CN107726990A - The collection of dot matrix grid image and recognition methods in a kind of Sheet metal forming strain measurement - Google Patents

Abstract

The invention discloses a method for collecting and identifying lattice grid images in the strain measurement of sheet metal forming, which belongs to the technical field of optical three-dimensional non-contact measurement. The implementation steps of the method include 1) shooting multi-exposure lattice grid images; 2) Calculation of weight map; 3) Tower decomposition of weight map; 4) Tower decomposition of multi-exposure lattice grid image; 5) Laplacian pyramid weighting; 6) Laplacian pyramid fusion; 7) Lattice Grid image reconstruction; 8) Lattice grid image binarization; 9) Sub-pixel boundary recognition; 10) Ellipse fitting. Through the operation steps, the present invention can eliminate the influence of light reflection on the surface of the metal sheet, improve the quality of dot grid image collection, and increase the recognition rate of the dot grid image.

Description

A Method for Acquisition and Recognition of Lattice Grid Image in Strain Measurement of Sheet Metal Forming

technical field

The invention belongs to the technical field of optical three-dimensional non-contact measurement, and relates to a lattice grid image acquisition and recognition method in metal sheet metal forming strain measurement. It further relates to a dot grid image acquisition and recognition method based on multi-exposure image fusion.

Background technique

Sheet metal forming is one of the important branches of material processing technology, and has been widely used in various sectors of the national economy such as aerospace, automobiles, equipment manufacturing, and electrical appliances. In the sheet metal forming process, it is necessary to analyze the forming condition by measuring the three-dimensional strain on the surface of the sheet metal, so as to monitor critical deformation parts, solve complex forming, and optimize the stamping process. The means and methods of three-dimensional strain measurement on the surface of metal sheet mainly include mechanical method, electrical measurement method and optical three-dimensional non-contact measurement method. Among them, the optical three-dimensional non-contact measurement method is currently the main means of three-dimensional full-field strain measurement in sheet metal forming.

Chinese invention patent ZL201110263622.X discloses a method for measuring strain in sheet metal forming based on optical three-dimensional non-contact measurement. In this method, a digital camera or an industrial CCD camera is used to sample and shoot the dot grid pattern prepared on the surface of the metal sheet, and two or more grid images are obtained for three-dimensional reconstruction of the grid. The three-dimensional full-field strain on the surface of the metal sheet is calculated based on the size change of the lattice. Compared with the mechanical method and the electrical measurement method, this method has the advantages of non-contact, high measurement efficiency, high measurement accuracy, three-dimensional full-field strain can be obtained, and it can be applied to large-sized sheet metal forming parts. However, due to the strong reflective properties of the metal sheet surface, usually, there is a large difference between light and dark in the collected lattice grid image, especially in some areas with serious reflection, the imaging of the grid nodes in the image only retains a small part or It cannot be imaged, which makes the identification rate of grid nodes low, which in turn affects the measurement of three-dimensional strain.

Contents of the invention

In order to solve the above-mentioned defects in the prior art, the present invention provides a dot grid image acquisition and recognition method based on multi-exposure image fusion, which can eliminate the influence of strong reflection on the surface of the metal sheet during the sampling and shooting process, Improve the shooting quality of dot grid images and improve the recognition rate of dot grid images.

For achieving above object, the technical scheme that the present invention takes is:

A method for collecting and recognizing dot grid images in sheet metal forming strain measurement, comprising the following steps:

Step 1: Take a multi-exposure lattice image

Keep the camera still, and shoot several dot matrix images with different exposures in the black and white shooting mode;

Step 2, calculate the weight map

Calculate the contrast and exposure of each multi-exposure dot grid image taken in step 1, and multiply the calculated contrast and exposure of each multi-exposure dot grid image to obtain each A weight map for a multi-exposure lattice grid image;

Step 3, weight map tower decomposition

Carry out Gaussian pyramid decomposition to the weight map of each multi-exposure lattice grid image calculated in step 2;

Step 4, multi-exposure lattice grid image pyramid decomposition

Perform Laplacian pyramid decomposition on each multi-exposure lattice grid image taken in step 1, and the number of layers decomposed is the same as the number of layers of the weight map Gaussian pyramid decomposed in step 3;

Step 5, Laplacian Pyramid Empowerment

The image on each layer of the Laplacian pyramid of each multi-exposure lattice image decomposed in step 4 and the weight map Gaussian pyramid corresponding layer of the multi-exposure lattice image decomposed in step 3 The images of are multiplied to obtain the empowered Laplacian pyramid;

Step 6, Laplacian Pyramid Fusion

Add the images on the same layer of the Laplacian pyramid of several pieces of multi-exposure lattice grid images that are empowered in step 5;

Step 7, lattice grid image reconstruction

Carry out inverse pyramid transformation to the Laplacian pyramid fused in step 6, and reconstruct a new lattice grid image;

Step 8, binarize the lattice image

Perform local adaptive binarization processing on the lattice grid image reconstructed in step 7 to obtain a binary image;

Step 9, sub-pixel boundary recognition

Using the 8-connected domain rule to identify the entire pixel boundary of the lattice grid node in the binary image obtained in step 8; on this basis, using the space moment method to identify the sub-pixel boundary of the lattice grid node;

Step ten, ellipse fitting

For the sub-pixel boundaries of the lattice grid nodes identified in step 9, the coordinates of the center of the ellipse are iteratively fitted using the least squares method, thereby obtaining the coordinates of the lattice grid nodes.

Further, the method for collecting and recognizing lattice grid images in sheet metal forming strain measurement also includes, before performing step 1 to take a multi-exposure lattice grid image, first adjusting the focal length so that the lattice on the surface of the metal sheet The grid can be clearly imaged, and the focus is locked after adjustment.

Further, when taking multi-exposure dot matrix images in the first step, adjust the camera to black and white shooting mode, keep the camera stable, adjust the exposure, and shoot 3 or more images from underexposure, normal exposure to overexposure Bitmap grid image.

Further, the calculation method of the weight map in the step 2 is as follows:

2.1) For each multi-exposure lattice image taken in step 1, use the Laplacian operator to filter, and take the absolute value of the filtering result to obtain the contrast of each multi-exposure lattice image factor;

2.2) For each multi-exposure lattice grid image taken in step 1, the exposure factor is calculated using the following formula:

Wherein, e represents a natural constant, k represents the kth multi-exposure lattice image, (i, j) represents the pixel point position, and I _k (i, j) represents the kth multi-exposure lattice image in ( The gray value of the pixel at position i, j), E _k (i, j) represents the exposure factor of the k-th multi-exposure lattice grid image at position (i, j), and σ represents the width of the kernel function ;

2.3) On the basis of the contrast factor calculated in step 2.1) and the exposure factor calculated in 2.2), the weight factor of each multi-exposure lattice grid image is calculated by the following formula:

in, Represents the weight factor of the kth multi-exposure lattice image at the position (i, j), C _k (i, j) represents the kth multi-exposure lattice image calculated in step 2.1) at (i , j) the contrast factor at the position;

2.4) normalize the weight factor of each multi-exposure lattice image calculated in step 2.3), obtain the weight map of each multi-exposure lattice image, and the normalization method is as follows:

Among them, W _k (i, j) represents the gray value of the pixel at the position (i, j) of the weight map of the kth multi-exposure lattice grid image, and N represents the multi-exposure lattice captured in step 1 The number of grid images, s represents the sth multi-exposure lattice grid image, Indicates the weight factor of the s-th multi-exposure lattice grid image at position (i, j) calculated in step 2.3).

Further, the Laplacian h used in the step 2.2) is:

Further, the calculation method of the tower-shaped decomposition layer number L in the step 3 is as follows:

Among them, r and c represent the number of rows and columns of the multi-exposure lattice grid image respectively, ln() represents the natural logarithmic function with the natural constant e as the base, min(r,c) represents the multi-exposure lattice grid The number of rows and columns of the image takes the minimum value function.

Further, the local adaptive binarization method of the lattice grid image in the eighth step is as follows:

8.1) Use the window-by-window binarization method to quickly binarize the lattice grid image reconstructed in step 7 to obtain a binary image;

8.2) Mark the connected region on the binary image obtained in step 8.1), and further use the pixel-by-pixel binarization method to re-binarize all pixels in the connected region for the connected region whose number of pixels is greater than a given threshold τ :

a) Calculate a binarization threshold for any pixel in the connected region greater than a given threshold τ:

Among them, (i, j) represents the position of the pixel point, T(i, j) represents the binarization threshold of the pixel point at the position (i, j) in the connected area, Indicates the average gray value of the pixels in the u×v local window centered on (i, j) in the lattice grid image reconstructed in step 7, where u and v represent the length and height of the window, respectively, m and n represent two integer variables, and I′ ₀ (m, n) represents the gray value of the pixel at the (m, n) position of the lattice grid image reconstructed in step 7, Represent the standard deviation value of the gray value distribution of the pixels in the u×v local window centered on (i, j) in the lattice grid image reconstructed in step 7, and γ and offset are two constants;

b) re-binarize the pixels according to the binarization threshold calculated in step a):

Among them, I'(i, j) represents the gray value of the pixel at the position (i, j) of the binary image obtained in step 8.1) after re-binarization, and I' ₀ (i, j) represents the step The gray value of the pixel at the (i, j) position of the reconstructed lattice grid image.

Further, the value of the threshold τ in the step 8.2) is τ=10000.

Further, during each iteration in step ten, the distance from each boundary point to the fitted ellipse and the standard deviation value sigma are calculated, and boundary points with a distance greater than 3 sigma are not involved in the next iterative calculation.

Compared with the prior art, the inventive method has the following advantages:

(1) The method of the present invention can eliminate the influence of strong reflection on the surface of the sheet metal, and collect high-quality lattice grid images;

(2) the dot grid image quality that the inventive method collects is high, so the recognition rate of dot grid image also improves accordingly;

(3) the quality of the dot grid image collected by the inventive method is high, so the recognition accuracy of the dot grid image is also improved accordingly;

(4) The method of the present invention improves the recognition rate and recognition accuracy of the lattice grid image, thus indirectly improving the reconstruction accuracy of the three-dimensional grid and the measurement accuracy of the three-dimensional strain.

Description of drawings

Fig. 1 is a flowchart of specific operation steps of the present invention.

Figure 2a is an underexposed lattice grid image of a cupping specimen.

Figure 2b shows the normal exposure dot grid image of a cupping specimen.

Figure 2c is an overexposed dot grid image of a cupping specimen.

Fig. 3 Reconstructed dot grid image of a cupping specimen.

Fig. 4 Detection results of underexposed lattice grid image of a cupping specimen.

Fig. 5 Detection results of dot matrix image of a cupping specimen under normal exposure.

Fig. 6 Detection results of over-exposed dot matrix image of a cupping specimen.

Figure 7. Reconstructed dot matrix image detection results of a cupping specimen.

Detailed ways

The invention will be further described in detail below in conjunction with the accompanying drawings and embodiments, but it is not used as a basis for any limitation on the invention.

The present invention proposes a dot matrix image acquisition and recognition method in sheet metal forming strain measurement, as shown in FIG. 1 . When collecting and recognizing a dot matrix image on the surface of a metal sheet from a certain viewing angle, the first step is to shoot a multi-exposure dot grid image. Before shooting, adjust the focal length first, so that the dot grid on the surface of the metal sheet can be clearly imaged, and lock the focal length after the adjustment is completed. When shooting, in order to reduce noise, adjust the camera to black and white shooting mode, keep the camera steady, adjust the exposure, and shoot 3 or more dot matrix images from underexposure, normal exposure to overexposure.

The second step is to calculate the weight map. Calculate the weight map of each multi-exposure lattice grid image taken in the first step. The weight map calculation process includes the following steps:

1) Calculate the contrast factor. For each multi-exposure lattice image taken in the first step, the Laplacian operator h is used to filter, and the absolute value of the filtering result is obtained to obtain the contrast of each multi-exposure lattice image factor. The Laplacian operator h used is:

2) Calculate the exposure factor. For each multi-exposure lattice image taken in the first step, the exposure factor is calculated using the following formula:

Among them, e represents a natural constant, k represents the kth multi-exposure lattice image, (i, j) represents the pixel position, E _k (i, j) represents the kth multi-exposure lattice image in ( Exposure factor at position i, j), I _k (i, j) represents the gray value of the pixel at position (i, j) of the k-th multi-exposure lattice grid image, σ represents the width of the kernel function , its value range is 0<σ≤0.5, generally choose σ=0.2.

3) Calculate the weight factor. On the basis of the contrast factor calculated in step 1) and the exposure factor calculated in step 2), the weight factor of each multi-exposure lattice grid image is calculated using the following formula:

in, Indicates the weight factor of the k-th multi-exposure lattice image at position (i, j), C _k (i, j) represents the k-th multi-exposure lattice image calculated in step 1) at ( i, contrast factor at position j).

4) Normalization of weight factors. The weight factor of each multi-exposure lattice grid image calculated in step 3) is normalized to obtain the weight map of each multi-exposure lattice grid image, and the normalization formula adopted is as follows:

Among them, W _k (i, j) represents the gray value of the pixel at the position (i, j) of the weight map of the kth multi-exposure lattice grid image, and N represents the multi-exposure point taken in the first step The number of array grid images, s represents the sth multi-exposure lattice grid image, Indicates the weight factor of the sth multi-exposure lattice grid image at the position (i, j) calculated in step 3).

The third step is the tower decomposition of the weight map. Gaussian pyramid decomposition is performed on the weight map of each multi-exposure lattice grid image calculated in the second step. The decomposition process includes the following steps:

1) Calculate the number of tower decomposition layers. The calculation method of tower-shaped decomposition layer number L is as follows:

Among them, r and c represent the number of rows and columns of the multi-exposure lattice grid image respectively, ln() represents the natural logarithmic function with the natural constant e as the base, min(r,c) represents the multi-exposure lattice grid The number of rows and columns of the image takes the minimum value function.

2) Construct the 0th layer of the weight map Gaussian pyramid. The weight map of each multi-exposure lattice grid image calculated in the second step is used as the image on the 0th layer of the weight map Gaussian pyramid:

G _k,0 (i,j)=W _k (i,j),

Among them, k represents the k-th multi-exposure lattice image, (i, j) represents the pixel position, G _k,0 (i, j) represents the weight map Gaussian gold of the k-th multi-exposure lattice image The gray value of the pixel of the image on the 0th layer of the tower at the position (i, j), W _k (i, j) represents the weight map of the kth multi-exposure lattice grid image calculated in the second step The gray value of the pixel at position (i, j).

3) Construct the tth layer of the weight map Gaussian pyramid. Convolve the image on the t-1 layer of the weight map Gaussian pyramid with a 5×5 window function, and then down-sample the convolution result every row and column to get the image on the t-th layer of the weight map Gaussian pyramid:

Among them, t represents the tth layer of the pyramid, and G _{k, t} (i, j) represents the weight map of the kth multi-exposure lattice grid image of the image on the tth layer of the Gaussian pyramid at the (i, j) position The gray value of the pixel, m and n represent two integer variables, G _{k, t-1} (2i+m, 2j+n) represents the weight map of the kth multi-exposure lattice grid image Gaussian pyramid t- The gray value of the pixel at the position (2i+m,2j+n) of the image on layer 1, λ(m,n) represents the 5×5 window function:

On the basis of the 0th layer of the weight map Gaussian pyramid of the kth multi-exposure lattice grid image constructed in step 2), according to the t-layer construction method of the weight map Gaussian pyramid, the kth multi-exposure lattice image can be constructed The first layer of the Gaussian pyramid of the weight map of the exposure lattice image, and so on, can construct the L layer of the weight map of the k-th multi-exposure lattice image of the Gaussian pyramid.

The fourth step is to decompose the multi-exposure lattice grid image into pyramids. Laplacian pyramid decomposition is performed on each multi-exposure lattice image captured in the first step. The decomposition process includes the following steps:

1) Multi-exposure lattice image Gaussian pyramid decomposition. According to the weight map pyramid decomposition method described in the third step, each multi-exposure lattice grid image taken in the first step is decomposed into a Gaussian pyramid, and the image on the tth layer of the decomposed Gaussian pyramid is recorded Be G′ _k,t , wherein, k represents the kth multi-exposure lattice grid image;

2) Construct the Lth layer of the Laplacian pyramid. The image on the L layer of the Gaussian pyramid decomposed in step 1) is used as the image on the L layer of the Laplacian pyramid:

LP _k,L (i,j)=G′ _k,L (i,j),

Among them, (i, j) represents the pixel position, L represents the number of tower decomposition layers calculated in the third step, LP _k,L (i, j) represents the k-th multi-exposure lattice grid image Laplacian The gray value of the pixel at the position (i, j) of the image on the L layer of the pyramid, G′ _k,L (i, j) represents the kth multi-exposure lattice grid decomposed in step 1) The gray value of the pixel point of the image on the L layer of the image Gaussian pyramid at (i, j) position;

3) Construct the tth layer of the Laplacian pyramid. The construction method of the tth layer of the Laplacian pyramid is as follows:

a) Interpolate and enlarge the image on the t+1th layer of the Gaussian pyramid decomposed in step 1):

Among them, t+1 represents the t+1th layer of the pyramid, Indicates the gray value of the pixel at the (i, j) position of the interpolated enlarged image, m and n represent two integer variables, Indicates that the image on the t+1th layer of the Gaussian pyramid of the kth multi-exposure lattice grid image decomposed in step 1) is in The gray value of the pixel at the position, Represents a transition variable, λ(m,n) represents the 5×5 window function described in the third step;

b) Subtract the image on the tth layer of the Gaussian pyramid decomposed in step 1) and the interpolated and enlarged image in step a):

Among them, LP _{k, t} (i, j) represents the gray value of the pixel at the (i, j) position of the image on the tth layer of the Laplacian pyramid of the kth multi-exposure lattice grid image, G ' _k,t (i,j) represents the gray value of the pixel at (i,j) position of the image on the tth layer of the Gaussian pyramid of the kth multi-exposure lattice image decomposed in step 1) ;

According to the construction method of the t-th layer of the Laplacian Pyramid, the L-1 to 0-th layers of the Laplacian Pyramid can be sequentially constructed.

The fifth step is the empowerment of Laplace Pyramid. For the Laplacian pyramid of each multi-exposure lattice image decomposed in the fourth step, the following method is used for weighting:

LP′ _k,t (i,j)=G _k,t (i,j)×LP _k,t (i,j),

Among them, k represents the k-th multi-exposure lattice grid image, t represents the t-th layer of the pyramid, (i, j) represents the pixel position, LP′ _k,t (i, j) represents the weighted k-th Multi-exposure lattice grid image The gray value of the pixel at the (i, j) position of the image on the tth layer of the Laplacian pyramid, G _k,t (i, j) represents the decomposition in the third step The weight map of the k-th multi-exposure lattice grid image The gray value of the pixel at the (i, j) position of the image on the t-th layer of the Gaussian pyramid, LP _k,t (i, j) represents the fourth step The decomposed k-th multi-exposure lattice grid image is the gray value of the pixel at the (i, j) position of the image on the t-th layer of the Laplacian pyramid, and the symbol × represents the gray value of two pixels values are multiplied.

The sixth step, Laplacian pyramid fusion. The Laplacian pyramid fusion method is as follows:

Wherein, k represents the kth multi-exposure lattice image, t represents the tth layer of the pyramid, (i, j) represents the pixel position, and N represents the number of multi-exposure lattice images taken in the first step , Indicates the gray value of the pixel at the (i, j) position of the image on the tth layer of the fused Laplacian pyramid, and LP′ _k,t (i, j) represents the weighted value of the fifth step The gray value of the pixel at the position (i, j) of the image on the tth layer of the Laplacian pyramid of k multi-exposure lattice images.

The seventh step is to reconstruct the lattice grid image. Perform inverse pyramid transformation on the Laplacian pyramid fused in the sixth step to construct a new lattice image. The inverse tower transformation includes the following steps:

1) Restore the Lth layer of the Gaussian pyramid. The image on the L layer of the Laplacian pyramid fused in the sixth step is used as the image on the L layer of the Gaussian pyramid to be restored:

Among them, (i, j) represents the pixel position, L represents the number of tower decomposition layers calculated in the third step, I′ _L (i, j) represents the image on the Lth layer of the restored Gaussian pyramid in (i, j) the gray value of the pixel at the position, Represent the gray value of the pixel at the (i, j) position of the image on the L layer of the Laplacian pyramid that is fused in the sixth step;

2) Restore the tth layer of the Gaussian pyramid. The recovery method of the tth layer of the Gaussian pyramid is as follows:

a) The image on the t+1th layer of the Laplacian pyramid fused in the sixth step is interpolated and enlarged:

Among them, t+1 represents the t+1th layer of the pyramid, Indicates the gray value of the pixel at the (i, j) position of the interpolated enlarged image, m and n represent two integer variables, and λ(m, n) represents the 5×5 window function described in the third step , Indicates that the image on the t+1th layer of the Laplacian pyramid fused in the sixth step is at The gray value of the pixel at the position, represents a transition variable;

b) Add the interpolated and enlarged image in step a) to the image on the tth layer of the Laplacian pyramid fused in the sixth step:

Among them, I′ _t (i, j) represents the gray value of the pixel at the (i, j) position of the restored image on the tth layer of the Gaussian pyramid, Represent the gray value of the pixel at the (i, j) position of the image on the t-th layer of the Laplacian pyramid that is fused in the sixth step;

According to the recovery method for the t layer of the Gaussian pyramid, the Gaussian pyramid layer L-1 to the 0th layer can be restored successively, and the image I'0 on the _0th layer of the Gaussian pyramid restored is the lattice grid image to be reconstructed .

The eighth step is adaptive binarization of the lattice grid image. The adaptive binarization process includes the following steps:

1) Use the window-by-window binarization method described in the literature "Circular grid pattern based surface strain measurement system for sheet metal forming" (Shi B and Liang J, OPT LASER ENG, 2012, 50(9): 1186-1195) to Fast binarization is performed on the lattice grid image reconstructed in step 7 to obtain a binary image.

2) Mark the connected region on the binary image obtained in step 1), and for the region where the number of pixels is greater than a given threshold τ, further use the pixel-by-pixel binarization method to re-binarize all pixels in the connected region :

a) Calculate a binarization threshold for any pixel in the connected region greater than a given threshold τ:

Among them, (i, j) represents the position of the pixel point, T(i, j) represents the binarization threshold of the pixel point at the position (i, j) in the connected area, Represents the average gray value of pixels in the u×v local window centered on (i, j) in the lattice grid image reconstructed in the seventh step, m and n represent two integer variables, u and v Denote the length dimension and the height dimension of the window respectively, usually take u=v=23, and I' ₀ (m, n) represents the pixels of the lattice grid image reconstructed in step 7 at the (m, n) position grayscale value, Indicates the standard deviation value of the gray value distribution of pixels in the u×v local window centered on (i, j) in the lattice grid image reconstructed in the seventh step, γ and offset are two constants, generally Choose γ=0.1, offset=5.0, threshold τ is a positive integer, generally take τ=10000;

b) re-binarize the pixels according to the binarization threshold calculated in step a):

Among them, I'(i, j) represents the gray value of the pixel at the position (i, j) of the binary image obtained in step 8.1) after re-binarization, and I' ₀ (i, j) represents the step The gray value of the pixel at the (i, j) position of the reconstructed lattice grid image, .

The ninth step is sub-pixel boundary recognition. The sub-pixel boundary recognition method is as follows:

1) Using the 8-connected domain rule to identify the integer pixel boundaries of the lattice grid nodes in the binary image obtained in the eighth step;

2) On the basis of the whole-pixel boundary of the lattice grid node identified in step 1), the space moment method is used to further identify the sub-pixel boundary of the lattice grid node.

The tenth step, ellipse fitting. For the sub-pixel boundaries of the lattice grid nodes in the binary image identified in the ninth step, the document "Least squares fitting of circles and ellipses" (W.Gander, G.H. Goluband R. Strebel, BIT Numerical Mathematics, 1994, 34:558-578) to fit the coordinates of the center of the ellipse using the iterative method of least squares. In order to reduce noise interference, the distance from each boundary point to the fitted ellipse and the standard deviation value sigma are calculated during each iteration, and the boundary points with a distance greater than 3 sigma will not participate in the next iteration. The final fitted ellipse center coordinates are the coordinates of the lattice grid nodes.

Below in conjunction with concrete simulation experiment, the present invention is described, and wherein the present invention method realizes corresponding algorithm on VS2010 and opengl platform and runs on the PC of Intel i7-4770CPU 3.4GHz, 16GB internal memory.

The three multi-exposure lattice grid images of a certain cupping specimen taken are shown in Figure 2a-2c, among which, Figure 2a shows the underexposure lattice grid image of a certain cupping specimen, and Figure 2b shows Figure 2c shows the overexposure dot grid image of a cupping test piece taken under normal exposure. Fig. 3 shows a new dot grid image reconstructed by the method of the present invention based on 3 multi-exposure dot grid images of a certain cupping test piece. By comparing Figure 2a-Figure 2c with Figure 3, it can be seen that the lattice grid in the reconstructed lattice image is clearer. Figures 4 to 7 show the underexposed lattice grid image, the normal exposure lattice grid image, the overexposed lattice grid image and the lattice grid in the reconstructed lattice grid image of a cupping specimen According to the node detection results, the recognition rate of lattice grid nodes is 49.6%, 77.78%, 72.3%, and 86.87%. Wherein, the white cross lines represent the coordinates of the detected lattice grid nodes. By comparing Fig. 4-Fig. 7, it can be seen that in the reconstructed lattice image, the detected number of lattice lattice nodes is the largest, and the recognition rate of lattice lattice nodes is the highest, which is 86.87%. It can also be shown through this embodiment that the method of the present invention can eliminate the influence of strong reflection on the surface of the metal sheet, improve the quality of the dot grid image collection, and increase the recognition rate of the dot grid image.

The embodiments of the present invention have been described above in conjunction with the accompanying drawings and Examples, but the present invention is not limited to the above embodiments, within the knowledge of those of ordinary skill in the art, it is also possible to Make various changes below.

Claims (9)

1. a dot grid image acquisition and recognition method in sheet metal forming strain measurement, is characterized in that, comprises the following steps: Step 1: Take a multi-exposure lattice image Keep the camera still, and shoot several dot matrix images with different exposures in the black and white shooting mode; Step 2, calculate the weight map Calculate the contrast and exposure of each multi-exposure dot grid image taken in step 1, and multiply the calculated contrast and exposure of each multi-exposure dot grid image to obtain each A weight map for a multi-exposure lattice grid image; Step 3, weight map tower decomposition Carry out Gaussian pyramid decomposition to the weight map of each multi-exposure lattice grid image calculated in step 2; Step 4, multi-exposure lattice grid image pyramid decomposition Perform Laplacian pyramid decomposition on each multi-exposure lattice grid image taken in step 1, and the number of layers decomposed is the same as the number of layers of the weight map Gaussian pyramid decomposed in step 3; Step 5, Laplacian Pyramid Empowerment The image on each layer of the Laplacian pyramid of each multi-exposure lattice image decomposed in step 4 and the weight map Gaussian pyramid corresponding layer of the multi-exposure lattice image decomposed in step 3 The images of are multiplied to obtain the empowered Laplacian pyramid; Step 6, Laplacian Pyramid Fusion Add the images on the same layer of the Laplacian pyramid of several pieces of multi-exposure lattice grid images that are empowered in step 5; Step 7, lattice grid image reconstruction Carry out inverse pyramid transformation to the Laplacian pyramid fused in step 6, and reconstruct a new lattice grid image; Step 8, binarize the lattice image Perform local adaptive binarization processing on the lattice grid image reconstructed in step 7 to obtain a binary image; Step 9, sub-pixel boundary recognition Using the 8-connected domain rule to identify the entire pixel boundary of the lattice grid node in the binary image obtained in step 8; on this basis, using the space moment method to identify the sub-pixel boundary of the lattice grid node; Step ten, ellipse fitting For the sub-pixel boundaries of the lattice grid nodes identified in step 9, the coordinates of the center of the ellipse are iteratively fitted using the least squares method, thereby obtaining the coordinates of the lattice grid nodes. 2. The method for collecting and identifying lattice grid images in sheet metal forming strain measurement according to claim 1, further comprising: before performing step 1 to take a multi-exposure lattice grid image, first adjust The focal length enables the dot grid on the surface of the metal sheet to be clearly imaged, and the focal length is locked after adjustment. 3. The method for collecting and identifying lattice grid images in a metal sheet forming strain measurement according to claim 1, wherein in said step 1, when the multi-exposure lattice grid images are taken, the camera is adjusted to Go to the black and white shooting mode, keep the camera steady, adjust the exposure, and shoot 3 or more dot matrix images from underexposure, normal exposure to overexposure. 4. a kind of lattice grid image acquisition and recognition method in the sheet metal forming strain measurement according to claim 1, it is characterized in that, in described step 2, weight map calculation method is as follows: 2.1) For each multi-exposure lattice image taken in step 1, use the Laplacian operator to filter, and take the absolute value of the filtering result to obtain the contrast factor of each multi-exposure lattice image ; 2.2) For each multi-exposure lattice grid image taken in step 1, the exposure factor is calculated using the following formula: Wherein, e represents a natural constant, k represents the kth multi-exposure lattice grid image, (i, j) represents the pixel point position, and I _k (i, j) represents the k-th multi-exposure lattice grid image in ( The gray value of the pixel at position i, j), E _k (i, j) represents the exposure factor of the k-th multi-exposure lattice grid image at position (i, j), and σ represents the width of the kernel function ; 2.3) On the basis of the contrast factor calculated in step 2.1) and the exposure factor calculated in step 2.2), the weight factor of each multi-exposure lattice grid image is calculated by the following formula: in, Represents the weight factor of the kth multi-exposure lattice image at the position (i, j), C _k (i, j) represents the kth multi-exposure lattice image calculated in step 2.1) at (i , j) the contrast factor at the position; 2.4) normalize the weight factor of each multi-exposure lattice image calculated in step 2.3), obtain the weight map of each multi-exposure lattice image, and the normalization method is as follows: Among them, W _k (i, j) represents the gray value of the pixel at the position (i, j) of the weight map of the kth multi-exposure lattice grid image, and N represents the multi-exposure lattice captured in step 1 The number of grid images, s represents the sth multi-exposure lattice grid image, Indicates the weight factor of the s-th multi-exposure lattice grid image at position (i, j) calculated in step 2.3). 5. a kind of lattice grid image acquisition and recognition method in sheet metal forming strain measurement according to claim 4, is characterized in that, the Laplacian operator h that adopts in described step 2.2) is: 6. a kind of lattice grid image acquisition and recognition method in sheet metal forming strain measurement according to claim 1, it is characterized in that, the calculation method of tower shape decomposition layer number L in the described step 3 is as follows: Among them, r and c represent the number of rows and columns of the multi-exposure lattice grid image respectively, ln() represents the natural logarithmic function with the natural constant e as the base, min(r,c) represents the multi-exposure lattice grid The number of rows and columns of the image takes the minimum function. 7. A kind of lattice grid image acquisition and recognition method in a kind of sheet metal forming strain measurement according to claim 1, it is characterized in that, in described step 8, local self-adaptive binarization method of lattice grid image is as follows : 8.1) Use the window-by-window binarization method to quickly binarize the lattice grid image reconstructed in step 7 to obtain a binary image; 8.2) Mark the connected region on the binary image obtained in step 8.1), and further use the pixel-by-pixel binarization method to re-binarize all pixels in the connected region for the connected region whose number of pixels is greater than a given threshold τ : a) Calculate a binarization threshold for any pixel in the connected region greater than a given threshold τ: Among them, (i, j) represents the position of the pixel point, T(i, j) represents the binarization threshold of the pixel point at the position (i, j) in the connected area, Represents the average gray value of the pixels in the u×v local window centered on (i, j) in the lattice grid image reconstructed in step 7, u and v represent the length and height dimensions of the window respectively, m and n represent two integer variables, and I′ ₀ (m, n) represents the gray value of the pixel at the (m, n) position of the lattice grid image reconstructed in step 7, Represent the standard deviation value of the gray value distribution of the pixels in the u×v local window centered on (i, j) in the lattice grid image reconstructed in step 7, and γ and offset are two constants; b) re-binarize the pixels according to the binarization threshold calculated in step a): Among them, I'(i, j) represents the gray value of the pixel at the position (i, j) of the binary image obtained in step 8.1) after re-binarization, and I' ₀ (i, j) represents the step The gray value of the pixel at the (i, j) position of the reconstructed lattice grid image. 8 . The method for collecting and identifying images of lattice grids in sheet metal forming strain measurement according to claim 7 , wherein the value of the threshold τ in the step 8.2) is τ=10000. 9. A kind of lattice grid image acquisition and recognition method in sheet metal forming strain measurement according to claim 1, it is characterized in that, in each iteration process in described step ten, calculate each boundary point to The distance between the fitted ellipse and the standard deviation value sigma, for the boundary points whose distance is greater than 3sigma, do not participate in the next iterative operation.