CN114743252A - Feature point screening method, device and storage medium for head model - Google Patents

Abstract

The invention relates to the technical field of measurement, and discloses a method and equipment for screening feature points of a head model and a storage medium. The method comprises the following steps: preprocessing an original image of a preset head model to obtain a target image, wherein the preset head model is used for an automobile collision dummy; determining a head reference plane and a reference line based on the target image; combining a head reference surface and a reference line, and dividing a face region according to the proportional relation of the three parts, the three parts and the five parts to obtain a plurality of face subregions; determining initial feature points from the plurality of face sub-regions according to a preset selection condition; and screening the initial feature points at least by combining facial parts missing from a preset head model to obtain a plurality of feature point sets, wherein different feature point sets are used for describing features of different parts of the head and the face. The embodiment can quickly obtain the head characteristic points required by measurement, and provides guidance for designing the bionic dummy head model with the face and the size according with the Chinese characteristics.

Description

Feature point screening method, device and storage medium for head model

technical field

The invention relates to the technical field of measurement, and in particular, to a feature point screening method, device and storage medium for a head model.

Background technique

With the continuous advancement of technology, the application fields of human measurement data are becoming more and more extensive. For example, the head design of a car crash dummy is closely related to the measurement data of human head and face. The head and face have unique and diverse features, which are the most direct and critical external features to distinguish different people. As an important bionic device for testing vehicle safety, the design of the crash dummy head requires accurate head and face size data as a design reference to describe the complex human head and face shape.

However, the current crash dummies used in my country's automobile safety regulations are all developed based on American anthropometric data, which makes the protection effectiveness of the existing automobile safety design on the Chinese human body debatable. The facial measurement data is applied to the dummy head design to meet the Chinese head and facial features.

In view of this, the present invention is proposed.

SUMMARY OF THE INVENTION

In order to solve the above technical problems, the present invention provides a feature point screening method, device and storage medium for a head model, which realizes the feature point screening for the head model and provides a reference for the simplified design of the bionic head model. .

An embodiment of the present invention provides a feature point screening method for a head model, the method comprising:

Perform preprocessing on an original image of a preset head model to obtain a target image, wherein the preset head model is used for a car crash dummy;

determining a head reference plane and a reference line based on the target image;

Combining the head datum plane and the datum line, the facial area is divided according to the proportional relationship of three stops and five parts, and a plurality of facial sub-areas are obtained;

Determine initial feature points from the plurality of face sub-regions according to preset selection conditions;

The initial feature points are screened at least in combination with the missing facial parts of the preset head model to obtain multiple feature point sets, wherein the different feature point sets are used to describe the features of different parts of the head and face.

An embodiment of the present invention provides an electronic device, and the electronic device includes:

processor and memory;

The processor is configured to execute the steps of the feature point screening method for a head model described in any one of the embodiments by invoking a program or an instruction stored in the memory.

An embodiment of the present invention provides a computer-readable storage medium, where the computer-readable storage medium stores a program or an instruction, and the program or instruction causes a computer to perform the feature point screening for a head model described in any one of the embodiments steps of the method.

The embodiment of the present invention has the following technical effects:

The feature point screening method provided by the embodiment of the present invention can quickly and simply screen out representative feature points from the numerous feature points on the head of the collision dummy, avoiding excessive description, and the screened feature points can reflect the main head and face information , describe facial features. The principle of the method is clear and easy to understand, the operation method is relatively simple, no expensive equipment is required, the time period is short, the engineering cost is low, and the head feature points required for the measurement can be quickly obtained. The dummy head model provides guidance.

Description of drawings

In order to illustrate the specific embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that need to be used in the description of the specific embodiments or the prior art. Obviously, the accompanying drawings in the following description The drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative efforts.

1 is a flowchart of a feature point screening method for a head model provided by an embodiment of the present invention;

2 is a schematic diagram of a Hybrid III dummy head model provided by an embodiment of the present invention;

3 is a schematic diagram of a head reference plane provided by an embodiment of the present invention;

4 is a schematic diagram of a part of a head reference line provided by an embodiment of the present invention;

5 is a schematic diagram of a part of a head reference line provided by an embodiment of the present invention;

6 is a schematic diagram of a head and face area division provided by an embodiment of the present invention;

7 is a schematic diagram of feature points of a dummy head model provided by an embodiment of the present invention;

8 is a schematic diagram of a feature point of a dummy head model provided by an embodiment of the present invention;

9 is a schematic diagram of a feature point of a dummy head model provided by an embodiment of the present invention;

10 is a schematic diagram of a feature point of a dummy head model provided by an embodiment of the present invention;

11 is a schematic diagram of a local feature point search process provided by an embodiment of the present invention;

12 is a schematic diagram of a feature point of a dummy head profile provided by an embodiment of the present invention;

13 is a schematic diagram of a feature point of a dummy head profile provided by an embodiment of the present invention;

14 is a schematic diagram of feature points of facial features on a dummy head provided by an embodiment of the present invention;

15 is a schematic diagram of a feature point screening process according to an embodiment of the present invention;

16 is a schematic flowchart of a feature point extraction provided by an embodiment of the present invention;

FIG. 17 is a schematic diagram of screening head profile features and main features of the nose according to an embodiment of the present invention;

18 is a schematic diagram of a technical route for feature point screening provided by an embodiment of the present invention;

FIG. 19 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be described clearly and completely below. Obviously, the described embodiments are only some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present invention.

The feature point screening method for the head model provided by the embodiment of the present invention is mainly suitable for overcoming the problem that the simplified head model feature point of the collision dummy head cannot be quickly and easily screened in China at present. The embodiment of the present invention The purpose is to provide a method with reasonable logic, simple operation, high economical applicability, and satisfying the feature point screening of this type of head model at this stage, and provide a reference for the simplified design of the bionic head model.

FIG. 1 is a flowchart of a feature point screening method for a head model provided by an embodiment of the present invention. Referring to Figure 1, the feature point screening method for the head model specifically includes:

S110. Perform preprocessing on an original image of a preset head model to obtain a target image, where the preset head model is used for a car collision dummy.

The present invention will be further described in detail with the Hybrid III dummy head model, so that those skilled in the art can refer to the description text to implement accordingly. Among them, the schematic diagram of the Hybrid III dummy head model can be referred to as shown in Figure 2.

The acquisition devices used for the original images of the head model are generally high-definition cameras, digital cameras, mobile phone cameras, etc., which will lead to different quality and clarity of the original images collected, and some original images are even accompanied by strong noise, color, etc. The brightness will also be different, which will make the feature information of the image difficult to process, which will affect the extraction and accuracy of image features and bring difficulties to the analysis. Therefore, the image preprocessing work is more meaningful in the feature point screening. The color photos are processed into grayscale images through grayscale and other preprocessing methods to ensure high-quality and efficient recognition. By performing edge detection on the image, you can To achieve the purpose of reducing the time spent in subsequent feature point screening.

In general, the preprocessing is performed on the original image of the preset head model to obtain the target image, including:

Grayscale processing is performed on the original image to obtain a grayscale image corresponding to the original image; edge detection is performed on the grayscale image to obtain the target image.

Specifically, the color in a color image is composed of three basic pigment components, R, G, and B. According to the different proportions assigned to each basic pigment component, the brightness of the three basic pigments is different, thereby forming different colors. Therefore, when grayscale processing is performed on the original image, the grayscale value of the pixel point is first calculated according to the three primary color pigment components of the pixel point in the original image, and then the grayscale value is assigned to each basic pigment component, and finally the grayscale image is obtained. . The calculation formula of the gray value of the pixel point is shown in the following formula (1).

（1）

(1)

Among them, g ( x, y ) is the gray value obtained by converting a pixel in the original image, R ( x, y ) is the red component of a pixel in the original image, G ( x, y ) is the is the green component of a pixel in the original image, and B ( x, y ) is the blue component of a pixel in the original image.

The edge refers to the set of pixels whose surrounding pixels change sharply, and it is the basic feature of the image. The edge exists between the target, the background and the area, which is a sign of the position and is not sensitive to the change of grayscale. The method of edge detection on the grayscale image of the head model first, and then the feature point extraction can be achieved to reduce the feature. The effect of matching the amount of calculation, improving processing efficiency and accuracy. Optionally, use the Sobel operator edge detection method to determine the edge of the grayscale image, and then obtain the target image.

S120. Determine a head reference plane and a reference line based on the target image.

Optionally, the determining the head reference plane and the reference line based on the target image includes:

According to the geometric features and prior knowledge of the preset head model, determine the first feature point of the set position of the preset head model based on the target image; determine the head according to the first feature point Datum planes and datum lines.

Optionally, according to the intuitive geometric features and prior knowledge, the feature points of the key positions can be preliminarily marked by manual methods, or based on the preset features, the feature points of the key positions can be determined through feature matching, and then placed on the dummy head. Determine the head datum and datum line.

Wherein, the head reference plane includes sagittal plane, coronal plane, midsagittal plane, horizontal plane and Frankfurt plane. Referring to a schematic diagram of a head reference plane as shown in FIG. 3, the sagittal plane 310 refers to a longitudinal section that divides the head into left and right parts according to the direction of the sagittal axis, which is perpendicular to the horizontal plane and the coronal plane. For the midsagittal plane, the head is divided into left and right parts. The coronal plane 320 refers to a longitudinal section that divides the head into anterior and posterior parts according to the direction of the coronal axis, which is perpendicular to the horizontal plane and the sagittal plane. The horizontal plane 330 refers to a transverse plane that is perpendicular to both the coronal plane and the sagittal plane, dividing the head into upper and lower parts. The Frankfurt plane 340 refers to a plane passing through the left and right infraorbital points and parallel to the head manufacturing datum.

A schematic diagram of a part of the head reference line shown in FIG. 4, the reference line includes at least a sagittal axis 410, a coronal axis 420 and a vertical axis 430, wherein the sagittal axis 410 is from front to back and the vertical axis and the coronal axis The coronal axis 420 is the axis perpendicular to the sagittal and coronal axes from left to right; the vertical axis 430 is the axis perpendicular to the sagittal and coronal axes and perpendicular to the horizontal plane.

As shown in FIG. 5 , the head reference line may also include: midline 510, line between outer canthus 520, line 530 under the nose, vertical line 540 between inner canthus, line between the corners of the mouth (oral cleft line) 550 and chin Dotted line 560 below. Wherein, the median line 510 is a straight line formed by the point between the eyebrows-the point under the nose-the point under the chin; the line 520 between the outer corner points is the straight line formed by the outer corner point of the left eye-the outer corner point of the right eye; the subnasal point line 530 is a line passing under the nose The line 520 between the point and the outer corner point is parallel to the midline 510; the inner corner point vertical line 540 starts from the inner corner of the left eye and the inner corner of the right eye, and is parallel to the midline 510 and between the outer corner of the eye. Line 520 is a vertical line; the line between the corners of the mouth (orifice fissure line) 550 is a line connecting the corner points on both sides, and is perpendicular to the midline 510 and parallel to the line 520 between the outer corners of the eye; the submental line 560 is a line through the A line that is perpendicular to the midline 510 and parallel to the line 550 between the points of the mouth.

S130. Combine the head reference plane and the reference line, and divide the face area according to the proportional relationship of three stops and five parts, so as to obtain a plurality of face sub-regions.

By dividing the face into regions, it is beneficial to better determine the feature points that can better express the facial features.

Referring to the schematic diagram of a head and face area division shown in FIG. 6, the three stops refer to the upper stop area 610 between the mid-frontal point and the point between the eyebrows, and the stop area between the point between the eyebrows and the point below the nose. 620 and the stop area 630 between the subnasal and submental points. The five parts refer to the four parts 640 between the left cranial point and the protruding point of the left eyebrow arch, and the five parts 650 between the right cranial point and the protruding point of the right eyebrow arch, and between the protruding point of the left eyebrow arch and the inner corner of the left eye There are two parts 660, three parts 670 between the protruding point of the right eyebrow arch and the inner corner of the right eye, and one part 680 between the inner corner of the left eye and the inner corner of the right eye.

S140. Determine initial feature points from the multiple face sub-regions according to preset selection conditions.

Specifically, referring to the "Anthropometric Method" and the simplified head model, the feature points that meet the following requirements are determined as the initial feature points: in key positions, such as the facial features, and in places with large curvature, such as the corners of the eyes, etc. There should be calibration; the feature points should be distributed as evenly as possible; the density of the feature point distribution should be appropriate, if the density is too small, it will affect the accuracy of the model, and if the density is too large, it will affect the measurement efficiency; there should be a one-to-one correspondence between the corresponding feature points in different images, that is, Corresponding feature points in different samples should be marked with the same image features; covering the main measurement points specified in the "Anthropometric Method" and the key measurement points in the main measurement items.

The initial feature points determined from the plurality of face sub-regions according to preset selection conditions include the following points:

Head vertex P1, forehead midpoint P2, front chimney point P3, origin point/hairline point P4, eyebrow point P5, upper eyebrow point P6, nose root point P7, nasal dorsum point/nasal concave point P8, middle nose Point P9, nose tip point P10, nose point P11, alar point P12, gum point P13, upper lip point/upper lip midpoint P14, mouth cleft point P15, lower lip point/lower lip midpoint P16, mouth corner point P17, upper chin point P18, premental point P19, submental point P20, pupil point P21, protuberance point P22, inner corner point P23, outer corner point P24, infraorbital point P25, supraorbital point P26, tragus point P27, supraauricular attachment point / supra-auricular point/ supra-ear base point P28, below-ear attachment point/infra-auricular point/ below-ear base point P29, pre-auricular point P30, post-ear point P31, supra-ear point P32, sub-ear point P33, ear knot point P34, Frontotemporal point/temporal crest point P35, zygomatic point P36, mandibular angle point P37, mastoid point P38, cranial point/cranial lateral point P39, posterior occipital point/posterior point P40, occipital apex P41 and external occipital eminence point P42.

S150. Screen the initial feature points in combination with at least the missing facial parts of the preset head model to obtain multiple feature point sets, wherein different feature point sets are used to describe the features of different parts of the head and face.

Optionally, according to the "Anthropometric Method", combined with the simplified features of the Hybrid III dummy head model, remove the feature points at the missing parts of the Hybrid III dummy head model. Among them, the simplified characteristics of the Hybrid III dummy head model are shown in Table 1 below.

Table 1 Simplified features of Hybrid III dummy head model

Because the Hybrid III dummy head model lacks the coronal suture and sagittal suture of the skull, the bony chimney point P3 is eliminated according to feature 6 shown in Table 1 above. Because the head model of the Hybrid III dummy lacks the forehead hairline, the hairline point P4 is eliminated according to feature 3 shown in Table 1 above. Because the upper edge of the left and right eyebrows is missing based on the Hybrid III dummy head model, the upper point P6 between the eyebrows is eliminated according to feature 2 shown in Table 1 above. Since the transition point between the nasal dorsum and the forehead cannot be determined based on the Hybrid III dummy head model, the dorsum of the nose point P8 is excluded according to feature 5 shown in Table 1 above. Due to the lack of gingiva based on the Hybrid III dummy head model, the gingival point P13 was removed according to feature 4 shown in Table 1 above. Due to the lack of eyes based on the Hybrid III dummy head model, the pupil point P21 and the eye protrusion point P22 are eliminated according to feature 2 shown in Table 1 above. Due to the lack of ears based on the Hybrid III dummy head model, the tragus point P27, the attachment point P28 above the ear, the attachment point P29 under the ear, the preauricular point P30, the postauricular point P31, the The upper point P32, the lower ear point P33, and the ear knot node P34. Due to the lack of temporal ridge based on the Hybrid III dummy head model, the frontotemporal point/temporal ridge point P35 was excluded according to feature 6 shown in Table 1 above. Due to the absence of the mastoid based on the Hybrid III dummy head model, the mastoid point P38 was removed according to feature 6 shown in Table 1 above. Because the head model based on the Hybrid III dummy lacks the external occipital protuberance, the posterior head vertex P41 and the external occipital protuberance point P42 are removed according to feature 6 shown in Table 1 above.

The first feature point set is obtained after the first round of screening of the initial feature points by combining the missing facial parts and their characteristics of the Hybrid III dummy head model, and the feature points included in the first feature point set are as follows in Table 2 Feature points C1~C23 shown.

Table 2 The first feature point set of the dummy head model

According to the definition of the feature points, the feature points of the dummy head model are roughly marked by manual methods, refer to the schematic diagrams of the feature points of the dummy head model shown in Figure 7, Figure 8, Figure 9 and Figure 10.

Further, the feature points in the first feature point set are continuously screened according to the head contour of the dummy head model, and the feature points describing the head contour are counted into the second feature point set. Exemplarily, the statistics of the feature points describing the outline of the head to the second feature point set include:

Generate a scale space based on the Hessian matrix according to the pixel values of the feature points in the first feature point set, and determine the local extremum points in the scale space; determine candidate feature points according to the local extremum points based on a preset strategy; For the candidate feature points, the candidate feature points in the required continuous scale space are determined by interpolation; if the distance between the candidate feature points and the interpolation center point exceeds the threshold, the candidate feature points are removed from the A feature point set is eliminated; the remaining feature points in the first feature point set are counted into a second feature point set.

Specifically, combined with the simplified dummy head model, the Speeded UP Robust Features (SURF) algorithm is used to continue the feature point screening. The algorithm is fast and robust. It can be seen from Table 2 above that the screening area only needs to be Near the midline of one, at the edges of four and five.

，即任何一个特征点均可定义为一个二维函数

，其中x和y是特征点对应的空间（平面）坐标，而在任何一对空间坐标（x，y）处的函数值

一般称为该特征点的像素值。海塞矩阵H是由

和其偏导数共同构成。特征点的海塞矩阵如下式（2）所示：First, the scale space is generated using the Hessian matrix. The scale of the image refers to the thickness of the image content, and the concept of scale is used to simulate the distance between the observer and the object. The scale space of an image is a collection of blurred pictures formed by an image after passing through several different Gaussian kernels, which is used to simulate the distance and blur of objects seen by the human eye. The SURF algorithm selects the approximate value of the determinant eigenvalues of the Hessian matrix to generate the scale space, and the eigenpoint function is

, that is, any feature point can be defined as a two-dimensional function

, where x and y are the spatial (plane) coordinates corresponding to the feature points, and the function value at any pair of spatial coordinates (x, y)

Generally referred to as the pixel value of the feature point. The Hessian matrix H is given by

together with its partial derivatives. The Hessian matrix of the feature points is shown in the following formula (2):

（2）

(2)

Each feature point can get its own H matrix, and the discriminant of the H matrix is shown in the following formula (3):

（3）

(3)

The result obtained by the above formula (3) represents the eigenvalues of the Hessian matrix. All the feature points are divided according to the positive and negative of the result obtained by the above formula (3), and then it is determined whether the corresponding feature point is in the scale space. the local extreme point.

＞0，且det（H）＞0，则矩阵为正定矩阵。Specifically, when the Hessian matrix of the feature point is a positive definite matrix, the feature point is a local minimum point. Among them, the positive definite matrix discrimination method is: if the main subform of each order is greater than 0, that is

> 0, and det(H) > 0, the matrix is a positive definite matrix.

＜0，且det（H）＞0，则矩阵为负定矩阵。When the Hessian matrix of a feature point is a negative definite matrix, the feature point is a local maximum point. Among them, the negative definite matrix discrimination method is: the even-order main subform is positive, and the odd-order main subform is negative, that is,

<0, and det(H)>0, the matrix is negative definite.

It can be understood that the purpose of generating the scale space is to extract local extreme points (ie, maximum and minimum values) from all the feature points in the first feature point set, and mark the extracted local extreme points as candidates. point.

）做对比。该26 个候选点为提取的关键点（该关键点是指所述候选点中的任意一个）和其所在的层级邻域内其它8个候选点以及与该关键点所在层级相邻的上下两个层级内的共18个候选点（相邻的上层级内的9个候选点和相邻的下层级内的9个候选点，如图11所示），将这些区域内（即图11所示的27个候选点）像素值表现为最大值和最小值的候选点筛选出来，并将其确定为候选特征点。Next, find the best extreme point based on the generated scale space. Using a filter of size 3 × 3, the pixel values of all the candidate points approximated by the Hessian matrix determinant are compared with the pixel values of the 26 candidate points in the same scale space neighborhood and the corresponding neighborhood in the adjacent scale space ( which is

) for comparison. The 26 candidate points are the extracted key points (the key point refers to any one of the candidate points), the other 8 candidate points in the neighborhood of the level where it is located, and the upper and lower two adjacent to the level where the key point is located. There are a total of 18 candidate points in the level (9 candidate points in the adjacent upper level and 9 candidate points in the adjacent lower level, as shown in Figure 11). The 27 candidate points) whose pixel values represent the maximum and minimum candidate points are screened out and determined as candidate feature points.

Finally, the final feature points are determined based on the determined candidate feature points. Use the interpolation method to filter out the feature points in the actual required continuous scale space, and obtain the quadratic equation of the approximate value D(x) according to the Taylor series operation formula, as shown in the following formula (4):

（4）

(4)

得出极值，如下式（5）所示。Among them, D(x) represents the approximate value of the extreme value of the candidate feature point in the continuous scale space, x is the coordinate of the candidate feature point in the scale space, and the difference operation of the extreme value of each corresponding candidate feature point in the two adjacent scale spaces can be calculated. Get the various coefficients of D(x). Derivative with respect to D(x), let

The extreme value is obtained, as shown in the following formula (5).

（5）

(5)

表示尺度空间内候选特征点与插值中心点相偏移程度，若

的值过大（比如大于阈值），代表着极值点已经偏离候选特征点所在尺度空间内的轨迹，这种情况应该将对应候选特征点删除。通过此种多次迭代的方式筛选出特征点的精准定位及尺度，校核并调整第一特征点集中的特征点，从一部正中线附近、四部和五部边缘处将能描述头部轮廓的特征点存入第二特征点集。示例性的，参考如表3示出的第二特征点集所包括的特征点。同时，参考图12和图13所示的假人头部轮廓特征点的示意图。In the formula,

Indicates the degree of offset between the candidate feature points in the scale space and the interpolation center point, if

The value of is too large (for example, greater than the threshold), which means that the extreme point has deviates from the trajectory in the scale space where the candidate feature point is located. In this case, the corresponding candidate feature point should be deleted. Through this method of multiple iterations, the precise positioning and scale of the feature points are screened, and the feature points in the first feature point set are checked and adjusted. The feature points of are stored in the second feature point set. Exemplarily, refer to the feature points included in the second feature point set as shown in Table 3. Meanwhile, refer to the schematic diagrams of the feature points of the dummy head profile shown in FIG. 12 and FIG. 13 .

Table 3 The second feature point set

Further, continue to screen the feature points in the first feature point set according to the facial features, count the feature points describing the eye features to the third feature point set, and count the feature points describing the nose feature to the fourth feature point set. , the feature points describing the features of the mouth are counted into the fifth feature point set, the feature points describing the features of the eyebrows are counted into the sixth feature point set, and the feature points describing the features of the ear are counted into the seventh feature point set.

Specifically, combined with the simplified head model, the SURF algorithm is used to continue screening, and the feature points that can describe the main features of the eye are stored in the third feature point set from the second, third, fourth, and fifth parts of the pause; The feature points that can describe the main features of the nose are stored in the fourth feature point set from the first, second, and third parts; the feature points that can describe the main features of the mouth are stored in the fifth set from the first, second, and third parts. The feature point set; the feature points that can describe the main features of the eyebrows are stored in the sixth feature point set; the feature points that can describe the main features of the ear are stored in the seventh feature point set. Among them, the feature points that describe the main features refer to the feature points that can determine the position and scale of the facial features.

Among them, the feature point reference in the third feature point set is shown in Table 4 below, the feature point reference in the fourth feature point set is shown in Table 5 below, and the feature point reference in the fifth feature point set is shown in Table 6 below.

Table 4: The third feature point set

Table 5: Fourth Feature Point Set

Table 6: Fifth feature point set

In combination with the above Table 4, Table 5 and Table 6, refer to the schematic diagram of the facial features of a dummy head as shown in FIG. 14 .

In general, the initial feature points are screened at least in combination with the missing facial parts of the preset head model to obtain multiple feature point sets, including:

Eliminating the feature points belonging to the facial parts missing from the preset head model in the initial feature points, and counting the remaining initial feature points to the first feature point set;

According to the preset head model, the feature points in the first feature point set are continuously screened by the accelerated robust feature SURF algorithm, so as to count the feature points describing the outline of the head to the second feature point set, which will describe the eye contour. The initial feature points of the facial features are counted to the third feature point set, the initial feature points describing the nose features are counted to the fourth feature point set, and the initial feature points describing the mouth features are counted to the fifth feature point set. The initial feature points of the eyebrow feature are counted to the sixth feature point set, and the initial feature points describing the ear feature are counted to the seventh feature point set. Optionally, you can also refer to a schematic diagram of a feature point screening process as shown in Figure 15, which specifically includes: selecting a head feature point-determining whether a specific feature point meets the measurement requirements, if not, then the feature point. If it is satisfied, it will be further judged whether the feature point is located in the missing part. If so, the feature point will be eliminated. If not, the feature point will be stored in the first feature point set (that is, the feature point shown in Figure 15). On the other hand, it is further judged whether the feature point can describe the outline of the head. If so, the feature point is stored in the second feature point set (that is, the feature point set 2 shown in Figure 15). If not, Then continue to judge whether the feature point can describe the facial features, if not, remove the feature point, and if so, store the feature point in the third feature point set (ie, feature point set 3 shown in Figure 15).

In summary, the filtered feature points are classified into different feature point sets. The first feature point set (C1~C23) can describe the geometric features of facial features, location and contour features of the head and face. The second feature point set (L1~L11) can describe the main contour features of the head. The third feature point set (E1~E4) can describe the main features of the eye. The fourth feature point set (N1~N4) can describe the main features of the nose. The fifth feature point set 5 (M1~M3) can describe the main features of the mouth. The sixth feature point set (∅) can indicate that the eyebrow features are simplified; the seventh feature point set (∅) can indicate that the ear features are all simplified.

Further, the method also includes:

According to the target requirement, a union operation is performed on part of the feature point sets in the plurality of feature point sets to obtain a target feature point set that meets the target requirement. Correspondingly, refer to a schematic flowchart of a feature point extraction as shown in FIG. 16 , which specifically includes: input requirements - selecting feature point sets - intersection and union processing - extracting feature points.

That is, according to the measurement requirements, the feature point sets that can describe different head and face features are obtained by merging the sets. For example, if you want to measure the contour features of the head and the main features of the nose, take the union of the second feature point set and the fourth feature point set to get the selected head feature points. Measure the characteristic dimensions of head length, nose length, head width, nose width, head height, nose height, head coronal circumference, etc., so as to obtain the head contour features and main features of the nose, and then prove that the bionic dummy head features conform to the corresponding race facial features. Correspondingly, refer to a schematic diagram of screening of head contour features and main features of the nose as shown in FIG. 17 .

This embodiment has the following technical effects:

The representative feature points can be quickly and simply screened out from the numerous feature points on the head of the collision dummy, avoiding excessive description, and the screened feature points can reflect the main head and face information and describe the facial features. The principle of the method is clear and easy to understand, the operation method is relatively simple, no expensive equipment is required, the time period is short, the engineering cost is low, and the head feature points required for the measurement can be quickly obtained. The dummy head model provides guidance.

In general, refer to the schematic diagram of a technical route for feature point screening as shown in Figure 18, which specifically includes the following general steps: simplified head model - image preprocessing - determination of head reference plane - drawing head reference Line - divide head and face area - input missing parts - filter head feature points - extract feature points on demand - get head feature points. Specifically, by studying the simplified head model, determining its head reference plane and reference line, studying its face division method, screening head feature points according to the "Anthropometric Method" and the facial features of the dummy head, and classifying them into different features point set, and then merge the sets according to the requirements to obtain the feature point set that can describe the geometric features of the head, so as to use the simplest feature point set to describe the main features of the head model, and finally obtain a complete set of head and face feature point screening methods, which are The simplified design of the bionic head model provides reference.

Image preprocessing: Preprocess images by means of grayscale, edge detection, etc. to ensure high-quality and efficient identification, reduce the time spent on subsequent feature point screening, and improve the accuracy of image feature extraction.

Determining the head datum plane: Determine the sagittal, coronal, midsagittal, horizontal and Frankfurt planes on the simplified head model.

Draw the reference line of the head: Determined by a green light level, and draw the reference line: sagittal axis, coronal axis, vertical axis. Midline: point between the eyebrows - point under the nose - point under the chin; line between the outer canthus points: point between the outer corner of the eye - point of the outer corner of the eye; point line under the nose: parallel to the line between the points under the nose and the outer corner of the eye, and perpendicular to the midline ;Inner corner point vertical line: It starts from the left and right inner corner points, parallel to the midline, and perpendicular to the line between the outer corner points. The line between the corners of the eyes is parallel; the submental line: This line is perpendicular to the midline through the submental point and parallel to the corner of the mouth.

Divide the facial area: There are commonalities in the positions of the facial features of different people. Generally speaking, they generally conform to the proportional relationship in the "Three Stops and Five Parts". "Three stops and five parts" is the general standard ratio of a person's face length to face width. The simple three stops and five parts divide the head into three parts up and down, and divide the head into five parts from left to right. The so-called "three stops" is to describe the length relationship of the face, and the distance from the chin to the hairline is divided into three equal parts. The hairline from the chin to the bottom of the nose, from the bottom of the nose to the eyebrows, and from the eyebrows to the center of the forehead is a total of "three stops". The "five parts" describe the width relationship of the human face. It divides the distance between the left and right hairlines into 5 equal parts: between the two eyes, the left and right eyes, the outer corners of the left and right eyes to the hairline on the same side. as a "department". The rule of "three stops and five parts" has great reference value for the face region division of the simplified head model, and this embodiment uses similar rules to divide the face region. Since the positional relationship of the facial features of some head models is somewhat different from that of "Three Stops and Five Parts", but generally not too extreme, this is also considered in this embodiment. When locating the facial features of the head model, relative to " Three stops and five departments” feature, the location range has been adjusted. According to the following method, the face is divided into regions to better describe the facial features.

Three stops: the midpoint of the forehead - the point between the eyebrows - the point under the nose - the point under the chin.

Five parts: head lateral point - eyebrow arch protruding point - inner canthus point - inner canthus point - eyebrow arch protruding point - head lateral point.

Screening of head feature points: At present, the determination of domestic human head and face measurement points is mainly based on the 42 feature points described in "Anthropometric Methods" and GB/T 38131-2019 "Method for Obtaining Anthropometric Reference Points for Clothing" The 2 feature points described are simplified and analyzed, the feature points are extracted and the related measurement items are determined. "Methods for Obtaining Human Body Measurement Reference Points for Clothing" involves too few feature points to describe the geometric structure and contour size of the collision dummy's head. "Anthropometric Methods" involves too many feature points and complicated operations. The calculation is cumbersome, and some features of the head and face of the collision dummy are missing, and it is not necessary to cover all the feature points, so the feature points should be screened.

(1) Screening according to the measurement requirements. The screening of feature points follows the following principles: select key feature points, such as facial features, and also have calibration in places with large curvature, such as the corners of the eyes; the key feature points should be distributed as evenly as possible; The density of the distribution of feature points should be appropriate. If the density is too small, the accuracy of the model will be affected. If the density is too large, the measurement efficiency will be affected. There should be a one-to-one correspondence between the corresponding feature points in different images, that is, the corresponding feature points in different samples should be Mark the same image features; cover the main measurement points specified in the "Anthropometric Method" and the key measurement points in the main measurement items.

(2) Screening according to the missing parts, according to the "Anthropometric Method", combined with the simplified head model, screen the cephalometric feature points, remove the feature points at each missing part, and store the screened feature points in the first A set of feature points.

(3) Screening according to the head profile, combined with the simplified head model, continue screening, and store the feature points that can describe the head profile into the second feature point set.

(4) Screening according to the facial features, combined with the simplified head model, continue screening, and store the feature points that can describe the main features of the eyes into the third feature point set, and store the feature points that can describe the main features of the nose into the In the fourth feature point set, the feature points that can describe the main features of the mouth are stored in the fifth feature point set, the feature points that can describe the main features of the eyebrows are stored in the sixth feature point set, and the feature points that can describe the main features of the ear are stored in the sixth feature point set. The feature points are stored in the seventh feature point set.

Finally, according to the measurement requirements, merge the sets to obtain feature point sets that can describe different head and face features.

The feature point screening method provided by the embodiment of the present disclosure divides the region by three stops and five parts, and uses simplified features, head contour features, and facial features geometric features as the judgment basis; for the new requirement of simplified head model feature point screening, The new method is used to extract the head and face information, which has lower complexity and lower engineering cost than the existing technology; through this method, various feature points can be screened, and the head and face features can be assigned to the bionic dummy head model through the feature points, which is Provide guidance for designing a bionic dummy head model whose appearance and size are in line with Chinese characteristics.

FIG. 19 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. As shown in FIG. 19 , electronic device 400 includes one or more processors 401 and memory 402 .

Processor 401 may be a central processing unit (CPU) or other form of processing unit having data processing capabilities and/or instruction execution capabilities, and may control other components in electronic device 400 to perform desired functions.

Memory 402 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and/or cache memory (cache). The non-volatile memory may include, for example, a read only memory (ROM), a hard disk, a flash memory, and the like. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 401 may execute the program instructions to implement the feature points for the head model in any of the embodiments of the present invention described above Screening methods and/or other desired functions. Various contents such as initial extrinsic parameters, thresholds and the like may also be stored in the computer-readable storage medium.

In one example, the electronic device 400 may also include an input device 403 and an output device 404 interconnected by a bus system and/or other form of connection mechanism (not shown). The input device 403 may include, for example, a keyboard, a mouse, and the like. The output device 404 can output various information to the outside, including early warning information, braking force, and the like. The output devices 404 may include, for example, displays, speakers, printers, and communication networks and their connected remote output devices, among others.

Of course, for simplicity, only some of the components in the electronic device 400 related to the present invention are shown in FIG. 19 , and components such as buses, input/output interfaces, and the like are omitted. Besides, the electronic device 400 may also include any other appropriate components according to the specific application.

In addition to the methods and apparatuses described above, embodiments of the present invention may also be computer program products comprising computer program instructions that, when executed by a processor, cause the processor to perform the functions provided by any of the embodiments of the present invention. Steps of the feature point screening method for the head model.

The computer program product may be written in any combination of one or more programming languages, including object-oriented programming languages, such as Java, C++, etc. , also includes conventional procedural programming languages, such as "C" language or similar programming languages. The program code may execute entirely on the user computing device, partly on the user device, as a stand-alone software package, partly on the user computing device and partly on a remote computing device, or entirely on the remote computing device or server execute on.

In addition, an embodiment of the present invention may also be a computer-readable storage medium on which computer program instructions are stored, the computer program instructions, when executed by a processor, cause the processor to execute the functions provided by any embodiment of the present invention. The steps of the feature point screening method for the head model.

The computer-readable storage medium may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may include, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses or devices, or a combination of any of the above. More specific examples (non-exhaustive list) of readable storage media include: electrical connections with one or more wires, portable disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disk read only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

Claims (9)

1. a feature point screening method for head model, is characterized in that, comprises: Perform preprocessing on an original image of a preset head model to obtain a target image, wherein the preset head model is used for a car crash dummy; determining a head reference plane and a reference line based on the target image; Combining the head datum plane and the datum line, the facial area is divided according to the proportional relationship of three stops and five parts, and a plurality of facial sub-areas are obtained; Determine initial feature points from the plurality of face sub-regions according to preset selection conditions; The initial feature points are screened at least in combination with the missing facial parts of the preset head model to obtain multiple feature point sets, wherein the different feature point sets are used to describe the features of different parts of the head and face. 2. The method according to claim 1, wherein the preprocessing is performed on the original image of the preset head model to obtain the target image, comprising: Performing grayscale processing on the original image to obtain a grayscale image corresponding to the original image; Perform edge detection on the grayscale image to obtain the target image. 3. The method according to claim 1, wherein the determining a head reference plane and a reference line based on the target image comprises: According to the geometric features and prior knowledge of the preset head model, based on the target image, determine the first feature point of the set position of the preset head model; The head reference plane and the reference line are determined according to the first feature point. 4. The method according to claim 3, wherein the head reference plane includes sagittal plane, coronal plane, midsagittal plane, horizontal plane and Frankfurt plane, and the reference line at least includes sagittal axis, coronal plane axis and vertical axis. 5. The method according to claim 1, wherein the initial feature points are screened at least in combination with the missing facial parts of the preset head model to obtain multiple feature point sets, comprising: Eliminating the feature points belonging to the facial parts missing from the preset head model in the initial feature points, and counting the remaining initial feature points to the first feature point set; According to the preset head model, the feature points in the first feature point set are continuously screened by the accelerated robust feature SURF algorithm, so as to count the feature points describing the outline of the head to the second feature point set, which will describe the eye contour. The initial feature points of the facial features are counted to the third feature point set, the initial feature points describing the nose features are counted to the fourth feature point set, and the initial feature points describing the mouth features are counted to the fifth feature point set. The initial feature points of the eyebrow feature are counted to the sixth feature point set, and the initial feature points describing the ear feature are counted to the seventh feature point set. 6 . The method according to claim 5 , wherein the statistics of the feature points describing the outline of the head to the second feature point set comprises: 6 . Generate a scale space based on the Hessian matrix according to the pixel values of the feature points in the first feature point set, and determine the local extreme point in the scale space; Determine candidate feature points according to the local extreme points based on a preset strategy; According to the candidate feature points, the candidate feature points in the required continuous scale space are determined by interpolation; If the distance between the candidate feature point and the interpolation center point exceeds a threshold value, remove the candidate feature point from the first feature point set; The remaining feature points in the first feature point set are counted into the second feature point set. 7. The method according to any one of claims 1-6, further comprising: According to the target requirement, a union operation is performed on part of the feature point sets in the plurality of feature point sets to obtain a target feature point set that meets the target requirement. 8. An electronic device, characterized in that the electronic device comprises: processor and memory; The processor is configured to execute the steps of the feature point screening method for a head model according to any one of claims 1 to 7 by invoking a program or an instruction stored in the memory. 9. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a program or an instruction, and the program or instruction causes a computer to execute the method for the head according to any one of claims 1 to 7 The steps of the feature point screening method of the model.

Images