CN104809738A

CN104809738A - Airbag overall dimension detection method based on binocular vision

Info

Publication number: CN104809738A
Application number: CN201510250862.4A
Authority: CN
Inventors: 王晓东; 董博; 孙锐; 张炜
Original assignee: Changchun University of Technology
Current assignee: Changchun University of Technology
Priority date: 2015-05-18
Filing date: 2015-05-18
Publication date: 2015-07-29
Anticipated expiration: 2035-05-18
Also published as: CN104809738B

Abstract

The invention relates to an airbag overall dimension detection method based on binocular vision. The airbag overall dimension detection method comprises the steps of: (1) calibrating a binocular vision system by using a calibration board; (2) simultaneously shooting a to-be-detected airbag by using a binocular camera, preprocessing two images, then extracting edges, thinning into single pixel, and filling breakpoints; (3) segmenting and matching a contour curve of the to-be-detected airbag; (4) reconstructing a three-dimensional curve segment, and optimizing the curve segment; (5) performing coordinate transformation on the three-dimensional curve segment of the to-be-detected airbag, and matching the three-dimensional curve segment of the to-be-detected airbag with a three-dimensional digital-analog curve segment of the edge contour of a standard airbag; (6) calculating position tolerances Vx,Vy and Vz of detection points and a form tolerance K of the contour curve segment in which the points are positioned, so as to respectively determine whether the detection points are in the ranges of the position tolerances and the form tolerance. Traditional airbag overall dimension detection methods including a three-coordinate detection method and a test tool detection method can be simultaneously replaced by the airbag overall dimension detection method, the cost can be reduced, the efficiency can be improved, and the airbag overall dimension detection method has the advantages of automation, non contact, high precision, high adaptability and the like and can be used for effectively detecting the overall dimensions of various airbags.

Description

Airbag contour size detection method based on binocular vision

Technical Field

The invention belongs to the technical field of computer vision and image measurement, and particularly relates to a binocular vision safety airbag contour dimension detection method.

Background

Along with the high-speed development of scientific technology, the assembly process requirements of the safety air bag are more and more rigorous, but the safety air bag detection technology at the present stage is not perfect, if the detection quality is not over-limit, the attractiveness and the service performance are influenced, more importantly, safety hazards are brought inevitably, the existing safety air bag assembly detection method has a three-coordinate detection method and a detection tool detection method, has the defects of high detection cost, poor detection result repeatability, high detection error rate, low efficiency and the like, is not suitable for the online full detection of a production line, and along with the continuous development of image processing, computer technology and industrial camera manufacturing levels, the computer vision technology is also rapidly developed, the shape, position and size of an object in a three-dimensional space can be measured, and compared with a contact detection system, the intelligent detection system, the flexibility and the detection speed convenience are more superior, will gradually become an important means for the on-line detection of the size of industrial products and the future development trend.

Disclosure of Invention

The invention provides a method for detecting the outline dimension of an airbag based on binocular vision, which combines the binocular vision with the outline dimension detection, designs a binocular vision measurement system according to the characteristics of outline dimension parameters, utilizes the technologies of image acquisition and preprocessing, curve segmentation and stereo matching, three-dimensional reconstruction, coordinate transformation, registration and the like to realize the automatic measurement of the outline dimension parameters of the airbag, solves the defects of high detection cost, poor repeatability of detection results, low efficiency and the like, has the characteristics of automation, non-contact, high precision and universality, and can effectively judge whether the airbag is qualified or not.

In order to achieve the purpose, the invention adopts the following technical scheme that the binocular vision safety airbag contour size detection method comprises the following steps:

(1) calibrating a binocular stereoscopic vision system: respectively calibrating the two cameras by using an MATLAB calibration tool box to obtain internal parameters and external parameters of each camera, and then performing three-dimensional calibration on the calibrated cameras by using OpenCV in VS2010 to obtain a rotation matrix R and a translation matrix T of the position relationship between the two cameras;

(2) image acquisition and preprocessing: two cameras are used for shooting the safety air bag to be detected simultaneously to obtain a left image I and a right image I₁、I₂Carrying out image preprocessing such as Gaussian filtering, contrast enhancement, Canny operator edge detection and the like on a visual image of the safety airbag to obtain a continuous and closed safety airbag edge profile two-dimensional curve;

(3) segmentation and three-dimensional matching of the contour curve of the safety airbag: the method comprises the steps of segmenting an air bag contour curve, roughly matching characteristic points by utilizing a normalized cross correlation coefficient and a brightness mean square error method, and finally optimally matching curve segments between matched point pairs by utilizing a dynamic programming algorithm and taking boundary potential energy as a measurement standard.

(4) Three-dimensional reconstruction of airbag contour curve: determining the intersection line of the space curved surface where the matching curve section is located according to the optical center of the camera by using the internal and external parameters, the translation matrix and the rotation matrix obtained in the step (1) and the matching curve section obtained in the step (3), and then determining the space curve section corresponding to the curve section on the image according to the end point of the curve section on the image and the distance from the point of any mark point on the curve in the three-dimensional space to the candidate space curve.

(5) And (3) three-dimensional coordinate transformation and registration of the outline of the air bag to be detected and the outline of the standard air bag: extracting three-dimensional digital-analog edge information of a CATIA (computer aided three-dimensional Interactive application) of a standard safety airbag, performing three-dimensional matching on a space curve to determine common points of the standard safety airbag and a safety airbag to be detected, performing space coordinate transformation by adopting a seven-parameter model of Boolean Sha, fitting corresponding seven parameters by using a least square method according to the common point coordinates of the standard safety airbag and the safety airbag to be detected, converting the space coordinates of the safety airbag to be detected into the space coordinates of the standard safety airbag by using the obtained seven parameters, and further completing the three-dimensional coordinate transformation and registration of the outline of the safety airbag to be detected and the standard airbag.

(6) Calculating the position tolerance V of the detection points_x、V_y、V_zAnd the shape tolerance K of the contour curve section where the point is located, and whether the point is within the range of the position tolerance and the shape tolerance is judged, so that whether the safety airbag to be detected is qualified is judged.

Compared with the prior art, the invention has the beneficial effects that: the method can automatically complete the calculation of the contour dimension parameters by only utilizing the images of the to-be-detected safety air bag provided by the CCD camera and the three-dimensional digital analogy of the standard safety air bag, judge whether the to-be-detected safety air bag is qualified, can simultaneously replace two devices of three-coordinate detection and detection of the detection tool, can reduce the cost and improve the efficiency, and has the advantages of automation, non-contact, high precision, strong adaptability and the like.

Drawings

FIG. 1 is a flow chart of the binocular vision based airbag contour dimension detection method of the present invention;

FIG. 2 is a flow chart of image acquisition and pre-processing of the present invention;

FIG. 3 is a schematic representation of a three-dimensional reconstruction of a curve segment of the present invention;

Detailed Description

Embodiments of the present invention will now be described in more detail with reference to the accompanying drawings.

FIG. 1 is a flow chart of the binocular vision-based airbag contour dimension detection method of the present invention, which comprises the following steps:

(1) and (5) calibrating the binocular stereoscopic vision system. The calibration of the cameras is to obtain internal and external parameters of the cameras, perform single-camera calibration on the two cameras respectively by using an MATLAB calibration tool box to obtain the internal parameter and the external parameter of each camera, perform three-dimensional calibration on the calibrated cameras by using OpenCV in VS2010, and obtain a rotation matrix R and a translation matrix T of a position relationship between the two cameras.

(2) And (5) image acquisition and preprocessing. As shown in fig. 2, includes the following substeps:

(2.1) shooting the same scene by using two cameras simultaneously to obtain a left image I and a right image I₁、I₂；

(2.2) denoising the two images by adopting Gaussian filtering;

(2.3) carrying out contrast enhancement and gray level normalization on the two images to enhance the outline edge of the safety airbag;

(2.4) extracting the edge of the image by adopting a Canny operator;

(2.5) adopting a mathematical morphology method to refine the edge into a single pixel, carrying out boundary tracking on a breakpoint generated on the boundary in the edge refining process, filling the breakpoint on the boundary to obtain vectorization representation of an image, and further obtaining a continuous and closed airbag contour edge two-dimensional curve;

(3) the segmentation and the three-dimensional matching of the contour curve of the safety air bag comprise the following substeps:

and (3.1) segmenting the contour curve of the airbag. Let n sequence points describe the contour curve L of the airbag, i.e.:

L＝{p_i＝(x_i,y_i),i＝1,2,...,n}

wherein (x)_i,y_i) Is a point p on the boundary L_iCoordinate of (a), p_i+1Is p_iAn abutment point. S_k(p_i) As boundary L with p_iA small curve segment centered, namely:

S_k(p_i)＝{p_j|j＝i-k,i-k+1,...,i+k-1,i+k}

curve segment S_k(p_i) Point p on_iThe covariance matrix C of (a) is defined as follows:

C = [\begin{matrix} c_{11} & c_{12} \\ c_{21} & c_{22} \end{matrix}]

wherein,

c_xand c_yIs a curved line segment S_k(p_i) I.e.:

eigenvalues λ of the covariance matrix C_LAnd λ_SComprises the following steps:

the formula of the characteristic value shows lambda_L≥λ_S，

The product formula of the characteristic values of different support areas is as follows:

<math> <mrow> <mi>P</mi> <mrow> <mo>(</mo> <msub> <mi>p</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> <mo>=</mo> <munder> <mi>Π</mi> <mrow> <mi>s</mi> <mo>&Element;</mo> <mi>Ω</mi> </mrow> </munder> <msub> <mi>λ</mi> <mi>s</mi> </msub> <mrow> <mo>(</mo> <msub> <mi>p</mi> <mi>i</mi> </msub> <mo>,</mo> <mi>s</mi> <mo>)</mo> </mrow> </mrow> </math>

wherein Ω represents the product of the eigenvalues of the different support regionsThe characteristics of a point on the curve can be measured, and if the point is a distinct salient characteristic point, the characteristic value product P (P) of the store_i) The threshold value is larger and gives the product of the characteristic valuesIf the eigenvalue product P (P) of a point_i) Is greater thanThe point is a characteristic point, otherwise, the point is regarded as a general point on the curve, and the airbag contour curve can be segmented.

And (3.2) carrying out three-dimensional matching on the airbag contour curve segment.

Firstly, roughly matching feature points by utilizing a normalized cross correlation coefficient NCC and a brightness mean square error ASD method, and defining the potential energy of one point on a curve as CEPF (L):

wherein (x)_i,y_i) And (x)_j,y_j) The coordinates of the ith point on the curve segment L and the jth point in the range of L × k around the ith point are respectively I (x)_j,y_j) And I (x)_i,y_i) For their gray values, n is the number of points on the curve segment, and m is the number of pixel points within the range of l × k. For the left view and the right view I (x, y) and I '(x, y), L and L' are corresponding curve segments on the two views, respectively, and the matching metric formula is as follows:

<math> <mrow> <mi>ρ</mi> <mrow> <mo>(</mo> <mi>L</mi> <mo>,</mo> <msup> <mi>L</mi> <mo>′</mo> </msup> <mo>)</mo> </mrow> <mo>=</mo> <mo>|</mo> <mover> <mrow> <mi>CEPF</mi> <mrow> <mo>(</mo> <mi>L</mi> <mo>)</mo> </mrow> </mrow> <mo>&OverBar;</mo> </mover> <mo>-</mo> <mover> <mrow> <mi>CEPF</mi> <mrow> <mo>(</mo> <msup> <mi>L</mi> <mo>′</mo> </msup> <mo>)</mo> </mrow> </mrow> <mo>&OverBar;</mo> </mover> <mo>|</mo> </mrow> </math>

andis the average of the potential energies of the curve segments L and L', i.e.:

<math> <mrow> <mover> <mrow> <mi>CEPF</mi> <mrow> <mo>(</mo> <mi>L</mi> <mo>)</mo> </mrow> </mrow> <mo>&OverBar;</mo> </mover> <mo>=</mo> <mfrac> <mrow> <mi>CEPF</mi> <mrow> <mo>(</mo> <mi>L</mi> <mo>)</mo> </mrow> </mrow> <mi>n</mi> </mfrac> </mrow> </math>

the value of rho is larger than 0, the smaller the value is, the more similar the two curve segments are, therefore, rho can measure the similarity degree of the curve segments, and then the rho is used as a measurement criterion to optimally match the curve segments between the matching point pairs by using a dynamic programming algorithm.

(4) The three-dimensional reconstruction of the airbag contour curve, as shown in fig. 3, comprises the following substeps:

(4.1) fitting the matching curve segments in (3) respectively by using a least square method, and obtaining the image I₁、I₂Upper quadratic curve segment C₁、C₂Can be respectively expressed as:

X_{1}^{T} C_{1} X_{1} = 0

X_{2}^{T} C_{2} X_{2} = 0

wherein,

X₁＝(x₁,y₁,1)^T

X₂＝(x₂,y₂,1)^T

X₁、X₂are respectively an image I₁、I₂Upper position of quadratic curve C₁And C₂Homogeneous coordinates of the image of the upper arbitrary point, C₁And C₂A symmetric matrix of 33. From the projection equation:

s_{1} [\begin{matrix} x_{1} \\ y_{1} \\ 1 \end{matrix}] = [\begin{matrix} {\hat{p}}_{11} \\ {\hat{p}}_{12} \\ {\hat{p}}_{13} \end{matrix}] X

s_{2} [\begin{matrix} x_{2} \\ y_{2} \\ 1 \end{matrix}] = [\begin{matrix} {\hat{p}}_{21} \\ {\hat{p}}_{22} \\ {\hat{p}}_{23} \end{matrix}] X

wherein,andare respectively a phaseProjector matrixThe row vector of (2). By introducing projection equations intoObtaining:

X^{T} {\hat{P}}_{1}^{T} C_{1} {\hat{P}}_{1} X = 0

X^{T} {\hat{P}}_{2}^{T} C_{2} {\hat{P}}_{2} X = 0

let the optical centers of the left and right cameras be O₁、O₂Then this formula is the optical center O₁、O₂Corresponding curve segment C₁、C₂The intersection of the determined conical surfaces.

And (4.2) determining the end points of the curve segment in the (4.1) and the three-dimensional coordinates of any characteristic point pair on the corresponding curve segment according to the principle of triangulation and the internal and external parameters, the rotation matrix R and the translation matrix T which are obtained in the step (1).

And (4.3) calculating the sum of Euclidean distances from the three-dimensional space point calculated in the step (4.2) to each intersection line of the two conical surfaces in the step (4.1), and determining the minimum intersection line as a space curve segment corresponding to the curve segment on the two images.

(5) The three-dimensional coordinate transformation and registration of the outline of the air bag to be measured and the outline of the standard air bag comprise the following substeps:

(5.1) extracting three-dimensional digital-analog edge information of the CATIA of the standard safety airbag, and respectively calculating equal-parameter-interval sampling points C of edge contour space curves of the standard safety airbag and the safety airbag to be detected¹，C²。

Wherein,

C^{1} = {P_{0}^{1}, {P_{1}}^{1}, . . ., P_{m}^{1}}

C^{2} = {P_{0}^{2}, {P_{1}}^{1}, . . ., P_{n}^{2}}

are respectivelyAndof (c) is performed. The curvature may act as a local shape label for the discrete points. Judging whether the discrete points are paired, wherein a calculation formula is as follows:

in the formula, ξ is a threshold value.

This patent uses the following equation to determine the value of ξ:

in the formula,are respectively a curve C¹、C²Maximum and minimum of curvature.

Obtained from the above formulaAndthe matching point may be a true matching point or not, and the local shapes are similar, the matching errors are very large, the optimal matching point pair is easy to be mistaken, and the point pair serial number (i, j) of the matching point pair meeting the formula and the average curvature of the matching point pair are stored

k = (k_{i}^{1} + k_{j}^{2}) / 2

The distance squared sum of the matching point pairs is used as the measurement of the matching error, a proper threshold value is selected, unreal matching points are eliminated, and then the optimal matching point pair is obtained, wherein the calculation formula is as follows:

<math> <mrow> <mi>D</mi> <mo>=</mo> <mfrac> <mn>1</mn> <mrow> <mn>2</mn> <mi>w</mi> <mo>+</mo> <mn>1</mn> </mrow> </mfrac> <munderover> <mi>Σ</mi> <mrow> <mi>k</mi> <mo>=</mo> <mo>-</mo> <mi>w</mi> </mrow> <mi>w</mi> </munderover> <msup> <mrow> <mo>(</mo> <mo>|</mo> <mo>|</mo> <mi>M</mi> <mrow> <mo>(</mo> <msubsup> <mi>P</mi> <mrow> <mi>i</mi> <mo>&CirclePlus;</mo> <mi>k</mi> </mrow> <mn>1</mn> </msubsup> <mo>)</mo> </mrow> <mo>-</mo> <msubsup> <mi>P</mi> <mrow> <mi>j</mi> <mo>&CirclePlus;</mo> <mi>k</mi> </mrow> <mn>2</mn> </msubsup> <mo>|</mo> <mo>|</mo> <mo>)</mo> </mrow> <mn>2</mn> </msup> </mrow> </math>

i＝0,1,…，m；j＝0,1，…，n

in the formula,respectively, standard and to-be-detected air bag edge space curve C¹、C²A point on; m is a matching matrix of a France internal standard frame; i. j, m and n are space curves C respectively¹、C²The serial number of the upper point and the length of the sampling point; k is the number of the corresponding point; l | · | | is the distance between two points;in order to sum the die and the die,is (i + k) mod m,is (j + k) modn; w is a matching window size control parameter; 2w +1 is the number of sample points within the window.

Setting different window parameters w, the algorithm can be applied to global curve matching or local curve matching. The setting of this patent w is determined by the mean curvature k of the corresponding point, which is mapped as a function of w:

in the formula, k_rNormalized curvature of k;the value is 1.0 for adjusting the weight.

The matching point pair coordinate P corresponding to the standard and the edge space curve of the air bag to be measured can be obtained by the process_i ¹(x_i ¹，y_i ¹，z_i ¹) And P_i ²(x_i ²，y_i ²，z_i ²) Wherein i is 1,2, …, n; n is the number of matching points.

(5.2) carrying out space coordinate transformation by adopting a seven-parameter model of the Boehringer, wherein the seven-parameter model of the Boehringer comprises 3 translation parameters (delta x, delta y and delta z) and 3 rotation parameters (a)_x，_y，_z) And 1 scale parameter k, the formula is as follows:

matching point pair coordinates P of the standard air bag and the air bag to be tested determined by the step (5.1)_i ¹(x_i ¹，y_i ¹，z_i ¹) And P_i ²(x_i ²，y_i ²，z_i ²) And fitting corresponding seven parameters by using a least square method, and then converting the space coordinate of the airbag to be tested into the space coordinate of the standard airbag by using the obtained seven parameters so as to complete the three-dimensional coordinate transformation and registration of the outline of the airbag to be tested and the outline of the standard airbag.

(6) Calculating the position tolerance V of the detection points_x、V_y、V_zAnd the shape tolerance K of the contour curve section where the point is located, and comprises the following substeps:

(6.1) calculating the point with the shortest distance to the outline curve of the air bag to be detected through the detection points on the standard air bag, namely the corresponding detection point on the air bag to be detected, and further calculating the position tolerance of the detection points:

V_x＝X_measuring-X_{Sign board}

V_y＝Y_Measuring-Y_{Sign board}

V_z＝Z_Measuring-Z_{Sign board}

(6.2) calculating the Hausdorff distance between the standard and the corresponding profile curve section of the corresponding detection point of the profile of the air bag to be detected, and further solving the shape tolerance of the profile curve section to be detected:

K＝K_measuring-K_{Sign board}

And (6.3) comparing the required shape tolerance and position tolerance with the shape tolerance and position tolerance in the standard three-dimensional digital-analog engineering drawing of the safety air bag to judge whether the shape tolerance and the position tolerance are within the error range, and further judging whether the safety air bag to be detected is qualified.

Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications or variations may be made without inventive faculty, based on the technical solutions of the present invention.

Claims

1. A method for detecting the outline size of an air bag based on binocular vision is characterized in that,

the method comprises the following specific steps:

(2) image acquisition and preprocessing:two cameras are used for shooting the airbag to be tested simultaneously to obtain a left image and a right image I₁、Ι₂Carrying out image preprocessing such as Gaussian filtering, contrast enhancement, Canny operator edge detection and the like on a visual image of the safety airbag to obtain a continuous and closed safety airbag edge profile two-dimensional curve;

(3) segmentation and three-dimensional matching of the contour curve of the safety airbag: firstly, segmenting an air bag contour curve, then carrying out three-dimensional matching on the air bag contour curve segment, firstly carrying out coarse matching on characteristic points by utilizing a normalized cross correlation coefficient and brightness mean square error method, and then carrying out optimized matching on the curve segment between matched point pairs by utilizing a dynamic programming algorithm and taking boundary potential energy as a measurement standard;

(4) three-dimensional reconstruction of airbag contour curve: determining the intersection line of the space curved surface where the matching curve section is located according to the optical center of the camera by using the internal and external parameters, the translation matrix and the rotation matrix obtained in the step (1) and the matching curve section obtained in the step (3), and then determining the space curve section corresponding to the curve section on the image according to the end point of the curve section on the image and the distance from the point of any mark point on the curve in the three-dimensional space to the candidate space curve;

(5) and (3) three-dimensional coordinate transformation and registration of the outline of the air bag to be detected and the outline of the standard air bag: extracting three-dimensional digital-analog edge information of a CATIA (computer aided three-dimensional Interactive application) of a standard safety airbag, performing three-dimensional matching on a three-dimensional curve of a profile of the standard airbag and a three-dimensional curve of a profile of the standard airbag to determine matching points of the standard safety airbag and the safety airbag to be tested, performing space coordinate transformation by adopting a seven-parameter model of Boolean Sha, fitting corresponding seven parameters by using a least square method according to the matching point coordinates of the standard safety airbag and the safety airbag to be tested, converting the space coordinates of the safety airbag to be tested into the space coordinates of the standard safety airbag by using the obtained seven parameters, and further completing the three-dimensional coordinate transformation and registration;

V_x＝X_measuring-X_{Sign board}

V_y＝Y_Measuring-Y_{Sign board}

V_z＝Z_Measuring-Z_{Sign board}

K＝K_measuring-K_{Sign board}；

2. The binocular vision-based airbag contour dimension detection method as recited in claim 1, wherein: (2) image acquisition and preprocessing, comprising the following substeps:

(2.1) shooting the same scene by using two cameras simultaneously to obtain a left image and a right image I₁、Ι₂；

(2.2) denoising the two images by adopting Gaussian filtering;

(2.4) extracting the edge of the image by adopting a Canny operator;

and (2.5) adopting a mathematical morphology method to refine the edge into a single pixel, carrying out boundary tracking on a breakpoint generated on the boundary in the edge refining process, filling the breakpoint on the boundary, obtaining vectorization representation of the image, and further obtaining a continuous and closed airbag contour edge two-dimensional curve.

3. The binocular vision-based airbag contour dimension detection method as recited in claim 1, wherein: (3) the segmentation and the three-dimensional matching of the contour curve of the safety air bag comprise the following substeps:

(3.1) segmenting the airbag contour curve: let n sequence points describe the contour curve L of the airbag, i.e.:

L＝{p_i＝(x_i,y_i),i＝1,2,...,n}

S_k(p_i)＝{p_j|j＝i-k,i-k+1,...,i+k-1,i+k}

C = [\begin{matrix} c_{11} & c_{12} \\ c_{21} & c_{22} \end{matrix}]

wherein,

c_xand c_yIs a curved line segment S_k(p_i) I.e.:

the formula of the characteristic value can know lambda_L≥λ_S，

where Ω represents a collection of different support areas, the product of the eigenvalues of the different support areas may measure the characteristic of a point on the curve, the eigenvalue product P (P) of the store if the point is a distinct salient eigenvalue_i) The threshold value is larger and gives the product of the characteristic valuesIf the eigenvalue product P (P) of a point_i) Is greater thanIf the point is a characteristic point, otherwise, the point is regarded as a general point on the curve, and the safety airbag contour curve can be segmented;

(3.2) carrying out three-dimensional matching on the airbag contour curve segment:

firstly, a normalized cross correlation coefficient NCC and a brightness mean square error ASD method are utilized to carry out rough matching on the characteristic points,

defining the potential energy of a point on the curve as CEPF (L):

4. The binocular vision-based airbag contour dimension detection method as recited in claim 1, wherein: (4) three-dimensional reconstruction of airbag contour curves, comprising the following substeps:

X_{1}^{T} C_{1} X_{1} = 0

X_{2}^{T} C_{2} X_{2} = 0

wherein,

X₁＝(x₁,y₁,1)^T

X₂＝(x₂,y₂,1)^T

X₁、X₂are respectively an image I₁、I₂Upper position of quadratic curve C₁And C₂Homogeneous coordinates of the image of the upper arbitrary point, C₁And C₂A symmetric matrix of 33, represented by the projection equation:

s_{1} = [\begin{matrix} x_{1} \\ y_{1} \\ 1 \end{matrix}] = [\begin{matrix} {\hat{p}}_{11} \\ {\hat{p}}_{12} \\ {\hat{p}}_{13} \end{matrix}] X

s_{2} = [\begin{matrix} x_{2} \\ y_{2} \\ 1 \end{matrix}] = [\begin{matrix} {\hat{p}}_{21} \\ {\hat{p}}_{22} \\ {\hat{p}}_{23} \end{matrix}] X

wherein,andrespectively projecting matrices for camerasA row vector of (a); by introducing projection equations into Obtaining:

X^{T} {\hat{P}}_{1}^{T} C_{1} {\hat{P}}_{1} X = 0

X^{T} {\hat{P}}_{2}^{T} C_{2} {\hat{P}}_{2} X = 0

let the optical centers of the left and right cameras be O₁、O₂Then this formula is the optical center O₁、O₂Corresponding curve segment C₁、C₂Determining the intersection line of the conical surfaces;

(4.2) determining the end points of the curve segments in (4.1) and the three-dimensional coordinates of any characteristic point pair on the corresponding curve segments according to the principle of triangulation and the internal and external parameters, the rotation matrix R and the translation matrix T which are obtained in the step (1);

5. The binocular vision-based airbag contour dimension detection method as recited in claim 1, wherein: (5) the three-dimensional coordinate transformation and registration of the outline of the air bag to be measured and the outline of the standard air bag comprise the following substeps:

(5.1) extracting three-dimensional digital-analog edge information of the CATIA of the standard safety airbag, and respectively calculating equal-parameter-interval sampling points C of edge contour space curves of the standard safety airbag and the safety airbag to be detected¹，C²；

Wherein,

C^{1} = {P_{0}^{1}, P_{1}^{1}, . . ., P_{m}^{1}}

C^{2} = {P_{0}^{2}, P_{1}^{1}, . . ., P_{n}^{2}}

are respectivelyAndthe curvature can be used as a local shape label of the discrete point, whether the discrete point is paired or not is judged, and a calculation formula is provided:

in the formula, xi is a threshold value;

this patent uses the following equation to determine the value of ξ:

in the formula,are respectively a curve C¹、C²Maximum and minimum of curvature;

obtained from the above formulaAndthe matching point pairs may be true matching points or not, and only have similar local shapes, the matching errors of the matching points are very large, the optimal matching point pairs are easily mistaken, and the point pair serial numbers (i, j) meeting the matching point pairs of the above formula and the average curvature of the matching point pairs are stored:

k = (k_{i}^{1} + k_{j}^{2}) / 2

the sum of squared distances of the matching points is used as a measure of the matching error, a suitable threshold is selected,

rejecting unreal matching points to obtain an optimal matching point pair, wherein the calculation formula is as follows:

i＝0,1,…，m；j＝0,1，…，n

in the formula, P_i ¹、P_j ²Respectively, standard and to-be-detected air bag edge space curve C¹、C²A point on; m is a matching matrix of a France internal standard frame; i. j, m and n are space curves C respectively¹、C²The serial number of the upper point and the length of the sampling point; k is the number of the corresponding point; l | · | | is the distance between two points;in order to sum the die and the die,is (i + k) mod m,is (j + k) modn; w is a matching window size control parameter; 2w +1 is the number of sampling points in the window;

k_r∈[0,1]

in the formula, k_rNormalized curvature of k;for adjusting the weight, the value is 1.0;

the matching point pair coordinate P corresponding to the standard and the edge space curve of the air bag to be measured can be obtained by the process_i ¹(x_i ¹，y_i ¹，z_i ¹) And P_i ²(x_i ²，y_i ²，z_i ²) Wherein i is 1,2, …, n; n is the number of the matching points;