An inkjet printing defect recognition method and system

Abstract

The invention relates to an inkjet printing defect recognition method and system, which belongs to the technical field of image processing. The method comprises the following steps: collecting the image of the printing to be detected and the standard image of the corresponding printing template for template matching; calculating the color difference value between each pair of matched pixel points, obtaining the color difference matrix and dividing the color difference range to obtain the high color difference area and the low color difference area; based on the high color difference area, the obvious ink dot defect area is recognized; based on the low color difference area, the defect probability of each pixel point is calculated; based on the defect probability, the corrected Euclidean distance between the pixel points is obtained, the non-obvious ink dot defect area is clustered and recognized; the detection result and distribution of the ink dot defect are counted to realize the feedback control of the inkjet system. Through the pixel point template matching, clustering and color difference correction technology, the problem of missing non-obvious ink dot defects is solved, the accurate recognition of the ink dot defect is realized, the accuracy and reliability of the product quality detection are enhanced, and the printing quality of the product is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a method and system for identifying defects in inkjet printing. BACKGROUND

[0002] Inkjet printing technology is essentially a technology that uses printer principles to dye cloth, abandoning the complex process of making a plate required by traditional printing, and enabling direct inkjet printing on fabric. In the production scene of inkjet printing, the inkjet system is controlled by digital technology, the digital image is scanned and color separation and color management are performed, and according to the digital image data of the corresponding color of each step, different colors of ink are gradually jetted onto the corresponding position of the fabric, one color is jetted each time, and finally a complex color and pattern texture is drawn. Due to the adaptability between ink and fabric material and the control precision of the nozzle, ink dot defects may occur on the surface of the fabric, and the ink dot defects on the surface of the fabric directly affect the determination of the product quality level. Therefore, it is necessary to detect the ink dot defects of the fabric after each step of corresponding color jetting, and to adjust the inkjet system in a timely manner according to the detection results to ensure that the jetting operation of each color meets the standard, thereby ensuring that the final inkjet fabric product is qualified.

[0003] However, when detecting ink dot defects on the fabric, the existing quality detection method based on machine vision adopts a scheme of processing RGB images by threshold segmentation or morphological operation, which excessively relies on fixed or simple adaptive color difference threshold. When dealing with complex multi-color patterns, even under constant illumination and flat fabric conditions, low-contrast non-obvious ink dot defects may be easily missed, for example: a light-colored ink stain on a complex color texture may be ignored when detecting ink dot defects due to the color difference not meeting the threshold. Moreover, missing non-obvious ink dot defects may reduce the detection accuracy of ink dot defects, making it difficult to accurately feedback and control the inkjet system, causing the number and area of ink dot defects to gradually increase in the subsequent printing process of other colors, and seriously affecting the printing quality of the final product. SUMMARY

[0004] In order to improve the detection accuracy of ink dot defects and solve the problem that traditional methods are prone to miss non-obvious ink dot defects, the present application provides a method and system for identifying defects in inkjet printing, and the technical scheme is as follows:

[0005] In a first aspect, the present application provides a method for identifying defects in inkjet printing, comprising the following steps: collecting an image of the printing to be detected, and performing template matching with a standard image of the corresponding printing template to obtain a plurality of pairs of matching pixel points; calculating the color difference value between each pair of matching pixel points to obtain a color difference matrix and dividing the color difference range to obtain a high color difference region and a low color difference region; identifying the obvious ink dot defect region based on the pixel point data of the high color difference region; calculating the defect probability of each pixel point based on the pixel point data of the low color difference region; obtaining the modified Euclidean distance between the pixel points based on the defect probability, clustering and identifying the non-obvious ink dot defect region according to the modified Euclidean distance; and counting the detection results and distribution of the ink dot defects to realize feedback control of the inkjet system.

[0006] The calculation process of the defect probability is as follows: obtaining the mean value of the RGB color value data of all pixel points in the obvious ink dot defect region as the mean value of the obvious defect color, taking the Euclidean distance between the RGB color value of each pixel point in the low color difference region and the mean value of the obvious defect color as the ink dot color difference value, taking the absolute value of the difference between the color difference value of the pixel point and the ink dot color difference value as the corrected color difference value, setting the hyperparameter to scale the corrected color difference value, adding the ink dot color difference value, and then performing reverse mapping to obtain the correction coefficient of each pixel point in the low color difference region, and taking the product of the color difference value of the pixel point and the corresponding correction coefficient as the defect probability of the pixel point.

[0007] Preferably, after the inkjet printing process of each color corresponding ink is completed, the image of the printing to be detected is collected; the standard image of the printing template corresponding to the current printing color in the inkjet system is obtained, the RGB color value and the corresponding coordinate data of all pixel points in the detected printing image and the corresponding standard image are extracted, the detected printing image and the corresponding standard image are matched based on the coordinate data of the pixel points, the pixel point coordinates of the two images are aligned, the pixel points of the two images are one-to-one corresponding, and a plurality of pairs of matching pixel points are obtained.

[0008] Preferably, all pixel points on the detected printing image and the corresponding standard image are traversed by using a nested loop structure, the distance of the RGB color value between the matching pixel points is calculated by using the Euclidean distance formula as the color difference value between the matching pixel points, and after all pixel points are traversed, the color difference matrix of the detected printing image is obtained.

[0009] Preferably, the color difference value in the color difference matrix, the RGB color value data and the corresponding coordinate data of the pixel points in the detected printing image are extracted, the threshold parameter is set based on the distribution of the color difference value, the pixel points with color difference value greater than the threshold parameter are divided into the high color difference region, and the pixel points with color difference value less than or equal to the threshold parameter are divided into the low color difference region.

[0010] Preferably, the color difference value data, RGB color value data and coordinate data corresponding to all pixel points in the high color difference region are extracted, the coordinate, color value and color difference value corresponding to each pixel point in the high color difference region are merged, and a six-dimensional feature vector corresponding to each pixel point is constructed The six-dimensional feature vectors corresponding to all pixel points in the high color difference region are normalized to obtain a normalized feature vector corresponding to each pixel point, and the Euclidean distance between the normalized feature vectors of all pixel points in the high color difference region is calculated using the Euclidean distance formula; the DBSCAN clustering algorithm is executed based on the Euclidean distance and noise points are excluded to obtain a plurality of effective clustering clusters, each of which corresponds to an obvious ink dot defect region.

[0011] Preferably, the color difference value data, RGB color value data and coordinate data corresponding to all pixel points in the low color difference region are extracted, the ink dot color difference value and the corrected color difference value corresponding to each pixel point in the low color difference region are calculated, the super parameter for adjusting the weight of the corrected color difference value is set based on the degree of color difference between the ink dot color and the printing ground color, the product of the super parameter and the corrected color difference value is taken as the corrected color difference value, the sum of the ink dot color difference value and the corrected color difference value is inversely mapped using the natural exponential function to obtain the correction coefficient of each pixel point in the low color difference region, and the product of the color difference value of each pixel point in the low color difference region and the corresponding correction coefficient is taken as the defect probability of the corresponding pixel point.

[0012] Preferably, the coordinate and RGB color value corresponding to each pixel point in the low color difference region are merged to construct a five-dimensional feature vector corresponding to each pixel point in the low color difference region The five-dimensional feature vectors corresponding to all pixel points in the low color difference region are normalized to obtain a normalized feature vector corresponding to each pixel point, the original Euclidean distance between the normalized feature vectors of all pixel points in the low color difference region is calculated using the Euclidean distance formula; the absolute value of the defect probability difference between all pixel points in the low color difference region is taken as the probability difference between the corresponding two pixel points, the probability difference between the two pixel points is positively mapped using the natural exponential function, and the product of the mapping value and the corresponding original Euclidean distance is taken as the correction value of the original Euclidean distance between the two pixel points, and the sum of the correction value and the corresponding original Euclidean distance is taken as the corrected Euclidean distance between the two pixel points.

[0013] Preferably, based on the corrected Euclidean distance between all pixel points in the low color difference area, the DBSCAN clustering algorithm is executed and the noise points are excluded to obtain a plurality of clustering clusters, the mean value of the color difference value data of all pixel points in each clustering cluster is calculated, and the average color difference value is recorded, based on the color difference between the ink point color and the printing ground color, the color difference threshold T is set, when the average color difference value of a certain clustering cluster is greater than T, it is determined that the clustering cluster is an obvious ink point defect area; when the average color difference value of a certain clustering cluster is less than or equal to T, it is determined that the clustering cluster is a normal area.

[0014] Preferably, based on the identification and detection results of the obvious ink point defect area and the non-obvious ink point defect area, the number and area of the ink point defect area generated after the current color jet printing are counted, the number threshold parameter and the area proportion threshold parameter are set, and it is determined whether the printing quality of the current color jet printing is qualified; when the printing quality is determined to be unqualified, based on the coordinate data of all ink point defect area pixel points, the distribution of the ink point defect area is analyzed, and feedback control of the inkjet system is realized in the subsequent jet printing process.

[0015] In the second aspect, the present application provides an inkjet printing defect identification system for realizing the above-mentioned inkjet printing defect identification method, comprising a processor, a memory, a communication interface, an image acquisition device, the processor stores computer program instructions for realizing the above-mentioned inkjet printing defect identification method, the image acquisition device can adjust and fix the shooting angle and the shooting height, the image acquisition device includes a standard light source and a high-resolution camera, the standard light source is a uniform light source with constant color temperature, and the communication interface is in communication connection with the image acquisition device and the inkjet printing equipment.

[0016] Compared with the prior art, the present application has the following advantages:

[0017] The present application effectively solves the problem of non-obvious ink point defect missed detection caused by excessive dependence on fixed threshold in traditional methods through pixel point template matching, clustering and color difference correction technology, can significantly improve the accuracy and reliability of inkjet printing ink point defect detection, and is beneficial to improve the printing quality of products; Moreover, the ink point defects are divided into obvious ink point defects and non-obvious ink point defects, which can facilitate the determination of product quality grade, and through the analysis of the distribution of the ink point defect area, feedback control of the inkjet system can be realized in the subsequent jet printing process, thereby targeted production optimization is realized, and fine management of printing quality and cost is realized. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The flow chart of the inkjet printing defect identification method is shown in the figure.

[0019] Figure 2 The structure block diagram of the inkjet printing defect identification system is shown in the figure. DETAILED DESCRIPTION

[0020] The technical features of the present application will be further described in detail below with reference to the accompanying drawings, so that those skilled in the relevant art can understand.

[0021] An inkjet printing defect recognition method, the implementation process is as shown in Figure 1 The specific implementation steps are as follows:

[0022] Step S1: Collect the image of the printing to be detected, and perform template matching with the standard image of the corresponding printing template to obtain a plurality of pairs of matching pixel points;

[0023] Specifically, a high-resolution camera is deployed inside the inkjet system and is matched with a uniform light source with constant color temperature. After the printing process of each color corresponding ink is completed, the printing fabric to be detected is placed flat, and the image of the printing to be detected is collected by the high-resolution camera under the irradiation of the uniform light source with constant color temperature. At the same time, the standard image of the printing template corresponding to the current printing color in the inkjet system is obtained, the RGB color values and the corresponding coordinate data of all pixel points in the image of the printing to be detected and the corresponding standard image are extracted, the image of the printing to be detected and the corresponding standard image are matched based on the coordinate data of the pixel points, and the pixel point coordinates of the two images are aligned to make the pixel points of the two images one-to-one corresponding, thereby obtaining a plurality of pairs of matching pixel points.

[0024] In addition, the photoelectric sensor can be triggered to shoot synchronously with the production line. After the printing process of each color corresponding ink is completed, the RGB image of the printing fabric to be detected needs to be collected at the same shooting angle, under the same lighting and object distance conditions, in order to eliminate lighting and position interference and ensure the comparability between images.

[0025] Step S2: Calculate the color difference value between each pair of matching pixel points, obtain the color difference matrix, and divide the color difference range to obtain the high color difference area and the low color difference area;

[0026] Specifically, all pixel points on the image of the printing to be detected and the corresponding standard image are traversed by using a nested loop structure. Since the RGB color value data of each pixel point can be regarded as a three-dimensional vector, the Euclidean distance formula is used to calculate the distance of the RGB color value between the matching pixel points as the color difference value between the matching pixel points, which is used to quantify the color deviation degree between each pixel point on the image of the printing to be detected and the corresponding pixel point on the corresponding standard image. After traversing all pixel points, the color difference matrix of the image of the printing to be detected is obtained.

[0027] In addition, the color difference values ​​in the color difference matrix and the RGB color values ​​and corresponding coordinate data of the pixels in the printed image to be detected are extracted. A threshold parameter is set based on the distribution of the color difference values. Pixels with color difference values ​​greater than the threshold parameter are divided into high color difference regions, and pixels with color difference values ​​less than or equal to the threshold parameter are divided into low color difference regions. For example, the top 70% of the pixels with color difference values ​​from largest to smallest in the color difference matrix are extracted as high color difference regions, and the remaining 30% of the pixels are divided into low color difference regions. This is used to initially screen the area range of obvious ink spot defects and the area range of potential inconspicuous ink spot defects.

[0028] Step S3: Identify areas with obvious ink spot defects based on pixel data in areas with high color difference;

[0029] Specifically, the color difference value data, RGB color value data, and coordinate data corresponding to all pixels in the high color difference region are extracted. The coordinates, color values, and color difference values ​​corresponding to each pixel in the high color difference region are merged to construct a six-dimensional feature vector corresponding to each pixel. The six-dimensional feature vectors corresponding to all pixels in the high color difference region are standardized to obtain a normalized feature vector for each pixel. The Euclidean distance formula is used to calculate the pairwise Euclidean distance between the normalized feature vectors of all pixels in the high color difference region. The purpose of standardizing the six-dimensional feature vectors corresponding to all pixels is to eliminate the difference in dimensions and avoid large numerical dimensions such as coordinates from dominating the Euclidean distance, so that the features represented by each dimension in the vector contribute equally to the clustering result.

[0030] Furthermore, based on the Euclidean distance between each pair of normalized feature vectors of all pixels within the high color difference region, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm is executed, and noise points are eliminated to obtain several effective clusters. Each effective cluster corresponds to a region with a significant ink spot defect. The specific clustering process is as follows: [The following is a separate, unrelated section:] ...set according to the specific implementation scenario and the experience of the implementers... Domain radius and The value of the minimum cluster, for example: can be set The radius of the domain is 0.2. The minimum cluster size is 7, when the Euclidean distance between the feature vectors of pixels in the high color difference region is less than or equal to 7. When the corresponding pixels are considered as neighbors, several clusters are formed through density connectivity. Noise points marked as -1 during the DBSCAN clustering process are excluded, and several effective clusters can be obtained.

[0031] Step S4: Calculate the defect probability of each pixel point based on the pixel point data in the low color difference region;

[0032] Since the obviousness of the same color ink dot is different under different fabric printing ground colors, the color difference correction of the ink dot defects on different ground colors can make the not obvious ink dot defects relatively obvious. Therefore, the defect probability of each pixel point in the low color difference region needs to be color difference corrected to enlarge the obviousness of the ink dot.

[0033] Specifically, the mean value of the RGB color value data of all pixel points in the obvious ink dot defect region is obtained as the obvious defect color mean value. The color difference value data, RGB color value data and coordinate data of all pixel points in the low color difference region are extracted. The distance between the RGB color value of each pixel point in the low color difference region and the obvious defect color mean value is calculated as the ink dot color difference value of each pixel point in the low color difference region using the Euclidean distance formula. The smaller the color difference value of the pixel point in the low color difference region, the more likely it is that the color of the pixel point itself is lighter, or the closer the ink color is to the ground color of the printing template. The absolute value of the difference between the color difference value of each pixel point in the low color difference region and the ink dot color difference value is taken as the corrected color difference value of each pixel point. The smaller the corrected color difference value, the more likely it is that the pixel point is a hidden ink dot defect, and the more it should be enlarged to highlight the difference between the ink color and the ground color of the printing template, so that the not obvious ink dot is as obvious as possible to avoid missing the not obvious ink dot.

[0034] In addition, based on the obviousness of the color difference between the ink color and the printing ground color, an hyperparameter for adjusting the weight of the corrected color difference value is set. The product of the hyperparameter and the corrected color difference value is taken as the corrected color difference value. The sum of the ink dot color difference value and the corrected color difference value is inversely mapped using a natural exponential function to obtain the correction coefficient of each pixel point in the low color difference region. The product of the color difference value of each pixel point in the low color difference region and the corresponding correction coefficient is taken as the defect probability of the corresponding pixel point. The defect probability represents the probability of the corresponding pixel point belonging to the not obvious ink dot defect.

[0035] The defect probability calculation formula of the pixel point in the low color difference region is:

[0036]

[0037] In the formula, represents the defect probability of the i-th pixel point in the low color difference region, represents the color difference value of the i-th pixel point in the low color difference region, represents the ink dot color difference value of the i-th pixel point in the low color difference region, represents the corrected color difference value of the i-th pixel point in the low color difference region, denotes the natural exponential function. denotes the natural exponential function.

[0038] Step S5: obtaining the corrected Euclidean distance between the pixel points based on the defect probability, clustering and identifying the non-obvious ink dot defect region according to the corrected Euclidean distance;

[0039] When the defect probabilities of two pixel points are similar, maintaining the original distance relationship can protect the natural focusing of homogeneous ink dots, and when the defect probabilities of two pixel points are significantly different, the original distance between the two pixel points needs to be significantly enlarged, that is, even if the position coordinates of the two pixel points are similar, the two pixel points must be forced to separate to avoid the error clustering of the high-probability pixel points and the low-probability pixel points, therefore, before clustering the pixel points in the low-color-difference region, the original Euclidean distance between the pixel points needs to be corrected;

[0040] Specifically, the coordinates and RGB color values corresponding to each pixel point in the low-color-difference region are merged to construct a five-dimensional feature vector corresponding to each pixel point in the low-color-difference region The five-dimensional feature vectors corresponding to all pixel points in the low-color-difference region are standardized to obtain a normalized feature vector corresponding to each pixel point, and the original Euclidean distance between the normalized feature vectors of all pixel points in the low-color-difference region is calculated by using the Euclidean distance formula; the absolute value of the defect probability difference between all pixel points in the low-color-difference region is taken as the probability difference between the corresponding two pixel points, the probability difference between the two pixel points is positively mapped by using the natural exponential function, and the product of the mapping value and the corresponding original Euclidean distance is taken as the correction value of the original Euclidean distance between the two pixel points, and the sum of the correction value and the corresponding original Euclidean distance is taken as the corrected Euclidean distance between the two pixel points;

[0041] The specific calculation formula of the corrected Euclidean distance between two pixel points in the low-color-difference region is as follows:

[0042]

[0043] In the formula, p i and p j denote the defect probabilities of the pixel points i and j in the low-color-difference region respectively, denotes the corrected Euclidean distance between the pixel points i and j in the low-color-difference region, denotes the original Euclidean distance between the pixel points i and j in the low-color-difference region, denotes the probability difference between the pixel points i and j in the low-color-difference region, denotes the defect probability of the pixel point i in the low-color-difference region, denotes the defect probability of the pixel point j in the low-color-difference region, denotes the natural exponential function.

[0044] In addition, based on the corrected Euclidean distance between all pixel points in the low color difference region, the DBSCAN clustering algorithm is executed and noise points are excluded to obtain a plurality of clustering clusters, and the clustering process is similar to that in step S3; the mean value of the color difference value data of all pixel points in each clustering cluster is calculated, and the mean color difference value is recorded, and based on the degree of color difference between the ink dot color and the printing ground color, a color difference threshold T is set, wherein the color difference threshold T can be an empirical value of 20; when the mean color difference value of a certain clustering cluster is greater than T, it is determined that the clustering cluster is an obvious ink dot defect region; when the mean color difference value of a certain clustering cluster is less than or equal to T, it is determined that the clustering cluster is a normal region.

[0045] Step S6: The detection results and distribution of the ink dot defects are counted to realize feedback control of the inkjet system.

[0046] Specifically, based on the identification and detection results of the obvious ink dot defect region and the non-obvious ink dot defect region, the number and area of the ink dot defect region generated after the current color printing are counted, the ratio of the sum of the areas of all ink dot defect regions to the area of the printing image to be detected is calculated to obtain the area ratio of the ink dot defect region, and the number threshold parameter and the area ratio threshold parameter are set; when the number and area ratio of the ink dot defect region do not exceed the corresponding threshold parameter, it is determined that the printing quality of the current color printing is qualified; when either the number or the area ratio of the ink dot defect region exceeds the corresponding threshold parameter, it is determined to be unqualified; when the printing quality is determined to be unqualified, the distribution of the ink dot defect region is analyzed based on the coordinate data of all ink dot defect region pixels, and feedback control of the inkjet system is realized in the subsequent printing process.

[0047] The application further discloses an inkjet printing defect identification system for realizing the inkjet printing defect identification method, which has the structure as shown in the figure. Figure 2 The application further discloses an inkjet printing defect identification system for realizing the inkjet printing defect identification method, which has the structure as shown in the figure.

[0048] The embodiments contained in the application only describe the preferred embodiments of the application, and are not limited to the precise structure described above and shown in the drawings, and various modifications and changes can be made without departing from the protection scope thereof; various modifications and changes made by the engineering technicians in the field to the technical solutions of the application without departing from the design idea of the application shall fall within the protection scope of the application.

Claims

1. A method for identifying defects in inkjet printing, characterized in that: The image of the to-be-detected stamp is collected, and template matching is performed with the standard image of the corresponding stamp template to obtain a plurality of pairs of matching pixel points; the color difference value between each pair of matching pixel points is calculated to obtain a color difference matrix and divide the color difference range to obtain a high color difference region and a low color difference region; based on the pixel point data of the high color difference region, an obvious ink dot defect region is identified; Based on the pixel point data of the low color difference region, the defect probability of each pixel point is calculated; Based on the defect probability, the corrected Euclidean distance between the pixel points is obtained, and the non-obvious ink dot defect region is clustered and identified according to the corrected Euclidean distance; The detection results and distribution of the ink dot defects are counted to realize feedback control of the inkjet system. The calculation process of the defect probability is as follows: the mean value of the RGB color value data of all pixel points in the obvious ink dot defect region is obtained as the mean value of the obvious defect color, the Euclidean distance between the RGB color value of each pixel point in the low color difference region and the mean value of the obvious defect color is taken as the ink dot color difference value, the absolute value of the difference between the color difference value of the pixel point and the ink dot color difference value is taken as the corrected color difference value, the corrected color difference value is scaled by setting a hyperparameter and added to the ink dot color difference value, and then reverse mapping is performed to obtain the correction coefficient of each pixel point in the low color difference region, and the product of the color difference value of the pixel point and the corresponding correction coefficient is taken as the defect probability of the pixel point.

2. The method according to claim 1, wherein The image of the to-be-detected stamp is collected, and template matching is performed with the standard image of the corresponding stamp template to obtain a plurality of pairs of matching pixel points; the color difference value between each pair of matching pixel points is calculated to obtain a plurality of pairs of matching pixel points; the color difference value between each pair of matching pixel points is calculated to obtain a color difference matrix and divide the color difference range to obtain a high color difference region and a low color difference region; based on the pixel point data of the high color difference region, an obvious ink dot defect region is identified; 3. The method according to claim 1, wherein Based on the pixel point data of the low color difference region, the defect probability of each pixel point is calculated; 4. The method according to claim 1, wherein Based on the defect probability, the corrected Euclidean distance between the pixel points is obtained, and the non-obvious ink dot defect region is clustered and identified according to the corrected Euclidean distance; The detection results and distribution of the ink dot defects are counted to realize feedback control of the inkjet system. The calculation process of the defect probability is as follows: the mean value of the RGB color value data of all pixel points in the obvious ink dot defect region is obtained as the mean value of the obvious defect color, the Euclidean distance between the RGB color value of each pixel point in the low color difference region and the mean value of the obvious defect color is taken as the ink dot color difference value, the absolute value of the difference between the color difference value of the pixel point and the ink dot color difference value is taken as the corrected color difference value, the corrected color difference value is scaled by setting a hyperparameter and added to the ink dot color difference value, and then reverse mapping is performed to obtain the correction coefficient of each pixel point in the low color difference region, and the product of the color difference value of the pixel point and the corresponding correction coefficient is taken as the defect probability of the pixel point. The image of the to-be-detected stamp is collected, and template matching is performed with the standard image of the corresponding stamp template to obtain a plurality of pairs of matching pixel points; the color difference value between each pair of matching pixel points is calculated to obtain a plurality of pairs of matching pixel points; the color difference value between each pair of matching pixel points is calculated to obtain a color difference matrix and divide the color difference range to obtain a high color difference region and a low color difference region; based on the pixel point data of the high color difference region, an obvious ink dot defect region is identified; The calculation process of the defect probability is as follows: the mean value of the RGB color value data of all pixel points in the obvious ink dot defect region is obtained as the mean value of the obvious defect color, the Euclidean distance between the RGB color value of each pixel point in the low color difference region and the mean value of the obvious defect color is taken as the ink dot color difference value, the absolute value of the difference between the color difference value of the pixel point and the ink dot color difference value is taken as the corrected color difference value, the corrected color difference value is scaled by setting a hyperparameter and added to the ink dot color difference value, and then reverse mapping is performed to obtain the correction coefficient of each pixel point in the low color difference region, and the product of the color difference value of the pixel point and the corresponding correction coefficient is taken as the defect probability of the pixel point.

5. The method according to claim 1, wherein the method is characterized by, The identifying the obvious ink dot defect region comprises: extracting color difference value data, RGB color value data and coordinate data corresponding to all pixel points in the high color difference region, merging the coordinates, RGB color values and color difference values corresponding to each pixel point in the high color difference region, and constructing a six-dimensional feature vector corresponding to each pixel point The six-dimensional feature vectors corresponding to all pixel points in the high color difference region are standardized to obtain a normalized feature vector corresponding to each pixel point, and the Euclidean distance between the normalized feature vectors of all pixel points in the high color difference region is calculated by using a Euclidean distance formula; a DBSCAN clustering algorithm is executed based on the Euclidean distance and noise points are excluded to obtain a plurality of effective clustering clusters, and each effective clustering cluster corresponds to an obvious ink dot defect region.

6. The method according to any one of claims 1 to 5, wherein, The calculating the defect probability of each pixel point comprises: extracting color difference value data, RGB color value data and coordinate data corresponding to all pixel points in the low color difference area, calculating the ink dot color difference value and the corrected color difference value corresponding to each pixel point in the low color difference area, setting an hyperparameter for adjusting the weight of the corrected color difference value based on the obviousness of the color difference between the ink dot color and the printing ground color, taking the product of the hyperparameter and the corrected color difference value as the corrected color difference value, using a natural exponential function to inversely map the sum of the ink dot color difference value and the corrected color difference value, obtaining the correction coefficient of each pixel point in the low color difference area, and taking the product of the color difference value of each pixel point in the low color difference area and the corresponding correction coefficient as the defect probability of the corresponding pixel point.

7. The method according to claim 6, wherein the step of identifying the defect is performed by using a neural network. The corrected Euclidean distance between the pixel points is obtained based on the defect probability, including: merging the coordinates and RGB color values of each pixel point in the low color difference area to construct a five-dimensional feature vector corresponding to each pixel point in the low color difference area The five-dimensional feature vectors corresponding to all pixel points in the low color difference area are normalized to obtain a normalized feature vector corresponding to each pixel point, and the original Euclidean distance between the normalized feature vectors of all pixel points in the low color difference area is calculated by using the Euclidean distance formula; the absolute value of the defect probability difference between all pixel points in the low color difference area is taken as the probability difference between the corresponding two pixel points, the probability difference between the two pixel points is positively mapped by using a natural exponential function, and the product of the mapping value and the corresponding original Euclidean distance is taken as the correction value of the original Euclidean distance between the two pixel points, and the sum of the correction value and the corresponding original Euclidean distance is taken as the corrected Euclidean distance between the two pixel points.

8. The method according to claim 7, wherein the step of identifying the defect is performed by using a neural network. The clustering and identifying the non-obvious ink dot defect area according to the corrected Euclidean distance comprises: based on the corrected Euclidean distance between all pixel points in the low color difference area, executing a DBSCAN clustering algorithm and excluding noise points to obtain a plurality of clustering clusters, calculating the mean value of the color difference value data of all pixel points in each clustering cluster, denoted as the average color difference value, setting a color difference threshold T based on the obviousness of the color difference between the ink dot color and the printing ground color, and determining that a clustering cluster is a non-obvious ink dot defect area when the average color difference value of the clustering cluster is greater than T; and determining that a clustering cluster is a normal area when the average color difference value of the clustering cluster is less than or equal to T.

9. The method according to claim 1, wherein The counting the detection result and the distribution of the ink dot defect and realizing feedback control of the inkjet system comprises: based on the identification and detection result of the obvious ink dot defect area and the non-obvious ink dot defect area, counting the number and area of the ink dot defect area generated after the current color is printed, setting a number threshold parameter and an area proportion threshold parameter, and determining whether the printing quality of the current color is qualified; and when the printing quality is determined to be unqualified, analyzing the distribution of the ink dot defect area based on the coordinate data of all pixel points of the ink dot defect area, and realizing feedback control of the inkjet system in the subsequent printing process.

10. An inkjet printing defect recognition system, characterized by: The device comprises a processor, a memory, a communication interface, and an image acquisition device, the processor stores computer program instructions for implementing the inkjet printing defect identification method of any one of claims 1 to 9, the image acquisition device can adjust and fix the shooting angle and the shooting height, the image acquisition device comprises a standard light source and a high-resolution camera, the standard light source is a uniform light source with constant color temperature, and the communication interface is in communication connection with the image acquisition device and the inkjet printing equipment.

Images