A battery separator automated assembly positioning method based on machine vision

Abstract

The present invention relates to the field of automated assembly of battery diaphragms, and in particular to a method for automated assembly and positioning of battery diaphragms based on machine vision. A visual positioning method applied to battery diaphragms, the method comprising the following steps: S100, a truss machine drives a vacuum adsorption plate to absorb the diaphragm from a silo, and slowly lowers the diaphragm to contact with optical white glass; S200, an industrial camera collects grayscale images, and transmits them to upper computer software via TCP / IP protocol; S300, non-target area suppression processing is performed on the collected grayscale images. The benefits of the present invention are: (1) a scheme for vacuum adsorption plate to absorb diaphragm feeding is proposed, which ensures the overall flatness of the diaphragm during the feeding process; (2) a scheme for optical white glass pressing diaphragm is proposed, which reduces the warping of the diaphragm edge and reduces the positioning error caused by local tensile deformation of the diaphragm.

Description

Technical Field

[0001] The invention relates to the field of automatic assembly of battery diaphragms, and in particular to an automatic assembly positioning method of battery diaphragms based on machine vision. Background Art

[0002] As an important part of energy storage battery production, high-precision automated assembly of diaphragms is of great significance. Traditional manual assembly and positioning methods have disadvantages such as low efficiency and insufficient precision, which makes it difficult to meet the needs of automated production.

[0003] At present, the automated assembly positioning method based on machine vision has the advantages of good real-time performance, high stability, and low long-term cost. However, there are some problems with the automated assembly of battery separators:

[0004] (1) During the diaphragm cutting stage, due to the limitations of its material properties and cutting process, local tensile deformation and dimensional fluctuations are prone to occur;

[0005] (2) During the diaphragm feeding stage, due to its large area, thin thickness, soft material and certain air permeability, the overall flatness cannot be guaranteed;

[0006] (3) In the visual positioning stage, traditional algorithms are easily affected by background light spots and interference inside the diaphragm, and it is difficult to meet the real-time and accuracy requirements at the same time;

[0007] The above problems pose certain challenges to the visual positioning method of the automatic assembly of the diaphragm. It is difficult to use the existing machine vision positioning method to simultaneously meet the application scenarios under the above working conditions. Therefore, there is an urgent need for a machine vision positioning method for the automatic assembly task of the battery diaphragm to meet the following task requirements:

[0008] (1) Identify the diaphragm size and remove the diaphragms that do not meet the requirements;

[0009] (2) Ensure positioning accuracy for diaphragms in non-flat states;

[0010] (3) Adapt to precise positioning in complex backgrounds and meet real-time requirements while ensuring positioning accuracy. Summary of the invention

[0011] In order to solve the above technical problems, the present invention discloses a visual positioning method applied to a battery separator, the method comprising the following steps:

[0012] S100, the truss machine drives the vacuum adsorption plate to absorb the diaphragm from the silo and slowly lower it to the diaphragm to contact the optical clear glass;

[0013] S200, industrial camera collects grayscale images and transmits them to the host computer software through TCP / IP protocol;

[0014] S300, performing non-target area suppression processing on the collected grayscale image;

[0015] S400, based on the center extension search method, interval sampling to obtain the pixel-level contour edge coordinates of the diaphragm object;

[0016] S500, obtaining sub-pixel level contour edge coordinates of the diaphragm object based on a quadratic polynomial interpolation formula;

[0017] S600, based on the refraction correction formula, the above sub-pixel level contour edge coordinates are converged inwardly to obtain the corrected contour edge coordinates of the diaphragm object;

[0018] S700, based on the least square method, adopting a weighted fitting strategy to obtain a fitting straight line of the edge of the diaphragm contour;

[0019] S800, calculating the actual size of the diaphragm based on the fitted straight line, and putting the diaphragm with unqualified size into a waste frame;

[0020] S900, based on the offset calculation formula, the current diaphragm position is compared with the pre-calibrated position, the position parameters of the current diaphragm are calculated, and transmitted to the truss machine, thereby completing the diaphragm positioning task;

[0021] Further, in step S100, one side of the adsorption surface of the vacuum adsorption plate is provided with adsorption holes, and the inside of the vacuum adsorption plate is provided with main cavities and branch cavities; further, the branch cavities correspond to the adsorption holes one by one; further, the S100 is also characterized in that the diaphragm is pressed onto the surface of the optical white glass to resist local tensile deformation of the diaphragm; further, the optical white glass is also characterized in that the surface is provided with exhaust holes, which reduce the contact area with the diaphragm by destroying the continuity of the light surface, thereby reducing the effect of vacuum negative pressure and surface adhesion on the diaphragm when the vacuum adsorption plate is lifted;

[0022] Furthermore, the S300 includes the following sub-steps:

[0023] S301, performing a Zero Padding operation on the image to prevent subsequent operations from changing the size of the original image; preferably, the Padding operation may also use Reflection Padding, which can prevent occasional highlight pixels on the edge of the original image from affecting the suppression effect by mirroring the pixel values ​​of the existing edge area;

[0024] S302, performing a traversal operation on the non-filled points in the image at intervals S according to the following formula:

[0025]

[0026] Where Vm,n is the original gray value of the current operation point, V m,n ' is the gray value after operation of the current operation point, 2N + 1 is the size of the current operation area, and Num(Value<T) is the number of pixels with gray values less than the threshold T in the current operation area; Preferably, the traversal interval S = 2N; Further, the feature of the present invention is that through the above traversal operation, on the premise of retaining the important information area near the diaphragm object, the interference of non-important pixel areas such as the background and the inside of the diaphragm is suppressed as much as possible; Further, this operation sharpens the contour edge features by compressing the edge gray level length of the diaphragm, providing a more ideal gray level model for the subsequent edge extraction operation;

[0027] S303. Based on the connected component analysis method, obtain the number of connected components in the current image;

[0028] S304. If the number of connected components in the current image is 0, it is determined that there is no high-brightness target in the image, that is, there is no diaphragm object, and an ERROR alarm signal is output to terminate the current detection task; The feature of the present invention is that taking the number of connected components as the determination condition can not only evaluate the suppression effect of the non-target area in this round, but also detect whether the image contains an effective diaphragm object;

[0029] Preferably, after step S304, the following steps are further included:

[0030] S305. If the number of connected components in the current image is greater than 1, repeat steps S301 to S303;

[0031] S306. If the number of connected components in the current image is equal to 1, there is no bright spot interference in the non-target area, and the current loop ends; The feature of the present invention is that by repeatedly performing the non-target area suppression process and taking the number of connected components in the image as the determination condition, it is ensured that all non-diaphragm objects in the image are effectively suppressed, and the gray feature details related to the diaphragm object are effectively retained;

[0032] Further, the S400 includes the following steps:

[0033] S401. Starting from the center point of the image as the search starting point, perform single-row / column search traversal in the up, down, left, and right four directions respectively until a gray value mutation point is retrieved;

[0034] S402. Taking the gray value mutation point as the secondary search starting point, traverse along both sides perpendicular to the first search direction, and apply the Sobel operator to obtain the pixel-level contour edge P of the diaphragm object O; Further, the secondary search traversal adopts an interval sampling method; Further, the present invention is characterized in that, according to the characteristics of the diaphragm image, an edge search strategy is designed with the image center as the starting point center, reducing the invalid traversal area by more than 85%, and effectively reducing the algorithm time consumption and improving the detection performance while ensuring the edge detection accuracy;

[0035] Furthermore, the step S500 includes the following steps:

[0036] S501, traverse and calculate the contour edge P O The gray gradient R nearby i-1 and R i+1 ;

[0037] S502: Calculate the sub-pixel level contour edge P of the diaphragm object according to the following quadratic polynomial interpolation formula: A :

[0038]

[0039] In the formula, R i-1 For point (x i ,y i )The gray gradient of the previous point, R i+1 For point (x i ,y i ) is the grayscale gradient of the next point, w is the grayscale gradient of the adjacent point to the point (x i ,y i ), α is the gradient direction angle; preferably, a calculation method such as a moment method and a fitting method can also be used to achieve the purpose of calculating the sub-pixel edge coordinates;

[0040] Furthermore, the step S600 includes the following steps:

[0041] S601, traverse and calculate the above sub-pixel level contour edge P A The distance to the center of the image is l i and its horizontal angle θ with the image i ;

[0042] S602: Based on the assumption that the center of the image coordinate system is aligned with the center of the camera coordinate system, calculate the corrected contour edge P of the diaphragm object according to the following correction formula: K , to offset the imaging error caused by refraction of optical white glass:

[0043]

[0044] In the formula, h is the thickness of the optical white glass, H is the camera object distance, and f is the refractive index of the optical white glass; further, the present invention is characterized in that the image imaging error caused by the refraction of the optical white glass is offset by the correction formula, thereby effectively improving the contour edge coordinate accuracy of the diaphragm object;

[0045] Further, the S700 includes the following steps:

[0046] S701, based on the following least squares formula, fit the diaphragm edge straight line L O :

[0047]

[0048] In the formula, k and b are the slope and intercept of the least squares fitting line, respectively, and N is the number of edge points to be fitted;

[0049] S702: Calculate the distance between the edge points and the fitting line L according to the following formula: O The degree of dispersion D 2 :

[0050]

[0051] S703, if the dispersion D 2 Less than the preset threshold The deformation degree of the edge to be fitted is low, and the straight line fitted by S701 is used as the approximate contour edge straight line L of the diaphragm. K ;

[0052] S704, if the dispersion D 2 Greater than the preset threshold The fitting straight line L of the diaphragm contour edge is calculated according to the following formula: K :

[0053]

[0054] In the formula, ε is a weight parameter, and M is the number of inner edge points of the fitting line; further, the present invention is characterized in that the lightening degree of the area near the edge of the diaphragm is evaluated according to the discrete degree of the extracted contour edge points, and a weighted fitting strategy is adopted for fitting the diaphragm edge with a high degree of discreteness, so as to enhance the weight of the inner edge points, offset the positioning error caused by the outward deformation of the edge points, and improve the edge line fitting accuracy;

[0055] Further, the S800 includes the following steps:

[0056] S801, calculate the edge fitting straight line L of the diaphragm K Intersection point P r ;

[0057] S802, according to the intersection point P described in S801 r , calculate the length r of each edge of the diaphragm;

[0058] S803, if there is an edge length that exceeds the preset range R, the size of the diaphragm is determined to be unqualified, and the truss machine drives the vacuum adsorption plate to put it into the waste frame, ending this round of loading and re-taking the material;

[0059] Furthermore, the S900 includes the following steps:

[0060] S901. Calculate the position parameters of the diaphragm object in the image coordinate system according to the following formula:

[0061]

[0062] Where △X, △Y, and △Z are the translation and angular displacement of the diaphragm in the image coordinate system, respectively. is the pre-calibrated standard offset, k1, k2, k3, k4 are the slopes of the four edge fitting lines, b1, b2, b3, b4 are the intercepts of the four edge fitting lines;

[0063] S902. Calculate the position parameters of the diaphragm object in the truss machine coordinate system according to the following formula:

[0064]

[0065] Where △O, △U, and △V are the translation and angular offset of the diaphragm in the truss machine coordinate system, k is the camera calibration value, and α is the angle between the image coordinate system and the truss machine coordinate system.

[0066] The present invention is beneficial in that:

[0067] (1) A vacuum adsorption plate is proposed to absorb the diaphragm for loading, which ensures the overall flatness of the diaphragm during the loading process;

[0068] (2) Propose an optical white glass pressed diaphragm solution to reduce the warping of the diaphragm edge and reduce the positioning error caused by local tensile deformation of the diaphragm;

[0069] (3) A non-target area suppression processing algorithm is proposed to effectively suppress the interference of non-membrane edge areas, compress the edge grayscale gradient, and effectively improve the positioning accuracy;

[0070] (4) An edge extraction algorithm based on the center extension search method is proposed. While ensuring pixel-level detection accuracy, it avoids the invalid calculation caused by the traditional traversal method and effectively improves the algorithm detection speed.

[0071] (5) A correction algorithm is proposed to offset the imaging error caused by the refraction of optical white glass, further improving the sub-pixel edge point accuracy and providing effective data for edge line fitting;

[0072] (6) A least squares method using a weighted fitting strategy to fit edge lines is proposed. Based on the cause of small-scale deformation of the diaphragm edge, the discreteness of the edge points is evaluated and edge fitting is performed in the form of assigned weights, which reduces the error caused by the traditional least squares fitting method and effectively improves the positioning accuracy of the algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below.

[0074] Figure 1 It is a flow chart of a first embodiment of a battery diaphragm positioning method based on machine vision of the present invention;

[0075] Figure 2 yes Figure 1 A schematic diagram of an embodiment of step S100;

[0076] Figure 3 yes Figure 1 A flowchart of an embodiment of step S300;

[0077] Figure 4 yes Figure 1 Schematic diagram of an embodiment of step S400 in FIG. DETAILED DESCRIPTION

[0078] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application; obviously, the embodiments described are only part of the embodiments of the present application, rather than all of the embodiments.

[0079] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0080] See also Figure 1 , is a flow chart of an embodiment of a battery diaphragm positioning method based on machine vision of the present application, which specifically includes:

[0081] S100, truss machine drives the vacuum adsorption plate to absorb the diaphragm from the silo, slowly lowering the diaphragm to contact the optical clear glass (see Figure 2 ); Preferably, the gap between the diaphragm and the optical white glass is 0.1 mm;

[0082] S200, industrial camera collects grayscale images and transmits them to the host computer software through TCP / IP protocol;

[0083] S300. Perform non-target area suppression processing on the collected grayscale image; please refer to Figure 3 , and the non-target area suppression processing includes the following steps:

[0084] S301. Perform Zero Padding operation on the image; preferably, the size N of the Padding is 3pix; preferably, the Padding operation can adopt Reflection Padding;

[0085] S302. Traverse the non-padding points in the image at intervals of S according to the following formula:

[0086]

[0087] In the formula, V m,n is the original grayscale value of the current operation point, V m,n ' is the grayscale value after operation of the current operation point, 2N + 1 is the size of the current operation area, and Num(Value<T) is the number of pixel points with grayscale values less than the threshold T in the current operation area; preferably, the traversal interval S = 2pix, and the threshold T = 200;

[0088] S400. Based on the center extension search method, obtain the pixel-level contour edge coordinates of the diaphragm object by interval sampling; please refer to Figure 4 , and the center extension search method includes the following steps:

[0089] S401. Take the center point of the image as the search starting point, and perform single-row / column search traversal in the four directions of up, down, left, and right respectively until a grayscale mutation point is retrieved;

[0090] S402. Take the grayscale mutation point as the secondary search starting point, traverse along both sides in the direction perpendicular to the first search, and apply the Sobel operator to obtain the pixel-level contour edge P O of the diaphragm object; further, the secondary search traversal adopts an interval sampling method; preferably, the sampling interval is 5pix;

[0091] S500. Based on the quadratic polynomial interpolation formula, obtain the sub-pixel-level contour edge P A of the diaphragm object; specifically, the sub-pixel edge calculation includes the following steps:

[0092] S501. Traverse and calculate the grayscale gradients R O near the contour edge P i-1 and R i+1 ;

[0093] S502. According to the following quadratic polynomial interpolation formula, calculate the sub-pixel-level contour edge P A:

[0094]

[0095] In the formula, R i-1 For point (x i ,y i )The gray gradient of the previous point, R i+1 For point (x i ,y i ) is the grayscale gradient of the next point, w is the grayscale gradient of the adjacent point to the point (x i ,y i ), α is the gradient direction angle; preferably, a calculation method such as a moment method and a fitting method can also be used to achieve the purpose of calculating the sub-pixel edge coordinates;

[0096] S600: Based on the refraction correction formula, the above sub-pixel level contour edge coordinates are converged inwardly to obtain the corrected contour edge coordinates of the diaphragm object; specifically, the correction calculation includes the following steps:

[0097] S601, traverse and calculate the above sub-pixel level contour edge P A The distance to the center of the image is l i and its horizontal angle θ with the image i ;

[0098] S602: Based on the assumption that the center of the image coordinate system is aligned with the center of the camera coordinate system, calculate the corrected contour edge P of the diaphragm object according to the following correction formula: K , to offset the imaging error caused by refraction of optical white glass:

[0099]

[0100] Where, h is the thickness of optical white glass, H is the camera object distance, and f is the refractive index of optical white glass;

[0101] S700, based on the least squares method, adopting the inner extreme point fitting strategy to obtain the fitting straight line of the edge of the diaphragm contour; specifically, comprising the following steps:

[0102] S701, based on the following least squares formula, fit the diaphragm edge straight line L O :

[0103]

[0104] In the formula, k and b are the slope and intercept of the least squares fitting line, respectively, and N is the number of edge points to be fitted;

[0105] S702: Calculate the edge point discreteness D according to the following formula: 2 :

[0106]

[0107] S703, if the dispersion D 2 Less than the preset threshold D 2 T , the deformation degree of the edge to be fitted is low, and the straight line fitted by S701 is used as the approximate contour edge straight line L of the diaphragm K ;

[0108] S704, if the dispersion D 2 Greater than the preset threshold D 2 T , then the fitting straight line L of the diaphragm contour edge is calculated according to the following formula K :

[0109]

[0110] Wherein, ε is a weight parameter, and M is the number of edge points inside the fitting line; preferably, the weight parameter ε1=0.78;

[0111] S800, calculating the actual size of the diaphragm based on the fitting straight line, and putting the diaphragm with unqualified size into a waste frame; specifically, the following steps are included:

[0112] S801, calculate the edge fitting straight line L of the diaphragm K Intersection point P r ;

[0113] S802, according to the intersection point P described in S801 r , calculate the length r of each edge of the diaphragm;

[0114] S803, if there is an edge length that exceeds the preset range R, the size of the diaphragm is determined to be unqualified, and the truss machine drives the vacuum adsorption plate to put it into the waste frame, ending this round of loading and re-taking the material;

[0115] S900, based on the offset calculation formula, the current diaphragm position is compared with the pre-calibrated position, the position parameters of the current diaphragm are calculated, and transmitted to the truss machine, thereby completing the diaphragm positioning task; specifically, the calculation of the diaphragm position parameters includes the following steps:

[0116] S901. Calculate the position parameters of the diaphragm object in the image coordinate system according to the following formula:

[0117]

[0118] Where △X, △Y, and △Z are the translation and angular displacement of the diaphragm in the image coordinate system, respectively. is the calibrated standard deviation, k1, k2, k3, k4 are the slopes of the four edge fitting lines, b1, b2, b3, b4 are the intercepts of the four edge fitting lines;

[0119] S902. Calculate the position parameters of the diaphragm object in the truss machine coordinate system according to the following formula:

[0120]

[0121] Where △O, △U, and △V are the translation and angular offset of the diaphragm in the truss machine coordinate system, k is the camera calibration value, and α is the angle between the image coordinate system and the truss machine coordinate system.

[0122] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only preferred examples of the invention and are not intended to limit the present invention. Without departing from the novel spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A visual positioning method for battery separators, the method comprising the following steps: S100, the truss machine drives the vacuum adsorption plate to absorb the diaphragm from the silo and slowly lower it to the diaphragm to contact the optical clear glass; S200, industrial camera collects grayscale images and transmits them to the host computer software through TCP / IP protocol; S300, performing non-target area suppression processing on the collected grayscale image; S400, based on the center extension search method, interval sampling to obtain the pixel-level contour edge coordinates of the diaphragm object; S500, obtaining sub-pixel level contour edge coordinates of the diaphragm object based on a quadratic polynomial interpolation formula; S600, based on the refraction correction formula, the above sub-pixel level contour edge coordinates are converged inwardly to obtain the corrected contour edge coordinates of the diaphragm object; S700, based on the least square method, adopting a weighted fitting strategy to obtain a fitting straight line of the edge of the diaphragm contour; S800, calculating the actual size of the diaphragm based on the fitted straight line, and putting the diaphragm with unqualified size into a waste frame; S900, based on the offset calculation formula, the current diaphragm position is compared with the pre-calibrated position, the position parameters of the current diaphragm are calculated, and transmitted to the truss machine, thereby completing the diaphragm positioning task; The S700 includes the following steps: S701, based on the following least squares formula, fit the diaphragm edge straight line : ； In the formula, are the slope and intercept of the least squares fitting line, is the number of edge points to be fitted; S702: Calculate the distance between the edge points and the fitting straight line according to the following formula: The degree of dispersion : ； S703, if the discrete Less than the preset threshold , the deformation degree of the edge to be fitted is low, and the straight line fitted by S701 is used as the approximate contour edge straight line of the diaphragm ; S704, if the discrete Greater than the preset threshold , then the fitting straight line of the diaphragm contour edge is calculated according to the following formula : ； In the formula, is the weight parameter, The number of edge points on the inner side of the fitting line; the lightening degree of the area near the edge of the diaphragm is evaluated according to the discreteness of the extracted contour edge points. For the diaphragm edge with a high degree of discreteness, a weighted fitting strategy is used to fit, enhance the weight of the inner edge points, offset the positioning error caused by the outward deformation of the edge points, and improve the edge straight line fitting accuracy; The S900 includes the following steps: S901. Calculate the position parameters of the diaphragm object in the image coordinate system according to the following formula: ； In the formula, are the translation and angular offset of the diaphragm in the image coordinate system, is the pre-calibrated standard offset, are the slopes of the four edge fitting lines, are the intercepts of the four edge fitting lines respectively; S902. Calculate the position parameters of the diaphragm object in the truss machine coordinate system according to the following formula: ； In the formula, are the translation and angular offset of the diaphragm in the truss machine coordinate system, Camera calibration values, is the angle between the image coordinate system and the truss machine coordinate system.

2. The visual positioning method for battery separator according to claim 1, characterized in that: In the step S100, adsorption holes are distributed on one side of the adsorption surface of the vacuum adsorption plate, and a main cavity and a branch cavity are distributed inside the plate body of the vacuum adsorption plate; the branch cavities correspond to the adsorption holes one by one; the S100 is also characterized in that the local tensile deformation of the diaphragm is resisted by pressing the diaphragm to the surface of the optical white glass; the optical white glass is also characterized in that exhaust holes are distributed on its surface, which destroy the continuity of the smooth surface and reduce the contact area with the diaphragm, thereby reducing the vacuum negative pressure and surface adhesion force on the diaphragm when the vacuum adsorption plate is lifted.

3. The visual positioning method for battery separator according to claim 1, characterized in that: The S300 includes the following sub-steps: S301, performing a Zero Padding operation on the image to prevent subsequent operations from changing the size of the original image; the Padding operation adopts Reflection Padding, which copies the pixel values ​​of the existing edge area by mirroring, to prevent occasional highlight pixels on the edge of the original image from affecting the suppression effect; S302: According to the following formula, the non-filled points in the image are spaced Perform traversal operations: ； In the formula, is the original gray value of the current operating point, is the gray value after operation at the current operating point, is the current operating area size, The gray value in the current operation area is less than the threshold The number of pixels in the traversal interval; The number of pixels in the traversal interval ; Through the above traversal operation, the interference of the background and the non-important pixel area inside the diaphragm is suppressed as much as possible while retaining the important information area near the diaphragm object; this operation compresses the edge grayscale length of the diaphragm and sharpens the contour edge features, providing a more ideal grayscale model for subsequent edge extraction operations; S303, based on the connected domain analysis method, obtaining the number of connected domains in the current image; S304. If the number of connected domains in the current image is 0, it is determined that there is no high-brightness target and no barrier object in the image, an ERROR alarm signal is output, and the detection task is terminated; the number of connected domains is used as the judgment condition to evaluate the suppression effect of non-target areas in this round, and detect whether the image contains valid barrier objects.

4. The visual positioning method for battery separator according to claim 1, characterized in that: The step S304 further includes the following steps: S305, if the number of connected domains in the current image is greater than 1, repeat steps S301 to S303; S306. If the number of connected domains in the current image is equal to 1, there is no bright spot interference in the non-target area, and the current loop ends; the non-target area suppression processing is executed cyclically, with the number of connected domains in the image as the judgment condition, to ensure that all non-membrane objects in the image are effectively suppressed and the grayscale feature details related to the membrane objects are effectively retained.

5. The visual positioning method for battery separator according to claim 1, characterized in that: The S400 includes the following steps: S401, taking the center point of the image as the search starting point, performing single row / column search traversal in four directions, namely, up, down, left, and right, until a grayscale mutation point is retrieved; S402, taking the grayscale mutation point as the starting point of the second search, traversing along both sides perpendicular to the first search direction, and applying the Sobel operator to obtain the pixel-level contour edge of the diaphragm object ; The secondary search traversal adopts the interval sampling method; according to the characteristics of the diaphragm image, the edge search strategy is designed with the image center as the starting point, which reduces the invalid traversal area by more than 85%. Under the premise of ensuring the edge detection accuracy, it effectively reduces the algorithm time consumption and improves the detection performance.

6. The visual positioning method for battery separator according to claim 1, characterized in that: The S500 includes the following steps: S501, traverse and calculate contour edges Grayscale gradient nearby and ; S502: Calculate the sub-pixel level contour edge of the diaphragm object according to the following quadratic polynomial interpolation formula: : ； In the formula, For point The grayscale gradient of the previous point, For point The grayscale gradient of the next point, Adjacent point to point The distance is the gradient direction angle; the moment method and fitting method are used to achieve the purpose of calculating the sub-pixel edge coordinates.

7. The visual positioning method for battery separator according to claim 1, characterized in that: The S600 includes the following steps: S601, traverse and calculate the above sub-pixel level contour edge Distance to the center of the image and its horizontal angle with the image ; S602: Based on the assumption that the center of the image coordinate system is aligned with the center of the camera coordinate system, the corrected contour edge of the diaphragm object is calculated according to the following correction formula: , to offset the imaging error caused by refraction of optical white glass: ； In the formula, is the thickness of optical clear glass, is the camera object distance, is the refractive index of optical white glass; through the correction formula, the image imaging error caused by the refraction of optical white glass is offset, and the contour edge coordinate accuracy of the diaphragm object is effectively improved.

8. The visual positioning method for battery separator according to claim 1, characterized in that: The S800 includes the following steps: S801, calculate the edge fitting straight line of the diaphragm Intersection ; S802, according to the intersection point described in S801 , calculate the length of each edge of the diaphragm ; S803: If the edge length exceeds the preset range , the diaphragm size is judged to be unqualified, and the truss machine drives the vacuum adsorption plate to put it into the waste frame, ending this round of loading and re-taking the material.

Images