A method and system for detecting production quality of glass fiber plain weave cloth
By performing white spot segmentation and defect reliability analysis on the image sequence of glass fiber plain jerseys, the error detection problem caused by reflection in glass fiber plain jerseys is solved, and the detection accuracy is improved.
Patent Information
- Application Number
- CN202411287836.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-14
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-09-14
AI Technical Summary
The existing quality detection method for glass fiber jersey production quality detection methods lead to false detection of normal tissue due to the reflective characteristics of glass fibers, and the detection accuracy is low.
By obtaining the image sequence of glass fiber plain cloth during uniform movement in a certain direction, white spot segmentation and defect reliability probability analysis are performed, and defect reliability is determined by using the change law of suspected defect positions in the image sequence to improve detection accuracy.
It effectively overcomes the misdetection problem caused by the reflective characteristics of glass fibers and improves the accuracy of the production quality inspection of glass fiber jerseys.
Smart Images

Figure CN118794885B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of image processing, and in particular to a method and system for detecting production quality of glass fiber plain weave cloth. Background Art
[0002] Glass fiber plain weave fabric is a fabric woven from untwisted roving in plain weave. The surface quality of the fabric directly affects the economic benefits of the enterprise. Therefore, it is very important to test the production quality of glass fiber plain weave fabric.
[0003] In the existing method, the production quality detection of glass fiber plain weave cloth is realized by directly analyzing the abnormal point features in the glass fiber plain weave cloth image.
[0004] However, due to the reflective properties of glass fiber, it is easy to cause misdetection of normal tissues in glass fiber plain weave cloth, resulting in low accuracy of glass fiber plain weave cloth production quality inspection. Summary of the invention
[0005] The embodiment of the present invention provides a method and system for detecting the production quality of glass fiber plain woven fabric, which can improve the accuracy of detecting the production quality of glass fiber plain woven fabric.
[0006] One aspect of an embodiment of the present invention provides a method for detecting the production quality of glass fiber plain weave cloth, comprising:
[0007] Acquire an image sequence of the glass fiber plain woven fabric in a process of uniformly moving along a first direction, the image sequence comprising a plurality of detection images of the glass fiber plain woven fabric, and the detection images in the image sequence are sorted in the order of acquisition time;
[0008] Performing white spot segmentation on a first detection image in the image sequence to determine a first suspected defect position of the first detection image;
[0009] Determine the defect reliability probability of the first suspected defect position according to the first suspected defect position of the first detection image and the second suspected defect position of each second detection image in the image sequence, where the second detection image is N detection images after the first detection image in the image sequence, where N is a positive integer;
[0010] When the defect reliability probability of the first suspected defect position is greater than or equal to the first probability threshold, it is determined that the glass fiber plain weave cloth has a production quality abnormality at the first suspected defect position.
[0011] In one aspect of an embodiment of the present invention, a glass fiber plain weave fabric production quality inspection system is provided, including an image acquisition device, a motion platform, and a controller;
[0012] An image acquisition device, used for acquiring surface feature images of the glass fiber plain weave cloth;
[0013] The motion platform is used to control the glass fiber plain weave fabric to move at a uniform speed along a first direction;
[0014] The controller is used to execute the glass fiber plain weave fabric production quality detection method provided in any one of the above aspects.
[0015] In the glass fiber plain cloth production quality detection method provided by the embodiment of the present invention, the first suspected defect position of the first detection image in the image sequence is determined according to the uniformity law of the white spots reflected by the glass fiber plain cloth by performing white spot segmentation on the first detection image. It can overcome the situation that the normal tissue in the glass fiber plain cloth is misdetected due to the reflective characteristics of the glass fiber. At the same time, according to the first suspected defect position of the first detection image and the second suspected defect position of each second detection image after the first detection image in the image sequence, the defect reliability of the first suspected defect position is determined by using the change law of the suspected defect position during the movement of the glass fiber plain cloth. By considering the position change between continuous images during the movement, the accuracy of the detection is further improved. In this way, the embodiment of the present invention can improve the accuracy of the production quality detection of glass fiber plain cloth by white spot segmentation and analysis from the image sequence of glass fiber plain cloth. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] 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.
[0017] Figure 1 A schematic diagram of a process for detecting the production quality of a first glass fiber plain woven fabric provided by an embodiment of the present invention;
[0018] Figure 2 A schematic diagram of a detection image of a glass fiber plain woven fabric provided by an embodiment of the present invention;
[0019] Figure 3 A schematic flow chart of a second method for detecting the production quality of glass fiber plain weave cloth provided by an embodiment of the present invention;
[0020] Figure 4 A schematic diagram of a binarized image provided by an embodiment of the present invention;
[0021] Figure 5A schematic flow chart of a third method for detecting the production quality of glass fiber plain weave fabric provided by an embodiment of the present invention;
[0022] Figure 6 A schematic flow chart of a fourth method for detecting the production quality of glass fiber plain weave cloth provided by an embodiment of the present invention;
[0023] Figure 7 A schematic diagram of the structure of a glass fiber plain weave fabric production quality inspection system provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0024] 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 plain weave fabric production quality detection method and system proposed by the present invention, its specific implementation method, structure, features and effects, in conjunction 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.
[0025] 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.
[0026] It should be noted that the acquisition, storage, use, and processing of data in the technical solution of the present invention are in compliance with the relevant provisions of national laws and regulations.
[0027] It should be noted that in the embodiments of the present invention, certain software, components, models and other existing solutions in the industry may be mentioned, which should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present invention, but it does not mean that the applicant has or will necessarily use the solution.
[0028] In the existing method, the production quality inspection of glass fiber plain cloth is realized by directly analyzing the abnormal point features in the glass fiber plain cloth image. However, due to the reflective characteristics of glass fiber, it is easy to misdetect the normal tissue in the glass fiber plain cloth, resulting in low accuracy of glass fiber plain cloth production quality inspection.
[0029] The object of the present invention is to provide a method and system for detecting the production quality of glass fiber plain weave cloth. In the method for detecting the production quality of glass fiber plain weave cloth provided by the embodiment of the present invention, by performing white spot segmentation on the first detection image in the image sequence, the first suspected defect position of the first detection image is determined according to the uniformity law of the white spots reflected by the glass fiber plain weave cloth. It can overcome the situation that the normal tissue in the glass fiber plain weave cloth is misdetected due to the reflective characteristics of the glass fiber. At the same time, according to the first suspected defect position of the first detection image and the second suspected defect position of each second detection image after the first detection image in the image sequence, the defect reliability of the first suspected defect position is determined by using the change law of the suspected defect position during the movement of the glass fiber plain weave cloth. By considering the position change between continuous images during the movement, the accuracy of the detection is further improved. In this way, the embodiment of the present invention can improve the accuracy of the production quality detection of glass fiber plain weave cloth by white spot segmentation and analysis from the image sequence of glass fiber plain weave cloth.
[0030] The following describes a specific embodiment of a glass fiber plain weave fabric production quality detection method and system provided by an embodiment of the present invention.
[0031] Figure 1 A schematic flow chart of a glass fiber plain fabric production quality detection method is provided. The glass fiber plain fabric production quality detection method can be applied to a controller in a glass fiber plain fabric production quality detection system. The glass fiber plain fabric production quality detection method can include the following S101 to S104.
[0032] S101, obtaining an image sequence of the glass fiber plain woven fabric in a process of uniformly moving along a first direction, wherein the image sequence includes multiple detection images of the glass fiber plain woven fabric, and the detection images in the image sequence are sorted in the order of acquisition time.
[0033] In this embodiment, the glass fiber plain weave fabric moves at a uniform speed along a first direction. For example, the first direction may be a horizontal direction.
[0034] The image sequence stores multiple inspection images of glass fiber plain weave cloth, and the inspection images are sorted according to the order of image acquisition time. Figure 2 As shown, a schematic diagram of a detection image of glass fiber plain cloth is provided.
[0035] As an example, the controller sends an image capture instruction to a high-definition camera, and then uses the high-definition camera to capture multiple detection images of the glass fiber plain cloth, and sorts the detection images in chronological order of image capture time to obtain an image sequence corresponding to the glass fiber plain cloth.
[0036] S102, performing white spot segmentation on a first detection image in the image sequence to determine a first suspected defect position of the first detection image.
[0037] In this embodiment, the first detection image is any detection image in the image sequence.
[0038] White spot segmentation is the process of analyzing the detection image, identifying and analyzing the white spot area in the detection image.
[0039] As an example, glass fiber plain weave cloth is composed of horizontal and vertical fibers. The fibers in the two directions are intertwined and overlapped, thus producing a series of local tiny protrusions. When these protrusions are exposed to light, they are closer to the light source and have a smoother surface, so they tend to produce more obvious reflections than non-protrusions. Therefore, a series of white spots will appear in the image, that is, each white spot represents an overlapping fiber.
[0040] Therefore, after obtaining the image sequence corresponding to the glass fiber plain weave cloth, the controller selects an inspection image from the image sequence as the first inspection image. Then, the first inspection image is segmented for white spots to obtain a binary image corresponding to the inspection image. The distribution of black pixels and white pixels in the binary image is analyzed to obtain the first suspected defect position in the first inspection image.
[0041] Specifically, in the binary image corresponding to the detection image, the normal place is alternately distributed with black pixels and white pixels, and there is obvious interlacing; while the gray value of the defective place is low and exists in patches, that is, the defective place has fewer white pixels and no interlacing. Therefore, the area with fewer white pixels and no interlacing in the binary image corresponding to the detection image can be determined as the first suspected defect position.
[0042] S103, determining the defect reliability probability of the first suspected defect position according to the first suspected defect position of the first detection image and the second suspected defect position of each second detection image in the image sequence, wherein the second detection image is N detection images after the first detection image in the image sequence, and N is a positive integer.
[0043] In this embodiment, the second detection image is a continuous N images acquired after the first detection image, where N is a positive integer, for example, N may be 5.
[0044] Since the glass fiber plain weave cloth is constantly moving along the first direction, the position displayed at different times should be different according to the detection images taken by the high-definition camera.
[0045] As an example, when the movement speed of the glass fiber plain cloth and the shooting frequency of the high-definition camera remain unchanged, if the change of the suspected defect position in the first detection image and the subsequent N consecutive second detection images is more uniform, the defect reliability probability of the first suspected defect position is greater; if the change of the suspected defect position in the first detection image and the subsequent N consecutive second detection images is more uneven, the defect reliability probability of the first suspected defect position is smaller.
[0046] As another example, when the movement speed of the glass fiber plain cloth is kept unchanged and the shooting frequency of the high-definition camera is changed exponentially, the more the change in the suspected defect position in the first detection image and the subsequent N consecutive second detection images is changed exponentially, the greater the defect reliability probability of the first suspected defect position; if the change in the suspected defect position in the first detection image and the subsequent N consecutive second detection images is less changed exponentially, the smaller the defect reliability probability of the first suspected defect position.
[0047] S104: When the defect reliability probability at the first suspected defect position is greater than or equal to a first probability threshold, it is determined that the glass fiber plain weave fabric has a production quality abnormality at the first suspected defect position.
[0048] In this embodiment, the first probability threshold is a preset probability threshold for determining whether a suspected defect position actually has a production quality abnormality. For example, the first probability threshold may be preset to 0.73.
[0049] As an example, when the defect reliability probability at the first suspected defect position is greater than or equal to 0.73, it is determined that the glass fiber plain cloth has a production quality abnormality at the first suspected defect position; when the defect reliability probability at the first suspected defect position is less than 0.73, it is determined that the glass fiber plain cloth does not have a production quality abnormality at the first suspected defect position.
[0050] Through the glass fiber plain cloth production quality detection method provided by this embodiment, the first detection image in the image sequence is segmented for white spots, and the first suspected defect position of the first detection image is determined according to the uniformity law of the white spots reflected by the glass fiber plain cloth. It can overcome the situation where the normal tissue in the glass fiber plain cloth is misdetected due to the reflective characteristics of the glass fiber. At the same time, according to the first suspected defect position of the first detection image and the second suspected defect position of each second detection image after the first detection image in the image sequence, the defect reliability of the first suspected defect position is determined by using the change law of the suspected defect position during the movement of the glass fiber plain cloth. By considering the position change between continuous images during the movement, the accuracy of the detection is further improved. In this way, the embodiment of the present invention can improve the accuracy of the production quality detection of glass fiber plain cloth by segmenting white spots and analyzing the image sequence of glass fiber plain cloth.
[0051] As an optional embodiment, S101 may specifically include:
[0052] Acquire multiple surface feature images of the glass fiber plain weave fabric moving at a uniform speed along a first direction according to a preset acquisition frequency by using an image acquisition device;
[0053] Preprocessing each surface feature image to obtain a detection image corresponding to each surface feature image, wherein the preprocessing includes at least one of a denoising process and a gray value equalization process;
[0054] The detected images are sorted in the order of image acquisition time to obtain an image sequence of the glass fiber plain weave cloth.
[0055] In this embodiment, the image acquisition device is used to acquire images of the glass fiber plain weave cloth. As an example, the image acquisition device may be a camera.
[0056] The preset acquisition frequency is a preset frequency at which the image acquisition device acquires images. As an example, the preset acquisition frequency may be 30 Hz.
[0057] As an example, the controller sends an image capture instruction to the image acquisition device, and then the image acquisition device periodically captures images of the glass fiber plain weave cloth moving at a uniform speed along a first direction according to a preset capture frequency, thereby obtaining multiple surface feature images of the glass fiber plain weave cloth.
[0058] Then, each surface feature image of the glass fiber plain weave cloth is subjected to denoising and gray value equalization processing, thereby obtaining multiple detection images of the glass fiber plain weave cloth. Finally, the detection images are sorted in the order of image acquisition time to obtain an image sequence of the glass fiber plain weave cloth.
[0059] Through this embodiment, the image acquisition device is used to acquire images of the glass fiber plain woven fabric at a fixed acquisition frequency. At the same time, the acquired images are further subjected to denoising and gray value equalization processing, thereby obtaining an image sequence of the glass fiber plain woven fabric sorted in the order of acquisition time. In this way, by accurately obtaining the image sequence of the glass fiber plain woven fabric, it is helpful to determine the defect reliability probability by using the image sequence later, thereby improving the accuracy of the production quality inspection of the glass fiber plain woven fabric.
[0060] As an optional embodiment, Figure 3 As shown, S102 may specifically include the following S301 to S305.
[0061] S301, performing white spot segmentation on a first detection image in an image sequence to generate a binary image corresponding to the first detection image;
[0062] S302, using the average length and average width of the white spots in the binary image, stratifying the binary image to obtain a plurality of horizontal layers and a plurality of vertical layers corresponding to the binary image;
[0063] S303, determining the defect position probability of each horizontal layer according to the distribution of black pixels and white pixels in each horizontal layer, and determining the defect position probability of each vertical layer according to the distribution of black pixels and white pixels in each vertical layer;
[0064] S304, determining a horizontal layer whose defect position probability is greater than or equal to a second probability threshold as a defective horizontal layer, and determining a vertical layer whose defect position probability is greater than or equal to the second probability threshold as a defective vertical layer;
[0065] S305: Determine a first suspected defect position of the first detection image according to each defect transverse layer and each defect longitudinal layer.
[0066] In this embodiment, if Figure 4 As shown, a schematic diagram of a binary image is provided. It can be seen that in the binary image corresponding to the glass fiber plain weave cloth, black pixels and white pixels are alternately distributed.
[0067] The second probability threshold is a preset probability threshold for determining whether each layer in the binary image belongs to a defect layer. For example, the second probability threshold can be preset to 0.65.
[0068] As an example, after obtaining the image sequence of the glass fiber plain weave cloth, the controller performs white spot segmentation on the first detection image in the image sequence to obtain the corresponding binary image, and then calculates the mean value of the size of the white spots in the binary image to obtain the average length and width of the white spots.
[0069] Then, the average width of the white spots is used as the stratification index, and the binary image is stratified along the horizontal axis to obtain multiple horizontal layers corresponding to the binary image; then, the average length of the white spots is used as the stratification index, and the binary image is stratified along the vertical axis to obtain multiple vertical layers corresponding to the binary image.
[0070] Then, for each horizontal layer, the alternating distribution of black pixels and white pixels in the horizontal layer is counted to determine the defect location probability of each horizontal layer; and for each vertical layer, the alternating distribution of black pixels and white pixels in the horizontal layer is counted to determine the defect location probability of each vertical layer.
[0071] Then, the defect position probability of each horizontal layer is compared with the second probability threshold, and the horizontal layer whose defect position probability is greater than or equal to the second probability threshold is determined as a defective horizontal layer; the defect position probability of each vertical layer is compared with the second probability threshold, and the vertical layer whose defect position probability is greater than or equal to the second probability threshold is determined as a defective vertical layer.
[0072] Finally, the overlapping area of the defective transverse layer and the defective longitudinal layer is determined as the first suspected defect position of the first detection image. In the case of multiple overlapping areas at the same time, the defect position probability of the defective transverse layer and the defective longitudinal layer corresponding to the overlapping area are further accumulated and averaged, and the overlapping area with the largest average is determined as the first suspected defect position of the first detection image.
[0073] Through this embodiment, the binary image corresponding to the first detection image is layered, and then the defect position probability of each layer is calculated respectively, and the first suspected defect position of the first detection image is accurately obtained according to the defect position probability of each layer. In this way, by layering the binary image and determining the defect position probability from each local layer respectively, the defect position probability can be calculated more accurately, thereby improving the accuracy of the first suspected defect position of the first detection image.
[0074] As an optional embodiment, S301 may specifically include:
[0075] Acquire a first detection image in an image sequence;
[0076] The first detection image is input into the white spot segmentation model to obtain a binary image corresponding to the first detection image. The white spot segmentation model is obtained by training a convolutional neural network model using image training samples. The image training samples include detection images of glass fiber plain cloth and corresponding binary labels.
[0077] In this embodiment, the controller first constructs an image training sample based on the detection image of the glass fiber plain cloth and the corresponding manually annotated binary label. The preset convolutional neural network model is trained through the image training sample until the training end condition is reached, thereby obtaining a trained white spot segmentation model. The training end condition may be that the preset number of training times is reached or the loss value of the model is less than the preset loss value.
[0078] When white spot segmentation is performed on the first detection image, the first detection image is directly input into the trained white spot segmentation model, so as to obtain a binary image corresponding to the first detection image.
[0079] Through this embodiment, the convolutional neural network model is trained using image training samples to construct a white spot segmentation model, so that the white spot segmentation model can be used to directly obtain a binary image corresponding to the first detection image, without the need for manual white spot segmentation, which can improve the efficiency of white spot segmentation.
[0080] As an optional embodiment, Figure 5 As shown, in S303, the defect position probability of each horizontal layer is determined according to the distribution of black pixels and white pixels in each horizontal layer, which may specifically include the following S501 to S503.
[0081] S501, creating an image detection window with an average size of white spots in the first detection image as the window size;
[0082] S502, traversing the horizontal layers to obtain the number of black pixels and the number of white pixels in each image detection window;
[0083] S503, determining the defect position probability of the transverse layer according to the number of black pixels and the number of white pixels in each image detection window.
[0084] In this embodiment, as an example, the controller first calculates the mean value of the size of the white spots in the binary image to obtain the average length and average width of the white spots, and creates an image detection window according to the average length and average width of the white spots.
[0085] Then use the image detection window, starting from the leftmost side of the horizontal layer, and slide the image detection window to the right until it reaches the rightmost side of the horizontal layer, and then count the number of black pixels and the number of white pixels in the image detection window each time.
[0086] Finally, according to the statistical results, when the total number of black pixels is equal to the total number of white pixels, the probability of determining the defect position of the transverse layer is 0; when the difference between the total number of black pixels and the total number of white pixels is greater, the probability of determining the defect position of the transverse layer is greater.
[0087] Through this embodiment, an image detection window is created according to the average size of white spots in the first detection image, and then the image detection window is used to traverse the horizontal layer. Thus, the number of black pixels and the number of white pixels in each image detection window in the horizontal layer can be obtained, and then the defect position probability of the horizontal layer can be accurately calculated according to the number of black pixels and the number of white pixels in each image detection window, and the defect position probability can be calculated more accurately, thereby improving the accuracy of the first suspected defect position of the subsequent first detection image.
[0088] As an optional embodiment, S503 may specifically include:
[0089] Determine, based on the number of black pixels and the number of white pixels in each image detection window, a first variance of the ratio of the number of black pixels to the number of white pixels in the odd-numbered image detection windows and a second variance of the ratio of the number of black pixels to the number of white pixels in the even-numbered image detection windows;
[0090] The first variance is added to the second variance and then normalized to obtain the defect location probability of the transverse layer.
[0091] In this embodiment, the defect location probability of the lateral layer can be determined by the following formula 1:
[0092] Formula 1
[0093] In the formula, It is used to represent the defect location probability of the i-th transverse layer, and norm is used to represent the normalization function. It is used to represent the first variance of the ratio of the number of black pixels to the number of white pixels in the image detection window of odd times in the i-th horizontal layer, The second variance is used to represent the ratio of the number of black pixels to the number of white pixels in the even-numbered image detection windows in the i-th horizontal layer.
[0094] in, It represents the sum of the first variance and the second variance. The variance represents the fluctuation of the ratio of black and white pixels. The larger the value is, the greater the fluctuation of the black-white pixel ratio in the odd-numbered image detection window in the i-th transverse layer is, which means that the regularity of the staggered distribution in the odd-numbered image detection window in the i-th transverse layer is worse, and the higher the probability of defect location is. The larger it is, the greater the fluctuation of the black-white pixel ratio in the even-numbered image detection window in the i-th transverse layer, which means that the regularity of the staggered distribution in the even-numbered image detection window in the i-th transverse layer is worse, and the higher the probability of the defect location is.
[0095] Defect location probability of the i-th transverse layer The larger the value is, the worse the overall staggered distribution pattern in the i-th transverse layer is, that is, the i-th transverse layer is more likely to be a defective position.
[0096] Through this embodiment, the image detection window is used to traverse the horizontal layer, and the number of black pixels and the number of white pixels in each image detection window in the horizontal layer are obtained. Therefore, the defect position probability of the horizontal layer is accurately calculated according to the number of black pixels and the number of white pixels in each image detection window, and the defect position probability can be calculated more accurately, thereby improving the accuracy of the first suspected defect position of the subsequent first detection image.
[0097] As an optional embodiment, S305 may specifically include:
[0098] Determine the overlapping area of the defective transverse layer and the defective longitudinal layer according to each defective transverse layer and each defective longitudinal layer;
[0099] The overlapping area is determined as the first suspected defect position of the first inspection image.
[0100] In this embodiment, according to the position information of each defective transverse layer and each defective longitudinal layer in the first detection image, the overlapping area of the defective transverse layer and the defective longitudinal layer can be obtained, and the overlapping area is the first suspected defect position in the first detection image.
[0101] When there are multiple overlapping areas at the same time, the defect position probability of the defect transverse layer and the defect position probability of the defect longitudinal layer corresponding to the overlapping area are further accumulated and the average is calculated, and the overlapping area with the largest average is determined as the first suspected defect position of the first detection image.
[0102] Through this embodiment, the overlapping area of the defective transverse layer and the defective longitudinal layer is obtained by using each defective transverse layer and each defective longitudinal layer, so as to determine the first suspected defect position of the first detection image. By accurately calculating the first suspected defect position of the first detection image, it is helpful to subsequently use the change law of the suspected defect position of the glass fiber plain weave cloth during the movement process to determine the production quality of the glass fiber plain weave cloth.
[0103] As an optional embodiment, Figure 6 As shown, S103 may specifically include the following S601 to S603.
[0104] S601, determining a coordinate value variation degree in the first direction according to first coordinate values of the first suspected defect position and each second suspected defect position in the first direction, where the coordinate value variation degree is used to characterize a uniform variation degree between each first coordinate value;
[0105] S602, determining a coordinate value variance in a second direction according to second coordinate values of the first suspected defect position and each second suspected defect position in the second direction, where the second direction is perpendicular to the first direction;
[0106] S603, determining the defect reliability probability of the first suspected defect position according to the coordinate value variation in the first direction and the coordinate value variance in the second direction.
[0107] In this embodiment, when the first direction is the horizontal direction, the coordinate value variation in the first direction is the coordinate value variation in the X-axis direction, and the second direction is the vertical direction, and the coordinate value variance in the second direction is the coordinate value variance in the Y-axis direction.
[0108] As an example, the defect reliability probability of the first suspected defect position can be determined by the following formula 2:
[0109] Formula 2
[0110] In the formula, It is used to represent the defect reliability probability of the first suspected defect position in the i-th image, Used to indicate the change in coordinate value in the X-axis direction. Used to represent the variance of coordinate values in the Y-axis direction.
[0111] in, Indicates the degree of consistency between the first suspected defect position and the motion trajectory of the glass fiber plain fabric. The smaller it is, that is, the better the uniformity of the coordinate values in the X-axis direction, the greater the consistency between the first suspected defect position and the motion trajectory of the glass fiber plain weave cloth, that is, the greater the defect reliability probability of the first suspected defect position in the i-th image.
[0112] Indicates the change of the coordinate value of the first suspected defect position in the Y-axis direction. The smaller it is, that is, the smaller the variance of the coordinate values in the Y-axis direction is, the smaller the vertical change of each suspected defect position in the continuous image is, that is, the greater the defect reliability probability of the first suspected defect position in the i-th image is.
[0113] Defect reliability probability of the first suspected defect position in the i-th image The larger the value is, the greater the possibility that the glass fiber plain weave cloth has production quality abnormality at the first suspected defect position corresponding to the i-th image, thereby indicating that the production quality of the glass fiber plain weave cloth is poor.
[0114] Through this embodiment, the first suspected defect position of the first detection image and the second suspected defect position of each second detection image in the image sequence are used to determine the coordinate value change degree in the first direction and the coordinate value variance in the second direction, so that the defect reliability probability of the first suspected defect position can be accurately calculated. This helps to accurately evaluate the production quality of glass fiber plain weave cloth and improve the accuracy of production quality evaluation of glass fiber plain weave cloth.
[0115] As an optional embodiment, S601 may specifically include:
[0116] Sorting the first coordinate value of the first suspected defect position and the first coordinate values of each second suspected defect position in the order of the corresponding image acquisition time to obtain a coordinate value sequence;
[0117] Determine the coordinate value difference between each two adjacent first coordinate values according to the coordinate value sequence;
[0118] The coordinate change degree in the first direction is determined according to the difference between the coordinate values.
[0119] In this embodiment, the coordinate change degree of the first direction can be determined by the following formula 3:
[0120] Formula 3
[0121] In the formula, Indicates the coordinate change degree in the X-axis direction, N indicates the number of coordinate value differences, Represents the coordinate value difference between two adjacent first coordinate values in the coordinate value sequence, Represents the mean of the differences between the coordinate values in the coordinate value sequence.
[0122] in, It indicates the degree of difference between the coordinate value difference between two adjacent first coordinate values in the coordinate value sequence and the mean of the coordinate value difference. The greater the degree of difference, the smaller the uniform change degree between the first coordinate values, that is, the greater the coordinate change degree.
[0123] Coordinate change in the X-axis direction When the value is larger, the uniformity of the coordinate value in the X-axis direction is worse, the consistency between the first suspected defect position and the motion trajectory of the glass fiber plain weave cloth is more inconsistent, and the defect reliability probability of the corresponding first suspected defect position is lower.
[0124] Through this embodiment, the coordinate value difference between each two adjacent first coordinate values is used to calculate the coordinate change degree in the first direction, thereby facilitating the subsequent determination of the defect reliability probability of the first suspected defect position according to the coordinate value change degree in the first direction, so as to accurately evaluate the production quality of the glass fiber plain weave cloth.
[0125] Based on the glass fiber plain woven fabric production quality detection method, accordingly, the present application also provides a specific embodiment of the glass fiber plain woven fabric production quality detection system.
[0126] As an optional embodiment, a glass fiber plain weave fabric production quality inspection system includes an image acquisition device, a motion platform, and a controller;
[0127] An image acquisition device, used for acquiring surface feature images of the glass fiber plain weave cloth;
[0128] A motion platform, used for controlling the glass fiber plain weave fabric to move at a uniform speed along a first direction;
[0129] A controller is used to execute the glass fiber plain weave fabric production quality detection method provided in any one of the above aspects.
[0130] In this embodiment, if Figure 7 As shown, a structural schematic diagram of a glass fiber plain weave cloth production quality inspection system is provided. The glass fiber plain weave cloth is placed on a motion platform, and the motion platform controls the glass fiber plain weave cloth to move at a uniform speed along a first direction, and the first direction is the cloth running direction of the glass fiber plain weave cloth.
[0131] The image acquisition device is installed directly above the motion platform to collect the surface feature image of the glass fiber plain weave cloth. At the same time, in order to obtain a more accurate image, the glass fiber plain weave cloth can also be illuminated by light to obtain an image with a greater exposure. Among them, the light irradiation direction is aimed at the image acquisition point of the glass fiber plain weave cloth.
[0132] The controller is connected to the image acquisition device and the motion platform, and is used to send control instructions to the image acquisition device and the motion platform, so as to execute the glass fiber plain weave cloth production quality detection method provided in any one of the above aspects.
[0133] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.
[0134] It should also be noted that the exemplary embodiments mentioned in the present invention describe some methods or systems based on a series of steps or devices. However, the present invention is not limited to the order of the above steps, that is, the steps can be performed in the order mentioned in the embodiments, or in a different order from the embodiments, or several steps can be performed simultaneously.
[0135] The above is only a specific implementation of the present invention. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the system, module and unit described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. It should be understood that the protection scope of the present invention is not limited to this. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be covered within the protection scope of the present invention.
Claims
1. A method for detecting the production quality of glass fiber plain weave cloth, characterized in that: The method comprises: Acquire an image sequence of the glass fiber plain woven fabric in a process of uniformly moving along a first direction, wherein the image sequence includes a plurality of detection images of the glass fiber plain woven fabric, and the detection images in the image sequence are sorted in the order of acquisition time; Performing white spot segmentation on a first detection image in the image sequence to determine a first suspected defect position of the first detection image; Determine the defect reliability probability of the first suspected defect position according to the first suspected defect position of the first detection image and the second suspected defect position of each second detection image in the image sequence, wherein the second detection image is N detection images after the first detection image in the image sequence, where N is a positive integer; When the defect reliability probability of the first suspected defect position is greater than or equal to a first probability threshold, determining that the glass fiber plain weave cloth has a production quality abnormality at the first suspected defect position; The step of acquiring an image sequence of the glass fiber plain weave fabric in a process of uniformly moving along a first direction comprises: Acquire multiple surface feature images of the glass fiber plain weave fabric moving at a uniform speed along a first direction according to a preset acquisition frequency by an image acquisition device; Preprocessing each of the surface feature images to obtain the detection image corresponding to each of the surface feature images, wherein the preprocessing includes at least one of denoising processing and gray value equalization processing; Arrange the detection images in the order of image acquisition time to obtain an image sequence of the glass fiber plain weave fabric; The step of performing white spot segmentation on the first detection image in the image sequence to determine a first suspected defect position of the first detection image includes: Performing white spot segmentation on a first detection image in the image sequence to generate a binary image corresponding to the first detection image; Using the average length and average width of the white spots in the binary image, the binary image is layered to obtain a plurality of horizontal layers and a plurality of vertical layers corresponding to the binary image; Determine the defect position probability of each of the transverse layers according to the distribution of black pixels and white pixels of each of the transverse layers, and determine the defect position probability of each of the longitudinal layers according to the distribution of black pixels and white pixels of each of the longitudinal layers; Determine the transverse layer whose defect position probability is greater than or equal to the second probability threshold as a defective transverse layer, and determine the longitudinal layer whose defect position probability is greater than or equal to the second probability threshold as a defective longitudinal layer; Determine a first suspected defect position of the first detection image according to each of the defect transverse layers and each of the defect longitudinal layers; The determining, according to the first suspected defect position of the first detection image and the second suspected defect position of each second detection image in the image sequence, the defect reliability probability of the first suspected defect position includes: Determine, according to the first coordinate values of the first suspected defect position and each of the second suspected defect positions in the first direction, a coordinate value variation degree in the first direction, wherein the coordinate value variation degree is used to characterize a uniform variation degree between each of the first coordinate values; Determine a coordinate value variance in a second direction according to the second coordinate values of the first suspected defect position and each of the second suspected defect positions, wherein the second direction is perpendicular to the first direction; The defect reliability probability of the first suspected defect position is determined according to the coordinate value variation in the first direction and the coordinate value variance in the second direction.
2. A method for detecting production quality of glass fiber plain weave fabric according to claim 1, characterized in that: The step of performing white spot segmentation on the first detection image in the image sequence to generate a binary image corresponding to the first detection image includes: Acquire the first detection image in the image sequence; The first detection image is input into the white spot segmentation model to obtain the binary image corresponding to the first detection image. The white spot segmentation model is obtained by training a convolutional neural network model using image training samples. The image training samples include the detection image of the glass fiber plain cloth and the corresponding binary label.
3. A method for detecting production quality of glass fiber plain weave fabric according to claim 1, characterized in that: The determining the defect position probability of each of the transverse layers according to the distribution of black pixels and white pixels of each of the transverse layers respectively includes: Creating an image detection window using the average size of white spots in the first detection image as the window size; Traversing the horizontal layer, obtaining the number of black pixels and the number of white pixels in the image detection window each time; The defect position probability of the transverse layer is determined according to the number of black pixels and the number of white pixels in the image detection window each time.
4. A method for detecting production quality of glass fiber plain weave fabric according to claim 3, characterized in that: Determining the defect position probability of the transverse layer according to the number of black pixels and the number of white pixels in the image detection window each time includes: Determine, according to the number of black pixels and the number of white pixels in the image detection window each time, a first variance of the ratio of the number of black pixels to the number of white pixels in the image detection window for odd times, and a second variance of the ratio of the number of black pixels to the number of white pixels in the image detection window for even times; The first variance is added to the second variance and then normalized to obtain the defect position probability of the transverse layer.
5. A method for detecting production quality of glass fiber plain weave fabric according to claim 1, characterized in that: The determining, according to each of the defect transverse layers and each of the defect longitudinal layers, a first suspected defect position of the first detection image comprises: Determining, according to each of the defective transverse layers and each of the defective longitudinal layers, an overlapping area between the defective transverse layers and the defective longitudinal layers; The overlapping area is determined as a first suspected defect position of the first detection image.
6. A method for detecting production quality of glass fiber plain weave fabric according to claim 1, characterized in that: The determining the change degree of the coordinate value in the first direction according to the first suspected defect position and the first coordinate values of each of the second suspected defect positions in the first direction includes: Sorting the first coordinate value of the first suspected defect position and the first coordinate value of each of the second suspected defect positions in the order of the corresponding image acquisition time to obtain a coordinate value sequence; Determine, according to the coordinate value sequence, a coordinate value difference between two adjacent first coordinate values; The coordinate change degree of the first direction is determined according to the coordinate value differences.
7. A glass fiber plain weave cloth production quality inspection system, characterized in that: The system includes an image acquisition device, a motion platform and a controller; The image acquisition device is used to acquire the surface feature image of the glass fiber plain weave cloth; The motion platform is used to control the glass fiber plain weave cloth to move at a uniform speed along a first direction; The controller is used to execute the glass fiber plain weave fabric production quality detection method described in any one of claims 1 to 6.
