruction sub-model is based on occupant physical parameters, hip joint characteristic parameters and eye point characteristic parameters to solve spine characteristic parameters. The limb analysis sub-model is based on occupant physical parameters, vehicle interior parameters, hip joint characteristic parameters and spine characteristic parameters to solve shoulder joint characteristic parameters and limb joint characteristic parameters. The application has the beneficial effect of decoupling posture prediction into anchor point positioning, spine reconstruction and limb analysis, and connecting them in series through clear mathematical mapping, thereby improving posture prediction accuracy and efficiency.

Description

Technical Field

[0001] This application relates to the field of vehicle ergonomics technology, and more specifically, to a posture prediction method and apparatus for constructing a digital human driving model. Background Technology

[0002] With the widespread application of 3D computer-aided design technology in the automotive industry, digital human models have become a core tool for vehicle interior development. Currently, the methods for determining the pose of digital human models used in vehicle interior design mainly suffer from the following shortcomings: (1) Poor prediction accuracy Current mainstream posture prediction methods are significantly inadequate in characterizing driver adaptive behavior. On the one hand, most posture prediction models pre-set elbow angles, knee angles, and trunk tilt angles based on empirical values ​​of "comfort angles." However, experimental observations show that drivers employ complex joint coordination strategies when faced with different vehicle layouts. For example, some studies indicate that when the steering wheel position changes, drivers primarily adapt by adjusting their knee joints (approximately 16°) and elbow joints (approximately 26°), while the trunk tilt angle only changes by about 3°. This nonlinear joint trade-off mechanism is difficult to reproduce using fixed empirical angle values, leading to a systematic deviation between predicted and actual postures. On the other hand, physiological variables such as age and body mass index are generally excluded from posture prediction models. The modeling foundation is mostly based on fixed percentile human templates (such as the 50th percentile of males), which essentially assumes that "people of the same height will have the same sitting posture." However, existing studies have shown that as drivers age from 20 to 80, the lumbar spine shifts posteriorly by an average of approximately 22 mm, the eyepoint height decreases by approximately 12 mm, and the pelvis tends to tilt posteriorly. Other studies have demonstrated a clear correlation between body mass index and pelvic rotation angle and lumbar support belt position. The failure to consider these physiological variations leads to a significant decrease in the accuracy of existing posture prediction models when targeting older and obese individuals.

[0003] (2) The algorithm is not transparent While some commercial software has built-in posture prediction functions, its algorithms are based on proprietary data and closed optimization logic. Users cannot obtain its mathematical expression, making it difficult to verify its prediction accuracy in specific vehicle models or populations in specific regions. At the same time, its algorithms are deeply bound to the kinematic chain of specific human models, and the prediction results cannot be ported to other simulation platforms, resulting in a technological lock-in.

[0004] (3) Low computational efficiency Some existing studies solve multiple joint degrees of freedom, such as elbows, knees, and torsos, simultaneously through multi-objective optimization. However, this method has a slow convergence speed and is prone to getting trapped in local optima, with a single prediction typically taking more than a second. This efficiency bottleneck makes it difficult to support the tens of thousands of Monte Carlo simulations required in the early stages of vehicle development, which greatly limits the practical value of this type of method in engineering practice.

[0005] In summary, current technologies lack a comprehensive solution that can balance prediction accuracy, algorithmic transparency, and computational efficiency. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a posture prediction method and apparatus for constructing a digital human driving model.

[0007] The technical solution adopted by this invention is as follows: Firstly, a posture prediction method for constructing a driving digital human model is provided, the posture prediction method comprising calculating an occupant posture dataset using a posture prediction model based on a prediction sample dataset; wherein: The prediction sample dataset includes occupant vital signs parameters and vehicle interior parameters; The occupant posture dataset includes eye point feature parameters, shoulder joint feature parameters, spine feature parameters, hip joint feature parameters, and limb joint feature parameters. The posture prediction model includes an anchor point positioning sub-model, a spine reconstruction sub-model, and a limb analysis sub-model. The anchor point positioning sub-model is configured to calculate the hip joint feature parameters and the eye point feature parameters based on the occupant's vital signs parameters and the vehicle interior parameters. The spine reconstruction sub-model is configured to calculate the spine feature parameters based on the occupant's vital signs parameters, the hip joint feature parameters, and the eye point feature parameters. The limb analysis sub-model is configured to calculate the shoulder joint feature parameters and the limb joint feature parameters based on the occupant's vital signs parameters, the vehicle interior parameters, the hip joint feature parameters, and the spine feature parameters.

[0008] Furthermore, the hip joint characteristic parameters include the position of the geometric center of the hip joint; The eyepoint feature parameters include the eyepoint position and the relative offset of the eyepoint, wherein the relative offset of the eyepoint is the positional offset of the eyepoint relative to the geometric center of the hip joint; The anchor point positioning sub-model includes a first solution mapping, a second solution mapping, and a third solution mapping. The first solution mapping is based on statistical regression calculation, using a first variable as input, to map the position of the geometric center of the hip joint. The second solution mapping is based on statistical regression calculation, using a second variable as input, to map the relative offset of the eye point. The third solution mapping is based on vector offset calculation, using the position of the geometric center of the hip joint and the relative offset of the eye point as input, to map the position of the eye point. The first variable independently includes at least a portion of the occupant vital signs parameters and at least a portion of the vehicle interior parameters; The second variable independently includes at least a portion of the occupant vital signs parameters and at least a portion of the vehicle interior parameters.

[0009] Furthermore, the predicted sample dataset also includes solution constraint parameters, including the load application time; The anchor point positioning sub-model also includes a correction sub-mapping, which is based on statistical regression and uses a third variable as input to map the seat cushion compression deformation variable. The third solution mapping is based on vector offset operation, taking the geometric center position of the hip joint, the relative offset of the eye point, and the compression deformation of the seat cushion as inputs to map the eye point position; The third variable independently includes at least a portion of the occupant vital signs parameters, at least a portion of the vehicle interior parameters, and the solution constraint parameters.

[0010] Furthermore, the spinal characteristic parameters include spinal length, spinal angle, and spinal joint position; The spinal reconstruction sub-model includes a fourth solution mapping, a fifth solution mapping, and a sixth solution mapping. The fourth solution mapping is based on parametric scaling of the human spine, using a fourth variable as input to map the spine length. The fifth solution mapping is based on vector offset operations and spinal kinematic chain constraints, using the hip joint geometric center position, the eye point position, and the spine length as input to map the spine angle. The sixth solution mapping is based on vector offset operations, using the hip joint geometric center position, the spine length, and the spine angle as input to map the spine joint position. The fourth variable independently includes at least a portion of the occupant vital signs parameters.

[0011] Furthermore, the spinal length includes the length of the cervical vertebrae, thoracic vertebrae, lumbar vertebrae, and pelvic segments; The spinal angles include the overall trunk tilt angle, cervical spine tilt angle, thoracic spine tilt angle, lumbar spine tilt angle, and pelvic segment tilt angle; The spinal joint locations include the C7 / T1 position, the T12 / L1 position, and the L5 / S1 position; The fifth solution mapping includes: based on vector offset operation and the spinal kinematic chain constraint, mapping the overall trunk tilt angle using the geometric center position of the hip joint, the eye point position, and the spine length as input; and based on the spinal kinematic chain constraint, mapping the cervical spine tilt angle, the thoracic spine tilt angle, the lumbar spine tilt angle, and the pelvic segment tilt angle using the overall trunk tilt angle as input.

[0012] Furthermore, the statistical regression coefficients of the first solution mapping and the second solution mapping are determined by performing statistical regression fitting on the attitude sample dataset; The posture sample dataset includes at least prior occupant vital signs parameters, prior vehicle interior parameters, and prior sample occupant surface landmark parameters; the prior occupant surface landmark parameters include at least the positions of the left anterior superior iliac spine, the right anterior superior iliac spine, the pubic symphysis, the left posterior superior iliac spine, and the right posterior superior iliac spine, which are used to determine the prior geometric center position of the hip joint through the first center localization algorithm. The first center localization algorithm includes: establishing a pelvic coordinate system using the prior occupant's surface landmark parameters; mapping the prior occupant's skeletal joint parameters and skeletal pelvic coordinate system based on a first positioning space mapping, using the body mass index in the prior occupant's vital signs as input; and mapping the prior hip joint's geometric center position based on a first positioning scaling mapping, using the prior occupant's skeletal landmark parameters as input.

[0013] Furthermore, the spinal kinematic chain constraint of the fifth solution mapping is determined by the spinal motion allocation coefficient, which is determined by performing intra-individual statistical regression fitting on the posture sample dataset. The prior occupant surface landmark parameters include at least the C7 spinous process position, T8 spinous process position, T12 spinous process position, suprasternal notch position, and xiphoid process position, which are used to determine the prior spinal joint position through the second center localization algorithm. The prior spinal joint position includes the prior C7 / T1 position, the prior T12 / L1 position, and the prior L5 / S1 position. The second center localization algorithm includes: constructing a two-segment spinal kinematic model including the thoracic and lumbar vertebrae using the prior occupant's vital signs parameters and the prior occupant's surface landmark parameters; calculating the theoretical surface landmark parameters based on forward kinematics operations, using the two-segment spinal kinematic model as input; and calculating the prior spinal joint positions with the objective of minimizing the distance error between the theoretical surface landmark parameters and the prior occupant's surface landmark parameters. The method for determining the spinal motion allocation coefficient includes: calculating the prior overall trunk tilt angle and the prior spinal segment tilt angle under each posture sample based on the prior spinal joint position; taking an individual as a unit, determining the response ratio of the prior spinal segment tilt angle to the prior overall trunk tilt angle relative to the prior overall trunk tilt angle through statistical regression fitting based on the change relationship between the prior overall trunk tilt angle and the prior spinal segment tilt angle of each individual under each posture sample, and using this as the spinal motion allocation coefficient.

[0014] Furthermore, the shoulder joint characteristic parameters include the acromion point position; The characteristic parameters of the limb joints include wrist joint position, elbow joint position, elbow angle value, knee joint position, knee angle value, and ankle joint position; The limb analysis sub-model includes a seventh solution mapping, an eighth solution mapping, and a ninth solution mapping. The seventh solution mapping is based on vector offset operation, using the fifth variable as input, to map the acromion point position. The eighth solution mapping is based on parametric scaling of the human upper limb and upper limb kinematic chain constraints, using the sixth variable as input, to map the wrist joint position, the elbow joint position, and the elbow angle value. The ninth solution mapping is based on parametric scaling of the human lower limb and lower limb kinematic chain constraints, using the seventh variable as input, to map the knee joint position, the knee angle value, and the ankle joint position. The fifth variable independently includes at least a portion of the C7 / T1 position and the occupant vital signs parameters; The sixth variable independently includes the acromion point location, at least a portion of the occupant vital signs parameters, and at least a portion of the vehicle interior parameters; The seventh variable independently includes the geometric center position of the hip joint, at least a portion of the occupant vital signs parameters, and at least a portion of the vehicle interior parameters.

[0015] Furthermore, the occupant vital signs parameters include gender, height, sitting height, sitting height ratio, weight, body mass index, age, and an interactive item, wherein the interactive item is the product of the body mass index and the age; The vehicle interior parameters include seat height, seat cushion angle, seat cushion filling density, seat seating reference point position, steering wheel center position, steering wheel diameter, steering wheel tilt angle, pedal center position, and pedal tilt angle.

[0016] Secondly, a posture prediction device for constructing a driving digital human model is provided, wherein the posture prediction device is equipped with a computer program, which is executed by a processor to implement the above-described posture prediction method for constructing a driving digital human model.

[0017] The advantages and positive effects of this invention are due to the adoption of the above-mentioned technical solution: (1) By decoupling whole-body posture prediction into three progressive stages—anchor point localization, spinal reconstruction, and limb analysis—and connecting each stage with a clear mathematical mapping, the convergence instability and computational inefficiency problems caused by traditional multivariate coupled solutions are avoided. Specifically: macroscopic anchor point prediction uses cascaded regression to directly lock the geometric center position of the hip joint and the eye point position, providing reliable boundary conditions for subsequent posture calculation; spinal reconstruction introduces individualized motion allocation coefficients, reducing the dimensionality of multi-segment spinal collaborative adjustment to a single degree of freedom solution, significantly simplifying the calculation while ensuring biomechanical rationality; limb analysis is based on a two-link kinematic chain and closed geometric solution to quickly solve the pose of the limb joints. The output of each stage can be independently verified and traced, overcoming the defect that the prediction results of traditional black-box algorithms are difficult to interpret.

[0018] (2) This invention can effectively reflect the influence of different population characteristics on driving posture. Specifically: the anchor point positioning sub-model incorporates the interaction term between body mass index and age, and the sitting height ratio variable, so that the prediction results can capture the influence of changes in soft tissue distribution and differences in trunk and lower limb proportions caused by aging on sitting posture; the spinal reconstruction sub-model constrains the tilt angle response ratio of each segment of the spine with individualized motion allocation coefficients to ensure that the spinal curvature of occupants of different body types is reasonably reproduced. Therefore, this invention can maintain high prediction consistency when facing occupant groups of different ages and body types, and is applicable to diverse posture analysis scenarios including the elderly and obese people.

[0019] (3) In terms of engineering applicability, each step of this invention uses regression equations or closed analytical formulas, eliminating the need for multi-degree-of-freedom iterative optimization. The central processing unit's computation time for a single full-body posture prediction is less than 0.005s, which can meet the batch processing requirements of tens of thousands of Monte Carlo simulations in the early stages of vehicle development. Furthermore, this invention presents open mathematical equations, allowing all mapping relationships and coefficients to be independently reproduced and verified. The prediction results are output as standard skeletal joint coordinates, unrestricted by the kinematics chain definition of a specific digital human platform, facilitating cross-platform portability. In addition, this method introduces a coordinate compensation mechanism for seat compression deformation in real-world vehicle environments, ensuring consistency between laboratory predictions and actual road test results, further guaranteeing its engineering reliability.

[0020] By comprehensively applying the above-mentioned technical means, this invention achieves simultaneous improvement in three dimensions of attitude prediction method: prediction accuracy, algorithm transparency, and computational efficiency, providing reliable technical support for human-machine engineering design and safety simulation analysis in the early stages of vehicle development. Detailed Implementation

[0021] The technical solutions in the embodiments of this disclosure will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0022] This invention provides a method for constructing a posture prediction model, which can be used for posture prediction during the construction of a driving digital human model. The method for constructing the posture prediction model includes the following steps: S11. Collect diverse attitude sample data, and construct an attitude sample dataset after preprocessing the attitude sample data.

[0023] Specifically, the posture sample data includes prior occupant vital signs parameters, prior vehicle interior parameters, and prior occupant surface landmark parameters.

[0024] Prior occupant physical characteristics are used to characterize the inherent human attributes of prior occupants, including at least the prior occupant's height, sitting height, sitting-to-height ratio, weight, age, body mass index (BMI), and interaction terms, where the interaction term is the product of body mass index and age.

[0025] The interior parameters of the test vehicle are used to characterize the vehicle cabin environment in which the test occupants are located, including at least the seat height, seat cushion angle, seat reference point position, steering wheel center position, steering wheel diameter, steering wheel tilt angle, pedal center position, and pedal tilt angle of the test vehicle.

[0026] The prior occupant surface landmark parameters are the posture response data of the prior occupant under the prior vehicle interior parameters, used to quantitatively describe the actual posture distribution of the prior occupant in a specific vehicle cabin environment; the parameters include at least the three-dimensional coordinates of 14 surface landmarks of the prior occupant under the prior vehicle interior parameters, the coordinate values ​​are measured in the global spatial coordinate system, the 14 surface landmarks are: left canthus, right canthus, spinous process of the seventh cervical vertebra, spinous process of the eighth thoracic vertebra, spinous process of the twelfth thoracic vertebra, suprasternal notch, xiphoid process, left anterior superior iliac spine, right anterior superior iliac spine, left posterior superior iliac spine, right posterior superior iliac spine, pubic symphysis, left acromion, and right acromion.

[0027] The method of collecting parameters of prior occupant body surface markers is a well-known technology in the field. As a feasible implementation method, infrared reflective markers can be pasted at each body surface marker, and data collection can be completed with the help of an infrared motion capture system.

[0028] To facilitate understanding and clear articulation of the technical solutions, the meanings of the above terms are explained as follows: The global spatial coordinate system is structured as follows: the X-axis runs along the vehicle's length, the Y-axis along its width, and the Z-axis along its height. The origin of the coordinate system is the vehicle's design reference point. The X-axis position of this point is determined by the centerline of the front wheels, the Y-axis position by the vehicle's plane of symmetry, and the Z-axis position by the ground or the upper surface of the vehicle frame. The positive directions of each axis of the global spatial coordinate system can be selected according to specific application requirements. As an example, in this embodiment, the vehicle's forward direction is defined as the positive X-axis direction, the direction pointing to the right of the occupant is defined as the positive Y-axis direction, and the direction pointing to the top of the vehicle is defined as the positive Z-axis direction.

[0029] Height: The vertical distance from the top of the head to the horizontal plane of the feet when standing.

[0030] Sitting height: The vertical distance from the top of the head to the lowest point of the ischial tuberosities on both sides of the body when sitting.

[0031] Sitting height ratio: The ratio of sitting height to body height.

[0032] Body Mass Index (BMI): The ratio of weight to the square of height.

[0033] Seat height: The offset of the seat reference point (also known as the seat reference point or H point) from the overall vehicle design reference point along the vehicle height direction; expressed as the coordinate value of the seat reference point in the Z-axis direction of the global spatial coordinate system; unit: mm.

[0034] Seat seating reference point position: represented by the three-dimensional coordinates of the seat seating reference point in the global spatial coordinate system; unit: mm.

[0035] Steering wheel center position: represented by the three-dimensional coordinates of the geometric center point of the steering wheel in the global spatial coordinate system; unit: mm.

[0036] Pedal surface center position: represented by the three-dimensional coordinates of the pedal surface geometric center in the global spatial coordinate system; unit: mm.

[0037] Seat cushion angle: The angle between the upper surface of the seat cushion and the horizontal plane (XY plane in the global spatial coordinate system) in the vehicle's sagittal plane (i.e., the XZ plane). A positive value is taken when the front end of the seat cushion is higher than the rear end, and a negative value is taken otherwise; Unit: degrees.

[0038] Steering wheel tilt angle: The angle between the upper plane of the steering wheel and the horizontal plane (XY plane in the global spatial coordinate system) in the vehicle's sagittal plane (i.e., the XZ plane in the global spatial coordinate system). It is positive when the front end of the steering wheel is higher than the rear end of the steering wheel; unit: degree.

[0039] Pedal surface inclination angle: The angle between the upper surface of the pedal and the horizontal plane (i.e., the XZ plane in the global spatial coordinate system) in the vehicle's sagittal plane (i.e., the XZ plane in the global spatial coordinate system). It is positive when the front end of the pedal is higher than the rear end of the pedal; unit: degree.

[0040] Corner of the eye (left and right): Points on the unpressed skin surface at the junction of the outer sides of the upper and lower eyelids.

[0041] The spinous process of the seventh cervical vertebra: the depression on the skin surface corresponding to the most prominent spinous process on the midline of the neck root.

[0042] Spinous process of the eighth thoracic vertebra: The depression on the skin surface behind the spinous process of the eighth thoracic vertebra.

[0043] The spinous process of the twelfth thoracic vertebra: the point on the skin surface in the depression behind the spinous process of the twelfth thoracic vertebra.

[0044] Suprasternal notch: The depression in the middle of the upper edge of the manubrium of the sternum.

[0045] Xiphoid process: The skin surface point corresponding to the lowest protruding part of the sternum (triangular in shape, with its base connected to the body of the sternum).

[0046] Anterior superior iliac spine (left and right): Palpate forward along the iliac crest to its most prominent anterior point.

[0047] Posterior superior iliac spine (left and right): Palpate along the edge of the ilium to the most posterior protrusion.

[0048] Pubic symphysis: palpate down along the midline of the abdomen to the most prominent point at the upper edge of the pubic symphysis.

[0049] Acromion (left and right): Uncompressed skin surface points obtained by touching the anterior part of the lateral border of the acromion of the scapula.

[0050] The preprocessing steps for the above posture sample data include, but are not limited to: Coordinate System 1, data cleaning and interpolation, and feature standardization. Coordinate System 1 involves translating the origin of all samples to the vehicle design reference point, unifying the X, Y, and Z axis directions, and normalizing the coordinates of the seat seating reference point to the same reference position; Data cleaning and interpolation involves using the Z-score method to remove outliers, filling missing data with cubic spline interpolation, and supplementing with low-pass filtering for noise reduction; Feature standardization involves performing Z-score standardization on continuous variables such as height, BMI, and age, and constructing interaction terms such as BMI×age as needed; After processing, the data is divided into training and validation sets in an 8:2 ratio.

[0051] Due to the complexity of the soft tissue covering the human body and the complexity of the movement of the spinal joint chains, the prior occupant surface landmark parameters directly obtained through the infrared motion capture system cannot all be directly used for training the posture prediction model. For example, the spatial position of some surface landmarks is affected by factors such as individual subcutaneous fat thickness, soft tissue deformation, and offset of surface label placement, resulting in a systematic deviation from the skeletal joint positions that reflect the true skeletal posture. If the prior occupant surface landmark parameters are directly used as the training ground value, the model will learn non-skeletal displacements caused by soft tissue, thus affecting the biomechanical rationality of the predicted posture. For another example, the kinematic reference point for each segment of the human spine is located at the geometric center of the vertebral body, not the spinous process that can be touched on the body surface. The latter is only a bony protrusion on the side of the vertebral body. When describing changes in occupant trunk posture, the former must be used as the reference point for the bending deformation of each segment in order to accurately reflect the true geometric shape and kinematic characteristics of the spine. Therefore, the present invention sets step S12 to map the part of the prior occupant's body surface landmark parameters that needs to be converted to the corresponding prior occupant's skeletal joint parameters, so as to restore the prior occupant's real skeletal posture in the corresponding vehicle cabin environment.

[0052] S12. Using the body surface-skeleton mapping algorithm, the parameters of the prior occupant's body surface landmarks are converted into the parameters of the prior occupant's skeletal joints.

[0053] Specifically, the body surface-skeleton mapping algorithm includes a first center localization algorithm and a second center localization algorithm. The first center localization algorithm is used to reproduce the skeletal joint parameters of the pelvic region based on the surface landmark parameters of the prior occupant's pelvic region and determine the geometric center position of the prior hip joint, thereby generating ground truth data that can be used for model training. The second center localization algorithm is used to reproduce the skeletal joint parameters of the spine region based on the surface landmark parameters of the prior occupant's spine region, thereby generating ground truth data that can be used for model training.

[0054] The steps for reproducing the skeletal joint parameters of the prior occupant's pelvic region and determining the geometric center position of the prior hip joint using the first-center localization algorithm include: For any prior occupant, a pelvic coordinate system is established using the positions of the left anterior superior iliac spine, right anterior superior iliac spine, and pubic symphysis from the body surface landmark parameters. Based on the first positioning space mapping, and using the prior occupant's body mass index as input, the positions of the left anterior superior iliac spine, right anterior superior iliac spine, pubic symphysis, left posterior superior iliac spine, right posterior superior iliac spine, and the skeletal pelvic coordinate system of the prior occupant are mapped. Then, based on the first positioning scaling mapping, and using the positions of the left anterior superior iliac spine, right anterior superior iliac spine, left posterior superior iliac spine, right posterior superior iliac spine, and pubic symphysis of the prior occupant as input, the geometric center position of the prior hip joint is mapped.

[0055] It should be noted that, based on the common simplification practices in the field of ergonomics, the prediction of human torso posture is limited to analysis within the sagittal plane. Therefore, the prior geometric center of the hip joint, as referred to herein and hereinafter, is uniformly defined as the midpoint of the line connecting the center of the left and right acetabular balls.

[0056] The method for establishing the above-mentioned surface pelvic coordinate system includes: taking the direction of the line connecting the positions of the left and right anterior superior iliac spines as Y. S axial direction, Y S The positive direction of the axis points to the left side of the body; the plane containing the positions of the left and right anterior superior iliac spines is perpendicular to the Y-axis. S The axis and the direction pointing to the pubic symphysis is Z. S The negative direction of the axis; the direction determined by the three-dimensional orthogonal right-hand rule is X. S Axial direction, X S The positive axis points towards the front of the human pelvis; a surface pelvic coordinate system {O} is established with the pubic symphysis as the origin. S ,X S ,Y S Z S}

[0057] Because there are significant differences in soft tissue thickness at different BMI levels (p < 0.001), the first positioning spatial mapping provided by this invention uses the human anatomical vector direction as a reference and the soft tissue thickness vector size based on BMI stratification as a compensation basis to achieve accurate mapping from surface landmarks to skeletal joints. Specifically, the first positioning spatial mapping includes the following mapping relationships: (1) The position of the left anterior superior iliac spine is used as input, the first vector direction is used as spatial guide, and the first soft tissue thickness vector size based on BMI layering is used as offset amplitude to map the position of the left anterior superior iliac spine of the prior occupant; the mapping of the position of the right anterior superior iliac spine is performed in the same way, mirror symmetrically.

[0058] (2) Using the pubic symphysis position as input, the second vector direction as spatial guide, and the second soft tissue thickness vector size based on BMI layering as offset amplitude, the pubic symphysis bone position of the prior occupant is mapped.

[0059] (3) Using the position of the left posterior superior iliac spine as input, the third vector direction as spatial guide, and the third soft tissue thickness vector size based on BMI layering as offset amplitude, the position of the left posterior superior iliac spine of the prior passenger is mapped; the mapping of the position of the right posterior superior iliac spine is performed in the same way, mirror symmetrically.

[0060] In a preferred embodiment of the present invention, the first vector direction is defined as the local surface normal along the left anterior superior iliac spine, pointing from the outside to the inside. This surface normal is calculated based on local point cloud data within the target point's neighborhood (neighborhood radius 5mm~15mm, adaptively adjusted according to the surface curvature) using the least squares method to fit a local plane. The first soft tissue thickness vector size is set differently according to BMI stratification: when BMI < 25, the first soft tissue thickness vector size is 8mm; when 25 ≤ BMI < 30, the first soft tissue thickness vector size is 12mm; when BMI ≥ 30, the first soft tissue thickness vector size is 15mm.

[0061] In a preferred embodiment of the present invention, the second vector direction is defined as X in the body surface pelvic coordinate system. S -Z S In the plane (i.e., the sagittal plane of the human body), X S Negative direction of axis and Z S The direction of vector composition in the negative direction of the axis, and this direction of vector composition is the same as Z. S The angle between the negative axes is 15° to 25°; this angle range was obtained through morphological measurements and statistics from cadaveric anatomical data, reflecting the actual anatomical characteristic that the soft tissue at the pubic symphysis is mainly distributed on its posteroinferior side. The second soft tissue thickness vector dimension is set differently according to BMI stratification: when BMI < 25, the second soft tissue thickness vector dimension is 12 mm; when 25 ≤ BMI < 30, the second soft tissue thickness vector dimension is 18 mm; when BMI ≥ 30, the second soft tissue thickness vector dimension is 22 mm.

[0062] In a preferred embodiment of the present invention, the third vector direction is defined as the local surface normal along the left posterior superior iliac spine, pointing from the outside to the inside. This surface normal is calculated based on local point cloud data within the target point's neighborhood (neighborhood radius 5mm~15mm, adaptively adjusted according to the surface curvature), using the least squares method to fit a local plane. The third soft tissue thickness vector size is set differently according to BMI stratification: when BMI < 25, the third soft tissue thickness vector size is 10mm; when 25 ≤ BMI < 30, the third soft tissue thickness vector size is 15mm; when BMI ≥ 30, the third soft tissue thickness vector size is 20mm.

[0063] The values ​​of the vector directions and soft tissue thickness vector dimensions mentioned above are all obtained based on statistical analysis of cadaveric anatomical measurement data of a population in a specific region. In practical applications, the measurements can be re-performed and the corresponding mapping parameters determined according to the differences in body shape characteristics of the population in the target region. The scope of protection claimed by this invention should not be considered as limited to the values ​​specifically given above.

[0064] In this embodiment of the invention, the soft tissue thickness data at various surface landmarks are directly obtained using the acupuncture method. When both one-way ANOVA and linear regression tests show a significant correlation between BMI and soft tissue thickness (p < 0.001), the soft tissue thickness vector dimension is taken as the rounded integer value of the mean measurement within the corresponding BMI stratum as the standard value. During mapping, the determination methods for the first, second, and third vector directions differ: the first and third vector directions are taken as the local surface normal of each surface landmark, pointing from the outside to the inside, and vary independently with the individual sample and the body surface location; the second vector direction is a fixed direction in the posterolateral direction within the pelvic coordinate system, used uniformly for all samples and not varying with the individual. This method targets areas with significant individual differences in surface morphology, such as the anterior and posterior superior iliac spines. It employs a local normal that adaptively adjusts to the individual's surface curvature for mapping, accurately tracing the surface contours of bony prominences. Considering the anatomically consistent distribution of soft tissue at the pubic symphysis across the population, a fixed anatomical direction is used for mapping to penetrate along its inherent thickening direction. This differentiated mapping strategy balances individual surface adaptability with anatomical structural constancy, effectively reducing algorithm complexity while maintaining mapping accuracy.

[0065] Based on the aforementioned first positioning space mapping, the three-dimensional spatial coordinates of the prior occupant's left anterior superior iliac spine, right anterior superior iliac spine, pubic symphysis, left posterior superior iliac spine, and right posterior superior iliac spine in the global spatial coordinate system can be obtained. Subsequently, based on the prior occupant's left anterior superior iliac spine, right anterior superior iliac spine, and pubic symphysis positions, and following the aforementioned method for establishing a surface pelvic coordinate system, a skeletal pelvic coordinate system {O} is established. b ,X b ,Y b Z b}; Then, using the positions of the left anterior superior iliac spine, right anterior superior iliac spine, left posterior superior iliac spine, right posterior superior iliac spine, and pubic symphysis as inputs, the position of the geometric center of the prior hip joint is calculated through the first positioning scaling mapping.

[0066] In this embodiment of the invention, the first positioning scaling mapping is as follows: Using a linear scaling positioning method commonly used in the field of biomechanics, and based on the statistical analysis results of regionalized and large-scale cadaver anatomical measurement data, the prior geometric center position of the hip joint is established in the X-axis using geometric scale parameters of the pelvis in the transverse, longitudinal, and vertical dimensions. b Axis, Y b Axis, Z b The first positioning scaling equation in the three directions of the axis.

[0067] Specifically: The prior geometric center of the hip joint is located at X b The coordinates along the axis are HipXb=k b1 ×PD; The prior geometric center of the hip joint is located in Y. b The coordinates along the axis are HipYb=k b2 ×PW; The prior geometric center of the hip joint is located at Z. b The coordinates along the axis are HipZb=k b3 ×PH.

[0068] Among them, PD, PW, and PH are three scale parameters of the pelvis: pelvic depth PD represents the position of the ipsilateral anterior superior iliac spine, and posterior superior iliac spine represents the position of the ipsilateral anterior superior iliac spine within the X-ray range. b The distance along the axial direction, pelvic width PW, represents the position of the left anterior superior iliac spine and the position of the right anterior superior iliac spine in the Y-axis direction. b The distance along the axial direction. The pelvic height PH is the midpoint of the line connecting the left and right anterior superior iliac spines, and the location of the pubic symphysis at the Z-axis. b Distance along the axis; k b1 k is the scaling factor for the first hip positioning. b2 k is the scaling factor for the second hip positioning. b3 This is the scaling factor for the third hip positioning.

[0069] As a preferred embodiment of the present invention, k b1 Take 0.34, k b2 Take 0.38, k b3 The value is set to 0.36. The determination process for the aforementioned three coefficients is as follows: using a publicly available dataset of at least 65 cadaveric pelvic CT scans as samples, the fitted coordinates of the midpoint of the line connecting the left and right acetabular centers are used as the true value of the hip joint geometric center. The positions of the hip joint geometric center are established on the X-ray axis using three scale parameters: pelvic width (PW), pelvic depth (PD), and pelvic height (PH). b Axis, Y b Axis, Z b The linear regression equations in the three axes are used to obtain the standard regression coefficients of the independent variables in each equation through regression analysis, which are denoted as k. b1 k b2 k b3 The values ​​of are all p < 0.001. The method for obtaining the fitted coordinates of the unilateral acetabular center is as follows: Point clouds of the acetabular joint surface are extracted from the 3D pelvic model reconstructed using CT scan datasets. The least squares method is used to fit the optimal sphere, minimizing the sum of the squared distances from all points to the center of the sphere. These center coordinates are the fitted coordinates of the acetabular center. The fitting accuracy is evaluated using the coefficient of determination R², which is typically required to be no less than 0.90.

[0070] It should be noted that the above coefficient values ​​are obtained based on statistical analysis of publicly available anthropometric data of a population in a specific region. In practical applications, measurements can be re-conducted and the corresponding hip positioning scaling coefficients determined according to the differences in body shape characteristics of the population in the target region. The scope of protection claimed by this invention should not be considered as limited to the values ​​specifically given above.

[0071] Based on the aforementioned first positioning scaling mapping, the geometric center position of the prior occupant's hip joint in the skeletal pelvic coordinate system {O} can be obtained. b ,X b ,Y b Z b The three-dimensional coordinates in the coordinate system are then used to determine the coordinate offset between the pubic symphysis bone position and the vehicle design reference point in the global spatial coordinate system. This is achieved through coordinate system transformation, positioning the prior geometric center of the hip joint in the skeletal pelvic coordinate system {O}. b ,X b ,Y b Z b The coordinates in the previous equation are transformed to the global spatial coordinate system, yielding the coordinates HipX (X-axis) and HipZ (Z-axis) of the prior hip joint geometric center in the global spatial coordinate system. Following common simplification practices in ergonomics, the coordinates of the seat reference point in the global spatial coordinate system are taken as the Y-axis coordinates of the prior hip joint geometric center in the global spatial coordinate system. The aforementioned coordinate transformation process is well-known in the field, and its specific analytical formula will not be elaborated upon here. Therefore, even under practical conditions where anatomical and imaging examinations are not feasible, only the surface landmark parameters of the pelvic region of the prior occupant need to be obtained. The first center localization algorithm provided by this invention can accurately reproduce the position of the core skeletal joint, the geometric center of the hip joint, thus providing reliable ground truth data of the pelvic region for the training and validation of the posture prediction model.

[0072] The steps for reproducing the skeletal joint parameters of the spinal region of a prior occupant using the second-center localization algorithm include: For any prior occupant, based on the second positioning scaling mapping, using the prior occupant's height and sitting height as input, the lengths of each segment of the prior spine are mapped, including the prior cervical spine length, prior thoracic spine length, prior lumbar spine length, and prior pelvic segment length. Using the positions of the spinous processes of the seventh cervical vertebra, the eighth thoracic vertebra, the twelfth thoracic vertebra, the suprasternal notch, and the xiphoid process from the prior occupant's body surface landmarks, as well as the lengths of each segment of the prior spine, a thoracic-lumbar two-bar kinematic model is constructed. The model is used to calculate the theoretical surface landmark parameters based on inverse kinematics. The theoretical surface landmark parameters include the theoretical positions of the spinous processes of the seventh cervical vertebra, the eighth thoracic vertebra, the twelfth thoracic vertebra, the suprasternal notch, and the xiphoid process. With the goal of minimizing the distance error between the theoretical surface landmark parameters and the prior occupant surface landmark parameters, the prior spinal joint positions are calculated. The prior spinal joint positions include the prior C7 / T1 position, the prior T12 / L1 position, and the L5 / S1 position.

[0073] In this embodiment of the invention, the second positioning scaling mapping includes: using a linear scaling positioning method commonly used in the field of biomechanics, and based on the statistical analysis results of regionalized and large-scale cadaver anatomical measurement data, establishing a second positioning scaling equation between the length of each segment of the human spine and height and sitting height.

[0074] Specifically: Cervical spine length = K a1 × Sitting high; Thoracic vertebra length = K a2 × Sitting high; Lumbar spine length = K a3 × Sitting high; Pelvic segment length = K a4 × Height.

[0075] Among them, K a1 K is the scaling factor for the first spinal localization. a2 K is the scaling factor for second spinal positioning. a3 K is the scaling factor for the third spine localization. a4 This is the scaling factor for the fourth spine localization.

[0076] As a preferred embodiment of the present invention, K a1 Take 0.143, K a2 Take 0.300, K a3 Take 0.224, K a4The value is set to 0.105. It should be noted that the above coefficient values ​​are based on statistical analysis of cadaveric anatomical measurement data of a population in a specific region. In practical applications, measurements can be re-performed and the corresponding spinal positioning scaling factor can be determined according to the differences in body shape characteristics of the population in the target region. The scope of protection claimed by this invention should not be considered as limited to the values ​​specifically given above.

[0077] Based on the aforementioned second positioning scaling mapping, the prior cervical spine length, prior thoracic spine length, prior lumbar spine length, and prior pelvic segment length can be obtained. Subsequently, based on the positions of the spinous process of the seventh cervical vertebra and the suprasternal notch of the prior occupant in a standard driving posture on a hard reference seat, the prior C7 / T1 position is initially estimated; based on the positions of the spinous processes of the eighth and twelfth thoracic vertebrae, the suprasternal notch, and the xiphoid process of the prior occupant in a standard driving posture on a hard reference seat, the prior T12 / L1 position is initially estimated; based on the preliminary estimation results of the prior C7 / T1 position, the preliminary estimation results of the prior T12 / L1 position, the prior thoracic spine length, and the prior lumbar spine length, a thoracic-lumbar two-bar kinematic model is constructed. Based on the constructed thoracic-lumbar two-bar kinematic model and inverse kinematic algorithm, the theoretical positions of the spinous processes of the seventh cervical vertebra, the eighth thoracic vertebra, the twelfth thoracic vertebra, the suprasternal notch, and the xiphoid process are calculated.

[0078] Since the prior T12 / L1 position, which is directly estimated based solely on prior occupant surface landmark parameters, is affected by soft tissue coverage and has limited estimation accuracy, this invention is based on spinal kinematic chain constraints. It limits the physiological range of angles at each joint of the spine and aims to minimize the distance error between theoretical surface landmark parameters and prior occupant surface landmark parameters. For example, it aims to minimize the sum of squared distance errors between the theoretical and measured positions of five surface landmarks: the spinous processes of the seventh cervical vertebra, the eighth thoracic vertebra, the twelfth thoracic vertebra, the suprasternal notch, and the xiphoid process. This allows for the precise calculation of the prior C7 / T1, T12 / L1, and L5 / S1 positions, thereby obtaining a spinal model that best matches the measured positions of all surface landmarks in the spinal region. Therefore, under the practical condition that anatomical and imaging examinations cannot be performed, it is only necessary to obtain the surface landmark parameters of the spinal region of the prior occupant. The second center localization algorithm provided by this invention can reproduce the skeletal joint parameters of the spinal region of the prior occupant and reasonably reconstruct the spinal curvature of each prior occupant, thereby providing reliable true value data of the spinal region for the training and verification of the posture prediction model.

[0079] S13. Using the posture sample dataset and prior occupant skeletal joint parameters, construct the anchor point positioning sub-model, the spine reconstruction sub-model, and the limb analysis sub-model.

[0080] (1) The necessity of constructing the anchor point positioning sub-model lies in the fact that the pelvis and the eye point respectively constitute the base end and the visual constraint end of the seated human motion chain, and the two are connected through a multi-segment spine. The reconstruction of the spinal posture depends on the spatial boundary conditions at both ends of the spine - the geometric center of the hip joint provides a fixed anchor position for the pelvic end, and the position of the eye point reflects the overall spatial orientation of the head and upper cervical spine. The two together constitute the geometric constraints for solving the spinal deformation. At the same time, the analysis of the limb posture also needs to be based on the pelvic position, extending downward to the lower limbs through the hip joint, and transmitting to the upper limbs through the shoulder girdle. Therefore, it is necessary to use the occupant's vital signs parameters and vehicle interior parameters to prioritize the calculation of hip joint characteristic parameters and eye point characteristic parameters, providing two irreplaceable spatial positioning benchmarks for spinal reconstruction and limb analysis.

[0081] To establish a quantitative mapping relationship between occupant vital signs parameters, vehicle interior parameters, hip joint feature parameters, and eye point feature parameters, this embodiment of the invention constructs a first solution mapping, a second solution mapping, and a third solution mapping in the anchor point positioning sub-model. The first solution mapping is used to solve for the position of the geometric center of the hip joint in the hip joint feature parameters. It uses the occupant's height, body mass index, age, and interaction terms as independent variables on the vital signs side, and seat height, steering wheel center position, and seat cushion angle as independent variables on the interior side. It maps to the position of the geometric center of the hip joint as the dependent variable through a first regression equation. The second solution mapping is used to solve for the relative offset of the eye point in the eye point feature parameters. The relative offset of the eye point is defined as the positional offset of the eye point relative to the geometric center of the hip joint. This second solution mapping uses the occupant's height and seat height ratio as independent variables on the vital signs side, and the steering wheel center position and seat cushion angle as independent variables on the interior side. It maps to the relative offset of the eye point as the dependent variable through a second regression equation. The third solution mapping is used to solve for the eye point position in the eye point feature parameters. It takes the position of the geometric center of the hip joint and the relative offset of the eye point as inputs and maps them to the dependent variable of the eye point position through the first vector operation equation.

[0082] It should be noted that, based on the common simplification practices in the field of ergonomics, the prediction of the human eye position is limited to analysis within the sagittal plane. Therefore, the prior eye point position referred to herein and the eye point position referred to thereafter are uniformly defined as the position of the midpoint of the line connecting the left and right corners of the human eye.

[0083] When the model is deployed and used, the occupant vital signs parameters and vehicle interior parameters are sequentially processed through the first solution mapping, the second solution mapping, and the third solution mapping to complete the calculation of hip joint feature parameters and eye point feature parameters. However, during the model construction stage, the second solution mapping needs to be completed under the premise that the relative offset of the eye point is known. Therefore, the actual construction order of the anchor point positioning sub-model is: first construct the first solution mapping, then construct the third solution mapping, and then construct the second solution mapping.

[0084] When constructing the first solution mapping, the height, body mass index, age, and interaction terms from the prior occupant's physical characteristics parameters, as well as the seat height, steering wheel center position, and seat cushion angle from the prior vehicle interior parameters, are used as independent variables. The prior geometric center position of the hip joint calculated in step S12 is used as the dependent variable. Through multiple linear regression fitting, the first regression equation is obtained as follows: HipX=α0+α1·S+α2·BMI+α3·A+α4·H30+α5·L6_X+α6·L27+α7·(BMI×A) HipZ=α' 0+α' 1·S+α' 2·BMI+α' 3·A+α' 4·H30+α' 5·L6_Z+α' 6·L27+α'7·(BMI×A) Where HipX and HipZ are the coordinates of the prior geometric center of the hip joint in the X and Z axes of the global spatial coordinate system, respectively; S is height; BMI is body mass index; A is age; H30 is seat height; L6_X is the coordinate value of the center position of the steering wheel in the X axis of the global spatial coordinate system; L6_Z is the coordinate value of the center position of the steering wheel in the Z axis of the global spatial coordinate system; L27 is seat cushion angle; BMI×A is the interaction term, i.e., the product of body mass index and age; α0 and α'0 are the intercept terms in the corresponding analytical expressions; α1 and α'1 are the regression coefficients corresponding to height in the corresponding analytical expressions; α2 and α'2 are the regression coefficients corresponding to body mass index in the corresponding analytical expressions; α3 and α'3 are the regression coefficients corresponding to age in the corresponding analytical expressions; α4 and α'4 are the regression coefficients corresponding to seat height in the corresponding analytical expressions; α5 and α'5 are the regression coefficients corresponding to the center position of the steering wheel in the corresponding analytical expressions; α6 and α' α6 represents the regression coefficient corresponding to the seat angle in the corresponding analytical expression; α7 and α'7 represent the regression coefficients corresponding to the interaction term in the corresponding analytical expression.

[0085] When constructing the third solution mapping, the prior eye point position is first calculated using the left and right eye point positions from the prior occupant's surface landmark parameters. The prior eye point position is the midpoint of the line connecting the left and right eye point positions. Then, using the prior eye point position and the prior hip joint geometric center position calculated in step S12, the first vector operation equation is constructed according to the vector operation principle to calculate the relative offset of the prior eye point. The first vector operation equation is as follows: EyeX = HipX + EyeX rel EyeZ = HipZ + EyeZ rel Where EyeX and EyeZ are the coordinates of the prior eye point location along the X and Z axes in the global spatial coordinate system, respectively; EyeXrel EyeZ rel These are the components of the relative offset of the prior eye point in the X and Z axes of the global spatial coordinate system, respectively.

[0086] When constructing the second solution mapping, the height and seat height ratio from the prior occupant's physical characteristics, and the steering wheel center position and seat cushion angle from the prior vehicle interior parameters, are used as independent variables, and the relative offset of the prior eye point is used as the dependent variable. Through multiple linear regression fitting, the second regression equation is obtained as follows: EyeX rel =β0+β1·S+β2·(S / H)+β3·L6_X+β4·L27 EyeZ rel =β' 0+β' 1·S+β' 2·(S / H)+β' 3·L6_Z+β' 4·L27 Where S is height; (S / H) is the seat height ratio; L6_X is the coordinate value of the center position of the steering wheel in the X-axis direction of the global spatial coordinate system; L6_Z is the coordinate value of the center position of the steering wheel in the Z-axis direction of the global spatial coordinate system; L27 is the seat angle; β0 and β'0 are the intercept terms in the corresponding analytical expressions; β1 and β'1 are the regression coefficients corresponding to height in the corresponding analytical expressions; β2 and β'2 are the regression coefficients corresponding to the seat height ratio in the corresponding analytical expressions; β3 and β'3 are the regression coefficients corresponding to the center position of the steering wheel in the corresponding analytical expressions; β4 and β'4 are the regression coefficients corresponding to the seat angle in the corresponding analytical expressions.

[0087] Furthermore, based on the aforementioned common simplified processing methods in the fields of biomechanics and ergonomics, in this embodiment of the invention: the component of the relative offset of the prior eye point in the Y-axis direction of the global spatial coordinate system is taken as zero; the coordinates of the prior hip joint geometric center position and the prior eye point position in the Y-axis direction of the global spatial coordinate system are both directly adopted as the coordinate values ​​of the seat seating reference point position in the Y-axis direction of the global spatial coordinate system, and are uniformly incorporated into each calculation stage as fixed constants.

[0088] In the aforementioned anchor point positioning sub-model, by incorporating the interaction term of the product of body mass index and age into the first regression equation, the influence of changes in soft tissue distribution and muscle tension due to age on the offset of the geometric center of the hip joint is captured when the body mass index is the same. By incorporating the sitting height ratio into the second regression equation, the influence of the difference in the ratio of the trunk to the lower limbs on the positional offset of the eye point relative to the geometric center of the hip joint among occupants of the same height is captured. Through the above settings, a quantitative mapping relationship between occupant vital signs parameters, vehicle interior parameters, hip joint characteristic parameters, and eye point characteristic parameters can be established relatively accurately.

[0089] (2) The necessity of constructing a spinal reconstruction sub-model lies in the fact that the geometric center position of the hip joint and the eye point position output by the anchor point positioning sub-model only limit the boundary constraints at both ends of the spine, which is insufficient to describe the internal bending morphology of the spine as a multi-segment flexible structure. From an anatomical perspective, the human spine is composed of cervical, thoracic, lumbar, and pelvic segments connected in series. Each segment exhibits different degrees of flexion changes under driving posture, and the response of each segment to the human-machine arrangement of the vehicle is a coordinated adjustment rather than a rigid body rotation. Therefore, it is necessary to construct a separate spinal reconstruction sub-model. Based on the occupant's vital signs, hip joint characteristic parameters, and eye point characteristic parameters, the boundary constraints at both ends are mapped to the spatial pose and joint angles of each segment of the spine through the spinal motion distribution relationship. Then, the spinal characteristic parameters are calculated to achieve complete kinematic reconstruction from the anchor points at both ends to the multi-segment spinal morphology.

[0090] It should be noted that using the eyepoint as the basis for the upper spinal boundary constraint originates from the fixed geometric relationship between the eyepoint and the cervical pivot point established in SAE J941 / J1050. This simplification is a common processing method in the field, aiming to use standardized geometric mapping relationships to reasonably transform the eyepoint position into kinematic constraints for the upper spinal end, thereby supporting the boundary condition setting and kinematic solution of the spinal reconstruction sub-model.

[0091] To establish a quantitative mapping relationship between occupant vital signs, hip joint features, eyepoint features, and spinal features, this embodiment of the invention constructs a fourth, fifth, and sixth solution mapping within the spinal reconstruction sub-model. The fourth solution mapping takes the occupant's height and sitting height as input, and calculates the spinal length among the spinal features using a first solution scaling equation, specifically calculating the cervical, thoracic, lumbar, and pelvic segment lengths. The fifth solution mapping takes the geometric center position of the hip joint, eyepoint position, and spinal length as input, and calculates the spinal angles among the spinal features using a second vector operation equation established based on spinal kinematic chain constraints, specifically calculating the overall trunk tilt angle, cervical tilt angle, thoracic tilt angle, lumbar tilt angle, and pelvic segment tilt angle. The sixth solution mapping takes the geometric center position of the hip joint, the length of the spine, and the angle of the spine as inputs. It solves the position of the spinal joint points in the spinal feature parameters, especially the C7 / T1 position, the T12 / L1 position, and the L5 / S1 position, through the third vector operation equation established based on the constraints of the spinal kinematic chain.

[0092] In constructing the fourth solution mapping, a linear scaling positioning method commonly used in the field of biomechanics was adopted. Based on the statistical analysis results of regionalized and large-scale cadaver anatomical measurement data, the first solution scaling equation was established between the length of each segment of the human spine and the height and sitting height of the prior occupant. The first solution scaling equation is as follows: Cervical spine length = K c1 × Sitting high; Thoracic vertebra length = K c2 × Sitting high; Lumbar spine length = K c3 × Sitting high; Pelvic segment length = K c4 × Height.

[0093] Among them, K c1 Scaling factor for the first spine solution; K c2 Scaling factor for the second spine solution; K c3 Scaling factor for the third spine solution; K c4 The scaling factor is calculated for the fourth spine. It should be noted that the scaling principle of the first scaling equation is the same as that of the second positioning scaling equation. In practical applications, for the same target area population, the first scaling equation directly reuses the coefficient calibration results of the latter and does not repeat the fitting.

[0094] When constructing the fifth solution mapping, the positional offset of the eye point relative to the geometric center of the hip joint is attributed to the spatial extension of the spine. Specifically: in the X-axis direction of the global spatial coordinate system, the positional offset of the eye point relative to the geometric center of the hip joint corresponds to the horizontal projection length of the spine in the sagittal plane of the human body; in the Z-axis direction of the global spatial coordinate system, the positional offset of the eye point relative to the geometric center of the hip joint corresponds to the vertical projection length of the spine in the sagittal plane of the human body. Based on this, using the prior eye point position, the prior hip joint geometric center position, and the prior spine length as inputs, a second vector operation equation is constructed according to the vector operation principle to solve for the spinal angle. EyeX = HipX + L pelvis ·sin(θ pelvis )+L lumbar ·sin(θ pelvis +θ lumbar )+L thorax ·sin(θ pelvis +θ lumbar +θ thorax )+L neck ·sin(θ pelvis +θ lumbar +θ thorax +θ neck ) EyeZ = HipZ + L pelvis ·cos(θ pelvis )+L lumbar ·cos(θ pelvis +θ lumbar )+L thorax ·cos(θ pelvis +θ lumbar +θ thorax )+L neck·cos(θ pelvis +θ lumbar +θ thorax +θ neck ) Among them, L pelvis L is the length of the pelvic segment. lumbar L is the length of the lumbar spine. thorax L is the length of the thoracic vertebrae. neck θ is the length of the cervical vertebrae. pelvis θ is the pelvic segmental tilt angle. lumbar θ is the lumbar inclination angle. thorax θ is the thoracic vertebral inclination angle. neck This refers to the cervical spine tilt angle.

[0095] Since directly solving for the independent variables of each segment of the spine can easily lead to underdetermined equations, and since the overall posture change of the trunk can be further decomposed into the coordinated response of each segment of the spine, this invention establishes the analytical relationship between the overall trunk tilt angle and the tilt angle of each segment of the spine based on the kinematic chain constraints of the spine, as follows: θ pelvis =θ pelvis0 +γ pelvis (θ torso θ torso0 ) θ lumbar =θ lumbar0 +γ lumbar (θ torso θ torso0 ) θ thorax =θ thorax0 +γ thorax (θ torso θ torso0 ) θ neck =θ neck0 +γ neck (θ torso θ torso0 ) Where, θ torso θ is the overall tilt angle of the torso; torso0 θ is the overall reference attitude angle of the torso; pelvis0 θ lumbar0 θ thorax0 θ neck0 These are the baseline posture angles for the pelvic segment, lumbar spine, thoracic spine, and cervical spine, respectively; γ pelvis γ lumbar γ thorax γ neckThese are the spinal motion distribution coefficients for the pelvic segment, lumbar vertebrae, thoracic vertebrae, and cervical vertebrae, respectively, used to characterize the response proportion of each spinal segment to changes in tilt angle when the overall trunk tilt angle changes.

[0096] By substituting the analytical relationship between the overall trunk tilt angle and the tilt angles of each spinal segment into the analytical relationship between the eye point position, the geometric center position of the hip joint, the length of each spinal segment, and the tilt angle of each spinal segment, the unknown variable in the analytical expression is reduced from the tilt angle of each spinal segment to the overall trunk tilt angle. The overall trunk tilt angle can be calculated first using the eye point position, the geometric center position of the hip joint, and the length of each spinal segment. Then, the result of the overall trunk tilt angle calculation is substituted into the analytical relationship between it and the tilt angle of each spinal segment to calculate the tilt angle of each spinal segment, thereby reducing the difficulty of the solution and improving the computational efficiency.

[0097] During model training, θ torso0 θ pelvis0 θ lumbar0 θ thorax0 θ neck0 The mean values ​​of the overall trunk posture angles and the mean values ​​of the posture angles of each segment of the spine for the prior occupant under all their respective posture samples can be obtained separately. When using the model, because it is impossible to obtain multi-posture samples of the predicted occupant, and to avoid introducing additional surface measurement procedures, θ torso0 θ pelvis0 θ lumbar0 θ thorax0 θ neck0 Instead of being treated as independent individual input variables, these constants participate in the calculation as built-in baseline posture constants within the model. The values ​​of these constants are derived from the large-sample statistical mean obtained from a specific population matched with occupant vital signs (height, BMI, age, etc.) in the prediction sample dataset, under hard reference seats and standard driving postures. If refined stratified statistical data is lacking, the mean of the baseline posture angles of all samples in the posture sample dataset under standard driving postures can be used directly as the default calibration value. During model execution, the system automatically matches and calls the corresponding baseline posture constants based on the input occupant vital signs to participate in solving the second vector operation equation, without requiring any external measured data.

[0098] Experimental observations show that when occupants adjust their overall posture due to changes in vehicle interior parameters while in a driving position, the changes in the tilt angles of different spinal segments and the overall tilt angle of the torso exhibit an approximately linear coordinated response, and the response ratios of each segment show relatively stable numerical characteristics in population statistics. Based on this pattern, θ is abstracted at the model level. torsoAs an orientational parameter characterizing the overall tilt of the trunk, its change serves as a control variable driving the coordinated adjustment of each spinal segment. To establish a mathematical expression of this coordinated response relationship, each tilt angle needs to be quantitatively defined. Based on conventional simplification assumptions in biomechanics and ergonomics, all tilt angles and reference posture angles are quantified using the spatial orientation angle of the line connecting the spinal joints. The specific quantification method is as follows: In the sagittal plane of the human body, the line connecting the geometric center of the hip joint and the C7 / T1 position (or the acromion point) is selected as the equivalent axis characterizing the overall trunk tilt. The overall trunk tilt angle θ is defined by the angle between this equivalent axis and the horizontal (or vertical) direction. torso In the sagittal plane of the human body, the line connecting the geometric center of the hip joint and the L5 / S1 position is selected as the equivalent axis of pelvic segment tilt. The pelvic segment tilt angle θ is defined by the angle between this equivalent axis and the horizontal (or vertical) direction. pelvis Lumbar inclination angle θ lumbar Thoracic vertebral tilt angle θ thorax and cervical spine tilt angle θ neck The definition method is similar. The lines connecting L5 / S1 position and T12 / L1 position, T12 / L1 position and C7 / T1 position, and C7 / T1 position and eye point position (or craniocervical pivot point) are selected as the equivalent axes corresponding to each segment. The inclination angle of the corresponding segment is defined by the angle between each equivalent axis and the horizontal direction (or vertical direction).

[0099] The above θ torso The equivalent axis of θ has its upper endpoint at C7 / T1. Its geometric domain encompasses the segments from the geometric center of the hip joint to C7 / T1 (i.e., the pelvis, lumbar vertebrae, and thoracic vertebrae), excluding the cervical vertebrae above C7 / T1. This is because C7 / T1, as the intervertebral joint at the cervicothoracic junction, forms the kinematic boundary between the trunk and the head and neck; the segments below it constitute the main structure supporting the overall tilt of the trunk. The primary kinematic function of the cervical vertebrae above C7 / T1 is to maintain the horizontal field of vision of the head; their tilt angle changes are compensatory adjustments by the head and neck to the overall tilt of the trunk, rather than a direct component of the trunk's tilt. Therefore, the cervical vertebrae are excluded from θ. torso The geometric domain is outside of θ. However, based on the above-mentioned whole-segment coordinated response law, although the cervical spine is not within θ... torso Within its geometric domain, the change in its tilt angle can still be determined by γ. neck With θ torso The changes in the spinal motion distribution coefficient are used to establish a correlation between the changes in the spinal motion distribution coefficient and the overall orientation changes to each segment.

[0100] During the model training phase, all variables in the second vector operation equation can be obtained directly or indirectly through the parameters of prior occupant surface landmarks. Based on the intra-individual data of each prior occupant under different posture samples, the motion distribution coefficients of each spinal segment can be determined by fitting the response ratio of the tilt angle of each segment of the spine to the overall tilt angle of the trunk through intra-individual linear regression.

[0101] As a preferred embodiment of the present invention, γ pelvis Take 0.30, γ lumbar Take 0.15, γ thorax Take 0.50, γ neck The value is set to 0.05. It should be noted that the above spinal motion allocation coefficients are based on statistical analysis of measured posture data from a specific population area. Each spinal motion allocation coefficient represents the statistical mean of the corresponding sample, with a standard deviation between 0.03 and 0.08. In practical applications, sample collection and regression calibration can be performed again to determine the corresponding spinal motion allocation coefficients based on the body shape characteristics or sitting habits of the target population area. The scope of protection claimed in this invention should not be considered limited to the values ​​specifically given above.

[0102] When constructing the sixth solution mapping, the spatial vector operation principle of the fifth solution mapping is used. Taking the geometric center position of the hip joint, the length of each spinal segment, and the tilt angle of each spinal segment as input, the position of the spinal joint point in the spinal feature parameters is solved through the third vector operation equation established based on the spinal kinematic chain constraints. The position of each spinal joint point is solved by a segment-by-segment forward recursive method. L5 / S1_X=HipX+L pelvis ·sin(θ pelvis ) L5 / S1_Z=HipZ+L pelvis ·cos(θ pelvis ) T12 / L1_X=HipX+L pelvis ·sin(θ pelvis )+L lumbar ·sin(θ pelvis +θ lumbar ) T12 / L1_Z=HipZ+L pelvis ·cos(θ pelvis )+L lumbar ·cos(θ pelvis +θ lumbar ) C7 / T1_X=HipX+L pelvis ·sin(θ pelvis )+L lumbar ·sin(θ pelvis +θ lumbar )+Lthorax ·sin(θ pelvis +θ lumbar +θ thorax ) C7 / T1_Z=HipZ+L pelvis ·cos(θ pelvis )+L lumbar ·cos(θ pelvis +θ lumbar )+L thorax ·cos(θ pelvis +θ lumbar +θ thorax ) Wherein, L5 / S1_X and L5 / S1_Z are the coordinates of L5 / S1 in the X-axis and Z-axis directions of the global spatial coordinate system, respectively; T12 / L1_X and T12 / L1_Z are the coordinates of T12 / L1 in the X-axis and Z-axis directions of the global spatial coordinate system, respectively; C7 / T1_X and C7 / T1_Z are the coordinates of C7 / T1 in the X-axis and Z-axis directions of the global spatial coordinate system, respectively. The coordinates of L5 / S1, T12 / L1, and C7 / T1 in the Y-axis direction of the global spatial coordinate system are determined by the same principle as above, uniquely determined by the coordinates of the seat seating reference point in the Y-axis direction of the global spatial coordinate system.

[0103] In the aforementioned spinal reconstruction sub-model, by using the boundary constraints formed by the geometric center of the hip joint and the eye point, combined with the physiological proportions of the lengths of each segment in the spinal kinematic chain and the spinal motion distribution relationship calibrated by measured data, the positions of the three main spinal joints C7 / T1, T12 / L1, and L5 / S1 are accurately reconstructed in the sagittal plane of the human body. This allows for the prediction of the spatial orientation of the occupant's entire trunk and each segment in any vehicle cabin environment. Simultaneously, the aforementioned spinal motion distribution coefficients transform the original problem of solving the inclination angle of multiple spinal segments with four degrees of freedom into a single-degree-of-freedom problem with the overall trunk inclination angle as the only unknown. This significantly reduces the dimensionality of the equations, effectively simplifying the computational difficulty and improving computational efficiency while ensuring biomechanical rationality.

[0104] (3) The necessity of constructing the limb analysis sub-model lies in the fact that the spinal joint positions output by the spinal reconstruction sub-model limit the spatial pose of the trunk core, but the complete expression of the driving posture needs to be further extended to the extremities—the shoulder joint and upper limb need to form a geometric fit with the steering wheel and other control interfaces, and the hip joint and lower limb need to form a contact constraint with the seat cushion and pedals and other support interfaces. At the same time, the shoulder joint and hip joint are the starting points of the upper and lower limbs, respectively, and their spatial positions are determined by the pelvic segments and the thoracic vertebrae, respectively. The spatial distribution and angle changes of the joints of the limbs are jointly constrained by the proportion of the occupant's limb segment length, the range of motion of the joints, and the vehicle cabin environment. Therefore, it is necessary to construct a separate limb analysis sub-model, using the occupant's vital signs parameters, vehicle interior parameters, and the hip joint feature parameters and spinal feature parameters calculated by the previous sub-model as inputs, and calculating the shoulder joint feature parameters and limb joint feature parameters through the limb kinematic chain constraint to complete the final extension from the trunk posture to the complete whole body posture.

[0105] To establish a quantitative mapping relationship between occupant vital signs parameters, vehicle interior parameters, hip joint characteristic parameters, spine characteristic parameters, shoulder joint characteristic parameters, and limb joint characteristic parameters, this embodiment of the invention constructs a seventh, eighth, and ninth solution mapping in the limb analysis sub-model. The seventh solution mapping takes the C7 / T1 position and height from the occupant vital signs parameters as input. It solves for the acromion point position from the shoulder joint characteristic parameters using a fourth vector operation equation established based on the relative spatial relationship between the C7 / T1 joint and the acromion point in human anatomy. The eighth solution mapping takes the acromion point position, height from the occupant vital signs parameters, and the steering wheel center position, steering wheel diameter, and steering wheel tilt angle from the vehicle interior parameters as input. Based on the scaling principle of the human upper limb and the constraints of the upper limb kinematic chain, it solves for the wrist joint position, elbow joint position, and elbow angle from the limb joint characteristic parameters using a two-bar inverse kinematic geometry solution method. The ninth solution mapping takes the geometric center position of the hip joint, the height in the occupant's physical parameters, and the center position and inclination angle of the pedal surface in the vehicle interior parameters as inputs. Based on the scaling principle of the human lower limbs and the constraints of the lower limb kinematic chain, the ankle joint position, knee joint position, and knee angle value in the characteristic parameters of the limb joints are solved by the two-link inverse kinematic geometry solution method.

[0106] In constructing the seventh solution mapping, a linear scaling positioning method commonly used in biomechanics was adopted. Based on the statistical analysis results of regionalized and large-scale cadaveric anatomical measurement data, an offset vector between the C7 / T1 position and the acromion position was established. Specifically, under standard driving posture, based on anatomical statistical measurements, the offsets of the acromion position relative to the C7 / T1 position in the X, Z, and Y axes of the global spatial coordinate system were determined. According to statistical results, the offsets in the X and Z axes showed no significant correlation with height after linear regression testing; therefore, the anatomical statistical mean was taken as a fixed constant under standard driving posture, independent of individual height. The offset in the Y axis (i.e., the vehicle width direction) showed a significant linear correlation with height based on measured statistics. The offset was calculated by superimposing the height-based linear scaling offset onto the C7 / T1 position.

[0107] Based on the above, the three-dimensional coordinates of the acromion point in the global spatial coordinate system are represented by the fourth vector operation equation. Taking the C7 / T1 position as a reference, the offsets determined based on anatomical statistics are superimposed. Taking the left acromion point position as an example, the fourth vector operation equation includes: Shoulder_L_X=C7 / T1_X+ΔX shoulder _L Shoulder_L_Y=C7 / T1_Y+ΔY shoulder _L;ΔY shoulder _L=K d1 ·S Shoulder_L_Z=C7 / T1_Z+ΔZ shoulder _L Where Shoulder_L_X, Shoulder_L_Y, and Shoulder_L_Z are the coordinates of the left shoulder peak in the X, Y, and Z axes of the global spatial coordinate system, respectively; ΔX shoulder _L represents the offset of the left shoulder peak position relative to the C7 / T1 position along the X-axis of the global spatial coordinate system; ΔY shoulder _L represents the offset of the left shoulder peak position relative to the C7 / T1 position along the Y-axis in the global spatial coordinate system, S represents height, and K represents… d1 ΔZ is the scaling factor for the shoulder peak point; shoulder _L represents the offset of the left shoulder peak position relative to the C7 / T1 position along the Z-axis in the global spatial coordinate system. The right shoulder peak position is calculated using the same method, executed symmetrically in mirror image.

[0108] In a preferred embodiment of the present invention, ΔX shoulder _L is set to +30mm, ΔZ shoulder _L is set to -80mm, K d1 Take 0.042.

[0109] When constructing the eighth solution mapping, its solution principle is first clarified as follows: the upper limb is regarded as a two-link kinematic chain consisting of the shoulder joint, elbow joint, wrist joint, upper arm, and forearm. Based on the upper limb kinematic chain constraints and the rotatable degree of freedom constraints of the elbow joint, the elbow joint position and elbow angle value are solved by jointly considering the acromion position, wrist joint position, upper arm length, and forearm length. Among them, the acromion position is determined by the seventh solution mapping, the wrist joint position depends on the steering wheel geometry parameters and the occupant's hand grip posture on the steering wheel, and the upper arm length and forearm length have a linear scaling relationship with the human body height. Therefore, it is necessary to establish the first joint solution relationship for solving the wrist joint position, the second solution scaling equation for solving the upper arm length and forearm length, and the second joint solution relationship for solving the elbow joint position and elbow angle value.

[0110] When constructing the first joint solution relationship, the origin O is first set at the center of the steering wheel surface. a With the normal of the steering wheel face as Z a The axis, the normal to the steering wheel surface is determined based on the steering wheel surface tilt angle, Z. a The positive direction of the axis points to the top of the vehicle. Within the steering wheel plane, the direction parallel to the Y-axis in the global coordinate system is defined as Y. a axis, Y a The positive direction of the axis points to the right side of the occupant, and is perpendicular to the Y-axis within the steering wheel plane. a The direction of the axis is X a Axis, X a The positive axis points forward of the vehicle; establish a local coordinate system {O} for the steering wheel. a ,X a ,Y a Z a Then, the steering wheel radius is calculated based on the steering wheel diameter, and the polar coordinate angle of the hand grip position in the local coordinate system of the steering wheel is determined based on the hand grip posture. The three-dimensional coordinates of the hand grip position in the local coordinate system of the steering wheel are calculated based on the steering wheel radius and the polar coordinate angle of the hand grip position. Then, a rotation matrix is ​​constructed based on the direction vectors of each axis of the local coordinate system of the steering wheel in the global coordinate system. Combined with the three-dimensional coordinates of the center position of the steering wheel surface in the global coordinate system as the translation vector, the three-dimensional coordinates of the hand grip position in the local coordinate system of the steering wheel are transformed into the three-dimensional coordinates of the hand grip position in the global coordinate system through the rotation matrix and translation transformation.

[0111] In a preferred embodiment of the present invention, the hand grip posture is determined according to the standard driving posture, that is, the occupant holds the steering wheel at the 10 o'clock position and the 2 o'clock position respectively; taking the direction directly above the steering wheel plane (i.e., the 12 o'clock position) as the starting reference of the polar coordinate angle of 0°, and taking the clockwise direction as the direction of angle increase, the polar coordinate angle corresponding to the 2 o'clock position is determined to be 60°, and the polar coordinate angle corresponding to the 10 o'clock position is determined to be 120°.

[0112] In constructing the second scaling equation, a linear scaling positioning method commonly used in the field of biomechanics was adopted. Based on the statistical analysis results of regionalized and large-scale anthropometric data, the quantitative mapping relationship between the upper arm length, forearm length, and height was established as follows: Upper arm length = K e1 × Height; Forearm length = K e2 × Height.

[0113] Among them, K e1 K is the scaling factor for the first upper limb calculation. e2 The scaling factor is used to calculate the second upper limb.

[0114] In a preferred embodiment of the present invention, K e1 Take 0.186, K e2 The value is 0.146. It should be noted that the above coefficient value is based on statistical analysis of anthropometric data of a population in a specific region. In practical applications, measurements can be re-performed and the corresponding upper limb scaling factor can be determined according to the differences in body shape characteristics of the population in the target region. The scope of protection claimed by this invention should not be considered as limited to the values ​​specifically given above.

[0115] Based on the solution results of the first joint solution relationship and the second solution scaling equation, a second joint solution relationship is constructed to solve for the elbow joint position and elbow angle value. The core solution logic is as follows: Under the constraints of the upper limb kinematic chain, the elbow joint position simultaneously satisfies two geometric constraints—the elbow joint lies on both a sphere centered at the acromion and with the upper arm length as its radius, and a sphere centered at the wrist joint and with the forearm length as its radius; that is, the elbow joint rests on the first spatial circle formed by the intersection of the two spheres. Furthermore, a biomechanical constraint is introduced that the elbow joint's natural flexion angle is within a comfortable range. Candidate solutions on the first spatial circle are then screened to calculate the unique three-dimensional coordinates of the elbow joint position in the global spatial coordinate system. Subsequently, the elbow angle value is determined by the angle between the vector pointing from the elbow joint position to the acromion position and the vector pointing from the elbow joint position to the wrist joint position. As an engineering alternative, the supplementary angle is often used to characterize the degree of elbow flexion; the two expressions are equivalent in kinematic description.

[0116] In practice, this invention directly defines the three-dimensional coordinates of the hand grip point in the global spatial coordinate system as the wrist joint position. This simplification is based on the fact that, in a standard driving hand grip posture, the palm is in close contact with the steering wheel rim, and the wrist joint posture relative to the hand grip point remains approximately constant. The residual deviation between the two can be automatically compensated by angle optimization in the inverse kinematics solution of the elbow joint position, and the introduced error is within an acceptable engineering range.

[0117] In a specific embodiment of the present invention, the elbow angle is defined as the angle between the upper arm direction vector (pointing from the elbow joint position to the acromion position) and the forearm direction vector (pointing from the elbow joint position to the wrist joint position), which is the angle value naturally output in the geometry of the intersection of two spherical surfaces. The rotational degrees of freedom of the elbow joint are constrained as follows: the range of motion of the elbow joint is limited to the anatomical limit of 0. 145 Furthermore, a comfortable flex range of 80 is preferred for driving control. 120 During the solution process, this angular interval is used as a boundary constraint to eliminate outlier solutions on the first spatial circle. After this filtering, two candidate elbow joint position solutions usually remain on the first spatial circle, located on either side of the line connecting the acromion and the wrist joint. Based on the anatomical characteristics of the upper limb naturally adducting in a standard driving posture, the candidate solution located on the inner side of the shoulder-wrist line (i.e., the side closer to the occupant's torso) is selected as the only valid elbow joint position, thus completing the final determination of the elbow joint position and elbow angle value.

[0118] When constructing the ninth solution mapping, the solution principle is first clarified as follows: the lower limb is regarded as a two-link kinematic chain consisting of the hip joint, knee joint, ankle joint, thigh, and calf. Based on the lower limb kinematic chain constraints and the rotational degree of freedom constraint of the knee joint, the knee joint position and knee angle value are solved by the geometric center position of the hip joint, the ankle joint position, the thigh length, and the calf length. Among them, the geometric center position of the hip joint is determined by the anchor point positioning sub-model, the ankle joint position depends on the pedal geometric parameters and the occupant's foot posture on the pedal, and the thigh length, calf length, and human height have a linear scaling relationship of the lower limb. Therefore, it is necessary to establish a third joint solution relationship for solving the ankle joint position, a third solution scaling equation for solving the thigh length and calf length, and a fourth joint solution relationship for solving the knee joint position and knee angle value.

[0119] When constructing the solution relationship for the third joint, the origin O is first set at the center of the pedal surface. c Let the normal direction of the pedal surface be Z. c The axis, the normal to the pedal surface is determined based on the pedal surface inclination angle, Z cThe positive direction of the axis points towards the occupant. Within the pedal surface, the direction parallel to the Y-axis in the global coordinate system is defined as X. c Axis, X c The positive direction of the axis points to the right side of the occupant, and a perpendicular line is taken within the pedal surface to X. c The direction of the axis is Y c axis, Y c The positive axis points forward of the vehicle; establish a local coordinate system {O} for the pedals. c ,X c ,Y c Z c Then, based on the foot posture and foot length, the three-dimensional coordinates of the ankle joint position in the local coordinate system of the pedal are calculated; then, based on the direction vectors of each axis of the local coordinate system of the pedal in the global coordinate system, a rotation matrix is ​​constructed, and combined with the three-dimensional coordinates of the pedal surface center position in the global coordinate system as the translation vector, the three-dimensional coordinates of the ankle joint position in the local coordinate system of the pedal are transformed into the three-dimensional coordinates of the ankle joint position in the global coordinate system through the rotation matrix and translation transformation.

[0120] In a preferred embodiment of the present invention, the foot position is determined based on the standard driving posture, specifically including the spatial position of the foot contact point and the dorsiflexion or plantarflexion state of the ankle joint. First, the connection between the lower leg and the foot is abstracted as a rotary joint connected by the ankle joint, and the center position of the pedal surface is taken as the foot contact point. The dorsiflexion or plantarflexion state of the ankle joint is quantified by the angle between the sole plane of the foot and the axis of the lower leg (i.e., the ankle angle): under the standard driving posture, the minimum value of this angle is defined as 90°, corresponding to the reference posture where the sole plane of the foot is perpendicular to the axis of the lower leg; the maximum value is defined as 110°, corresponding to the extreme posture where the sole plane of the foot deflects in the dorsiflexion direction; within this range, an ankle angle value that conforms to the natural sitting posture of the occupant is selected, for example, the middle value of 100° is taken, or it is optimized and determined based on actual vehicle measurement data.

[0121] Then, based on the determination of the spatial location of the foot contact point and the ankle angle value, and combined with the foot length, the ankle joint position is deduced through spatial geometric relationships. The specific method is as follows: First, based on the statistical analysis results of regionalized and large-scale anthropometric data, the quantitative relationship between human foot length and height is established as: Foot length = K f1 × Height, K f1 The scaling factor is calculated for the foot length. Since the local coordinate system of the pedal is established with the center of the pedal surface as the origin, the three-dimensional coordinates of the foot contact point in the local coordinate system of the pedal are (0,0,0). Then, based on the spatial constraint that the ankle joint position is located above and behind the foot contact point in the normal driving posture, the three-dimensional coordinates of the ankle joint position in the local coordinate system of the pedal can be derived according to the sine and cosine relationship of the ankle angle and the foot length: (0, -foot length × cos(ankle angle value), foot length × sin(ankle angle value)).

[0122] In a preferred embodiment of the present invention, K f1 Take 0.152.

[0123] In constructing the third scaling equation, a linear scaling positioning method commonly used in the field of biomechanics was adopted. Based on the statistical analysis results of regionalized and large-scale anthropometric data, the quantitative relationship between thigh length, calf length, and height was established as follows: Thigh length = K g1 × Height; Lower leg length = K g2 × Height.

[0124] Among them, K g1 K is the scaling factor for solving the first lower limb. g2 The scaling factor is used to calculate the second lower limb.

[0125] In a preferred embodiment of the present invention, K g1 Take 0.245, K g2 The value is set to 0.235. It should be noted that the above coefficient values ​​are based on statistical analysis of anthropometric data of a population in a specific region. In practical applications, measurements can be re-performed and the corresponding lower limb scaling coefficients determined according to the differences in body shape characteristics of the population in the target region. The scope of protection claimed by this invention should not be considered as limited to the values ​​specifically given above.

[0126] Based on the solution results of the third joint and the third solution scaling equation, a fourth joint solution is constructed to solve for the knee joint position and knee angle. The core solution logic is as follows: Under the constraints of the lower limb kinematic chain, the knee joint position simultaneously satisfies two geometric constraints—the knee joint lies on both a sphere centered at the geometric center of the hip joint and with a radius equal to the thigh length, and a sphere centered at the ankle joint and with a radius equal to the lower leg length; that is, the knee joint rests on the second spatial circle formed by the intersection of the two spheres. Furthermore, a biomechanical constraint ensuring the knee joint's natural flexion angle remains within a comfortable range is introduced to filter candidate solutions on the second spatial circle, thereby calculating the unique three-dimensional coordinates of the knee joint position in the global spatial coordinate system. Then, the knee angle value is determined by the angle between the vector pointing from the knee joint position to the geometric center of the hip joint and the vector pointing from the knee joint position to the ankle joint. As an engineering alternative, the supplementary angle is often used to characterize the degree of knee flexion; both expressions are equivalent in kinematic description.

[0127] Based on the fact that the knee joint is in a moderate flexion state during normal driving posture, and its adduction and extension range of motion in the coronal plane is extremely small, this invention simplifies knee joint movement to flexion and extension only in the sagittal plane when calculating the knee joint position, ignoring its vara / varus components in the coronal plane. This simplification conforms to biomechanical principles and meets the accuracy requirements of engineering applications. In a specific embodiment of this invention: the range of knee joint flexion angle in the sagittal plane is the anatomical limit, i.e., 0°. 130°, with an optimal comfort flex range of 80° for driving control. 120°. Solving at 0° 130° is used as a boundary constraint to eliminate outliers caused by extreme vehicle layouts that result in the knee joint exceeding the range of active human movement; simultaneously, 80° is used as a boundary constraint. A 120° range is designated as the preferred comfort zone, used to prioritize candidate solutions that better suit driving comfort among multiple feasible postures. After this selection, two candidate knee joint positions typically remain on the second spatial circle, located on either side of the line connecting the geometric center of the hip joint and the ankle joint. Based on the anatomical characteristics of the lower limbs naturally adducting under standard driving posture, the candidate solution located on the inner side of the hip-ankle line (i.e., closer to the occupant's torso) is chosen as the only valid knee joint position, thus completing the final determination of the knee joint position and knee angle.

[0128] Finally, the attitude sample dataset and the prior occupant skeletal joint parameters were divided into training and validation sets in an 8:2 ratio, respectively, for parameter fitting and generalization performance evaluation of the attitude prediction model. In the generalization performance evaluation stage, a systematic analysis of the distribution patterns and sources of the prediction residuals was conducted to identify the main factors contributing to model bias. Based on this, targeted modifications were made to the model structure or regression coefficients, effectively improving both the prediction accuracy and stability of the model under different input conditions.

[0129] In some specific implementation schemes, it was found that the predicted coordinates of the eyepoint position along the Z-axis of the global spatial coordinate system were generally lower than the measured values, exhibiting a consistent negative deviation. The main reason for this was that the compression deformation of the seat cushion under the long-term influence of the occupant's weight was not included in the model considerations. To address this issue, under the simulated condition of an elastic seat cushion, a correction sub-mapping was added to the third solution mapping of the anchor point positioning sub-model. This correction sub-mapping uses the occupant's weight, seat cushion filling density, and load duration as input variables to calculate the seat cushion compression deformation. Simultaneously, the original third solution mapping was expanded to use the geometric center position of the hip joint, the relative offset of the eyepoint, and the seat cushion compression deformation as inputs, and the eyepoint position was calculated using a modified first vector operation equation.

[0130] The revised first vector operation equation includes: EyeX = HipX + EyeXrel EyeZ = HipZ + EyeZ rel +δ Where δ is the seat cushion compression deformation, calculated using the following correction sub-mapping equation: δ=K h ·(W / W0)·(ρ0 / ρ)·ln(1+t / t0) Among them, K h ρ is the compression deformation coefficient of the seat cushion; W is the body weight; W0 is the standard reference body weight; ρ is the seat cushion filling density; ρ0 is the standard reference density; t is the load application time; t0 is the standard reference time.

[0131] During the model building phase, W0, ρ0, and t0 are taken as preset fixed reference values, while W, ρ, and t are determined from measured data. Based on this data, K is obtained through statistical regression fitting. h The value of K is determined and fixed as an embedded parameter in the model for subsequent prediction calculations. In a preferred embodiment of the present invention, K... h The value range is 0.025 to 0.040.

[0132] Another aspect of the present invention provides a posture prediction method for constructing a driving digital human model. This posture prediction method is executed based on the aforementioned constructed posture prediction model, and its implementation steps specifically include: S21. Construct the prediction sample dataset The prediction sample dataset includes occupant feature parameters, vehicle interior parameters, and solution constraint parameters. The occupant feature parameters include at least the occupant's height, sitting height, sitting-to-height ratio, weight, age, body mass index (BMI), and an interaction term (the product of BMI and age). Vehicle interior parameters include seat height, seat cushion angle, seat cushion fill density, seat reference point position, steering wheel center position, steering wheel diameter, steering wheel tilt angle, pedal center position, and pedal tilt angle. Solution constraint parameters include load application time.

[0133] S22. Using the anchor point positioning sub-model, solve for the hip joint feature parameters and eyepoint feature parameters. Hip joint characteristic parameters include the location of the hip joint geometric center.

[0134] The eye point feature parameters include the relative offset of the eye point and the position of the eye point.

[0135] (1) Using the first variable as input, calculate the position of the geometric center of the hip joint based on the first solution mapping.

[0136] The first variables include height, body mass index, age and interaction terms derived from occupant physical parameters, and seat height, steering wheel center position, seat cushion angle and seat reference point position derived from vehicle interior parameters.

[0137] The first solution mapping includes the first regression equation: HipX=α0+α1·S+α2·BMI+α3·A+α4·H30+α5·L6_X+α6·L27+α7·(BMI×A) HipZ=α' 0+α' 1·S+α' 2·BMI+α' 3·A+α' 4·H30+α' 5·L6_Z+α' 6·L27+α'7·(BMI×A) Where HipX and HipZ are the coordinates of the geometric center of the hip joint in the X-axis and Z-axis directions of the global spatial coordinate system, respectively; and HipY is the coordinate of the seat reference point in the Y-axis direction of the global spatial coordinate system. S represents height; BMI represents body mass index; A represents age; H30 represents seat height; L6_X represents the coordinates of the steering wheel center position along the X-axis in the global coordinate system; L6_Z represents the coordinates of the steering wheel center position along the Z-axis in the global coordinate system; L27 represents seat angle; BMI×A is the interaction term, i.e., the product of body mass index and age; α0 and α'0 are the intercept terms in the corresponding analytical expressions; α1 and α'1 are the regression coefficients corresponding to height in the corresponding analytical expressions; α2 and α'2 are the regression coefficients corresponding to body mass index in the corresponding analytical expressions; α3 and α'3 are the regression coefficients corresponding to age in the corresponding analytical expressions; α4 and α'4 are the regression coefficients corresponding to seat height in the corresponding analytical expressions; α5 and α'5 are the regression coefficients corresponding to the steering wheel center position in the corresponding analytical expressions; α6 and α'6 are the regression coefficients corresponding to seat angle in the corresponding analytical expressions; α7 and α'7 are the regression coefficients corresponding to the interaction term in the corresponding analytical expressions.

[0138] (2) Using the second variable as input, calculate the relative offset of the eye point based on the second solution mapping.

[0139] The second variable includes height and seat height ratio derived from occupant physical parameters, and steering wheel center position and seat cushion angle derived from vehicle interior parameters.

[0140] The second solution mapping includes the second regression equation: EyeX rel =β0+β1·S+β2·(S / H)+β3·L6_X+β4·L27 EyeZ rel=β' 0+β' 1·S+β' 2·(S / H)+β' 3·L6_Z+β' 4·L27 Among them, EyeX rel EyeZ rel These represent the components of the relative eye point offset along the X and Z axes in the global coordinate system, respectively; while the component of the relative eye point offset along the Y axis in the global coordinate system is EyeY. rel Take zero value. S is height; (S / H) is the seat height ratio; L6_X is the coordinate value of the center position of the steering wheel in the X-axis direction of the global spatial coordinate system; L6_Z is the coordinate value of the center position of the steering wheel in the Z-axis direction of the global spatial coordinate system; L27 is the seat cushion angle; β0 and β'0 are the intercept terms in the corresponding analytical expressions; β1 and β'1 are the regression coefficients corresponding to height in the corresponding analytical expressions; β2 and β'2 are the regression coefficients corresponding to the seat height ratio in the corresponding analytical expressions; β3 and β'3 are the regression coefficients corresponding to the center position of the steering wheel in the corresponding analytical expressions; β4 and β'4 are the regression coefficients corresponding to the seat cushion angle in the corresponding analytical expressions.

[0141] (3) Using the geometric center position of the hip joint and the relative offset of the eye point as input, the eye point position is calculated based on the third solution mapping.

[0142] In some embodiments of the present invention, the third solution mapping includes a first vector operation equation: EyeX = HipX + EyeX rel EyeZ = HipZ + EyeZ rel In other embodiments of the present invention, the third solution mapping includes a first vector operation equation corrected based on the correction sub-mapping. The correction sub-mapping takes a third variable as input and solves for the seat cushion compression deformation based on the correction sub-mapping equation.

[0143] Specifically: The third variables include body weight from occupant vital signs parameters, seat cushion filling density from vehicle interior parameters, and load application time from solution constraint parameters.

[0144] The analytical expression of the corrector mapping equation is: δ=K h ·(W / W0)·(ρ0 / ρ)·ln(1+t / t0), where δ is the compression deformation of the seat cushion; K h ρ is the compression deformation coefficient of the seat cushion; W is the body weight; W0 is the standard reference body weight; ρ is the seat cushion filling density; ρ0 is the standard reference density; t is the load application time; t0 is the standard reference time.

[0145] The first vector operation equation, corrected based on the sub-mapping correction, includes: EyeX = HipX + EyeX rel EyeZ = HipZ + EyeZ rel +δ In the first vector operation equation of the above embodiments: EyeX and EyeZ are the coordinates of the eye point position in the X-axis and Z-axis directions of the global spatial coordinate system, respectively; and the coordinates of the eye point position in the Y-axis direction of the global spatial coordinate system, EyeY, are the coordinates of the seat seating reference point position in the Y-axis direction of the global spatial coordinate system.

[0146] S23. Using the spinal reconstruction sub-model, solve for the spinal feature parameters. Spinal characteristic parameters include spinal length, spinal angle, and the location of spinal joints.

[0147] The length of the spine includes the length of the cervical spine, thoracic spine, lumbar spine, and pelvic segments.

[0148] Spinal angles include the overall trunk tilt angle, cervical spine tilt angle, thoracic spine tilt angle, lumbar spine tilt angle, and pelvic segment tilt angle.

[0149] The locations of spinal joints include the C7 / T1 position, the T12 / L1 position, and the L5 / S1 position.

[0150] (1) Using the fourth variable as input, calculate the spine length based on the fourth solution mapping.

[0151] The fourth variable includes height and sitting height from occupant vital signs parameters.

[0152] The fourth solution mapping includes the first solution scaling equation: Cervical spine length = K c1 × Sitting high; Thoracic vertebra length = K c2 × Sitting high; Lumbar spine length = K c3 × Sitting high; Pelvic segment length = K c4 × Height.

[0153] Among them, K c1 Scaling factor for the first spine solution; K c2 Scaling factor for the second spine solution; K c3 Scaling factor for the third spine solution; K c4 The scaling factor is used for the fourth spine solution.

[0154] (2) Using the geometric center position of the hip joint, the eye point position and the length of the spine as input, the spinal angle is calculated based on the fifth solution mapping.

[0155] The fifth solution mapping includes the second vector operation equation: EyeX=HipX+L pelvis ·sin(θ pelvis )+L lumbar ·sin(θ pelvis +θ lumbar )+L thorax ·sin(θ pelvis +θ lumbar +θ thorax )+L neck ·sin(θ pelvis +θ lumbar +θ thorax +θ neck ) EyeZ=HipZ+L pelvis ·cos(θ pelvis )+L lumbar ·cos(θ pelvis +θ lumbar )+L thorax ·cos(θ pelvis +θ lumbar +θ thorax )+L neck ·cos(θ pelvis +θ lumbar +θ thorax +θ neck ) θ pelvis =θ pelvis0 +γ pelvis (θ torso θ torso0 ) θ lumbar =θ lumbar0 +γ lumbar (θ torso θ torso0 ) θ thorax =θ thorax0 +γ thorax (θ torso θ torso0 ) θ neck =θ neck0 +γ neck (θ torso θ torso0 ) wherein, L pelvis is the pelvic segment length; L lumbar is the lumbar length; L thorax is the thoracic length; L neck is the cervical length; θtorso θ is the overall tilt angle of the torso; pelvis θ is the pelvic segmental tilt angle. lumbar θ is the lumbar inclination angle. thorax θ is the thoracic vertebral inclination angle. neck This refers to the cervical spine tilt angle.

[0156] θ torso0 θ is the overall reference posture angle of the spine. pelvis0 θ lumbar0 θ thorax0 θ neck0 The reference posture angles for the pelvic segment, lumbar spine, thoracic spine, and cervical spine are respectively used in the calculation as reference posture constants built into the model. Specifically, they can be the large sample statistical mean obtained by matching the occupant physical characteristics (height, BMI, age, etc.) in the prediction sample dataset with those of a specific population under hard reference seats and standard driving posture, or the mean of the reference posture angles of all samples in the posture sample dataset under standard driving posture can be used as the default calibration value.

[0157] γ pelvis γ lumbar γ thorax γ neck These are the spinal motion distribution coefficients for the pelvic segment, lumbar spine, thoracic spine, and cervical spine, respectively.

[0158] The solution order for the second vector operation equation is as follows: First, substitute the analytical relationship between the overall trunk tilt angle and the tilt angles of each segment of the spine into the analytical relationship between the eye point position, the geometric center position of the hip joint, the length of each segment of the spine, and the tilt angle of each segment of the spine, and transform the latter into a single unknown quantity to quickly solve the overall trunk tilt angle; then substitute the solution result of the overall trunk tilt angle into the analytical relationship between it and the tilt angles of each segment of the spine to solve the tilt angles of each segment of the spine.

[0159] (3) Using the geometric center of the hip joint, the length of the spine and the angle of the spine as inputs, the position of the spinal joint is calculated based on the sixth solution mapping.

[0160] The sixth solution mapping includes the third vector operation equation: L5 / S1_X=HipX+L pelvis ·sin(θ pelvis ) L5 / S1_Z=HipZ+L pelvis ·cos(θ pelvis ) T12 / L1_X=HipX+L pelvis ·sin(θ pelvis )+L lumbar ·sin(θ pelvis +θ lumbar ) T12 / L1_Z=HipZ+L pelvis ·cos(θ pelvis )+L lumbar ·cos(θ pelvis +θ lumbar ) C7 / T1_X=HipX+L pelvis ·sin(θ pelvis )+L lumbar ·sin(θ pelvis +θ lumbar )+L thorax ·sin(θ pelvis +θ lumbar +θ thorax ) C7 / T1_Z=HipZ+L pelvis ·cos(θ pelvis )+L lumbar ·cos(θ pelvis +θ lumbar )+L thorax ·cos(θ pelvis +θ lumbar +θ thorax ) Wherein, L5 / S1_X and L5 / S1_Z are the coordinates of L5 / S1 in the X-axis and Z-axis directions of the global spatial coordinate system, respectively; T12 / L1_X and T12 / L1_Z are the coordinates of T12 / L1 in the X-axis and Z-axis directions of the global spatial coordinate system, respectively; C7 / T1_X and C7 / T1_Z are the coordinates of C7 / T1 in the X-axis and Z-axis directions of the global spatial coordinate system, respectively.

[0161] The positions C7 / T1, T12 / L1, and L5 / S1 are the coordinates along the Y-axis in the global spatial coordinate system. C7 / T1_Y, T12 / L1_Y, and L5 / S1_Y are respectively equal to the coordinates of the seat seating reference point along the Y-axis in the global spatial coordinate system.

[0162] S24. Using the limb analysis sub-model, calculate the characteristic parameters of the shoulder joint and the characteristic parameters of the limb joints. Shoulder joint characteristic parameters include the location of the acromion.

[0163] The characteristic parameters of the limb joints include the position of the wrist joint, the position of the elbow joint, the elbow angle value, the position of the ankle joint, the position of the knee joint, and the knee angle value.

[0164] (1) Using the fifth variable as input, calculate the position of the shoulder peak based on the seventh solution mapping.

[0165] The fifth variable includes height derived from occupant vital signs parameters, and the C7 / T1 position in the spinal joint location.

[0166] The seventh solution mapping includes the fourth vector operation equation: Shoulder_L_X=C7 / T1_X+ΔX shoulder _L Shoulder_L_Y=C7 / T1_Y+ΔY shoulder _L;ΔY shoulder _L=K d1 ·S Shoulder_L_Z=C7 / T1_Z+ΔZ shoulder _L Where Shoulder_L_X, Shoulder_L_Y, and Shoulder_L_Z are the coordinates of the left shoulder peak in the X, Y, and Z axes of the global spatial coordinate system, respectively; C7 / T1_X, C7 / T1_Y, and C7 / T1_Z are the coordinates of the C7 / T1 position in the X, Y, and Z axes of the global spatial coordinate system, respectively; ΔX shoulder _L represents the offset of the left shoulder peak position relative to the C7 / T1 position along the X-axis of the global spatial coordinate system; ΔY shoulder _L represents the offset of the left shoulder peak position relative to the C7 / T1 position along the Y-axis in the global spatial coordinate system, S represents height, and K represents… d1 ΔZ is the scaling factor for the shoulder peak point; shoulder _L represents the offset of the left shoulder peak position relative to the C7 / T1 position along the Z-axis in the global spatial coordinate system. The right shoulder peak position is calculated using the same method, executed symmetrically in mirror image.

[0167] (2) Using the sixth variable as input, calculate the wrist joint position, elbow joint position and elbow angle value based on the eighth solution mapping.

[0168] The sixth variable includes height derived from occupant physical parameters, steering wheel center position, steering wheel diameter, steering wheel tilt angle, and shoulder acromion position derived from vehicle interior parameters.

[0169] The eighth solution mapping includes the first joint solution relation, the second solution scaling equation, and the second joint solution relation.

[0170] The first joint calculation relationship includes: establishing a local coordinate system for the steering wheel based on the center position of the steering wheel surface, using a preset hand grip posture as a constraint and the steering wheel diameter as input, and calculating the three-dimensional coordinates of the hand grip point in the local coordinate system; constructing a rotation matrix based on the direction vectors of each axis of the local coordinate system in the global coordinate system, combining the three-dimensional coordinates of the steering wheel surface center position in the global coordinate system as a translation vector, and through rotation matrix and translation transformation, calculating the three-dimensional coordinates of the hand grip point in the global coordinate system, using the three-dimensional coordinates of the hand grip point in the local coordinate system as input, and using this as the three-dimensional coordinates of the wrist joint position in the global coordinate system. The preset hand grip posture includes the hand grip point when the occupant operates the steering wheel in a standard driving posture.

[0171] The second solution scaling equation includes: upper arm length = K e1 × Height, Forearm Length = K e2 × Height. Among them, K e1 K is the scaling factor for the first upper limb calculation. e2 The scaling factor is used to calculate the second upper limb.

[0172] The second joint calculation relationship includes: taking the acromion position, wrist joint position, upper arm length, and forearm length as inputs, and based on the upper limb kinematic chain constraints (i.e., the elbow joint lies on both a sphere centered at the acromion and with the upper arm length as its radius, and a sphere centered at the wrist and with the forearm length as its radius), determining that the elbow joint position falls on the first spatial circle formed by the intersection of the two spheres); based on the biomechanical constraint that the elbow joint's natural flexion angle is within a comfortable range, candidate solutions on the first spatial circle are filtered to calculate the unique three-dimensional coordinates of the elbow joint position in the global spatial coordinate system; the mapping between the left and right elbow joint positions uses the same method, executed symmetrically. Subsequently, the elbow angle value is determined based on the angle between the vector pointing from the elbow joint position to the acromion position and the vector pointing from the elbow joint position to the wrist joint position.

[0173] In this method, the biomechanical constraint that keeps the elbow joint's natural flexion angle within a comfortable range is: the range of motion of the elbow joint is limited to its anatomical limit. 145 Furthermore, it prioritizes a comfortable flex range of 80 degrees during driving and handling. 120 This is used as a boundary constraint when solving the problem.

[0174] (3) Using the seventh variable as input, calculate the ankle joint position, knee joint position and knee angle value based on the ninth solution mapping.

[0175] The seventh variable includes height derived from occupant physical parameters, pedal surface center position and pedal surface inclination angle derived from vehicle interior parameters, and hip joint geometric center position.

[0176] The ninth solution mapping includes the third joint solution relation, the third solution scaling equation, and the fourth joint solution relation.

[0177] The third joint calculation relationship includes: establishing a local coordinate system for the pedal based on the center position of the pedal surface, using a preset foot posture as a constraint and foot length as input, to calculate the three-dimensional coordinates of the ankle joint position in the local coordinate system; constructing a rotation matrix based on the direction vectors of each axis of the local coordinate system in the global coordinate system, combining the three-dimensional coordinates of the pedal surface center position in the global coordinate system as a translation vector, and using the three-dimensional coordinates of the ankle joint position in the local coordinate system as input, to calculate the three-dimensional coordinates of the ankle joint position in the global coordinate system through rotation matrix and translation transformation. The foot length is calculated based on its quantitative relationship with height, specifically: Foot length = K f1 × Height, K f1 The scaling factor is calculated for foot length; the preset foot pedal position includes the foot contact point and ankle angle value formed on the pedal when the occupant operates the pedal in a standard driving posture.

[0178] The third solution scaling equation includes: Thigh length = K g1 × Height; Lower leg length = K g2 × Height. Among them, K g1 K is the scaling factor for solving the first lower limb. g2 The scaling factor is used to calculate the second lower limb.

[0179] The fourth joint solution involves: taking the geometric center of the hip joint, the ankle joint position, the thigh length, and the calf length as inputs; based on lower limb kinematic chain constraints—that is, the knee joint being located on both a sphere centered at the geometric center of the hip joint and with the thigh length as its radius, and a sphere centered at the ankle joint and with the calf length as its radius—the knee joint position is determined to lie on a second spatial circle formed by the intersection of these two spheres. Combined with the biomechanical constraint that the natural flexion angle of the knee joint is within a comfortable range, candidate solutions on the second spatial circle are filtered to calculate the unique three-dimensional coordinates of the knee joint position in the global spatial coordinate system. The mapping between the left and right knee joint positions uses the same method, executed symmetrically. Subsequently, the knee angle value is determined based on the angle between the vector pointing from the knee joint position to the geometric center of the hip joint and the vector pointing from the knee joint position to the ankle joint position.

[0180] In this method, the biomechanical constraint for the natural flexion angle of the knee joint to be within the comfortable range is: the range of flexion angles of the knee joint in the sagittal plane of the human body is the anatomical limit, i.e., 0°. 130°.

[0181] Ultimately, based on the posture prediction method provided by this invention, the complete posture response of an occupant in a specific vehicle layout can be predicted using occupant vital signs parameters, vehicle interior parameters, and solution constraint parameters as inputs. Specifically, the prediction results include at least the three-dimensional coordinates of 15 skeletal joints in the global spatial coordinate system, such as eye point position, C7 / T1 position, T12 / L1 position, L5 / S1 position, acromion position (left-right distinguishing), elbow joint position (left-right distinguishing), wrist joint position (left-right distinguishing), hip joint geometric center position, knee joint position (left-right distinguishing), and ankle joint position (left-right distinguishing), as well as joint angle parameters such as elbow angle and knee angle. This achieves posture output from the complete kinematic chain from pre-anchor point solution and spine reconstruction to limb analysis, providing comprehensive and quantifiable data support for posture construction of driving digital human models.

[0182] Another aspect of the present invention provides a posture prediction device for constructing a driving digital human model, the posture prediction device being configured with a computer program that is executed by a processor to implement the above-described posture prediction method for constructing a driving digital human model.

[0183] Compared with the prior art, the present invention has at least the following beneficial effects: (1) By decoupling whole-body posture prediction into three progressive stages—anchor point localization, spinal reconstruction, and limb analysis—and connecting each stage with a clear mathematical mapping, the convergence instability and computational inefficiency problems caused by traditional multivariate coupled solutions are avoided. Specifically: macroscopic anchor point prediction uses cascaded regression to directly lock the geometric center position of the hip joint and the eye point position, providing reliable boundary conditions for subsequent posture calculation; spinal reconstruction introduces individualized motion allocation coefficients, reducing the dimensionality of multi-segment spinal collaborative adjustment to a single degree of freedom solution, significantly simplifying the calculation while ensuring biomechanical rationality; limb analysis is based on a two-link kinematic chain and closed geometric solution to quickly solve the pose of the limb joints. The output of each stage can be independently verified and traced, overcoming the defect that the prediction results of traditional black-box algorithms are difficult to interpret.

[0184] (2) This invention can effectively reflect the influence of different population characteristics on driving posture. Specifically: the anchor point positioning sub-model incorporates the interaction term between body mass index and age, and the sitting height ratio variable, so that the prediction results can capture the influence of changes in soft tissue distribution and differences in trunk and lower limb proportions caused by aging on sitting posture; the spinal reconstruction sub-model constrains the tilt angle response ratio of each segment of the spine with individualized motion allocation coefficients to ensure that the spinal curvature of occupants of different body types is reasonably reproduced. Therefore, this invention can maintain high prediction consistency when facing occupant groups of different ages and body types, and is applicable to diverse posture analysis scenarios including the elderly and obese people.

[0185] (3) In terms of engineering applicability, each step of this invention uses regression equations or closed analytical formulas, eliminating the need for multi-degree-of-freedom iterative optimization. The central processing unit's computation time for a single full-body posture prediction is less than 0.005s, which can meet the batch processing requirements of tens of thousands of Monte Carlo simulations in the early stages of vehicle development. Furthermore, this invention presents open mathematical equations, allowing all mapping relationships and coefficients to be independently reproduced and verified. The prediction results are output as standard skeletal joint coordinates, unrestricted by the kinematics chain definition of a specific digital human platform, facilitating cross-platform portability. In addition, this method introduces a coordinate compensation mechanism for seat compression deformation in real-world vehicle environments, ensuring consistency between laboratory predictions and actual road test results, further guaranteeing its engineering reliability.

[0186] By comprehensively applying the above-mentioned technical means, this invention achieves simultaneous improvement in three dimensions of attitude prediction method: prediction accuracy, algorithm transparency, and computational efficiency, providing reliable technical support for human-machine engineering design and safety simulation analysis in the early stages of vehicle development.

[0187] The embodiments of the present invention have been described in detail above, but the content described is only a preferred embodiment of the present invention and should not be considered as limiting the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the patent coverage of the present invention.

Claims

1. A posture prediction method for constructing a digital human driving model, characterized in that, The attitude prediction method includes solving the occupant attitude dataset using an attitude prediction model based on a prediction sample dataset; wherein: The prediction sample dataset includes occupant vital signs parameters and vehicle interior parameters; The occupant posture dataset includes eye point feature parameters, shoulder joint feature parameters, spine feature parameters, hip joint feature parameters, and limb joint feature parameters. The posture prediction model includes an anchor point positioning sub-model, a spine reconstruction sub-model, and a limb analysis sub-model. The anchor point positioning sub-model is configured to calculate the hip joint feature parameters and the eye point feature parameters based on the occupant's vital signs parameters and the vehicle interior parameters. The spine reconstruction sub-model is configured to calculate the spine feature parameters based on the occupant's vital signs parameters, the hip joint feature parameters, and the eye point feature parameters. The limb analysis sub-model is configured to calculate the shoulder joint feature parameters and the limb joint feature parameters based on the occupant's vital signs parameters, the vehicle interior parameters, the hip joint feature parameters, and the spine feature parameters.

2. The posture prediction method for constructing a driving digital human model according to claim 1, characterized in that: The hip joint characteristic parameters include the position of the geometric center of the hip joint; The eyepoint feature parameters include the eyepoint position and the relative offset of the eyepoint, wherein the relative offset of the eyepoint is the positional offset of the eyepoint relative to the geometric center of the hip joint; The anchor point positioning sub-model includes a first solution mapping, a second solution mapping, and a third solution mapping. The first solution mapping is based on statistical regression calculation, with a first variable as input, to map the geometric center position of the hip joint. The second solution mapping is based on statistical regression calculation, using the second variable as input, to map and obtain the relative offset of the eye point; The third solution mapping is based on vector offset operation, taking the geometric center position of the hip joint and the relative offset of the eye point as input, and mapping to obtain the eye point position; The first variable independently includes at least a portion of the occupant vital signs parameters and at least a portion of the vehicle interior parameters; The second variable independently includes at least a portion of the occupant vital signs parameters and at least a portion of the vehicle interior parameters.

3. The posture prediction method for constructing a driving digital human model according to claim 2, characterized in that: The prediction sample dataset also includes solution constraint parameters, which include the load application time; The anchor point positioning sub-model also includes a correction sub-mapping, which is based on statistical regression and uses a third variable as input to map the seat cushion compression deformation variable. The third solution mapping is based on vector offset operation, taking the geometric center position of the hip joint, the relative offset of the eye point, and the compression deformation of the seat cushion as inputs to map the eye point position; The third variable independently includes at least a portion of the occupant vital signs parameters, at least a portion of the vehicle interior parameters, and the solution constraint parameters.

4. The posture prediction method for constructing a driving digital human model according to any one of claims 2-3, characterized in that: The spinal characteristic parameters include spinal length, spinal angle, and spinal joint position. The spinal reconstruction sub-model includes a fourth solution mapping, a fifth solution mapping, and a sixth solution mapping. The fourth solution mapping is based on parametric scaling of the human spine, using a fourth variable as input to map the spine length. The fifth solution mapping is based on vector offset operations and spinal kinematic chain constraints, using the hip joint geometric center position, the eye point position, and the spine length as input to map the spine angle. The sixth solution mapping is based on vector offset operations, using the hip joint geometric center position, the spine length, and the spine angle as input to map the spine joint position. The fourth variable independently includes at least a portion of the occupant vital signs parameters.

5. The posture prediction method for constructing a driving digital human model according to claim 4, characterized in that: The spinal length includes the length of the cervical spine, thoracic spine, lumbar spine, and pelvic segment. The spinal angles include the overall trunk tilt angle, cervical spine tilt angle, thoracic spine tilt angle, lumbar spine tilt angle, and pelvic segment tilt angle; The spinal joint locations include the C7 / T1 position, the T12 / L1 position, and the L5 / S1 position; The fifth solution mapping includes: based on vector offset operation and the spinal kinematic chain constraint, mapping the overall trunk tilt angle using the geometric center position of the hip joint, the eye point position, and the spine length as input; and based on the spinal kinematic chain constraint, mapping the cervical spine tilt angle, the thoracic spine tilt angle, the lumbar spine tilt angle, and the pelvic segment tilt angle using the overall trunk tilt angle as input.

6. The posture prediction method for constructing a driving digital human model according to claim 5, characterized in that, The statistical regression coefficients of the first solution mapping and the second solution mapping are determined by performing statistical regression fitting on the attitude sample dataset; The posture sample dataset includes at least prior occupant vital signs parameters, prior vehicle interior parameters, and prior sample occupant surface landmark parameters; the prior occupant surface landmark parameters include at least the positions of the left anterior superior iliac spine, the right anterior superior iliac spine, the pubic symphysis, the left posterior superior iliac spine, and the right posterior superior iliac spine, which are used to determine the prior geometric center position of the hip joint through the first center localization algorithm. The first center localization algorithm includes: establishing a body surface pelvic coordinate system using the prior occupant's body surface landmark parameters; and mapping the prior occupant's skeletal joint parameters and skeletal pelvic coordinate system based on the first localization space mapping, using the body mass index in the prior occupant's vital signs parameters as input. Based on the first positioning scaling mapping, the prior occupant skeletal landmark parameters are used as input to map the position of the prior hip joint geometric center.

7. The posture prediction method for constructing a driving digital human model according to claim 6, characterized in that, The spinal kinematic chain constraint of the fifth solution mapping is determined by the spinal motion assignment coefficient, which is determined by intra-individual statistical regression fitting of the posture sample dataset. The prior occupant surface landmark parameters include at least the C7 spinous process position, T8 spinous process position, T12 spinous process position, suprasternal notch position, and xiphoid process position, which are used to determine the prior spinal joint position through the second center localization algorithm. The prior spinal joint position includes the prior C7 / T1 position, the prior T12 / L1 position, and the prior L5 / S1 position. The second center localization algorithm includes: constructing a two-segment spinal kinematic model including the thoracic and lumbar vertebrae using the prior occupant's vital signs parameters and the prior occupant's surface landmark parameters; calculating the theoretical surface landmark parameters based on forward kinematics operations, using the two-segment spinal kinematic model as input; and calculating the prior spinal joint positions with the objective of minimizing the distance error between the theoretical surface landmark parameters and the prior occupant's surface landmark parameters. The method for determining the spinal motion allocation coefficient includes: calculating the prior overall trunk tilt angle and the prior spinal segment tilt angle under each posture sample based on the prior spinal joint position; taking an individual as a unit, determining the response ratio of the prior spinal segment tilt angle to the prior overall trunk tilt angle relative to the prior overall trunk tilt angle through statistical regression fitting based on the change relationship between the prior overall trunk tilt angle and the prior spinal segment tilt angle of each individual under each posture sample, and using this as the spinal motion allocation coefficient.

8. The posture prediction method for constructing a driving digital human model according to claim 5, characterized in that: The shoulder joint characteristic parameters include the acromion point position; The characteristic parameters of the limb joints include wrist joint position, elbow joint position, elbow angle value, knee joint position, knee angle value, and ankle joint position; The limb analysis sub-model includes a seventh solution mapping, an eighth solution mapping, and a ninth solution mapping. The seventh solution mapping is based on vector offset operation, using the fifth variable as input, to map the acromion point position. The eighth solution mapping is based on parametric scaling of the human upper limb and upper limb kinematic chain constraints, using the sixth variable as input, to map the wrist joint position, the elbow joint position, and the elbow angle value. The ninth solution mapping is based on parametric scaling of the human lower limb and lower limb kinematic chain constraints, using the seventh variable as input, to map the knee joint position, the knee angle value, and the ankle joint position. The fifth variable independently includes at least a portion of the C7 / T1 position and the occupant vital signs parameters; The sixth variable independently includes the acromion point location, at least a portion of the occupant vital signs parameters, and at least a portion of the vehicle interior parameters; The seventh variable independently includes the geometric center position of the hip joint, at least a portion of the occupant vital signs parameters, and at least a portion of the vehicle interior parameters.

9. The posture prediction method for constructing a driving digital human model according to any one of claims 1-3 and 5-8, characterized in that: The occupant vital signs parameters include gender, height, sitting height, sitting height ratio, weight, body mass index, age, and interactive items, wherein the interactive items are the product of the body mass index and the age. The vehicle interior parameters include seat height, seat cushion angle, seat cushion filling density, seat seating reference point position, steering wheel center position, steering wheel diameter, steering wheel tilt angle, pedal center position, and pedal tilt angle.

10. A posture prediction device for constructing a driving digital human model, characterized in that, The attitude prediction device is equipped with a computer program, which is executed by a processor to implement the attitude prediction method for constructing a driving digital human model as described in any one of claims 1-9.