A glass fiber partition defect detection method and system

Abstract

The present invention relates to the field of artificial intelligence technology, and in particular to a method and system for detecting defects in glass fiber partitions. The method obtains the defect probability through the difference between the pixel information of the glass fiber partition surface image and the standard average pixel value of the standard glass fiber partition surface image, and screens out abnormal partition images according to the defect probability. The gradient information of the abnormal partition is obtained, and the defect type is determined to be complete stratification or partial stratification according to the size of the gradient amplitude. If it is partial stratification, it is classified according to the gradient amplitude to obtain multiple gradient categories, and the gradient edge is obtained according to the coordinate information of the pixel points in the gradient category, and the stratification range is obtained according to the gradient edge and gradient direction. The present invention accurately identifies stratification defects based on the pixel features and gradient features of the image, and can achieve targeted subsequent processing of the product based on the stratification range.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and system for detecting defects in a glass fiber partition. Background Art

[0002] AGM is an ultra-fine glass fiber separator, an important component in battery production. Its quality directly affects the discharge capacity and charge-discharge cycle life of the battery. The multi-layer composite separator (mostly double-layer or triple-layer structure) has stronger capillary action and better liquid absorption ability than single-layer separators, which can eliminate the stratification of acid. And under low temperature conditions, batteries equipped with multi-layer separators have better high-current discharge performance.

[0003] When the raw materials and the pressing temperature are abnormal, the surface of the multi-layer partition will be delaminated, which seriously affects the product quality. However, due to the influence of light and the fact that the delamination phenomenon is inside the glass fiber partition, the gradient difference between the delaminated area and the normal area is small, resulting in poor detection effect when using traditional threshold segmentation technology or edge detection technology. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide a glass fiber separator defect detection method and system, and the technical solutions adopted are as follows:

[0005] The present invention provides a method for detecting defects in a glass fiber separator, the method comprising:

[0006] Obtaining multiple consecutive frames of glass fiber partition surface images; obtaining pixel value categories in the glass fiber partition surface images; obtaining a standard average pixel value of a standard glass fiber partition surface image; obtaining a defect probability based on the difference between the pixel value category and the standard average pixel value; if the defect probability is greater than a preset probability threshold, the corresponding glass fiber partition surface image is an abnormal partition image;

[0007] Obtaining gradient information at each pixel position of the abnormal partition image; the gradient information includes gradient amplitude and gradient direction; if the gradient amplitudes are all zero, it is considered that the defect type of the abnormal partition image is complete stratification; otherwise, it is considered that the defect type of the abnormal partition image is partial stratification;

[0008] According to the gradient magnitude classification, a plurality of gradient categories are obtained; according to the coordinate information of the pixel points in the gradient category, a plurality of gradient edges are obtained; and according to the gradient edge and the gradient direction, a layering range is determined.

[0009] Furthermore, the step of obtaining a plurality of continuous frames of the glass fiber separator surface image comprises:

[0010] Collecting a scene image of a glass fiber partition in a cutting process scene; removing background information in the scene image to obtain an initial glass fiber partition surface image;

[0011] The largest connected domain in the initial glass fiber partition board surface image is obtained; and the image in the largest connected domain is used as the glass fiber partition board surface image.

[0012] Further, the obtaining of pixel value categories in the glass fiber separator surface image includes:

[0013] Obtaining a first normalized grayscale histogram of the surface image of the glass fiber separator; the vertical axis of the first normalized grayscale histogram is grayscale proportion, and the horizontal axis is grayscale; taking multiple grayscales whose grayscale proportion in the first normalized grayscale histogram is not zero as the pixel value category;

[0014] Standard average pixel values ​​for obtaining a standard fiberglass partition surface image include:

[0015] A second normalized grayscale histogram of the standard glass fiber partition surface image is obtained; and the standard average pixel value is obtained according to the grayscale proportion and grayscale size in the second normalized grayscale histogram.

[0016] Further, obtaining the defect probability according to the difference between the pixel value category and the standard average pixel value includes:

[0017] The defect probability is obtained according to a defect probability formula, wherein the defect probability formula includes:

[0018]

[0019] Wherein, P is the defect probability, is the average pixel value of the glass fiber separator surface image, is the standard average pixel value, H i is the i-th pixel value category, and N is the number of pixel value categories.

[0020] Further, the obtaining of a plurality of gradient categories according to the gradient amplitude classification includes:

[0021] Classification is performed according to the gradient amplitude using a mean shift clustering algorithm to obtain a plurality of gradient categories.

[0022] Further, obtaining a plurality of gradient edges according to the coordinate information of the pixel points in the gradient category includes:

[0023] A density clustering algorithm is used to classify the pixels in the gradient category according to their coordinate information to obtain a plurality of clusters; the inter-class variance of each cluster is calculated, and the cluster corresponding to the largest inter-class variance is eliminated, and each of the remaining clusters is a gradient edge.

[0024] Further, determining the layer range according to the gradient edge and the gradient direction includes:

[0025] A plurality of regions are obtained according to the gradient edge, and the region pointed to by the gradient direction is used as the initial stratification region; and the range of the maximum horizontal coordinate and the minimum horizontal coordinate of the initial stratification region is used as the stratification range.

[0026] Furthermore, after determining the stratification range, the step further includes:

[0027] The defective glass fiber separator and the normal glass fiber separator are obtained by longitudinal cutting according to the delamination range; the defective glass fiber separator is removed and the normal glass fiber separator is retained.

[0028] The present invention also proposes a glass fiber partition defect detection system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any one of the steps of the glass fiber partition defect detection method when executing the computer program.

[0029] The present invention has the following beneficial effects:

[0030] The embodiment of the present invention determines the defect probability of the current glass fiber partition by comparing the pixel features of the collected glass fiber partition surface image with the pixel features of the qualified glass fiber partition surface image. The current glass fiber partition is preliminarily screened by the defect probability to obtain an abnormal partition image. The stratification of the current glass fiber partition is further determined based on the gradient information in the abnormal partition image. If partial stratification occurs, the stratification range is further determined based on the gradient direction, and the glass fiber partition can be targeted according to the stratification situation. Accurate stratification defect detection of glass fiber partitions is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0032] Figure 1A flow chart of a method for detecting defects in a glass fiber separator provided by one embodiment of the present invention;

[0033] Figure 2 A schematic diagram of a glass fiber partition processing scene provided by an embodiment of the present invention;

[0034] Figure 3 A schematic diagram of a glass fiber separator including a delamination defect provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0035] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of a glass fiber partition defect detection method and system proposed by the present invention, its specific implementation, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures, or characteristics in one or more embodiments may be combined in any suitable form.

[0036] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0037] The specific scheme of a glass fiber separator defect detection method and system provided by the present invention is described in detail below with reference to the accompanying drawings.

[0038] See also Figure 1 , which shows a flow chart of a glass fiber separator defect detection method provided by an embodiment of the present invention, the method comprising:

[0039] Step S1: obtain multiple consecutive frames of glass fiber partition surface images; obtain pixel value categories in the glass fiber partition surface images; obtain standard average pixel values ​​of standard glass fiber partition surface images; obtain defect probability based on the difference between the pixel value category and the standard average pixel value; if the defect probability is greater than a preset probability threshold, the corresponding glass fiber partition surface image is an abnormal partition image.

[0040] After the fiberglass partition is manufactured, it is usually cut into small pieces of the same or different specifications to facilitate transportation and use. Figure 2 , which shows a schematic diagram of a glass fiber partition processing scenario provided by an embodiment of the present invention. The white part in the figure is a glass fiber partition, which is fixed by two pressing rollers on a cutting machine. The outermost pressing roller is equipped with a cutting blade, and the glass fiber partition is cut by moving the glass fiber partition. The innermost pressing roller is used to fix and transport the glass fiber partition.

[0041] In order to obtain complete and clear image information of the fiberglass partition, a grayscale camera is deployed directly above the cutting device. The height of the camera is moderate to ensure that the camera field of view can completely include the area before the fiberglass partition is cut, and the camera sampling rate is adjusted to match the movement rate of the fiberglass cover, so that the camera can capture multiple consecutive frames of clear and complete scene images.

[0042] The scene image of the glass fiber partition in the cutting process scene is captured by a set camera. The scene image contains a large amount of background information, such as local information of the ground, pressure rollers or cutting machines. In order to facilitate subsequent detection, it is necessary to remove the background information to obtain the initial glass fiber partition surface image. The initial glass fiber partition surface image may include glass fiber partition information in two areas, namely the area to be cut and the area outside the innermost pressure roller. The glass fiber partition targeted by the embodiment of the present invention is the partition in the area to be cut, so the glass fiber partition in other areas needs to be removed. The maximum connected domain in the initial glass fiber partition surface image is obtained by the connected domain analysis method. The image in the maximum connected domain is used as the glass fiber partition surface image for the formal analysis.

[0043] In the embodiment of the present invention, because the surface of the glass fiber partition is white, its gray value is higher than other pixel information of the surrounding environment, so the scene image can be segmented using the threshold segmentation method to obtain the initial glass fiber partition. The segmentation threshold of the threshold segmentation can use the average pixel value in the scene image.

[0044] In the embodiment of the present invention, the connected domain analysis algorithm adopts a seed filling method.

[0045] See also Figure 3 , which shows a schematic diagram of a glass fiber separator containing delamination defects provided by an embodiment of the present invention. Figure 3 The pixel values ​​between the two areas are not very different, so it is impossible to use a fixed threshold to accurately segment the stratified area. However, taking the pixel information of the normal area as the standard, the pixel information of the stratified area can still represent the abnormality of the current glass fiber partition.

[0046] A first normalized grayscale histogram of the surface image of the fiberglass partition is obtained. The vertical axis of the first normalized grayscale histogram is the grayscale ratio, and the horizontal axis is the grayscale level. Multiple grayscale levels whose grayscale ratio in the first normalized grayscale histogram is not zero are used as pixel value categories. Similarly, a standard average pixel value of a standard fiberglass partition surface image is obtained: a second normalized grayscale histogram of a standard fiberglass partition surface image is obtained. The standard average pixel value is obtained according to the grayscale ratio and grayscale size in the second normalized grayscale histogram.

[0047] There is no delamination defect in the standard fiberglass partition surface image, so the standard average pixel value can be considered as the pixel value of the qualified area of ​​the fiberglass partition. The defect probability of the current fiberglass partition surface image can be obtained based on the difference between the pixel value category and the standard average pixel value, including:

[0048] The defect probability is obtained according to the defect probability formula, which includes:

[0049]

[0050] Where P is the defect probability, is the average pixel value of the glass fiber partition surface image, is the standard average pixel value, H i is the i-th pixel value category, and N is the number of pixel value categories.

[0051] In the defect probability formula, use Indicates the first offset degree between the current fiberglass partition surface image and the standard fiberglass partition surface image, The closer it is to 1, the closer the two are to each other. It should be noted that if the fiberglass partition is delaminated, the pixel value of the delaminated area will be greater than that of the normal area. will be greater than or equal to 1, The larger the value, the greater the degree of delamination and the greater the corresponding defect probability.

[0052] In the defect probability formula, The second offset degree of each gray level relative to the standard average pixel value is represented, and the first offset degree and the second offset degree jointly reflect the overall offset degree between the current fiberglass partition surface image and the standard fiberglass partition surface image.

[0053] If the defect probability is greater than the preset probability threshold, it means that the glass fiber partition board has abnormal information at this time, and the corresponding glass fiber partition board surface image is an abnormal partition board image. In the embodiment of the present invention, the probability threshold is set to 0.5.

[0054] Step S2: Obtain the gradient information at each pixel position of the abnormal partition image; the gradient information includes the gradient amplitude and the gradient direction; if the gradient amplitudes are all zero, the defect type of the abnormal partition image is considered to be complete stratification; otherwise, the defect type of the abnormal partition image is considered to be partial stratification.

[0055] The abnormal partition image indicates that the current fiberglass partition has pixel abnormalities, which may include two types of delamination: complete delamination and partial delamination. Complete delamination means that the entire current fiberglass partition is abnormal and there is no normal area. Partial delamination means that the current fiberglass partition has one or more delamination areas. There are gradient edges between delamination areas, so the gradient information can be used to determine the abnormality of the current fiberglass partition.

[0056] In the embodiment of the present invention, after Gaussian blur processing is performed on the abnormal partition image, the Sobel operator is used to perform gradient detection to obtain the gradient information at each pixel position. The gradient information includes gradient amplitude and gradient direction.

[0057] If the gradient amplitude is 0, it means that there is no edge information in the current abnormal partition image, that is, the pixel values ​​in the abnormal partition image are evenly distributed. Therefore, it is considered that the defect type corresponding to the abnormal partition image is complete delamination. In order to ensure product quality, all completely delaminated glass fiber partitions need to be cut off and removed.

[0058] If there is a non-zero gradient amplitude, it means that there is edge information in the abnormal partition image, which constitutes a delamination area. That is, the defect type at this time is partial delamination.

[0059] Step S3: obtaining multiple gradient categories according to gradient magnitude classification; obtaining multiple gradient edges according to coordinate information of pixel points in the gradient categories; and determining the stratification range according to the gradient edges and gradient directions.

[0060] Because the gradient magnitudes on the edges of the same layered regions have certain similarities, multiple gradient categories can be obtained based on the gradient magnitudes, including:

[0061] Classification is performed according to the gradient amplitude using a mean shift clustering algorithm to obtain a plurality of gradient categories.

[0062] Continuous pixels in each gradient category can form a gradient edge. Therefore, multiple gradient edges are obtained according to the coordinate information of the pixels in the gradient category, including:

[0063] According to the coordinate information of the pixel points in the gradient category, the density clustering algorithm is used for classification to obtain multiple clusters. In multiple clusters, discrete points such as noise will be classified into one category. Therefore, in order to prevent the influence of noise on detection, the inter-class variance of each cluster is calculated, and the cluster corresponding to the largest inter-class variance is eliminated. Each remaining cluster is a gradient edge.

[0064] The gradient edge can be enclosed into one or more areas. Because the degree of stratification is more complicated, the stratified area may not be a regular shape, and targeted cutting cannot be performed on the cutting equipment. Therefore, in order to facilitate the production process, it is necessary to determine the stratification range according to the gradient edge and gradient direction, so that the subsequent cutting equipment can effectively cut according to the stratification range and retain as many qualified areas as possible. The obtained stratification range specifically includes:

[0065] Multiple regions are obtained according to the gradient edge, and the region pointed to by the gradient direction is used as the initial stratification region. The range of the maximum horizontal coordinate and the minimum horizontal coordinate of the initial stratification region is used as the stratification range.

[0066] After determining the stratification range of each stratified area, the cutting equipment can be controlled to perform longitudinal cutting according to the stratification range to obtain defective fiberglass partitions and normal fiberglass partitions. The defective fiberglass partition contains most of the stratification area information, and the normal fiberglass partition only contains the normal area information. Therefore, the defective fiberglass partition is removed and only the normal fiberglass partition is retained to ensure product quality. The defective fiberglass partition can be recycled or the product can be repaired to maximize production efficiency.

[0067] In summary, the embodiment of the present invention obtains the defect probability through the difference between the pixel information of the glass fiber partition surface image and the standard average pixel value of the standard glass fiber partition surface image, and screens out abnormal partition images according to the defect probability. The gradient information of the abnormal partition is obtained, and the defect type is determined whether it is complete stratification or partial stratification according to the size of the gradient amplitude. If it is partial stratification, it is classified according to the gradient amplitude to obtain multiple gradient categories, and the gradient edge is obtained according to the coordinate information of the pixel points in the gradient category, and the stratification range is obtained according to the gradient edge and the gradient direction. The embodiment of the present invention accurately identifies stratification defects based on the pixel features and gradient features of the image, and can achieve targeted subsequent processing of the product based on the stratification range.

[0068] The present invention also proposes a glass fiber partition defect detection system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any one of the steps of a glass fiber partition defect detection method.

[0069] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0070] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0071] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting defects in a glass fiber separator, characterized in that: The method comprises: Obtaining multiple consecutive frames of glass fiber partition surface images; obtaining pixel value categories in the glass fiber partition surface images; obtaining a standard average pixel value of a standard glass fiber partition surface image; obtaining a defect probability based on the difference between the pixel value category and the standard average pixel value; if the defect probability is greater than a preset probability threshold, the corresponding glass fiber partition surface image is an abnormal partition image; Obtaining gradient information at each pixel position of the abnormal partition image; the gradient information includes gradient amplitude and gradient direction; if the gradient amplitudes are all zero, it is considered that the defect type of the abnormal partition image is complete stratification; otherwise, it is considered that the defect type of the abnormal partition image is partial stratification; According to the gradient magnitude classification, a plurality of gradient categories are obtained; according to the coordinate information of the pixel points in the gradient category, a plurality of gradient edges are obtained; according to the gradient edge and the gradient direction, a layering range is determined; The obtaining of the pixel value category in the glass fiber partition surface image comprises: Obtaining a first normalized grayscale histogram of the surface image of the glass fiber separator; the vertical axis of the first normalized grayscale histogram is grayscale proportion, and the horizontal axis is grayscale; taking multiple grayscales whose grayscale proportion in the first normalized grayscale histogram is not zero as the pixel value category; Standard average pixel values ​​for obtaining a standard fiberglass partition surface image include: Obtaining a second normalized grayscale histogram of the standard glass fiber separator surface image; obtaining the standard average pixel value according to the grayscale proportion and grayscale size in the second normalized grayscale histogram; The obtaining of the defect probability according to the difference between the pixel value category and the standard average pixel value comprises: The defect probability is obtained according to a defect probability formula, wherein the defect probability formula includes: Wherein, P is the defect probability, is the average pixel value of the glass fiber separator surface image, is the standard average pixel value, H i is the i-th pixel value category, and N is the number of pixel value categories.

2. A method for detecting defects in a glass fiber separator according to claim 1, characterized in that: The step of obtaining a plurality of continuous frames of glass fiber partition board surface images comprises: Collecting a scene image of a glass fiber partition in a cutting process scene; removing background information in the scene image to obtain an initial glass fiber partition surface image; The largest connected domain in the initial glass fiber partition board surface image is obtained; and the image in the largest connected domain is used as the glass fiber partition board surface image.

3. A glass fiber separator defect detection method according to claim 1, characterized in that: The obtaining of a plurality of gradient categories according to the gradient amplitude classification comprises: Classification is performed according to the gradient amplitude using a mean shift clustering algorithm to obtain a plurality of gradient categories.

4. A glass fiber separator defect detection method according to claim 1, characterized in that: The obtaining of a plurality of gradient edges according to the coordinate information of the pixel points in the gradient category comprises: A density clustering algorithm is used to classify the pixels in the gradient category according to their coordinate information to obtain a plurality of clusters; the inter-class variance of each cluster is calculated, and the cluster corresponding to the largest inter-class variance is eliminated, and each of the remaining clusters is a gradient edge.

5. A method for detecting defects in a glass fiber separator according to claim 1, characterized in that: Determining the layering range according to the gradient edge and the gradient direction comprises: A plurality of regions are obtained according to the gradient edge, and the region pointed to by the gradient direction is used as the initial stratification region; and the range of the maximum horizontal coordinate and the minimum horizontal coordinate of the initial stratification region is used as the stratification range.

6. A method for detecting defects in a glass fiber separator according to claim 5, characterized in that: The determination of the stratification scope further includes: The defective glass fiber separator and the normal glass fiber separator are obtained by longitudinal cutting according to the delamination range; the defective glass fiber separator is removed and the normal glass fiber separator is retained.

7. A glass fiber partition defect detection system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

Images