Visual inspection equipment for intelligent manufacturing industrial machine

Abstract

The invention discloses intelligent manufacturing industrial machine visual inspection equipment, which belongs to the technical field of visual inspection, and comprises the following structures: an equipment platform, which comprises a mechanical conveying table and a detection platform; the linear array industrial camera is located above the detection platform, and a strip-shaped detection light source is arranged above the linear array industrial camera; the detection system comprises a pulse encoder, an image acquisition card and an industrial processor, vamp images are scanned and acquired through a linear array industrial camera, algorithm detection and automatic identification are performed on image data through the image acquisition card and the industrial processor, qualified vamps are automatically detected and identified, and quantity statistics and identification are performed on unqualified vamps; manual visual inspection and statistical screening are replaced, so that full-automatic vamp defect detection is realized.

Description

Technical Field

[0001] The present invention discloses a visual inspection technology, and in particular relates to a visual inspection device for intelligent manufacturing industrial machines. Background Art

[0002] Shoes are designed to protect the feet, and their structure is extremely complex. A pair of shoes typically consists of an upper, sole, insole, and last. The upper is the primary component of the shoe's appearance, wrapping around the foot. When crafting the upper, factors such as material strength, wear resistance, breathability, and comfort must be considered.

[0003] Existing shoemaking technology has a low degree of automation, and many process flows still rely on manual operation and direct visual observation and judgment. This is especially true in the upper quality inspection process, where manual inspection is required to observe the uppers for defects in real time and to immediately mark, screen, and count them. This process is labor-intensive and has low inspection efficiency. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent manufacturing industrial machine vision inspection device in order to solve the above problems.

[0005] To achieve the above objectives, the present invention provides the following technical solutions: an intelligent manufacturing industrial machine vision inspection device, comprising the following structure:

[0006] Equipment platform, which includes a mechanical conveying platform and a detection platform;

[0007] A linear array industrial camera is located above the inspection platform and has a strip-shaped inspection light source above it;

[0008] A detection system including a pulse encoder, an image acquisition card and an industrial processor;

[0009] The detection system includes a counting algorithm and a defect detection algorithm. The counting algorithm and the defect detection algorithm adopt a template library model matching algorithm. The process is as follows:

[0010] S1. Load the image features of the sample library. A shoe upper design sample contains N basic sample images. Each basic sample image feature is a pixel array with a length and width of i×j, represented as a two-dimensional array. Connect the pixels row by row as a one-dimensional array, and create a new two-dimensional array R[n][m] to record the image features in the sample library, where m = i×j. In the new two-dimensional array, each row represents all the pixel information of a basic sample image, and each column represents the pixel information at the same position of different basic sample images.

[0011] S2, data decentralization processing, decentralized the two-dimensional array R, calculated the empirical mean along each row of the two-dimensional array R, and subtracted the calculated empirical mean from each column of the two-dimensional array R to obtain the data X;

[0012] S3, calculate the covariance matrix C of the data X and its eigenvalues and eigenvectors

[0013]

[0014] Among them, X T is the transposed vector;

[0015] C=U∑V T

[0016] Among them, U is an n-order square matrix that satisfies UU T =I unitary matrix, V is an m-order square matrix, satisfying VV T =I, ∑ is an n×m-order matrix. In the ∑ matrix, except for the diagonal elements, the remaining elements are 0, and the diagonal element values are the singular values σ i Arrange from largest to smallest;

[0017] CC T =U∑V T (U∑V T ) T =U∑ 2 U T

[0018] C T C=(U∑V T ) T U∑V T =V∑ 2 V T

[0019] Among them, U and V are C T C and CC T The eigenvector of 2 The target line element is C T C and CC T The characteristic value of

[0020] In U, define the column vector u i Is a left singular vector, in V, define the column vector v i is the right singular vector, and the following formula is obtained:

[0021] (CC T )u i =λ i u i

[0022] (C T C)v i =λ i v i

[0023] Among them, λi C T C and CC T The characteristic value of

[0024] At the same time, the formula Solve for the singular value σ i

[0025] S4, determine the principal components, sort the eigenvalues in the covariance matrix C in descending order, and select the eigenvectors corresponding to the first k eigenvalues as the principal components;

[0026] S5, load the eigenvector space, form an m×k matrix S by combining the two-dimensional array R and the eigenvectors containing the principal components in S4, and multiply R and S to obtain the n×k eigenvector space;

[0027] S6, solving the minimum Euclidean distance, collecting a basic sample image of the shoe upper to be tested, connecting the pixels row by row as the one-dimensional vector to be tested, calculating the Euclidean distance between the one-dimensional vector to be tested and each row vector in the feature vector space row by row, and arranging them in ascending order to obtain the minimum Euclidean distance;

[0028] S7, select the corresponding target sample. If the Euclidean distance between a certain one-dimensional vector to be tested and the characteristic row vector is the smallest, then the one-dimensional vector to be tested is selected as the target sample R0.

[0029] Preferably, the template library model matching algorithm is detected by normalized correlation matching method, and its process is as follows:

[0030] M1 scans the shoe surface to be inspected through a linear array industrial camera, outputs the captured image P to the industrial processor through the image acquisition card, performs a pixel-by-pixel measurement calculation on each position of the captured image P, and generates the matrix R according to the measurement value through S1. P ;

[0031] M2 sets the limit of successful matching based on the target sample R0 in S7, and then uses the minMaxLoc function to locate and output the matrix R P The position coordinates must be no less than R0, and the number of position coordinates that meet the conditions must be counted;

[0032] M3, through pixel multiplication operation on target samples R0 and R P Perform correlation calculation, set the pixel coordinate position of the target sample R0 to I(x,y) and the correlation result to R(x,y), and the correlation calculation formula is as follows

[0033]

[0034] Wherein, T(x',y') is the pixel position coordinate of the basic sample image, and I(x+x',y+y') is the pixel position coordinate of the acquired image P;

[0035] M4, if R(x,y)≥R0, it means that the captured image P meets the matching condition with the target sample R0, then the shoe uppers that meet the matching condition and the statistical number are recorded;

[0036] If R(x, y) < R0, it means that the captured image P does not meet the matching conditions with the target sample R0, and the shoe uppers that do not meet the matching conditions and the statistical number are recorded.

[0037] Preferably, the detection system controls the triggering of the linear array industrial camera through a pulse encoder, and collects the shoe upper to be detected through pulse frequency doubling and pulse discarding scanning of the pulse encoder. If the shoe upper to be detected exceeds the scanning height of the linear array industrial camera, the images scanned by the front and rear units of the linear array industrial camera are spliced row by row.

[0038] Preferably, the mechanical conveying platform includes several pallets that are lifted and lowered by a lifting frame, the pallets are movably connected to the lifting frame through rails, several separated compartments are provided on the pallets, the compartments are used to place the shoe uppers to be tested, the testing platform is provided with a manipulator that is retracted and extended by a cylinder, the manipulator is provided with a card block, and a card slot that cooperates with the card block is provided under the pallet.

[0039] Preferably, a marking device is provided on the side of the detection platform and is moved by a sliding platform and a lifting platform.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] The shoe upper image is scanned and collected by a linear array industrial camera, and the image data is subjected to algorithm detection and automatic recognition through an image acquisition card and an industrial processor. Qualified shoe uppers are automatically detected and identified, and unqualified shoe uppers are counted and marked. This replaces manual visual observation and statistical screening to achieve fully automated shoe upper defect detection.

[0042] The detection platform includes a gravity sensing system, which can perform preliminary screening of the weight of the shoe uppers. It can calibrate within an accuracy of ±5g to screen out unqualified shoe uppers from the same batch for rework or removal. It also has a batch identification function. The pallet is affixed with an identification QR code. The detection platform has a scanning device. When the pallet passes by, it can identify the batch, quantity, model and other information corresponding to the shoe uppers on the group of pallets, and select the corresponding sample library mode according to the preset parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart of the sample library model matching algorithm;

[0044] Figure 2This is a structural diagram of an intelligent manufacturing industrial machine vision inspection equipment;

[0045] Figure 3 It is a partial schematic diagram of the robot and pallet;

[0046] Figure 4 is a schematic diagram of the eigenvector space;

[0047] Figure 5 This is a demonstration of stitching images scanned by the front and rear units of a linear array industrial camera in rows;

[0048] Figure 6 This is a detection demonstration diagram of the sample library model matching algorithm using the normalized correlation matching method.

[0049] Figure numerals: 1. lifting frame; 2. tray; 3. track; 4. mechanical claw; 5. lifting platform; 6. sliding platform; 7. slot; 8. block. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, they cannot be understood as limitations on the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0051] Example 1

[0052] An intelligent manufacturing industrial machine vision inspection equipment, such as Figure 2-Figure 3 As shown, it includes the following structures:

[0053] Equipment platform, which includes a mechanical conveying platform and a detection platform;

[0054] A linear array industrial camera (not shown in the accompanying drawings) is located above the detection platform and has a strip-shaped detection light source above it;

[0055] A detection system including a pulse encoder, an image acquisition card and an industrial processor;

[0056] like Figure 5As shown, the inspection system controls the triggering of the linear array industrial camera through a pulse encoder, and scans and collects the shoe upper to be inspected by pulse frequency multiplication and pulse discarding of the pulse encoder. If the shoe upper to be inspected exceeds the scanning height of the linear array industrial camera, the images scanned by the front and rear units of the linear array industrial camera are spliced row by row. To avoid repeated matching, the previous shoe upper image is discarded, and the next shoe upper image is spliced with the next shoe upper image collected, and then matching is performed;

[0057] The mechanical conveyor platform includes a plurality of trays 2 that are raised and lowered by a lifting frame 1. The trays 2 are movably connected to the lifting frame 1 via a track 3. The trays 2 are provided with a plurality of compartments for placing the shoe uppers to be tested. The testing platform is provided with a manipulator 4 that is retracted and extended by a cylinder. The manipulator 4 is provided with a clamping block 8. A clamping slot 7 that cooperates with the clamping block 8 is provided under the tray 2.

[0058] During the inspection, the lifting frame 1 lifts the pre-inspected pallet 2 to the corresponding height, and the manipulator 4 extends to the slot 7 below the pallet 2 and drags it along the track 3 to the inspection platform, and the inspection system of the inspection platform starts working;

[0059] A marking device is provided on the side of the detection platform, which is moved by the sliding platform 6 and the lifting platform 5. The marking device will move to the shoe upper that fails the inspection to mark and issue an alarm.

[0060] The detection platform includes a gravity sensing system, which can perform preliminary screening of the weight of the shoe uppers. It can calibrate within an accuracy of ±5g to screen out unqualified shoe uppers from the same batch for rework or removal, and has a batch identification function. An identification QR code is affixed to tray 2. The detection platform has a scanning device. When tray 2 passes by, it can identify the batch, quantity, model and other information corresponding to the shoe uppers on this group of trays 2, and select the corresponding sample library mode according to the preset parameters.

[0061] Example 2

[0062] The detection system includes counting algorithm and defect detection algorithm, which adopts template library model matching algorithm, such as Figure 1 As shown, the process is as follows:

[0063] S1. Load the image features of the sample library. A shoe upper design sample contains N basic sample images. Each basic sample image feature is a pixel array with a length and width of i×j, represented as a two-dimensional array. Connect the pixels row by row as a one-dimensional array, and create a new two-dimensional array R[n][m] to record the image features in the sample library, where m = i×j. In the new two-dimensional array, each row represents all the pixel information of a basic sample image, and each column represents the pixel information at the same position of different basic sample images.

[0064] S2, data decentralization processing. Due to the large number of images in the sample library and the large value of N, in order to reduce the amount of calculation, the two-dimensional array R needs to be decentralized. The empirical mean is calculated along each row of the two-dimensional array R, and the calculated empirical mean is subtracted from each column of the two-dimensional array R to obtain the data X;

[0065] S3, calculate the covariance matrix C of the data X and its eigenvalues and eigenvectors

[0066]

[0067] Among them, X T is the transposed vector;

[0068] C=U∑V T

[0069] Among them, U is an n-order square matrix that satisfies UU T =I unitary matrix, V is an m-order square matrix, satisfying VV T =I, ∑ is an n×m-order matrix. In the ∑ matrix, except for the diagonal elements, the remaining elements are 0, and the diagonal element values are the singular values σ i Arrange from largest to smallest;

[0070] CC T =U∑V T (U∑V T ) T =U∑ 2 U T

[0071] C T C=(U∑V T ) T U∑V T =V∑ 2 V T

[0072] Among them, U and V are C T C and CC T The eigenvector of 2 The target line element is C T C and CC T The characteristic value of

[0073] In U, define the column vector u i Is a left singular vector, in V, define the column vector v i is the right singular vector, and the following formula is obtained:

[0074] (CC T )u i =λ i u i

[0075] (C T C)v i =λ i v i

[0076] Among them, λ i C T C and CC T The characteristic value of

[0077] At the same time, the formula Solve for the singular value σ i

[0078] S4, determining the principal components, arranging the eigenvalues in the covariance matrix C in descending order, and selecting the eigenvectors corresponding to the first k eigenvalues as the principal components, where k is 10% of the total number, and the sum of the first k eigenvalues can account for the majority of the sum of all eigenvalues;

[0079] S5, load the eigenvector space, form an m×k matrix S by combining the two-dimensional array R and the eigenvectors containing the principal components in S4, multiply R and S, and project the original target sample points onto the axes corresponding to the eigenvectors to obtain the PCA reduced dimensionality eigenvector space n×k. Any row vector in the reduced dimensionality eigenvector space n×k corresponds to a basic sample image, such as Figure 4 As shown;

[0080] S6, solving the minimum Euclidean distance, collecting a basic sample image of the shoe upper to be tested, connecting the pixels row by row as the one-dimensional vector to be tested, calculating the Euclidean distance between the one-dimensional vector to be tested and each row vector in the feature vector space row by row, and arranging them in ascending order to obtain the minimum Euclidean distance;

[0081] S7, select the corresponding target sample. The row vector in the feature vector space is represented as a basic sample image. If the Euclidean distance between the one-dimensional vector to be tested and the feature row vector is the smallest, it means that the shoe upper image to be tested is closest to the basic sample image corresponding to the feature row vector, and the two are the best match. Then the one-dimensional vector to be tested is selected as the target sample R0;

[0082] Example 3

[0083] like Figure 6 As shown, the template library model matching algorithm uses the normalized correlation matching method for detection, and its process is as follows:

[0084] M1 scans the shoe surface to be inspected through a linear array industrial camera, outputs the captured image P to the industrial processor through the image acquisition card, performs a pixel-by-pixel measurement calculation on each position of the captured image P, and generates the matrix R according to the measurement value through S1. P ;

[0085] M2 sets the limit of successful matching based on the target sample R0 in S7, and then uses the minMaxLoc function to locate and output the matrix R P The position coordinates must be no less than R0, and the number of position coordinates that meet the conditions must be counted;

[0086] M3, through pixel multiplication operation on target samples R0 and R P Perform correlation calculation, set the pixel coordinate position of the target sample R0 to I(x,y) and the correlation result to R(x,y), and the correlation calculation formula is as follows

[0087]

[0088] Where T(x',y') is the pixel position coordinate of the basic sample image, and I(x+x',y+y') is the pixel position coordinate of the collected image P. The larger the matching result R(x,y), the better the matching degree.

[0089] M4, if R(x,y)≥R0, it means that the captured image P meets the matching conditions with the target sample R0, then the shoe uppers that meet the matching conditions and the statistical number are recorded;

[0090] If R(x, y) < R0, it means that the captured image P does not meet the matching conditions with the target sample R0, and the shoe uppers that do not meet the matching conditions and the statistical number are recorded.

[0091] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

[0092] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. An intelligent manufacturing industrial machine vision inspection equipment, characterized in that: Includes the following structures: Equipment platform, which includes a mechanical conveying platform and a detection platform; A linear array industrial camera is located above the inspection platform and has a strip-shaped inspection light source above it; A detection system including a pulse encoder, an image acquisition card and an industrial processor; The detection system includes a counting algorithm and a defect detection algorithm. The counting algorithm and the defect detection algorithm adopt a template library model matching algorithm. The process is as follows: S1. Load the image features of the sample library. A shoe upper design sample contains N basic sample images. Each basic sample image feature is a pixel array with a length and width of i×j, represented as a two-dimensional array. Connect the pixels row by row as a one-dimensional array, and create a new two-dimensional array R[n][m] to record the image features in the sample library, where m = i×j. In the new two-dimensional array, each row represents all the pixel information of a basic sample image, and each column represents the pixel information at the same position of different basic sample images. S2, data decentralization processing, decentralized the two-dimensional array R, calculated the empirical mean along each row of the two-dimensional array R, and subtracted the calculated empirical mean from each column of the two-dimensional array R to obtain the data X; S3, calculate the covariance matrix C of the data X and its eigenvalues and eigenvectors Among them, X T is the transposed vector; C=UΣV T Among them, U is an n-order square matrix that satisfies UU T =I unitary matrix, V is an m-order square matrix, satisfying VV T =I, Σ is an n×m matrix. In the Σ matrix, except for the diagonal elements, the remaining elements are 0, and the diagonal element values are the singular values σ i Arrange from largest to smallest; CC T =U∑V T (U∑V T ) T =U∑ 2 U T C T C=(U∑V T ) T U∑V T =V∑ 2 V T Among them, U and V are C T C and CC T The characteristic vector of ∑2 object line element is C T C and CC T The characteristic value of In U, define the column vector u i Is a left singular vector, in V, define the column vector v i is the right singular vector, and the following formula is obtained: (CC T )u i =λ i u i (C T C)v i =λ i v i Among them, λ i C T C and CC T The characteristic value of At the same time, the formula Solve for the singular value σ i S4, determine the principal components, sort the eigenvalues in the covariance matrix C in descending order, and select the eigenvectors corresponding to the first k eigenvalues as the principal components; S5, load the eigenvector space, form an m×k matrix S by combining the two-dimensional array R and the eigenvectors containing the principal components in S4, and multiply R and S to obtain the n×k eigenvector space; S6, solving the minimum Euclidean distance, collecting a basic sample image of the shoe upper to be tested, connecting the pixels row by row as the one-dimensional vector to be tested, calculating the Euclidean distance between the one-dimensional vector to be tested and each row vector in the feature vector space row by row, and arranging them in ascending order to obtain the minimum Euclidean distance; S7, select the corresponding target sample. If the Euclidean distance between a certain one-dimensional vector to be tested and the characteristic row vector is the smallest, then the one-dimensional vector to be tested is selected as the target sample R0.

2. The intelligent manufacturing industrial machine vision inspection equipment according to claim 1, characterized in that: The template library model matching algorithm is detected by normalized correlation matching method, and its process is as follows: M1 scans the shoe surface to be inspected through a linear array industrial camera, outputs the captured image P to the industrial processor through the image acquisition card, performs a pixel-by-pixel measurement calculation on each position of the captured image P, and generates the matrix R according to the measurement value through S1. P ; M2 sets the limit of successful matching based on the target sample R0 in S7, and then uses the minMaxLoc function to locate and output the matrix R P The position coordinates must be no less than R0, and the number of position coordinates that meet the conditions must be counted; M3, through pixel multiplication operation on target samples R0 and R P Perform correlation calculation, set the pixel coordinate position of the target sample R0 to I(x,y) and the correlation result to R(x,y), and the correlation calculation formula is as follows Wherein, T(x',y') is the pixel position coordinate of the basic sample image, and I(x+x',y+y') is the pixel position coordinate of the acquired image P; M4, if R(x,y)≥R0, it means that the captured image P meets the matching condition with the target sample R0, then the shoe uppers that meet the matching condition and the statistical number are recorded; If R(x, y) < R0, it means that the captured image P does not meet the matching conditions with the target sample R0, and the shoe uppers that do not meet the matching conditions and the statistical number are recorded.

3. The intelligent manufacturing industrial machine vision inspection equipment according to claim 2, characterized in that: The detection system controls the triggering of the linear array industrial camera through a pulse encoder, and scans and collects the shoe upper to be inspected through pulse frequency doubling and pulse discarding of the pulse encoder. If the shoe upper to be inspected exceeds the scanning height of the linear array industrial camera, the images scanned by the front and rear units of the linear array industrial camera are spliced row by row.

4. The intelligent manufacturing industrial machine vision inspection equipment according to claim 3, characterized in that: The mechanical conveying platform comprises a plurality of pallets (2) which are lifted and lowered by a lifting frame (1); the pallets (2) are movably connected to the lifting frame (1) via a track (3); a plurality of separated compartments are provided on the pallets (2); the compartments are used to place shoe uppers to be tested; the testing platform is provided with a manipulator (4) which is retracted and extended by a cylinder; the manipulator (4) is provided with a card block (8); and a card slot (7) which cooperates with the card block (8) is provided below the pallet (2).

5. The intelligent manufacturing industrial machine vision inspection equipment according to claim 4, characterized in that: A marking device is provided on the side of the detection platform and is moved via a sliding platform (6) and a lifting platform (5).

Images