A method and apparatus for occupant injury prediction suitable for large angle seats

Abstract

The application relates to the technical field of collision test, in particular to a passenger injury prediction method and device suitable for large-angle seats. The method comprises the following steps: in a whole vehicle collision test, the relative spatial position change curve of a passenger and a seat belt fixing point is obtained through a machine vision technology; the relative spatial position change curve is input into a passenger injury prediction model to obtain a passenger head acceleration time sequence curve, a neck axial tensile force time sequence curve and a bending moment time sequence curve; and passenger head injury indexes and neck injury indexes are calculated according to the passenger head acceleration time sequence curve, the neck axial tensile force time sequence curve and the bending moment time sequence curve. The application realizes passenger injury prediction in a collision process at a low calculation cost, and forms a passenger injury prediction model with low cost, high efficiency and high precision.

Description

Technical Field

[0001] This application relates to the field of crash test technology, and more specifically, to a method and device for predicting occupant injury applicable to seats with large angles. Background Technology

[0002] Occupant injury in a collision is a crucial parameter for evaluating vehicle safety. With the development of new technologies such as large-angle seats, the relative spatial position of the dummy's H-point and the lower seatbelt anchorage point undergoes significant changes during a collision, greatly impacting head and neck injuries. Currently, posture analysis for occupants using large-angle seats relies on computer simulation technology, but the accuracy of these simulations has not yet been experimentally verified. Therefore, a method is needed to acquire the relative position of the dummy's H-point and the lower seatbelt anchorage point in real time and accurately predict occupant injury, thereby reducing testing and computational costs.

[0003] In view of the above, this application is hereby submitted. Summary of the Invention

[0004] The purpose of this application is to provide a method and device for predicting occupant injuries in a large-angle seat, so as to achieve occupant injury prediction during a collision with low computational cost and form a low-cost, high-efficiency, and high-precision occupant injury prediction model.

[0005] To achieve the above objectives, this application adopts the following technical solution:

[0006] Firstly, this application provides a method for predicting occupant injury applicable to seats with large angles, including:

[0007] In whole vehicle crash tests, machine vision technology is used to obtain curves showing the relative spatial position changes of occupants and seat belt anchorages.

[0008] The relative spatial position change curve is input into the occupant injury prediction model to obtain the occupant head acceleration time series curve, neck axial tensile force and bending moment time series curve;

[0009] Based on the time-series curves of occupant head acceleration, neck axial tensile force, and bending moment, the occupant's head injury index and neck injury index are calculated.

[0010] Secondly, this application provides an electronic device, comprising:

[0011] At least one processor, and a memory communicatively connected to at least one of the processors;

[0012] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the above-mentioned occupant injury prediction method applicable to a large-angle seat.

[0013] Compared with the prior art, the application has the following beneficial effects:

[0014] In the vehicle collision process, the application can track the positions of the dummy H point and the seat belt fixing point in real time and calculate the relative spatial position change curve. The relative spatial position change curve of the H point and the seat belt fixing point output in the actual collision is taken as a training data set, a mapping relationship between the vehicle seat parameters, the occupant position change and the occupant head injury and neck injury is established, the hyperparameters of the Bi-LSTM model are optimized through the particle swarm algorithm, a high-precision prediction model is obtained, and the prediction error of the model is minimized. Finally, combined with the machine vision technology and the prediction model, the occupant position and the seat belt fixing point position monitoring and the occupant injury prediction under the corresponding working condition can be realized under any actual collision working condition.

[0015] The application can solve the problem that the occupant position and the seat belt fixing point change cannot be detected in the collision process. Compared with the test method, the occupant injury prediction model can greatly reduce the test cost of the occupant injury research under multiple working conditions; compared with the computer simulation technology, the occupant injury prediction model can greatly reduce the calculation cost of the occupant injury research under multiple working conditions. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the specific embodiments or the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0017] Figure 1 is a flowchart of an occupant injury prediction method applicable to a large-angle seat provided by an embodiment of the application;

[0018] Figure 2 is a schematic diagram of a target provided by an embodiment of the application;

[0019] Figure 3 is a schematic diagram of the image state corresponding to each image processing step provided by an embodiment of the application;

[0020] Figure 4 is a structural schematic diagram of an electronic device provided by the application. DETAILED DESCRIPTION

[0021] The exemplary embodiments of this application are described below with reference to the accompanying drawings, which are meant to be exemplary in nature, and include various specific details to facilitate understanding, and should not be considered limiting. For the purpose of clarity and a concise description, architectural or details of construction known to those skilled in the art have not been described in detail.

[0022] The application is further described in detail below with reference to the embodiments.

[0023] Figure 1 A flowchart of a passenger injury prediction method suitable for large-angle seats provided by the embodiments of the application, the method can be executed by a computer program and integrated in an electronic device, which can be an electronic control unit (ECU) or a telematics box (T-box). The embodiments of the application predict passenger injuries in a vehicle crash test by the passenger injury prediction method suitable for large-angle seats integrated in the ECU or T-box. As shown in Figure 1 The embodiments of the application provide a passenger injury prediction method suitable for large-angle seats, which includes the following steps:

[0024] S110, in the vehicle crash test, the relative spatial position change curve of the passenger and the seat belt fixing point is obtained by machine vision technology.

[0025] In the vehicle crash test environment, a complete test system is established, including:

[0026] THOR-50M dummy model is used to adjust the sitting posture of the dummy according to SAE J826 standard. A large-angle seat is set, and the seat back angle is adjusted to 50-60 degrees. Six high-speed cameras are installed, with a sampling frequency of 1000Hz and a resolution of 1280x800. The cameras are arranged on both sides and the top of the vehicle to ensure full coverage of the seat belt fixing point and the dummy H point.

[0027] Camera calibration is performed by using a calibration board. Specifically, Zhang Zhengyou calibration method is used to control the re-projection error within 0.1 pixels. In the vehicle crash test, all cameras are triggered synchronously, and the time synchronization accuracy is ±0.1ms.

[0028] Optionally, in the whole vehicle crash test, the position changes of the lower belt anchor points L1 (non-buckle side lower belt anchor point), L2 (buckle side lower belt anchor point) and the dummy H point (the hinge center of the torso and thigh of the crash test dummy) during the crash process are tracked in real time through machine vision technology, and the coordinate time curves of the lower belt anchor points L1, L2 and the dummy H point are obtained respectively. The coordinate time curves represent the position changes of the lower belt anchor points L1, L2 and the dummy H point.

[0029] The coordinate time curves are converted into first angle time curves and second angle time curves; wherein the first angle represents the included angle between the line connecting the lower belt anchor point L1 and the H point and the horizontal line, and the second angle represents the included angle between the line connecting the lower belt anchor point L2 and the H point and the horizontal line, see the following formula:

[0030] ;

[0031] ;

[0032] wherein, is the first angle, is the second angle, , are the vertical and longitudinal coordinates of the H point respectively, , are the vertical and longitudinal coordinates of the lower belt anchor point L1 respectively, , are the vertical and longitudinal coordinates of the lower belt anchor point L2 respectively. This embodiment only considers the vertical and longitudinal displacements of the dummy during the crash process.

[0033] S120, input the relative spatial position change curve into the occupant injury prediction model to obtain the occupant head acceleration time curve, neck axial tensile force and bending moment time curve.

[0034] The occupant injury prediction model is a Bi-LSTM model optimized by a particle swarm algorithm. The structure of the Bi-LSTM model includes: 1) input layer: receiving the first angle time curve and the second angle time curve; 2) Bi-LSTM layer: a neural network layer that can understand current information from "past" and "future" at the same time, containing two independent LSTM layers, each with 128 neurons; 3) fully connected layer; 4) output layer: 3 neurons, corresponding to head acceleration, neck axial tensile force and bending moment.

[0035] The LSTM network controls the prior state, the current input and the current memory by input gate, forget gate and output gate, thereby improving the hidden layer of the conventional RNN, and using an activation function to calculate the state of the hidden layer. Among them, the input gate controls the information transmitted to the network at each time step, and the form calculation formula is:

[0036] ;

[0037] In the formula, is the input gate matrix at time t, is a nonlinear activation function, is the weight matrix of the input gate, is the bias matrix of the input gate, is the input at time t, is the hidden state at time t-1.

[0038] The forget gate decides whether to keep the information according to the previous output and the current input, and the form calculation formula is:

[0039] ;

[0040] In the formula, is the forget gate matrix at time t, is the weight matrix of the forget gate, is the bias matrix of the forget gate.

[0041] The output gate outputs information, and the form calculation formula is:

[0042] ;

[0043] In the formula, is the output gate matrix at time t, is the weight matrix of the output gate, is the bias matrix of the output gate.

[0044] The cell state represents the cell state, which is a candidate hidden state generated from the hidden state of the current input and the previous hidden state , and the form calculation formula is:

[0045] ;

[0046] ;

[0047] is the cell state at time t, is the weight matrix in the input data unit state, is the input data unit state bias, is an activation function.

[0048] hidden state is the time sequence feature extracted by LSTM, and its form calculation formula is:

[0049] ;

[0050] The Bi-LSTM model can learn in both directions for sequences, but its part of the structure parameters (number of hidden layer neurons) and training parameters (model learning rate) are difficult to determine, which can easily affect the model training fitting effect. Therefore, the particle swarm algorithm is needed to seek the best number of hidden layer neurons and model learning rate in the Bi-LSTM model, so the group dimension of the particle swarm algorithm is set to 2, the number of particle swarm algorithm groups is set to 50, the maximum inertia weight is 0.9, the minimum inertia weight is 0.3, the iteration number is 100, and the individual acceleration coefficient and group acceleration coefficient are both 2. The number of hidden layer neurons of the Bi-LSTM model ranges from 10 to 200, and the learning rate ranges from 0.01 to 0.2. The number of input layers is 6, the number of fully connected layers is 2, the dropout rate is 0.1, the gradient threshold is 1, and the training number is 100. Thus, the initialization parameter setting of the entire model is completed.

[0051] The fitness function is the root mean square error (RMSE), and its calculation formula is:

[0052] ;

[0053] wherein, is the predicted damage value, is the actual damage value, and m is the number of data participating in the calculation.

[0054] A large number of whole vehicle collision tests are performed in advance to obtain the relative spatial position change curve as the training sample. In order to eliminate the adverse effects of dimensional differences on prediction accuracy, the sample data is normalized. In one case, relevant piezoelectric sheets / accelerometers are installed on the head and neck to measure the true head acceleration and neck bending moment, and then the passenger head injury index and neck injury index are calculated as the label / actual damage value during training. In another case, the head acceleration and neck bending moment are obtained according to the simulation environment, and then the passenger head injury index and neck injury index are calculated as the label / actual damage value during training.

[0055] The steps of optimizing the Bi-LSTM model according to the particle swarm algorithm include:

[0056] Step 1, initialize the parameters of the particle swarm algorithm (including: population size, group dimension, maximum and minimum inertia weight, iteration number, individual acceleration coefficient, group acceleration coefficient), Bi-TSLM model hyperparameters (including: number of hidden layer neural units, learning rate, dropout rate, gradient threshold, training number, input layer number, and fully connected layer number).

[0057] Step 2, complete the particle velocity and position update of the particle swarm algorithm according to the initialization parameters and the following formula.

[0058] ;

[0059] ;

[0060] ;

[0061] wherein, is the velocity of the particle at time t, is the position of the particle at time t, is the velocity of the particle at time t-1. is the velocity of the particle at time t-1. is the individual acceleration coefficient, is the group acceleration coefficient, , are random numbers in the range of 0 to 1, is the individual optimal position of the i-th particle, is the global optimal position, is the inertia weight at iteration t-1, is the maximum inertia weight, is the minimum inertia weight, and n is the set iteration number.

[0062] Step 3, calculate the fitness RMSE, and read the individual extreme value and global extreme value of the particle according to the best fitness.

[0063] Step 4, if the iteration stopping condition is reached, output the global extreme value, that is, the best hidden layer neural unit number, learning rate in the Bi-TSLM model; if the iteration stopping condition is not met, return to step 2.

[0064] Based on the output best Bi-TSLM model, complete the occupant injury prediction model and verify the model accuracy of the verification set.

[0065] S130, according to the occupant head acceleration time curve, neck axial tensile force and bending moment time curve, calculate the occupant head injury index and neck injury index.

[0066] The occupant head injury index HIC is calculated by the following formula:

[0067] ;

[0068] wherein t1, t2 are any two time points in the collision acceleration curve, is the time sequence curve of the occupant head acceleration.

[0069] The neck injury index N is calculated using the following formula:

[0070] ;

[0071] wherein, is the axial tensile force of the neck, is the intercept value corresponding to the axial tensile force of the neck, is the bending moment, is the intercept value corresponding to the bending moment.

[0072] Optionally, in the whole vehicle collision test, the position changes of the lower belt fixing points L1 and L2 and the dummy H point during the collision process are tracked in real time by machine vision technology, and the coordinate time sequence curves of the lower belt fixing points L1 and L2 and the dummy H point are obtained, including:

[0073] Step 1, using a target recognition algorithm to track the targets pasted on the lower belt fixing points L1 and L2 and the H point. The specific target recognition algorithm is not limited in this embodiment, and the related algorithm of the YOLO series can be selected. The YOLO algorithm needs to be trained with images to achieve the purpose of high recognition accuracy, for example, 1000 images can be selected and divided into training and verification sets according to the ratio of 4:1 for model training. The identification target is as shown in the figure. The target is a square with a side length of d, and the radius of the circle in the middle is r. The target can be pasted on the lower belt fixing points L1 and L2 and the H point, and a camera is placed for video acquisition. Figure 2

[0074] The detection target format output by YOLO is [class, x, y, w, h], class is the classification number of the identified target, x is the normalized x direction coordinate of the center of the identified frame, y is the normalized y direction coordinate of the center of the identified frame, w is the normalized width of the identified frame, and h is the normalized height of the identified frame. In the actual shooting process, the resolution of the camera is , then the boundary point coordinates of the identified frame can be calculated by the following formula.

[0075] ;

[0076] ;

[0077] ;

[0078] ;​

[0079] wherein, is the left edge of the bounding box, is the right edge of the bounding box, is the top edge of the bounding box, is the bottom edge of the bounding box, is the original pixel width, is the original pixel height.

[0080] Step 2, obtain the real-time center coordinates of the target image through image processing technology.

[0081] The contrast adaptive histogram equalization (CLAHE) method is used to enhance the contrast of the image. The homomorphic filtering method is used to enhance the details in the dark place. Then, the processed image needs to be threshold segmented to become a visual analysis sample. In this embodiment, based on the known area ratio p of the circle and the square, the p parameter adaptive threshold segmentation method is used to obtain the center circle part. Specifically, first, the image region in the bounding box is extracted as the processing object, and the total number of pixels S_total of the region is recorded. The gray level histogram of the extracted region is calculated, and the histogram contains 256 gray levels, each gray level corresponds to the number of pixels, and the cumulative pixel distribution from the highest gray level to the lowest gray level is counted. The expected number of pixels of the target circular region: S_target = p × S_total; start from the highest end of the gray level histogram (255 level) and count down the number of pixels, when the cumulative pixel number first reaches or exceeds S_target, record the current gray level T, which is the optimal segmentation threshold. The determined threshold T is used for image binarization, and the pixels with a gray value greater than or equal to T are set as foreground (black), and the pixels with a gray value less than T are set as background (white), to obtain the preliminary segmentation result.

[0082] Next, the Mask mask technology is applied to eliminate the edge noise of the center circle part. Specifically, the pixels inside the mask retain the original value, and the pixels outside the mask are forced to be set as the background color, effectively eliminating the residual noise of the boundary box edge.

[0083] Next, the center circle part is subjected to edge detection through the Canny operator. The Candy operator has the advantages of good detection, high positioning, and only one use for a single edge. The specific method includes: smoothing the center circle part through a Gaussian filter; calculating the size and direction of the gradient; suppressing the gradient amplitude by non-maximum suppression method; performing edge detection and connection by double-threshold method; performing edge detection on the extracted image, and storing the target circle contour as a point set.

[0084] The target contour point set obtained from the Canny edge detection needs to be accurately fitted into an ellipse equation to obtain the center coordinate and the axis length parameter. The target of the Canny algorithm is to find the pixel points with the most dramatic intensity change in the image, and the line formed by connecting these points is the edge. The non-iterative least square ellipse fitting method is used to calculate the ellipse equation corresponding to the target contour point set, and the real-time center coordinate is output. After this operation, a unique ellipse cannot be identified, but multiple ellipses with similar centers, so it is necessary to limit the relative distance of each two ellipse centers, so that only one ellipse is finally presented. Based on this, an ellipse filtering threshold is set to filter the ellipses with a center coordinate distance less than the target pixel, and finally the unique ellipse center and the major and minor axes are output. The image states corresponding to each image processing step are shown in Figure 3

[0085] Step 3, convert the real-time center coordinate into actual displacement to obtain the coordinate time sequence curve.

[0086] The displacement calculation of machine vision is to convert the moving distance of the target in the image in pixel units into the actual motion distance in millimeter units, so the two key steps of displacement calculation are coordinate change and obtaining the scale factor. That is, the absolute pixel coordinates of the real-time center coordinate in the global image are calculated; the displacement of the absolute pixel coordinates is converted into actual displacement through the scale factor to obtain the coordinate time sequence curve.

[0087] First, the change amount of the center coordinates at different times is calculated. The absolute pixel coordinates of the center under the current image can be directly output after ellipse fitting, and for different images, the recognition box corresponding to each center also changes relatively. Relative to the start time 0, the absolute coordinate calculation formula of the center at time t relative to the global image is:

[0088] ;

[0089] ;

[0090] In the formula, is the absolute pixel coordinate of the center at time t relative to the global image, is the absolute pixel coordinate of the center at time t in the current image, is the normalized coordinate of the recognition box center x at time t output by YOLO, is the normalized width of the recognition box at time t output by YOLO, is the width pixel of the global image, is the absolute pixel coordinate of the center at time t relative to the global image, is the absolute pixel coordinate of the center at time t in the current image, is the normalized coordinate of the recognition box center y at time t output by YOLO, ​The normalized height of the recognition frame output by YOLO at time t, The height of the global image in pixels.

[0091] The scale factor is obtained by camera calibration. In this embodiment, an industrial camera is used for video shooting. When shooting, the camera optical axis is at least perpendicular to one target plane, and further ignores the influence of lens distortion. The scale factor can be obtained by the following formula:

[0092]

[0093] In the formula, s is the scale factor, D is the actual length of the object, and d is the pixel length of the object in the image.

[0094] The scale factor of the target perpendicular to the camera optical axis in each frame image is calculated respectively. That is, the ratio of the actual diameter of the circle to the pixel diameter is calculated in each image, and then the average of all calculation results is calculated to obtain the final average as the scale factor. At this time, the above formula is transformed into:

[0095]

[0096] In the formula, s is the scale factor, is the number of images contained in each video, is the actual diameter of the circle of the i-th image, is the pixel diameter of the circle of the i-th image.

[0097] According to the above, the displacement calculation formula of x and y directions is:

[0098]

[0099]

[0100] In the formula, is the displacement of the circle center in x direction, is the displacement of the circle center in y direction.

[0101] Combined with the whole vehicle coordinates of the safety belt lower fixed point and the dummy H point and the displacement change calculated by the video in the collision process, the whole vehicle coordinate time sequence curve can be obtained.

[0102] ​​​​The embodiment can track the positions of the dummy H point and the seat belt fixing point in real time during vehicle collision, and calculate the relative spatial position change curve. The relative spatial position change curve of the H point and the seat belt fixing point output in the actual collision is taken as a training data set, a mapping relationship between the vehicle seat parameters, the occupant position change and the occupant head injury and neck injury is established, the hyperparameters of the Bi-LSTM model are optimized through the particle swarm algorithm, a high-precision prediction model is obtained, and the prediction error of the model is minimized. Finally, combined with the machine vision technology and the prediction model, the occupant position and the seat belt fixing point position monitoring and the occupant injury prediction under the corresponding working condition can be realized under any actual collision working condition.

[0103] The application can solve the problem that the occupant position and the seat belt fixing point change cannot be detected during collision. Compared with the test method, the occupant injury prediction model can greatly reduce the test cost of occupant injury research under multiple working conditions; compared with the computer simulation technology, the occupant injury prediction model can greatly reduce the calculation cost of occupant injury research under multiple working conditions.

[0104] As shown in Figure 4 The embodiment provides an electronic device, which comprises:

[0105] at least one processor; and

[0106] a memory connected with the at least one processor in communication; wherein

[0107] The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the above method. The at least one processor in the electronic device can perform the above method, and thus has at least the same advantages as the above method.

[0108] Optionally, the electronic device further comprises an interface for connecting various components, including a high-speed interface and a low-speed interface. Various components are connected to each other by different buses, and can be installed on a common motherboard or in other ways as needed. The processor can process instructions executed within the electronic device, including graphical information stored in the memory or on the memory to display a GUI (Graphical User Interface) on an external input / output device such as a display device coupled to the interface. In other embodiments, multiple processors can be used with multiple memories, and / or multiple buses can be used with multiple memories, if necessary. Similarly, multiple electronic devices can be connected (for example, as a server array, a group of blade servers, or a multi-processor system), each device providing part of the necessary operations. Figure 4 The processor 301 is taken as an example in the embodiment.

[0109] The memory 302, as a computer readable storage medium, can be used to store software programs, computer executable programs and modules, such as program instructions / modules corresponding to the occupant injury prediction method applicable to large-angle seats in the embodiments of the present application. The processor 301 performs various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory 302, that is, implements the above-mentioned occupant injury prediction method applicable to large-angle seats.

[0110] The memory 302 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system and at least one application required by a function; the data storage area can store data created according to the use of the terminal, etc. In addition, the memory 302 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state memory device. In some examples, the memory 302 can further include a memory remotely arranged with respect to the processor 301, which can be connected to the device through a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0111] The electronic device can also include an input device 303 and an output device 304. The processor 301, the memory 302, the input device 303 and the output device 304 can be connected by a bus or other means, Figure 4 For example, by bus connection.

[0112] The input device 303 can receive input digital or character information, and the output device 304 can include a display device, an auxiliary lighting device (for example, an LED), a tactile feedback device (for example, a vibration motor), etc. The display device can include but is not limited to a liquid crystal display (LCD), a light emitting diode (LED) display and a plasma display. In some embodiments, the display device can be a touch screen.

[0113] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network or other programmable apparatus. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through a wired manner, for example, coaxial cable, optical fiber, digital subscriber line (DSL) or a wireless manner, for example, infrared, wireless, microwave, etc. The computer-readable storage medium can be any available medium accessible by a computer, or a data storage device such as a server, data center, etc. integrated with one or more available media. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium or a semiconductor medium, etc. It should be noted that the computer-readable storage medium mentioned in the embodiments of the present application can be a non-volatile storage medium, in other words, can be a non-transitory storage medium.

[0114] It should be understood that the above-mentioned various forms of processes can be reordered, added or deleted. For example, the steps described in the present application can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in the present application can be achieved, which are not limited herein.

[0115] The above detailed description does not constitute a limitation on the scope of protection of the present application. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A method for predicting occupant injury applicable to seats with large angles, characterized in that, include: In whole vehicle crash tests, machine vision technology is used to obtain curves showing the relative spatial position changes of occupants and seat belt anchorages. The relative spatial position change curve is input into the occupant injury prediction model to obtain the occupant head acceleration time series curve, neck axial tensile force and bending moment time series curve; Based on the time-series curves of occupant head acceleration, neck axial tensile force and bending moment, calculate the occupant head injury index and neck injury index. In the whole vehicle crash test, machine vision technology was used to obtain the relative spatial position change curves of the occupant and seat belt anchorage points, including: In the whole vehicle crash test, a target recognition algorithm was used to track the target attached to the fixed points L1, L2 and H under the seat belt; the real-time center coordinates of the target image were obtained through image processing technology; the real-time center coordinates were converted into actual displacement to obtain the coordinate time series curve. The coordinate time series curve is converted into a first angle time series curve and a second angle time series curve; wherein, the first angle represents the angle between the line connecting the lower fixing point L1 and point H of the seat belt and the horizontal line, and the second angle represents the angle between the line connecting the lower fixing point L2 and point H of the seat belt and the horizontal line.

2. The occupant injury prediction method applicable to seats with large angles according to claim 1, characterized in that, The occupant injury prediction model is a Bi-LSTM model optimized by the particle swarm optimization algorithm.

3. The occupant injury prediction method applicable to large-angle seats according to claim 1, characterized in that, Based on the aforementioned time-series curves of occupant head acceleration, neck axial tensile force, and bending moment, occupant head injury indices and neck injury indices are calculated, including: The HIC (Head Injury Index) for occupants is calculated using the following formula; ； Where t1 and t2 are any two time points in the collision acceleration curve. This is the timing curve of the occupant's head acceleration.

4. The occupant injury prediction method applicable to seats with large angles according to claim 1, characterized in that, Based on the aforementioned time-series curves of occupant head acceleration, neck axial tensile force, and bending moment, occupant head injury indices and neck injury indices are calculated, including: The neck injury index N is calculated using the following formula: ； in, For the axial tensile force of the neck, This is the intercept value corresponding to the axial tensile force in the neck. For bending moment, This is the intercept value corresponding to the bending moment.

5. The occupant injury prediction method applicable to seats with large angles according to claim 1, characterized in that, Using image processing techniques, the real-time coordinates of the target's center are obtained, including: Image contrast is enhanced by using a contrast-adaptive histogram equalization method. Use homomorphic filtering to enhance details in dark areas; Based on the known area ratio of the circle to the square, the central circular part is obtained by using the p-parameter adaptive threshold segmentation method. Apply masking technology to eliminate edge noise in the central circular area; Edge detection is performed on the central circular portion using the Canny operator; The equation of the ellipse is calculated using a non-iterative least squares ellipse fitting method, and the real-time coordinates of the center of the circle are output.

6. The occupant injury prediction method applicable to large-angle seats according to claim 1, characterized in that, The real-time center coordinates are converted into actual displacements to obtain a coordinate time-series curve, including: Calculate the absolute pixel coordinates of the real-time circle center in the global image; By using a scaling factor, the displacement of absolute pixel coordinates is converted into actual displacement, thus obtaining the coordinate time series curve.

7. An electronic device, characterized in that, include: A processor and a computer program stored in a memory, characterized in that, when the processor executes the computer program, it implements the occupant injury prediction method applicable to large-angle seats as described in any one of claims 1-6.

Images