A method and system for intelligently identifying production defects of seals

Abstract

The present invention belongs to the field of image processing technology, and specifically relates to a method and system for intelligently identifying production defects of seals, the method comprising: determining the overall grayscale retention rate of the pixel points according to the grayscale difference between the pixel points in the seal image and the corresponding pixel points in the low-frequency components of each level after wavelet transformation, determining the detail loss degree of the pixel points at different levels according to the element value change of the element corresponding to the pixel points in the detail components of different levels, determining the correction index of the element corresponding to the pixel points in the detail components of each level in combination with the overall grayscale retention rate, correcting the elements in the detail components to obtain the corrected detail components, reconstructing the seal enhanced image according to the low-frequency components of each level and the corrected detail components, thereby identifying the production defects of the seals. The present invention has a good effect on enhancing the details of defects that are not obvious on the seals, and improves the accuracy of identifying production defects of seals.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more specifically, to a method and system for intelligently identifying production defects of sealing parts. Background Art

[0002] As an important mechanical component, the production quality of seals directly affects the normal operation and service life of the equipment, so it is crucial to identify defects in the production process in real time. Traditional defect detection methods usually rely on basic image processing technology. However, the production process of seals often produces a variety of tiny defects, such as surface scratches, bubbles, cracks, glue deficiency, impurities, mold sticking, etc. These defects may be difficult to detect in the seal image. In order to improve the accuracy and automation level of defect detection, it is particularly necessary to enhance the seal image.

[0003] In the related art, for example, a Chinese patent application document with application publication number CN111583123A discloses an image enhancement algorithm based on wavelet transform that fuses high and low frequency information, including the following steps: Step S1: input a color image into a computer and convert it into a grayscale image, perform a two-layer wavelet decomposition of a two-dimensional signal on the grayscale image, and convert the image from the spatial domain to the frequency domain; Step S2: extract the low-frequency component of the image, and use homomorphic filtering to improve the high-frequency response in the low-frequency component; Step S3: transform the filtered low-frequency image into the spatial domain for grayscale transformation to enhance the image contrast and brightness; Step S4: reconstruct the image based on the enhanced low-frequency component and high-frequency component coefficients, perform threshold denoising based on wavelet transform on the reconstructed image, and finally obtain an enhanced image.

[0004] The above schemes can enhance the low-frequency information of the image, but the enhancement effect is poor for the tiny defects that are difficult to detect in the seal image, which may result in the inability to detect or identify production defects such as scratches, bubbles, cracks, glue deficiency, impurities, and mold sticking on the seal surface. Summary of the invention

[0005] In order to solve the technical problem that the production defects such as scratches, bubbles, cracks, glue deficiency, impurities, and mold sticking on the surface of the above-mentioned seals are difficult to detect in the image, which affects the defect identification, the present invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a method for intelligently identifying production defects of a seal, comprising:

[0007] Acquire a seal image; perform wavelet transform on the seal image to obtain low-frequency components and detail components at different levels; determine the overall grayscale retention rate of each pixel in the seal image based on the grayscale difference between the corresponding pixel in the low-frequency component at each level; determine the detail loss degree of each pixel at different levels based on the change in the element value of the corresponding element in the detail component at different levels; determine the correction index of the corresponding element of each pixel in the detail component at each level based on the overall grayscale retention rate and the detail loss; correct each element in the detail component based on all the correction indexes of each element to obtain the corrected detail component; reconstruct the seal enhanced image based on the low-frequency components at each level and the corrected detail components, and identify the production defects of the seal based on the seal enhanced image.

[0008] The present invention decomposes the seal image into low-frequency components and detail components of different levels, which can better extract the main features and minor differences of the seal image. The overall grayscale retention rate is determined by analyzing the grayscale difference between each pixel in the seal image and the corresponding pixel in the low-frequency component. The overall grayscale retention rate can be used to distinguish the main grayscale features in the seal image from textures, cracks, lack of glue, impurities, sticky molds, bubbles, scratches and other features with a relatively small grayscale ratio. The present invention calculates the changes of each pixel in the detail components of different levels and determines its detail loss, which can accurately identify the obvious degree of different details in the seal image. Subsequently, the correction index is set in combination with the detail loss and the overall grayscale retention rate, which can focus on enhancing the features such as inconspicuous textures, cracks, lack of glue, impurities, sticky molds, bubbles, scratches, etc., highlighting the minor defects in the seal image, reducing the difficulty of seal production defect detection, and improving the accuracy and efficiency of production defect detection.

[0009] Preferably, determining the overall grayscale retention rate of each pixel includes: negatively normalizing the grayscale difference between the pixel in the seal image and the corresponding pixel in the low-frequency component of each level as the grayscale retention rate of the pixel in the seal image in the low-frequency component of each level; weighted summing the grayscale retention rates of the pixel in the seal image in the low-frequency component of each level to obtain the overall grayscale retention rate of the pixel in the seal image.

[0010] The present invention uses the overall grayscale retention rate to reflect the grayscale information retention of each pixel in the seal image during the continuous downsampling process. The overall grayscale retention rate can be used to distinguish the main grayscale features in the seal image from texture, cracks, glue deficiency, impurities, mold sticking, bubbles, scratches and other features with a smaller grayscale ratio, providing a more accurate basis for subsequent image enhancement and defect identification.

[0011] Preferably, the grayscale retention rate of the pixel in the low-frequency component of each level satisfies the expression: ;in, Indicates the seal image The pixel at Grayscale retention rate in low-frequency components of each level; Indicates the seal image The gray value of each pixel; Indicates the seal image The pixel at The gray value of the corresponding pixel in the low-frequency component of each level; To represent the extreme grayscale in the seal image; is an exponential function with a natural constant as its base.

[0012] Preferably, the overall grayscale retention rate satisfies the expression: ;in, Indicates the seal image The overall grayscale retention rate of pixels; The serial number indicating the level; Indicates the total number of levels; Indicates the seal image The pixel at Grayscale retention rate in the low-frequency components of each level.

[0013] Preferably, the detail loss satisfies the expression:

[0014] ; The directions of the detail components are horizontal, vertical and diagonal, and any one of them is taken as the target direction. Indicates the seal image The pixel at The detail loss in the target direction at each level; , , Respectively represent the seal image Pixels at the first level, level, The corresponding elements in the detail components of the target direction at each level, , Indicates the level number. , , Indicates the total number of levels; represents the maximum value function, Represents the minimum function.

[0015] The present invention uses detail loss to quantify the loss of detail components at different levels for each pixel in the seal image. The detail loss reflects whether the details corresponding to the pixel in the seal image are clear, providing a basis for subsequent image enhancement and defect detection.

[0016] Preferably, the correction index of the corresponding element of each pixel point in the detail component of each level satisfies the expression: ;in, Indicates the seal image The pixel at The correction index of the corresponding element in the detail component of the target direction of each level; Indicates the seal image The pixel at The detail loss in the target direction at each level; Indicates the seal image The target direction is any one of the horizontal direction, the vertical direction and the diagonal direction.

[0017] The present invention sets a larger correction index for the elements corresponding to pixels with large detail loss and small grayscale retention rate in detail images at different levels, which can focus on enhancing subtle textures, cracks, glue deficiency, impurities, mold sticking, bubbles, scratches and other features, which is beneficial to improving the accuracy of subsequent production defect detection.

[0018] Preferably, the step of correcting each element according to all correction indices of each element in the detail component to obtain the corrected detail component includes: taking any one element in the detail component of the target direction of each level as the target element, the target element corresponding to pixels, then the target element has Correction index; correction value of the target element Satisfies the expression: ; Represents the element value of the target element; Indicates the target element A correction index, The serial number representing the correction index of the target element; the correction value of each element in the detail component of each direction at each level constitutes the corrected detail component of each direction at each level.

[0019] Preferably, identifying production defects of seals based on seal enhanced images includes: inputting the seal enhanced images into a trained neural network, and outputting production defects of seals, where types of production defects include cracks, glue deficiency, debonding, impurities, flash, and mold sticking.

[0020] The details in the enhanced seal image are clearer than those in the image before enhancement, and the results of identifying production defects based on the enhanced seal image are more accurate.

[0021] Preferably, the neural network is a YOLOv5 network.

[0022] In a second aspect, the present invention provides a system for intelligently identifying production defects of seals, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned method for intelligently identifying production defects of seals is implemented.

[0023] By adopting the above technical solution, the above-mentioned method for intelligently identifying production defects of a seal is generated into a computer program and stored in a memory so as to be loaded and executed by a processor, thereby making a terminal device based on the memory and the processor for easy use.

[0024] The beneficial effects of the present invention are as follows: the present invention can extract the main features and minor differences of the seal image, and distinguish the main grayscale features in the seal image from the texture, cracks, lack of glue, impurities, sticking mold, bubbles, scratches and other features with a relatively small grayscale ratio. The detail loss degree can be used to accurately identify the degree of prominence of different details in the seal image. By setting the correction index in combination with the detail loss degree and the overall grayscale retention rate, it is possible to focus on enhancing the unobvious features such as texture, cracks, lack of glue, impurities, sticking mold, bubbles, scratches, etc., highlighting the minor defects in the seal image, reducing the difficulty of detecting production defects of seals, and improving the accuracy and efficiency of production defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0026] Figure 1 is a flow chart schematically illustrating a method for intelligently identifying production defects of a seal in the present invention;

[0027] Figure 2 is a schematic diagram schematically showing an image of a seal;

[0028] Figure 3 is a flow chart schematically showing step S2 in a method for intelligently identifying production defects of a seal in the present invention;

[0029] Figure 4 is a schematic diagram schematically illustrating Haar wavelet transform decomposition;

[0030] Figure 5 is a visualization image schematically showing the low-frequency components of the first level;

[0031] Figure 6 is a visualization image schematically showing the low-frequency components of the second level;

[0032] Figure 7 is a visualization image schematically showing a detail component in the horizontal direction of the first level;

[0033] Figure 8 is a visualization image schematically showing a detail component in a vertical direction of a first level;

[0034] Fig. 9 is a visualization image schematically showing the detail components in the diagonal direction of the first level;

[0035] Fig.10 is a visualization image schematically showing a detail component in the horizontal direction of the second level;

[0036] Fig.11 is a visualization image schematically showing a detail component in a vertical direction of a second level;

[0037] Fig.12 is a visualization image schematically showing the detail components in the diagonal direction of the second level;

[0038] Fig.13 is a schematic diagram schematically showing the correspondence between pixels in a seal image and elements in a detail component;

[0039] Fig.14 is a visualized image schematically showing a corrected detail component in the horizontal direction of the second level;

[0040] Fig.15 is a schematic diagram schematically showing an enhanced image of a seal;

[0041] Fig.16 The figure schematically shows the results of identifying production defects of seals. DETAILED DESCRIPTION

[0042] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0043] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0044] The embodiment of the present invention discloses a method for intelligently identifying production defects of a seal, referring to Figure 1 , comprising steps S1 to S3:

[0045] S1. Collecting seal images.

[0046] Specifically, a camera disposed above the inspection platform is used to collect an image of the seal to be inspected from above. For ease of processing, the collected seal image is a grayscale image. Figure 2 A schematic diagram of a seal image.

[0047] S2. Decompose the seal image by wavelet transform, correct the decomposed detail components, and reconstruct the seal enhanced image according to the corrected detail components.

[0048] Specifically, the flowchart of step S2 refers to Figure 3 , including steps S201 to S206:

[0049] S201, performing wavelet transform on the seal image to obtain low-frequency components and detail components at different levels.

[0050] Specifically, the seal image is decomposed using the Haar wavelet transform function, and the total number of decomposition levels is , the implementer can set the total number of levels according to the actual implementation situation, for example During the decomposition process, each level can obtain a low-frequency component and three detail components in different directions, and the next level only decomposes the low-frequency component of the previous level. The directions of the detail components are horizontal, vertical and diagonal. Haar wavelet transform is a well-known technology and will not be described here. Figure 4 Schematic diagram of Haar wavelet transform decomposition. Figure 4 In the figure, a is the seal image, b is the low-frequency component of the first level, h1, v1, c1 are the detail components of the first level in the horizontal, vertical and diagonal directions, c is the low-frequency component of the second level, h2, v2, c2 are the detail components of the second level in the horizontal, vertical and diagonal directions, d is the low-frequency component of the third level, h3, v3, c3 are the detail components of the third level in the horizontal, vertical and diagonal directions. b, h1, v1, c1 are the decomposition results of a, c, h2, v2, c2 are the decomposition results of b, and d, h3, v3, c3 are the decomposition results of c.

[0051] Figure 5 , Figure 6 They are Figure 2 Visualization images of the first and second level low-frequency components, Figure 7 , Figure 8 , Fig. 9 They are Figure 2 The first-level horizontal, vertical, and diagonal detail components of Fig.10 , Fig.11 , Fig.12 They are Figure 2 The second level of detail components in the horizontal, vertical, and diagonal directions.

[0052] S202 , determining the overall grayscale retention rate of each pixel point according to the grayscale difference between each pixel point in the seal image and the corresponding pixel point in the low-frequency component of each level.

[0053] It should be noted that the low-frequency components at each level are essentially downsampled images of the seal image at different scales. Therefore, each pixel in the seal image has a corresponding pixel in the low-frequency components at each level, and multiple pixels in the seal image correspond to the same pixel in the low-frequency components at each level. For example Fig.13 a represents the seal image, b represents the low-frequency component of the first level, and the pixel point in a , , , After downsampling, it corresponds to b . Some features with a relatively small proportion of pixels in the seal image may be lost in the process of continuous downsampling, such as texture, crack defects, glue deficiency defects, impurity defects, mold sticking defects, etc. Some features with a relatively large proportion of pixels in the seal image are retained in the process of continuous downsampling, such as the seal background color, etc. Therefore, the present invention determines the grayscale retention rate of each pixel in the seal image in the low-frequency component by comparing the grayscale difference between each pixel in the seal image and its corresponding pixel in the low-frequency component.

[0054] Specifically, the grayscale retention rate of each pixel in the seal image in the low-frequency components at different levels satisfies the expression:

[0055] ;

[0056] in, Indicates the seal image The pixel at Grayscale retention rate in low-frequency components of each level; Indicates the seal image The gray value of each pixel; Indicates the seal image The pixel at The gray value of the corresponding pixel in the low-frequency component of each level; To represent the grayscale extreme value in the seal image, that is, the difference between the maximum grayscale value and the minimum grayscale value in the seal image, Perform normalization; is an exponential function with a natural constant as the base, used for Negative correlation mapping is performed. When the seal image Gray value of pixel With The gray value of the corresponding pixel in the low-frequency component of each level The difference between The more hours, the more The seal image is downsampled. The smaller the change in the gray value of each pixel, the The grayscale features represented by the pixels do not disappear. The pixel at The greater the grayscale retention rate in the low-frequency component of the first level, The grayscale features represented by the pixels are more likely to be the main grayscale features in the seal image, such as the background color of the seal; on the contrary, when the difference The bigger the The seal image is downsampled. The greater the change in the gray value of the pixel, the The grayscale features represented by the pixels may disappear. The pixel at The smaller the grayscale retention rate in the low-frequency component of the first level, The grayscale features represented by the pixels are more likely to be a minority of grayscale features in the seal image, such as texture, crack defects, glue deficiency defects, impurity defects, and mold sticking defects on the seal.

[0057] Furthermore, the overall grayscale retention rate of each pixel in the seal image is determined according to the grayscale retention rate of each pixel in the low-frequency components at different levels:

[0058] ;

[0059] in, Indicates the seal image The overall grayscale retention rate of pixels; The serial number indicating the level; Indicates the total number of levels; Indicates the seal image The pixel at The grayscale retention rate in the low-frequency component of each level; when the grayscale retention rate of the pixel point in the seal image in the low-frequency component with a larger level sequence number is larger, it means that the pixel point still retains its features after multiple downsampling. At this time, the pixel point is more likely to be the main feature in the seal image. Therefore, the present invention uses As a weight, when the level number is larger, the weight is larger, and more attention is paid to the grayscale retention rate of the pixel points in the seal image in the low-frequency component of the corresponding level.

[0060] S203 , determining the detail loss degree of each pixel at different levels according to the changes in the element values ​​of the elements corresponding to each pixel in the seal image in the detail components at different levels.

[0061] It should be noted that the detail component in the Haar wavelet transform reflects the details in the seal image. The larger the element value in the detail component, the clearer the details in the seal image. In the decomposition process of the Haar wavelet transform, details with poor clarity are gradually lost in detail components of different levels. Therefore, the present invention determines the detail loss degree of the corresponding element by comparing the changes in the element values ​​in detail components of different levels.

[0062] Specifically, the detail loss satisfies the expression:

[0063] ;

[0064] Among them, any one of the horizontal direction, vertical direction and diagonal direction is taken as the target direction. Indicates the seal image The pixel at The detail loss in the target direction at each level; Indicates the seal image The pixel at The corresponding elements in the detail components of the target direction at each level, Indicates the level number. ; Indicates the seal image The pixel at The corresponding elements in the detail components of the target direction at each level, , Indicates the total number of levels; Indicates the seal image The element corresponding to the pixel in the detail component of the target direction at the first level; represents the maximum value function, represents the minimum function, , It is used to limit the detail loss to the range of [0,1]. The element value of the element in the detail component reflects the details in the seal image. When the element value is larger, the details are more obvious, and when the element value is smaller, the details are more blurred. The pixel at The corresponding elements in the detail component of the target direction at each level Compared with the corresponding elements in the previous levels, the smaller the The pixel at The greater the loss of detail in the target direction of each level, the greater the loss of detail in the target direction. In the above expression, when hour, , i.e. the first The detail loss of each pixel in the first level target direction is 0.

[0065] S204, determining the correction index of the element corresponding to each pixel in the detail component at each level according to the detail loss degree of each pixel in the seal image at different levels and the overall grayscale retention rate of each pixel.

[0066] It should be noted that when the detail loss of a pixel in the seal image at a certain level is greater, the detail feature represented by the pixel is less clear, and conversely, when the detail loss is smaller, the detail feature represented by the pixel is clearer. When the grayscale retention rate of a pixel is smaller, the pixel is more likely to be a feature with a smaller grayscale ratio such as texture and crack defects. Conversely, when the grayscale retention rate of a pixel is larger, the pixel is more likely to be a feature such as the seal background color. Therefore, the present invention determines the correction index of the corresponding element in the detail component according to the detail loss and the grayscale retention rate, and assigns a larger correction index to features such as texture, crack defects, glue deficiency defects, impurity defects, sticky mold defects, bubble defects, scratch defects, etc. whose details are not clear, so that these features can be enhanced in the future to make them clearer and easier to identify defects.

[0067] Specifically, the correction index satisfies the expression:

[0068] ;

[0069] in, Indicates the seal image The pixel at The correction index of the corresponding element in the detail component of the target direction of each level; Indicates the seal image The pixel at The detail loss in the target direction at each level; Indicates the seal image The overall grayscale retention rate of pixels; when the detail loss The larger the correction index is, the The larger the seal, the more important it is to The pixel at The corresponding elements in the detail components of the target direction of each level are corrected, and their element values ​​are increased, so that the seal image restored according to the corrected detail components of each level is The more obvious the details represented by the pixels, the more obvious the overall grayscale retention rate. The smaller the value, the smaller the seal image. The features represented by the pixel points are textures, cracks, glue shortages, impurities, mold sticking, bubbles, scratches and other features with a small grayscale ratio. Therefore, the present invention uses gamma transformation to use Loss of detail To increase, when The smaller the size, the greater the loss of detail. The greater the increase in the seal image, the The pixel at Correction index of the corresponding element in the detail component of the target direction at each level The larger the value, the more emphasis will be placed on the unclear texture, crack defects, glue deficiency defects, impurity defects, and mold sticking defects. The larger the value, the more detail is lost. The smaller the increase in The value of Since the seal image The detail loss of pixels in the first level target direction To avoid , the power of 0 is meaningless, so it is stipulated that the seal image The correction index of the element corresponding to the pixel point in the detail component of the target direction of the first level is 0.

[0070] S205. Correct each element in the detail component according to all correction indices of each element to obtain a corrected detail component.

[0071] It should be noted that since Haar wavelet transform is a continuous decomposition process, the detail components of the next level are the detail components of the previous level. Therefore, four pixels in the seal image correspond to one element in the detail component of the first level, four pixels in the seal image correspond to one pixel in the low-frequency component of the first level, and four pixels in the low-frequency component of the first level correspond to one element in the detail component of the second level. Then, sixteen pixels in the seal image correspond to one element in the detail component of the second level, and so on. One element in the detail component of each level corresponds to multiple pixels in the seal image. The present invention corresponds to the first pixel in the seal image. The correction index of multiple pixels of the same element in the level is Correct the corresponding elements in each level and obtain their correction values.

[0072] Specifically, any element in the detail component of the target direction at each level is taken as the target element, and the target element corresponds to pixels, then the target element has A correction index.

[0073] Correct the target element according to all correction indices of the target element to obtain the correction value of the target element:

[0074] ;

[0075] in, Indicates the correction value of the target element; Represents the element value of the target element; Indicates the target element A correction index; The serial number representing the modified index of the target element; Indicates the number of pixels corresponding to the target element in the seal image, and also indicates the number of correction indexes of the target element; when the correction index of the target element is larger, the correction value of the target element increases more, and when the seal enhanced image is subsequently reconstructed based on the correction value of the target element, the details corresponding to the target element are clearer. Since all the correction indexes corresponding to each element in the detail component of the target direction of the first level are 0, the correction value of each element in the detail component of the target direction of the first level is the same as the value of each element before correction.

[0076] Similarly, the correction value of each element in the detail component of each direction at each level is obtained, and the correction value of each element in the detail component of each direction at each level constitutes the corrected detail component of each direction at each level. Fig.14 for Fig.10 The corresponding visualization image of the corrected detail component.

[0077] S206 , reconstructing a seal enhancement image according to the low-frequency components of each level and the corrected detail components.

[0078] Specifically, reconstructing an image based on detail components and low-frequency components is a well-known technique in Haar wavelet transform, which will not be described in detail here.

[0079] At this point, the enhancement of the seal image is achieved, and an enhanced seal image is obtained. Fig.15 This is a schematic diagram of the seal enhancement image. It can be seen Fig.15 Medium details compared to Figure 2 More obvious in.

[0080] S3. Identify production defects of the seal according to the seal enhancement image.

[0081] Specifically, a neural network is used to identify production defects in the enhanced image of the seal. The types of production defects include: cracks, glue deficiency, degumming, impurities, flash, mold sticking, scratches, and bubbles. The implementer can set the content of the neural network according to the actual implementation situation. The present invention uses the YOLOv5 network. Fig.16 for Fig.15 The recognition result of Fig.16 The bounding box in the figure selects the production defects in the seal enhanced image and identifies the types of production defects as impurities and sticking.

[0082] An embodiment of the present invention further discloses a system for intelligently identifying production defects of seals, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a method for intelligently identifying production defects of seals according to the present invention is implemented.

[0083] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, and their configuration and functions are known in the art, so they will not be described in detail here.

[0084] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0085] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for intelligently identifying production defects of seals, characterized in that: include: Acquiring an image of the seal; Perform wavelet transform on the seal image to obtain low-frequency components and detail components at different levels; Determining the overall grayscale retention rate of each pixel point according to the grayscale difference between each pixel point in the seal image and the corresponding pixel point in the low-frequency component of each level, including: Calculate the grayscale retention rate of pixels in the seal image in the low-frequency components of each level: ； in, Indicates the seal image The pixel at Grayscale retention rate in low-frequency components of each level; Indicates the seal image The gray value of each pixel; Indicates the seal image The pixel at The gray value of the corresponding pixel in the low-frequency component of each level; To represent the extreme grayscale in the seal image; is an exponential function with a natural constant as base; Calculate the overall grayscale retention rate of pixels in the seal image: ； in, Indicates the seal image The overall grayscale retention rate of pixels; The serial number indicating the level; Indicates the total number of levels; According to the changes in the element values ​​of the corresponding elements of the detail components of each pixel at different levels, the detail loss of each pixel at different levels is calculated: ； The directions of the detail components are horizontal, vertical, and diagonal, and any one of them is taken as the target direction. Indicates the seal image The pixel at The detail loss in the target direction at each level; , , Respectively represent the seal image Pixels at the first level, level, The corresponding elements in the detail components of the target direction at each level, , Indicates the level number. , ; represents the maximum value function, represents the minimum function; Determining, according to the overall grayscale retention rate and the detail loss degree, a correction index of a corresponding element of each pixel point in the detail component of each level; Correct each element according to all correction indexes of each element in the detail component to obtain a corrected detail component; The seal enhancement image is reconstructed according to the low-frequency components of each layer and the corrected detail components, and the production defects of the seal are identified according to the seal enhancement image.

2. The method for intelligently identifying production defects of a seal according to claim 1, characterized in that: The correction index of the corresponding element of each pixel in the detail component of each level satisfies the expression: ； in, Indicates the seal image The pixel at The correction index of the corresponding element in the detail component of the target direction of each level; Indicates the seal image The pixel at The detail loss in the target direction at each level; Indicates the seal image The target direction is any one of the horizontal direction, the vertical direction and the diagonal direction.

3. The method for intelligently identifying production defects of seals according to claim 1, characterized in that: The step of correcting each element in the detail component according to all correction indexes of each element to obtain a corrected detail component includes: Any element in the detail component of the target direction at each level is taken as the target element, and the target element corresponds to pixels, then the target element has A correction index; The corrected value of the target element Satisfies the expression: ; Represents the element value of the target element; Indicates the target element A correction index, The serial number representing the modified index of the target element; The corrected detail components in each direction of each level are constituted by the corrected values ​​of each element in the detail components in each direction of each level.

4. The method for intelligently identifying production defects of a seal according to claim 1, characterized in that: The method of identifying production defects of the seal according to the seal enhanced image comprises: The seal enhanced image is input into the trained neural network to output the production defects of the seal. The types of production defects include cracks, glue deficiency, debonding, impurities, flash and mold sticking.

5. The method for intelligently identifying production defects of seals according to claim 4, characterized in that: The neural network is a YOLOv5 network.

6. An intelligent identification system for production defects of seals, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a method for intelligently identifying production defects of a seal according to any one of claims 1 to 5 is implemented.

Images