A method for measuring the lower limb motion posture of a car collision dummy

By capturing and analyzing the edge profile of the lower limb of the car collision dummy, the problem of difficult measurement of the skeleton connection points in the existing technology is solved, efficient and accurate data collection is achieved, and the evaluation of automobile safety performance and design improvements are improved, and R&D is reduced. cost.

CN118518094BActive Publication Date: 2024-11-05CHINA AUTOMOTIVE TECH & RES CENT CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410995475.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-11-05
Estimated Expiration
2044-07-24

AI Technical Summary

Technical Problem

In the prior art, the skeleton connection point of the lower limb of a car crashing dummy is difficult to measure directly due to the dummy's skin obstruction, making it difficult to evaluate its movement, affecting the evaluation of the safety performance of the car.

Method used

By capturing the edge contour of the dummy's lower limbs and tracking and computing the position and angle changes of its skeleton connection points in real time, a two-dimensional dynamic model diagram is formed, which overcomes the problem of skin occlusion and improves the speed and accuracy of data collection.

Benefits of technology

It realizes more precise measurement of driver's lower limb movements, provides more reliable data to support the improvement of car safety performance, reduces R&D costs, and accelerates the process of car safety technology.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118518094B_ABST
    Figure CN118518094B_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of automobile collision dummies, and discloses a method for measuring the motion posture of the lower limbs of automobile collision dummies. In the actual vehicle collision test, the method continuously collects lower limb images of the automobile collision dummy during the collision process; identifies the edge contour in each lower limb image; samples multiple points on the edge contour for fitting, and respectively obtains the outer thigh expression, the inner thigh expression, the outer calf expression, and the inner calf expression; determines the position of each bone of the lower limb, the position of each joint, and the angle of each bone; and associates the position of each bone of the lower limb, the position of each joint, and the angles at both ends of the joint in multiple lower limb images in time sequence to form a series of two-dimensional human rod-shaped model images, wherein the two-dimensional human rod-shaped model images characterize the movement changes of the lower limbs, and overcome the problem in the prior art that the skeleton connection points are difficult to directly measure due to the occlusion of the dummy skin.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of automobile collision dummies, and in particular to a method for measuring the motion posture of lower limbs of an automobile collision dummy. Background Art

[0002] During the safety performance testing of automobiles, it is necessary to conduct collision tests on the automobiles. In the actual vehicle test, the dummy simulates a real person sitting in the seat of the car, and then the various data of the dummy are collected in the car collision test to simulate the value of the injury suffered by the occupants in the event of an accident. In these tests, the movement measurement of the driver's dummy's lower limbs is particularly important because it is directly related to the degree of injury that the driver may suffer in the accident and the evaluation of the vehicle's safety performance. In a collision accident, the movement state of the driver's lower limbs and the impact force directly affect the possibility and severity of his injury. Accurately measuring the movement of the driver's dummy's lower limbs through actual vehicle collision tests can help researchers analyze and evaluate the driver's risk of injury in different collision situations, and then optimize the design of the vehicle, such as the steering wheel, pedals, seats, etc., to increase the driver's survival space and reduce injuries.

[0003] In the prior art, the skeleton connection points of the dummy's lower limbs are blocked by the dummy's skin and cannot be directly measured, which makes it difficult to evaluate the movement of the dummy's lower limbs in a real vehicle collision test. In view of this, the present invention is proposed. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method for measuring the motion posture of the lower limbs of a car collision dummy. By capturing the edge contours of the dummy's lower limbs and tracking and calculating the position and angle changes of its skeleton connection points in real time, a two-dimensional dynamic model diagram of the dummy's lower limbs during the collision process can be obtained. This method is efficient and accurate, and overcomes the problem in the prior art that the skeleton connection points are difficult to measure directly due to the occlusion of the dummy's skin. This innovative method improves the speed and accuracy of data collection, thereby more effectively evaluating the possible injuries that drivers may suffer in accidents, and ultimately forming an evaluation system for car collision tests.

[0005] An embodiment of the present invention provides a method for measuring the motion posture of a lower limb of a car collision dummy, the method comprising:

[0006] In the actual vehicle collision test, the lower limb images of the vehicle collision dummy are continuously collected during the collision process.

[0007] Identify edge contours in each lower limb image;

[0008] Sampling multiple points on the edge contour for fitting, respectively obtaining the outer thigh expression, inner thigh expression, outer calf expression and inner calf expression;

[0009] According to the positional relationship between the femur, shin bone and the inner and outer contours, the femur expression is obtained according to the outer thigh expression and the inner thigh expression, and the shin bone expression is obtained according to the outer shin expression and the inner shin expression;

[0010] The positions of the bones, joints and angles of the bones in the lower limbs are determined based on the expressions of the thigh bones and the calf bones.

[0011] The positions of the bones, joints and angles at both ends of the joints of the lower limbs in the multiple lower limb images are associated in time sequence to form a two-dimensional human body rod model diagram, which represents the movement state of the lower limbs.

[0012] The embodiments of the present invention have the following technical effects:

[0013] (1) By capturing the edge contours of the dummy's lower limbs in real time and calculating the position and angle changes of the skeleton connection points, the movement of the driver's dummy's lower limbs during a collision can be measured more accurately. This high-precision data collection helps to more accurately simulate and evaluate the injuries that actual occupants may suffer in a collision, thereby providing reliable data support for improving the safety performance of the vehicle. Accurate lower limb movement data can help automakers and designers identify and improve the vehicle's safety design, such as seat structure, seat belt system, dashboard layout, etc. These improvements can provide better protection for the driver, reduce the risk of injury in a collision accident, and improve the overall safety of the vehicle.

[0014] (2) The high efficiency of this method means that more collision test data can be collected in a shorter period of time, which will accelerate the development of automotive safety technology. Rapidly obtained data feedback can be used in a timely manner for design improvements, shortening the time from design to market for products. Traditional collision tests are often costly, and the method of the present invention is expected to reduce the number of collision tests by improving the efficiency and accuracy of data collection, thereby reducing research and development costs.

[0015] (3) The present invention forms an expression based on the edge contour and obtains the position and angle of the bone according to the positional relationship between the contour and the bone and geometric principles, thereby overcoming the problem in the prior art that the skeleton connection points are difficult to measure directly due to the occlusion of the dummy skin. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 is a flow chart of a method for measuring the lower limb motion posture of a car collision dummy provided by an embodiment of the present invention;

[0018] Figure 2 is a schematic diagram of camera arrangement positions provided by an embodiment of the present invention;

[0019] Figure 3 is a structural diagram of a deep convolutional neural network provided by an embodiment of the present invention;

[0020] Figure 4 is a schematic diagram of the edge contour of the lower limbs of a car collision dummy provided by an embodiment of the present invention;

[0021] Figure 5 It is a two-dimensional human body rod model diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0023] The method for measuring the motion state of the lower limbs of a car crash dummy in a real car crash test provided by the embodiment of the present invention is mainly applicable to the application scenario of monitoring the motion of the lower limbs of a car crash dummy in a real car crash test. The car crash dummy can be configured on a real car or in a separate cockpit.

[0024] Embodiment 1

[0025] Figure 1 Flow chart of a method for measuring the lower limb motion posture of a car collision dummy provided by an embodiment of the present invention. Figure 1 , the method specifically comprises:

[0026] S110. In a real vehicle collision test, continuously collect images of the lower limbs of the vehicle collision dummy during the collision process.

[0027] As the collision progresses, the lower limbs of the car crash dummy will have a certain degree of movement, which can be captured by continuous shooting of the camera.

[0028] Optionally, in order to obtain a wider field of view and a higher spatial resolution, multiple cameras are arranged in the vehicle cockpit along the direction of the lower limbs of the automobile collision dummy, and the lower limb images of the automobile collision dummy during the collision are continuously collected by the multiple cameras simultaneously.

[0029] The camera is arranged in such a way that Figure 2 As shown in the figure, six cameras are arranged, namely D1, D2, D3, D4, D5, and D6. Short focal length cameras are preferred to obtain a wider field of view and higher spatial resolution, so as to capture subtle limb movements at the moment of collision. These cameras are precisely calibrated to ensure that the images they capture overlap to form a continuous field of view, covering all key areas of the dummy's lower limbs. During the collision test, multiple cameras will work simultaneously to capture the outline of the dummy's lower limbs from different angles and positions.

[0030] The lower limb images captured by multiple cameras at the same time are stitched together to form an image showing the complete lower limb. For example, if one image shows the calf and another image shows the thigh, then the stitching is a complete lower limb. Image stitching technology is to stitch together several overlapping images into a seamless panoramic lower limb image, which uses image registration technology and image fusion technology. For details, please refer to the description of the prior art, which will not be repeated here.

[0031] S120, identifying edge contours in each lower limb image.

[0032] Each lower limb image is input into a deep convolutional neural network, which outputs the location of the thigh and calf edge contours. Figure 3 is a structural diagram of a deep convolutional neural network provided by an embodiment of the present invention, defining is the convolutional layer, represents the downsampling layer, and i represents the layer number. Figure 3 middle , , represents three convolutional layers, , represents two pooling layers, Represents the fully connected layer, and the output layer outputs the final image recognition result: the location of the edge contours of the thigh and calf.

[0033] The mathematical formulas of the convolution layer and the pooling layer are introduced in detail below. In the convolution layer, each feature pixel value in the feature map is obtained by sliding convolution of the convolution kernel of this layer with the output image of the previous layer. Therefore, the feature pixel value The calculation formula can be expressed as:

[0034] ; (1)

[0035] in, , and Represent the first layer of convolution kernels Length and width. is the bias of the corresponding j-th feature map of layer l. and They represent the current vertical and horizontal feature pixels respectively, and the variables u and v represent the convolution kernels respectively. The vertical and horizontal step lengths of . p represents the pth training sample. Symbol Represents the activation function of layer l. The convolution operation occurs when the convolution kernel convolves the input feature map , l represents the number of convolutional layers, i represents the number of feature pixels, Represents the feature pixel in the jth feature map.

[0036] In the pooling layer, each pixel value of the feature map is sampled by sliding the pooling window over the convolution feature map output by the previous layer. The pooling layer used in this paper is average pooling, and the feature pixel value after pooling is The calculation formula can be expressed as:

[0037] ; (2)

[0038] in, , and are defined as 1-layer downsampling windows respectively. and They represent the current vertical and horizontal feature pixels respectively. and Represents the vertical and horizontal strides of the pooling window respectively. The average pooling operation is to add all feature pixel values ​​in the pooling window and then divide it by the total number of pixels in the pooling window to get the average value of the pooling area.

[0039] Feature pixel values ​​calculated by the fully connected layer The formula can be expressed as:

[0040] ; (3)

[0041] in, Represents the number of neurons in layer l-1. is the weight connecting neuron i in layer l-1 to neuron j in layer j. is the bias of the jth neuron in layer l, represents the activation function of layer l. This paper adopts the classic sigmoid function as the output function of the output neuron. After the feature extraction of the entire convolutional neural network, the original image is finally transformed into the feature vector of the output layer of the fully connected layer. These feature vectors are finally sent to the classifier for classification, and the deep edge features of the entire network can be extracted. Figure 4The edge contour lines of the outer side 4-1 of the dummy's thigh, the inner side 4-2 of the thigh, the outer side 4-3 of the calf, and the inner side 4-4 of the calf are shown.

[0042] S130, sampling multiple points on the edge contour for fitting, and obtaining an expression for the outer side of the thigh, an expression for the inner side of the thigh, an expression for the outer side of the calf, and an expression for the inner side of the calf, respectively.

[0043] from Figure 4 It can be seen from the above that the contour lines of the outer thigh 4-1, the inner thigh 4-2, the outer calf 4-3, and the inner calf 4-4 have their own characteristics and are suitable for establishing expressions separately.

[0044] Optionally, multiple points are sampled at equal intervals on the edge contours of the outer side of the thigh, the inner side of the thigh, the outer side of the calf and the inner side of the calf to perform polynomial curve fitting to obtain expressions of corresponding parts.

[0045] See also Figure 5 , taking the intersection of the seat back and the seat cushion as the origin, establish an xz rectangular coordinate system, with the x-axis along the seat cushion and the z-axis along the backrest. The coordinates (x, z) of each point on the edge contour line of the dummy's lower limbs after deep convolutional neural network processing can be obtained. The coordinates of the points on the edge line are taken every 1 mm and recorded as , i=1,2,……,n, where x is a natural number sequence, ranging from 1 to n. According to the coordinates of each point, a polynomial curve fitting is performed on each edge contour line to obtain a function expression.

[0046] The mathematical formula for the first-order polynomial curve fitting operation is as follows. Now the coordinates of each point are known , let the linear curve fitting function of these data be for:

[0047] ; (4)

[0048] in, and are the coefficients to be fitted.

[0049] According to the least squares polynomial fitting principle, the fitting formula is:

[0050] ; (5)

[0051] In formula (5), let , , , , then the formula can be expressed as:

[0052] ; (6)

[0053] in It is the cumulative sum of natural numbers. Set n to start from 1. According to the arithmetic sequence summation formula, we can get:

[0054] ; (7)

[0055] According to the formula of sum of squares of arithmetic sequence, we can get:

[0056] ; (8)

[0057] First, substitute formula (8) and formula (7) into formula (5) to obtain:

[0058] ; (9)

[0059] To reduce the number of operations, By reducing it, we can get the coefficient The optimization calculation formula is:

[0060] ; (10)

[0061] in, It is a product followed by an integral. Usually the value is relatively large, so it is best to perform the calculation first. , and then multiply by 2, so that shifting can be used for fast calculation.

[0062] get After the expression of , we can substitute formula (10) into formula (6) to get the coefficient The calculation formula is:

[0063] ; (11)

[0064] So far, the two coefficients in the fitting formula have been calculated. Similarly, the coefficients of other edge line fitting functions can be obtained. Suppose the expression of the outer side of the dummy thigh obtained by this method is , the inner thigh expression is , the expression for the outer side of the calf is , the expression for the inner side of the calf is .

[0065] S140. According to the positional relationship between the femur, calf bone and the inner and outer contours, obtain the femur expression according to the outer thigh expression and the inner thigh expression, and obtain the calf bone expression according to the outer calf expression and the inner calf expression.

[0066] In the car collision dummy, the femur and shin bone are set in the leg position. Therefore, in order to simplify the calculation, the femur and shin bone are simplified into lines. The expressions of the femur and shin bone can be determined according to the positional relationship between the femur, shin bone and the inner and outer contours.

[0067] Specifically, the femur and shin bones of the dummy are located in the middle of the upper and lower edges, so the average of the outer thigh expression and the inner thigh expression is used as the expression of the femur, and the average of the outer calf expression and the inner calf expression is used as the expression of the calf. The function expression of the femur of the dummy can be obtained as follows:

[0068] ; (12)

[0069] Similarly, the function expression of the shin bone of the dummy can be obtained as:

[0070] ; (13)

[0071] S150. Determine the positions of the bones, joints and bone angles of the lower limbs according to the thigh bone expression and the calf bone expression.

[0072] The lower limb bones include the thigh bone and shin bone, and the joints include the knee joint, ankle joint, and H point (the connection point between the human torso and the thigh). The angles of the bones include the angle between the thigh bone and the shin bone and the horizontal plane, and the angle between the thigh bone and the shin bone.

[0073] Specifically, the femur expression (12) and the shin bone expression (13) are combined to obtain the knee joint position, which is the connection point between the femur and the shin bone. Coordinates:

[0074] ; (14)

[0075] According to the length of the femur The H point position of the car collision dummy is obtained by combining formula (12) and formula (14) to obtain the coordinates of the H point:

[0076] ; (15)

[0077] According to the length of the calf bone , the knee joint position and the calf bone expression to obtain the ankle joint position, that is, by combining formula (13) and formula (14), we can get the connection point between the calf bone and the foot The coordinates are:

[0078] ; (16)

[0079] According to the expression of the femur, the expression of the shin bone and the position of the knee joint, we can get the angle between the femur and the horizontal plane and the angle between the femur and the shin bone. That is, by combining formula (12), formula (13) and formula (14), we can get the angle between the femur and the horizontal plane: The angle between the thigh bone and the shin bone They are:

[0080] ; (17)

[0081] In addition to calculating the position and angle of the thigh and calf, this embodiment also takes into account the position and angle of the dummy's shoes, thereby more comprehensively studying the movement state of the dummy's lower limbs.

[0082] Specifically, identifying the edge contour in each lower limb image includes: inputting each lower limb image into a deep convolutional neural network and outputting the position of the edge contour of the shoe. The deep convolutional neural network has the characteristic of extracting the edge contour of the object, and can extract the edge contours of the thigh, calf and shoe at the same time through image processing. On this basis, the shoe is simplified into a plane to determine the length and angle of the shoe plane. Figure 5 The XOZ plane is parallel to the shoe plane and perpendicular to the sole plane.

[0083] See also Figure 5 , identify the positions of the two endpoints in the length direction (that is, along the direction of the shoe plane) from the shoe contour of each lower limb image; the two endpoints include the lower edge point and the upper edge point . Construct an expression for the shoe plane based on the positions of the two endpoints:

[0084] ; (18)

[0085] Combining the expression (18) of the shoe plane and the expression (13) of the calf bone, we can get the angle between the shoe plane and the horizontal plane: , and the angle between the shoe plane and the calf :

[0086] ; (19)

[0087] S160, associating the positions of the bones, joints and angles at both ends of the joints of the lower limbs in the multiple lower limb images in chronological order to form a series of two-dimensional human body rod-shaped model images, wherein the two-dimensional human body rod-shaped model images represent the movement changes of the lower limbs.

[0088] Finally, the coordinates and angles of the lower limb bones and joints of each dummy are calculated to form a two-dimensional human rod model, such as Figure 5 During the collision between the dummy and the real car, each camera captures the real-time data. After the above analysis, a series of two-dimensional human body bar model diagrams are obtained, which can show the movement of the key points of the dummy's lower limbs in the real car collision.

[0089] It should be noted that the terms used in the present invention are only for describing specific embodiments, rather than limiting the scope of the present application. As shown in the present specification, unless the context clearly indicates an exception, the words "one", "a", "a kind of" and / or "the" do not specifically refer to the singular, but may also include the plural. The terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method or device. In the absence of further restrictions, the elements defined by the sentence "include one..." do not exclude the presence of other identical elements in the process, method or device including the elements.

[0090] It should also be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. Unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", etc. should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be a connection between the two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A method for measuring the lower limb motion posture of a car collision dummy. It is characterized in that include: In the real car collision test, the lower limb images of the car collision dummy during the collision are continuously collected; Identify edge contours in each lower limb image; A plurality of points are sampled at equal intervals on the edge contours of the outer side of the thigh, the inner side of the thigh, the outer side of the calf and the inner side of the calf to perform a polynomial curve fitting, and an expression of the outer side of the thigh, an expression of the inner side of the thigh, an expression of the outer side of the calf and an expression of the inner side of the calf are obtained respectively; According to the positional relationship between the femur, shin bone and the inner and outer contours, the femur expression is obtained according to the outer thigh expression and the inner thigh expression, and the shin bone expression is obtained according to the outer shin expression and the inner shin expression; the femur and shin bone are simplified into lines, and the expressions of the femur and shin bone are determined according to the positional relationship between the femur, shin bone and the inner and outer contours; the femur and shin bone of the dummy are located in the middle of the upper and lower edges, and the average of the outer thigh expression and the inner thigh expression is used as the femur expression, and the average of the outer shin expression and the inner shin expression is used as the shin bone expression; Determine the position of each bone, each joint position and each bone angle of the lower limbs according to the thigh bone expression and the calf bone expression; The positions of the bones, joints and angles at both ends of the joints of the lower limbs in multiple lower limb images are associated in chronological order to form a series of two-dimensional human body rod model diagrams, which represent the movement changes of the lower limbs.

2. The method according to claim 1, It is characterized in that Arrange multiple cameras in the vehicle cockpit along the direction of the lower limbs of the car crash dummy; In the real car collision test, the lower limb images of the car collision dummy during the collision are continuously collected, including: In the real car collision test, multiple cameras are used to continuously collect images of the lower limbs of the car collision dummy during the collision process; The lower limb images captured by multiple cameras at the same time are stitched together to form an image showing the complete lower limb.

3. The method according to claim 1, It is characterized in that The identifying edge contours in each lower limb image comprises: Each lower limb image is input into a deep convolutional neural network, which outputs the location of the thigh and calf edge contours.

4. The method according to claim 3, It is characterized in that Determining the positions of bones, joints and bone angles of the lower limbs according to the thigh bone expression and the calf bone expression includes: Combine the thigh bone expression and the calf bone expression to get the knee joint position; The H point position of the car collision dummy is obtained based on the femur length and knee joint position; The ankle joint position is obtained based on the calf bone length, knee joint position and calf bone expression; The angle between the femur and the horizontal plane and the angle between the femur and the shin bone are obtained according to the femur expression, the shin bone expression and the knee joint position.

5. The method according to claim 1, It is characterized in that The identifying edge contours in each lower limb image comprises: Each lower limb image is input into a deep convolutional neural network, which outputs the location of the shoe edge contour.

6. The method according to claim 5, It is characterized in that According to the femur expression and the calf bone expression, the positions of the bones, joints and bone angles of the lower limbs are determined, including: Identify the locations of two endpoints in the length direction from the shoe contour of each lower limb image; Constructing an expression of the shoe plane according to the positions of the two endpoints; By combining the expression of the shoe plane and the expression of the calf bone, the angle between the shoe plane and the horizontal plane, and the angle between the shoe plane and the calf are obtained.