A visual inspection method and system for digital printing quality of highly weather-resistant printed fabric

Abstract

The present invention relates to the technical field of image data processing, and in particular to a method and system for visually detecting the digital printing quality of highly weather-resistant printed fabrics. The method comprises the following steps: obtaining a surface image and a template image of the highly weather-resistant printed fabric, and processing the surface image through an image algorithm to obtain color features of different channels. The total value of the difference between each pixel value of pixels at the same position of the surface image and the template image is determined according to the color features of the different channels, the surface image and the template image. The method comprises the following steps: determining the possibility of color bleeding at the color level of each pixel point in the preset local window according to the preset local window and the total value of the pixel value difference. This method is helpful to avoid the possibility that the surface image is distorted and cannot highlight defects due to the influence of light and other conditions. The method comprises the following steps: determining a potential color bleeding area according to the possibility, judging the digital printing quality of the highly weather-resistant printed fabric, and improving the accuracy of the potential color bleeding area and the accuracy of quality detection.

Description

Technical Field

[0001] The invention relates to the technical field of image data processing, and in particular to a method and system for visually detecting the digital printing quality of highly weather-resistant printed cloth. Background Art

[0002] Digital printing is a printing method that uses digital technology. Digital printing technology is a high-tech product that integrates mechanical, computer and electronic information technology and has gradually developed with the continuous development of computer technology. The working principle of digital printing is basically the same as that of an inkjet printer. It uses digital patterns, measures and matches colors through computers, and prints the designed patterns directly on the fabric. However, since different colors of ink or dye may have differences in drying speed, viscosity and absorbency, some color layers dry faster while other color layers are still wet, resulting in color bleeding.

[0003] At present, the detection algorithm for color bleeding has low segmentation accuracy between two low-contrast colors due to the influence of color complexity in the pattern design process itself, and the directional features are used to significantly enhance the area within the image. Under the influence of light and other conditions, some defects may not be highlighted, resulting in low accuracy of the final quality detection. Summary of the invention

[0004] In order to solve the above technical problems, the purpose of the present invention is to provide a method and system for visual inspection of digital printing quality of highly weather-resistant printed fabrics. The technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present application provides a method for visually inspecting the quality of digital printing of highly weather-resistant printed fabrics, comprising:

[0006] Acquire a surface image and a template image of a highly weather-resistant printed fabric, and process the surface image by an image algorithm to obtain color features of different channels;

[0007] Determine the total value of the difference between each pixel value of the pixel points at the same position of the surface image and the template image according to the color features of different channels, the surface image and the template image;

[0008] Determine, based on the preset local window and the total value of the pixel value difference, the possibility of color bleeding at the color level within the preset local window for each pixel point;

[0009] According to the possibility, the potential bleeding area is determined to judge the digital printing quality of the highly weather-resistant printed fabric.

[0010] In one embodiment, the processing of the surface image by an image algorithm to obtain color features of different channels includes:

[0011] Filtering and downsampling the surface image by a Gaussian filter to obtain multi-channel images at different scales;

[0012] A color Gaussian pyramid is constructed according to the multi-channel image, and color features of different channels at different scales are determined from the color Gaussian pyramid.

[0013] In one embodiment, determining the total value of the difference between each pixel value of the pixel points at the same position of the surface image and the template image according to the color features of different channels, the surface image, and the template image comprises:

[0014] According to the color features of different channels, converting the surface image into a first four-channel image and converting the template image into a second four-channel image;

[0015] Determine, according to the first four-channel image and the second four-channel image, a difference value between each pixel value of a pixel point at the same position of the surface image and the template image in each channel;

[0016] The sum of the difference values ​​of all the channels corresponding to each pixel point at the same position is calculated respectively to obtain the total difference value of each pixel value of the pixel point at the same position between the surface image and the template image.

[0017] In one embodiment, determining the possibility of color bleeding at the color level of each pixel point within the preset local window according to the preset local window and the total value of the pixel value difference includes:

[0018] Determine a first pixel point from the surface image, and determine each second pixel point in the preset local window, except the first pixel point, with the first pixel point as the center of the preset local window;

[0019] respectively determining a difference between the total pixel value difference value of the first pixel point and the total pixel value difference value of each of the second pixel points, and determining an absolute value of each of the differences;

[0020] Determine a first sum of all the differences and a second sum of all the absolute values, and determine a specific trend of the first pixel point within the preset local window according to a ratio of the first sum to the second sum;

[0021] Determining, according to a specific trend of the first pixel point in the preset local window, a possibility that the first pixel point has color bleeding in the color layer in the preset local window;

[0022] Return to the step of determining the first pixel point from the surface image until a possibility of color bleeding at the color level within the preset local window is obtained for each pixel point.

[0023] In one embodiment, determining the possibility that the first pixel point has bleeding in the color level in the preset local window according to the specific trend of the first pixel point in the preset local window includes:

[0024] The possibility of color bleeding of the first pixel point in the preset local window at the color level is determined based on the product of the specific trend of the first pixel point in the preset local window and the total value of the pixel value difference of the first pixel point.

[0025] In one embodiment, determining the potential bleeding area according to the possibility includes:

[0026] Determine an edge image of the template image, and determine the directional authenticity of the color gradient of each pixel in the surface image according to the edge image, the surface image and the possibility;

[0027] According to the authenticity, determining the weight of each pixel point in the surface image on the color bleeding problem in terms of color features;

[0028] Determining a target saliency map according to the color features of different channels and the performance weights;

[0029] The target saliency map is subjected to threshold segmentation to determine potential bleeding areas.

[0030] In one embodiment, determining the directional authenticity of the color gradient of each pixel in the surface image according to the edge image, the surface image, and the possibility includes:

[0031] Determine a third pixel point from the surface image, and determine each fourth pixel point in the preset local window except the first pixel point, with the third pixel point as the center of the preset local window;

[0032] Obtaining respective possibility differences according to the difference between the possibility of the third pixel point and the possibility of the fourth pixel point, and using the absolute value of the possibility difference as the vector modulus between the third pixel point and each of the fourth pixel points to construct a plurality of initial vectors;

[0033] Accumulating all the initial vectors to obtain a first target vector passing through the third pixel point, selecting a first target fourth pixel point located in the edge image and having the shortest distance from the third pixel point along the reverse direction of the first target vector, selecting a preset number of second target fourth pixel points between the third pixel point and the first target fourth pixel point along the reverse direction of the first target vector, and arranging the third pixel point, the first target fourth pixel point, and a preset number of the second target fourth pixel points in sequence along the reverse direction of the first target vector to obtain a pixel point set;

[0034] Determine a second target vector between the first target fourth pixel and each of the second target fourth pixel in the pixel set, and determine a vector direction difference between two adjacent pixels according to the first target vector and each of the second target vectors;

[0035] According to the vector direction difference, the directional authenticity of the color gradient of the third pixel point in the surface image is determined, and the step of determining the third pixel point from the surface image is returned until the directional authenticity of the color gradient of each pixel point in the surface image is obtained.

[0036] In one embodiment, determining the performance weight of each pixel point in the surface image on the color bleeding problem in terms of color features according to the authenticity includes:

[0037] Respectively determine the range of the vector modulus length corresponding to each of the pixel point sets, determine the first arrangement sequence number of the third pixel point in each of the pixel point sets, and determine the second arrangement sequence number of the fifth pixel point with the largest vector modulus length in each of the pixel point sets;

[0038] Determine the sequence number difference between the first arrangement sequence number and the second arrangement sequence number respectively, and obtain the authenticity of the module length of each third pixel point in the color gradient according to the extreme difference of the vector module length corresponding to each pixel point set and the ratio of the sequence number difference;

[0039] Obtaining a weight of the third pixel in the surface image on the color bleeding problem in terms of color features according to the product of the direction authenticity of the third pixel and the modulus authenticity and a normalization function;

[0040] Return to the step of determining the third pixel point from the surface image until the weight of each pixel point in the surface image on the color bleeding problem in terms of color features is obtained.

[0041] In one implementation, the color feature includes a pixel value; and determining the target saliency map according to the color features of different channels and the representation weights includes:

[0042] In the multi-channel images at different scales, a difference operation is performed on the pixel value of the R channel and the pixel value of the G channel respectively to obtain a first difference feature of each pixel point corresponding to the multi-channel images at different scales, and a difference operation is performed on the pixel value of the B channel and the pixel value of the Y channel respectively to obtain a second difference feature of each pixel point corresponding to the multi-channel images at different scales;

[0043] Determine the product of the first differential feature and the performance weight of the pixel corresponding to the first differential feature, determine the color bleeding differential feature value of each pixel according to the sum of the product and the second differential feature, and determine the first saliency map of color bleeding at different scales according to the color bleeding differential feature value of the pixel;

[0044] According to the multi-channel images at different scales, construct the brightness Gaussian pyramid and direction pyramid;

[0045] Determine a first brightness channel image of the brightness Gaussian pyramid and a first directional characteristic image of the direction pyramid at a first scale, and determine a second brightness channel image of the brightness Gaussian pyramid and a second directional characteristic image of the direction pyramid at a second scale, wherein the first scale has a higher degree of fineness than the second scale;

[0046] interpolating the second brightness channel image to the first brightness channel image and performing matrix subtraction to obtain a second saliency map, and interpolating the second directional feature image to the first directional feature image and performing matrix subtraction to obtain a third saliency map;

[0047] A target saliency map is obtained according to the first saliency map, the second saliency map and the third saliency map.

[0048] In a second aspect, the embodiment of the present application provides a visual inspection system for digital printing quality of highly weather-resistant printed fabrics, comprising:

[0049] An acquisition module is used to acquire a surface image and a template image of the high-weather-resistant printed fabric, and process the surface image through an image algorithm to obtain color features of different channels;

[0050] A first determination module, configured to determine a total value of differences between pixel values ​​of pixels at the same position of the surface image and the template image according to the color features of different channels, the surface image and the template image;

[0051] A second determination module is used to determine the possibility of color bleeding at the color level of each pixel point within the preset local window according to the preset local window and the total value of the pixel value difference;

[0052] The third determination module is used to determine the potential bleeding area according to the possibility to judge the digital printing quality of the high weather-resistant printed cloth.

[0053] The present invention has the following beneficial effects:

[0054] By acquiring the surface image and template image of the highly weather-resistant printed fabric and processing the surface image through an image algorithm, the color characteristics of different channels are obtained. According to the color characteristics of different channels, the surface image and the template image, the total difference value of each pixel value of the pixel points at the same position of the surface image and the template image is determined. According to the preset local window and the total difference value of the pixel value, the possibility of color bleeding at the color level of each pixel point in the preset local window is determined. The total difference value of the pixel value between the color characteristics of the pixel points in different channels and the template image is analyzed. According to the difference in the total difference value of the pixel value of the local area of ​​the preset local window, the possibility of color bleeding at the color level of the pixel point in the local area is determined. This is conducive to avoiding the distortion of the surface image due to the influence of light and other conditions and the inability to highlight defects. According to the possibility, the potential color bleeding area is determined to judge the digital printing quality of the highly weather-resistant printed fabric, and improve the accuracy of the potential color bleeding area and the accuracy of quality detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] 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.

[0056] Figure 1 A schematic flow chart of the steps of a method for visually inspecting the quality of digital printing of highly weather-resistant printed fabric provided by one embodiment of the present invention;

[0057] Figure 2 A schematic diagram of a potential color bleeding area provided by one embodiment of the present invention;

[0058] Figure 3 This is a structural block diagram of a visual inspection system for digital printing quality of highly weather-resistant printed fabric provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0059] 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 the digital printing quality visual inspection method and system for highly weather-resistant printed fabrics proposed by the present invention, its specific implementation, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0060] 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.

[0061] It should be noted that the term “exemplary” in the embodiments of the present application refers to examples listed for the convenience of explanation, and other embodiments are not limited to the examples listed.

[0062] The specific scheme of the digital printing quality visual inspection method and system for highly weather-resistant printed fabric provided by the present invention is described in detail below with reference to the accompanying drawings.

[0063] See also Figure 1 , which shows a flow chart of a visual inspection method for digital printing quality of highly weather-resistant printed fabric provided by an embodiment of the present invention. The visual inspection method for digital printing quality of highly weather-resistant printed fabric may at least include steps S100-S400:

[0064] S100, obtaining a surface image and a template image of the high-weather-resistant printed fabric, and processing the surface image by an image algorithm to obtain color features of different channels.

[0065] S200, determining the total difference between each pixel value of pixel points at the same position of the surface image and the template image according to the color features of different channels, the surface image and the template image.

[0066] S300, determining the possibility of color bleeding at the color level within the preset local window for each pixel point according to the preset local window and the total value of the pixel value difference.

[0067] S400, based on the possibility, determine the potential bleeding area to judge the digital printing quality of high weather-resistant printed fabrics.

[0068] The technical solution of the embodiment of the present application obtains the surface image and template image of the highly weather-resistant printed cloth, and processes the surface image through an image algorithm to obtain color features of different channels. According to the color features of different channels, the surface image and the template image, the total difference value of each pixel value of the pixel points at the same position of the surface image and the template image is determined. According to the preset local window and the total difference value of the pixel value, the possibility of color bleeding at the color level of each pixel point in the preset local window is determined. The total difference value of the pixel value between the color features of the pixel points in different channels and the template image is analyzed, and according to the difference in the total difference value of the pixel value of the local area of ​​the preset local window, the possibility of color bleeding at the color level of the pixel point in the local area is judged, which is conducive to avoiding the distortion of the surface image due to the influence of light and other conditions and the inability to highlight defects. According to the possibility, the potential color bleeding area is determined to judge the digital printing quality of the highly weather-resistant printed cloth, and improve the accuracy of the potential color bleeding area and the accuracy of quality detection.

[0069] In one embodiment, according to the actual process of the actual high-weather-resistant printed fabric production process, the fabric needs to be stretched and tightened, and the printed fabric at this time is sampled and tested after the spraying is completed, for example, the surface image of the high-weather-resistant printed fabric is collected along the moving direction of the high-weather-resistant printed fabric on the production line at the same interval using an industrial camera, and the collected surface image is transmitted to the control terminal for subsequent processing. In addition, the control terminal obtains the template image of the high-weather-resistant printed fabric stored in advance, that is, the image of the high-weather-resistant printed fabric with normal detection quality and no bleeding, for subsequent processing, such as subsequent processing based on the itti (visual attention model) algorithm. The itti algorithm combines the human eye recognition preference, and at the same time, based on prior knowledge, digital printing adopts the principle of four-color separation, that is, through the combination of four colors of C (cyan), M (magenta), Y (yellow) and K (black) to achieve rich color expression. This technology can accurately control the color output, making the details of the pattern richer and more delicate.

[0070] In one embodiment, the surface image can be preprocessed. Since the possible bleeding area has a low contrast with the surrounding normal area, the embodiment of the present application does not use filtering and denoising to prevent the defects from being over-smoothed, but uses semantic segmentation technology to remove the background area and retain the fabric area to obtain a new surface image for subsequent processing.

[0071] In one embodiment, in step S100, the surface image is processed by an image algorithm to obtain color features of different channels, including:

[0072] First, the surface image is filtered and downsampled by a Gaussian filter to obtain multi-channel images at different scales σ. For example, the surface image is first represented as a Gaussian pyramid at the 0th layer, and the 1st to 8th layers are filtered and downsampled by a 3×3 Gaussian filter, respectively, with sizes of 1 / 2, 1 / 4, 1 / 8...1 / 256 of the surface image, and 9 scales are used as an example for explanation, so that multi-channel images at 9 different scales can be obtained, specifically three-channel images, with scales σ∈{0, 1, 2,...8}.

[0073] Secondly, according to the multi-channel image, a color Gaussian pyramid is constructed, and the color features of different channels at different scales are determined from the color Gaussian pyramid, specifically the four color features of RGBY, which are equivalent to the color features of four channels. The specific calculation process can be based on existing means and will not be repeated here.

[0074] It should be noted that, according to prior knowledge, since different colors of ink or dye may have differences in drying speed, viscosity and absorbency, some color layers dry faster while other color layers are still wet, resulting in bleeding; when bleeding occurs, a new outline is generated between the bleeding area and the invaded pattern area according to the principle of color superposition, which then presents unclear colors, seriously affecting the beauty and quality of the printed products. Existing visual detection methods, such as the conventional itti algorithm, can only extract areas with high contrast, while the bleeding area has a local penetration trend, and its color change amplitude is relatively gentle. Directly using this algorithm cannot significantly highlight the bleeding area, so it is necessary to identify the bleeding area; at the same time, due to the influence of light, the color itself in the image actually sampled from the printed cloth will change to a certain extent, and directly performing a differential operation based on the template and the surface image will be affected by light, resulting in a large error fluctuation in the difference, which indirectly affects the identification of the bleeding area. Therefore, according to prior knowledge, the color and shape of the actual pattern of digital printing will be generated and stored in the control device in advance, and the template image will be stored, and the printing accuracy is high.

[0075] In one implementation, step S200 includes steps S201-S203:

[0076] S201 . Convert a surface image into a first four-channel image and convert a template image into a second four-channel image according to color features of different channels.

[0077] Optionally, based on the RGBY color features of the four channels obtained above, the surface image is converted into a first four-channel image and the template image is converted into a second four-channel image.

[0078] S202: Determine, according to the first four-channel image and the second four-channel image, the difference between each pixel value of the pixel point at the same position of the surface image and the template image in each channel.

[0079] Optionally, according to the first four-channel image and the second four-channel image, the difference value between each pixel value of the pixel point at the same position of the surface image and the template image in each channel is determined, and the differential image is determined based on the difference value between each pixel point.

[0080] S203 , respectively calculating the sum of the difference values ​​of all channels corresponding to each pixel point at the same position, and obtaining the total difference value of each pixel value of the pixel point at the same position between the surface image and the template image.

[0081] Optionally, since light has different effects on different channels, the sum of the difference values ​​of all channels corresponding to each pixel at the same position is calculated separately. For example, the difference values ​​of the four channels of a pixel are a1, a2, a3, and a4. At this time, a1+a2+a3+a4 is added to obtain the total difference value of the pixel value of the pixel. Based on this principle, the total difference value of each pixel value of all pixels at the same position of the surface image and the template image can be obtained. , that is, the first The total value of the pixel value difference of pixels.

[0082] It should be noted that according to the above logic, the image is affected by light, and the impact of light on the local area is uniform, that is, the amplitude of the pixel value changes of each pixel point in a certain area is almost the same. Therefore, for any pixel point, the greater the total value of the pixel value difference compared with the surrounding pixels, the greater the possibility that the current pixel point will produce abnormal color changes.

[0083] In one implementation, step S300 includes steps S301-S305:

[0084] S301, determining a first pixel point from a surface image, and determining each second pixel point other than the first pixel point in the preset local window with the first pixel point as the center of the preset local window.

[0085] It should be noted that according to the above logic, the image is affected by light, and according to analysis, the influence of light on the local area is uniform, that is, the magnitude of the pixel value change of each pixel point in a certain area is almost the same. Therefore, for any pixel point, the greater the total value of the pixel value difference compared with the surrounding pixels, the greater the possibility that the current pixel point will produce abnormal color changes. Therefore, the first pixel point is randomly determined from the surface image (for example, at this time =1), and determine the first pixel point as the center of the preset local window, construct the preset local window, exemplarily preset the size D of the local window to be 7*7, and then determine each second pixel point in the preset local window except the first pixel point.

[0086] S302: respectively determine the difference between the total pixel value difference value of the first pixel point and the total pixel value difference value of each second pixel point , and determine the absolute value of each difference .

[0087] S303: Determine the first sum of all differences and the second sum of all absolute values , and determine the specific trend of the first pixel point in the preset local window according to the ratio of the first sum value to the second sum value .

[0088] Specifically, the formula is:

[0089] ;

[0090] In the formula, Indicates that the preset local window is =1, the serial number of the second pixel other than the first pixel, For the The specific trend of pixels in the preset local window, =1 indicates the specific trend of the current first pixel point within the preset local window, and 0.01 is used to prevent the denominator from being 0. The value range of is [-0.99, 0.99], and the approximate value range is [-1, 1]. The closer it is to 1, the greater the total difference in pixel value between the current first pixel and other surrounding pixels, which also indicates that the possibility of pixel value abnormality caused by color bleeding in the preset local window is higher.

[0091] S304: Determine the possibility of color bleeding of the first pixel point at the color level within the preset local window according to the specific trend of the first pixel point within the preset local window.

[0092] Optionally, according to the specific trend of the first pixel point in the preset local window Total difference between the pixel value and the first pixel The product of determines the possibility of color bleeding at the color level of the first pixel in the preset local window. , the specific formula is:

[0093] ;

[0094] In the formula, Indicates There is a possibility of color bleeding at the color level within the preset local window. =1 means that the current first pixel has the possibility of color bleeding in the color level within the preset local window. The larger the value, the higher the possibility that the current first pixel has color bleeding in the local area.

[0095] S305, returning to the step of determining the first pixel point from the surface image, until the possibility of color bleeding at the color level within the preset local window is obtained for each pixel point.

[0096] Optionally, return to the step of determining the first pixel point from the surface image, that is, return to step S301, determine a new first pixel point and a corresponding new second pixel point, until each pixel point (i.e. Specific trend within the preset local window when taking different values And the possibility of color bleeding in the preset local window .

[0097] It should be noted that, according to prior knowledge, the main reason for bleeding is that the dye in a certain fixed color area spreads to other surrounding color areas. However, due to the relatively complex design of the print, the original template pattern itself has a certain performance, so that some normal pixels are in a relatively complex local area, and the pixel value contrast between them and the surrounding pixels is relatively small. At this time, when there is bleeding, the performance of these pixels for bleeding features at the color level is not obvious; when the image of multiple color channels obtained by the itti algorithm is used to obtain the saliency map in the subsequent acquisition, the feature map in all directions will be calculated according to the division of coarse and fine scales. According to the above problem, the feature performance in the image is not obvious, which reduces the subsequent significant performance of defects. Therefore, it is necessary to judge the performance of the feature at any scale, and then determine the optimal coarse and fine scale division threshold. According to the analysis, for the pixel points that may show the bleeding area, they usually spread radially along the edge with a clear boundary determined by the template image to the other side of the edge, and in the spreading process, the saturation of the new color generated by the superposition of two colors is in a gradually decreasing state, so the actual color has a gradient characteristic.

[0098] In one implementation, step S400 includes steps S401-S404:

[0099] S401, determining an edge image of a template image, and determining the directional authenticity of the color gradient of each pixel in the surface image according to the edge image, the surface image and the possibility.

[0100] Optionally, a canny edge detection algorithm is used based on the template image to determine an edge image of the template image.

[0101] In one embodiment, determining the directional authenticity of the color gradient of each pixel in the surface image according to the edge image, the surface image, and the probability includes:

[0102] First, a third pixel point is determined from the surface image, and each fourth pixel point other than the first pixel point in the preset local window is determined with the third pixel point as the center of the preset local window. ( =1) pixel is the current third pixel, Indicates the sequence number of the fourth pixel.

[0103] Secondly, the difference between the probability of the third pixel and the probability of the fourth pixel is used to obtain the probability difference , and the probability difference The absolute value of is used as the vector modulus between the third pixel and each fourth pixel to construct several initial vectors, that is, each possible difference The corresponding initial vector between the third pixel and each fourth pixel. When it is positive, the vector direction of the corresponding initial vector is from the third pixel point to the fourth pixel point, and when it is negative, it is from the fourth pixel point to the third pixel point.

[0104] Furthermore, all the initial vectors are accumulated to obtain the first target vector passing through the third pixel point; along the reverse direction of the first target vector, the first target fourth pixel point (denoted as the fourth pixel point) located in the edge image and having the shortest distance to the third pixel point is selected. pixels), the shortest distance is ( =1), and select a preset number of second target fourth pixel points between the third pixel point and the first target fourth pixel point along the reverse direction of the first target vector, the preset number is 10 for example, that is, 10 second target fourth pixel points are obtained; the third pixel point, the first target fourth pixel point and the 10 second target fourth pixel points are arranged in sequence along the reverse direction of the first target vector to obtain a pixel point set, that is, there are a total of It should be noted that if there is a spreading trend, the difference between the vector directions corresponding to the pixels along the line is small.

[0105] Then, the second target vectors of the first target fourth pixel and each second target fourth pixel in the pixel set are determined, and the vector direction difference between two adjacent pixels is determined according to the first target vector and each second target vector. It should be noted that the principle of determining the second target vector is the same as that of the first target vector, and the direction of the vector is determined in the same way as that of the initial vector, which will not be described in detail. Represents the first pixel in the set pixels (e.g. =1 represents the vector direction of the third pixel point), Represents the first pixel in the set The vector direction of a pixel point (for example, a fourth pixel point adjacent to the third pixel point), so When taking different values, Represents the difference in vector direction between two adjacent pixels.

[0106] Finally, the directional authenticity of the color gradient of the third pixel in the surface image is determined based on the vector direction difference. , the specific formula is:

[0107] ;

[0108] In the formula, all the pixels in the pixel set are traversed, thus obtaining The smaller the value, the more continuity of the trend of the difference value (total pixel value difference) of each pixel point along the line. The higher the authenticity of the direction in which the color gradient of the third pixel point, that is, the third pixel point, is caused by the influence of color bleeding in its spreading direction.

[0109] Optionally, returning to the step of determining the third pixel point from the surface image, that is, based on the above principle, when When different values ​​are taken, the directional authenticity of the color gradient of each pixel in the surface image can be obtained. .

[0110] S402: Determine the weight of each pixel in the surface image on the color bleeding problem in terms of color features according to the authenticity.

[0111] First, the directional authenticity of the color gradient of each pixel in the surface image is obtained Based on the above steps, it can be seen that multiple pixel point sets can be obtained in this process, and the range of the vector modulus length corresponding to each pixel point set is determined respectively. , for example, to determine the pixel point set The extreme difference of the vector modulus length of the pixels, and then determine the first arrangement number of the third pixel in each pixel set and determining the second arrangement number of the fifth pixel with the largest vector modulus in each pixel set It should be noted that, during the gradient process, the possibility of bleeding should increase first and then decrease; and when the The closer a pixel is to the pixel corresponding to the maximum value of the modulus length, the greater the possibility that it is an abnormal pixel caused by color bleeding. Therefore, it is necessary to determine the second arrangement number. .

[0112] Secondly, determine the difference between the first sequence number and the second sequence number , respectively, according to the range of the vector modulus length corresponding to each pixel point set And the serial number difference The ratio of each third pixel ( Different values ​​represent different third pixels) in the color gradient modulus authenticity :

[0113] ;

[0114] In the formula, To prevent the denominator from being zero, =0.1; The smaller the value, the greater the possibility, so using fractional expression, The larger the value, the The higher the authenticity of the model length of the third pixel in the color gradient, the higher the authenticity of the model length of the third pixel in the color gradient.

[0115] Furthermore, according to the direction authenticity of the third pixel Authenticity with model length The product of and normalization function , we get the weight of the third pixel in the surface image on the color feature of the color bleeding problem:

[0116] ;

[0117] In the formula, The surface image The weight of the pixel point (specifically the corresponding third pixel point) on the color bleeding problem in terms of color features, The larger the value is, the higher the weight of the bleeding performance is, and the higher the weight is in the subsequent significance test.

[0118] Then, return to the step of determining the third pixel point from the surface image until the weight of each pixel point in the surface image on the color feature of the color bleeding problem is obtained. ,at this time For any surface image Pixels.

[0119] S403: Determine a target saliency map according to the color features and representation weights of different channels.

[0120] In one embodiment, the color feature includes a pixel value, specifically:

[0121] First, in multi-channel images at different scales, the pixel values ​​of the R channel and the pixel values ​​of the G channel are differentially operated to obtain the first differential features of each pixel corresponding to the multi-channel images at different scales. , that is, The first differential feature of each pixel point is obtained by performing differential operations on the pixel values ​​of the B channel and the Y channel respectively, and the second differential feature of each pixel point corresponding to the multi-channel images of different scales is obtained. , that is, The second differential feature of each pixel point; wherein, the calculation method of the differential feature is a well-known technology and will not be repeated here.

[0122] Secondly, determine the first differential features respectively The performance weight of the pixel corresponding to the first differential feature The product result of the second differential feature is used to determine the color bleeding differential feature value of each pixel. , the specific formula is:

[0123] ;

[0124] in, For the Next, the first salient image of color bleeding at different scales is determined according to the color bleeding difference eigenvalues ​​of the pixels. .

[0125] Furthermore, according to the multi-channel images at different scales, a brightness Gaussian pyramid and a direction pyramid are constructed, wherein, exemplarily taking nine scales as an example, the brightness Gaussian pyramid is averaged by three channels at the nine scales to determine the brightness channel images at the nine scales, and the direction pyramid may include The specific processes of the directional feature images of 0, 45, 90, and 135, the brightness Gaussian pyramid, and the directional pyramid are existing means and will not be repeated here.

[0126] It should be noted that the first scale The fineness is higher than the second scale , the first scale Also called fine scale, ∈{2,3,4}, the second scale Also called coarse scale = +φ, φ∈{3,4}. Then, determine the first brightness channel image of the brightness Gaussian pyramid at the first scale And the first direction feature image of the direction pyramid , and determine the second brightness channel image of the brightness Gaussian pyramid at the second scale And the second directional feature image of the directional pyramid , the first scale has higher fineness than the second scale.

[0127] In addition, the second brightness channel image Interpolate to the first brightness channel image Then perform matrix subtraction to obtain the brightness feature map , the second direction feature image Interpolate to the first direction feature image Then perform matrix subtraction to obtain the directional feature map :

[0128] ;

[0129] ;

[0130] in, It indicates that the matrix subtraction operation is performed after interpolation. It should be noted that the brightness feature map can be , Directional feature map The image is normalized, feature maps are superimposed, and other operations are performed to construct the final second saliency map and the third saliency map.

[0131] Finally, the target saliency map is obtained according to the first saliency map, the second saliency map and the third saliency map, that is, the final target saliency map includes the first saliency map based on the color bleeding at different scales. , the second saliency map based on brightness, and the third saliency map based on directional features are conducive to increasing the contrast between the bleeding area and the surrounding normal area, making the bleeding area in the image more obvious than the surrounding area.

[0132] S404: Perform threshold segmentation on the target saliency map to determine potential color bleeding areas.

[0133] In the embodiment of the present application, the Otsu method is used to perform threshold segmentation on the target saliency map, and then the possible bleeding area, that is, the potential bleeding area, is identified. The marking of the potential bleeding area facilitates the staff to analyze and judge the digital printing quality of the high weather-resistant printed fabric, which is conducive to improving the accuracy of quality detection. Figure 2 shown.

[0134] In some embodiments, the printing interval time between different main colors in the current printing process can be regulated according to the area ratio of the potential bleeding area in the surface image; specifically, a bleeding area ratio threshold can be set. For example, if the resolution of the collected surface image is 1920*1080, then when the area of ​​the existing bleeding area is set to 1 / 1000 of the total area, the device is regulated and the actual area ratio is recorded, and the interval time is appropriately extended based on the ratio.

[0135] The method of the embodiment of the present application is based on the itti algorithm, and verifies the characteristics of the bleeding itself by analyzing the possibility of bleeding in the local area of ​​the preset local window where each pixel point is located, the overall directional difference and the numerical difference in the possible spreading direction, and reconstructs a new bleeding feature map based on the feature, that is, the first salient map of bleeding, and finally constructs a new salient map, that is, the second salient map and the third salient map, in combination with the feature maps of brightness and direction, to identify the potential bleeding area, and then adjust the printing interval time in the device, thereby avoiding the poor ability to highlight the bleeding area when only based on the color channel feature map analysis, resulting in inaccurate identification of the bleeding area, thereby improving the accuracy of quality detection.

[0136] Reference Figure 3 , shows a structural block diagram of a digital printing quality visual inspection system for highly weather-resistant printed fabrics according to an embodiment of the present application, the system may include:

[0137] An acquisition module is used to acquire the surface image and template image of the high-weather-resistant printed fabric, and process the surface image through an image algorithm to obtain color features of different channels;

[0138] A first determination module is used to determine the total difference between each pixel value of the pixel points at the same position of the surface image and the template image according to the color features of different channels, the surface image and the template image;

[0139] A second determination module is used to determine the possibility of color bleeding at the color level of each pixel point in the preset local window according to the preset local window and the total value of the pixel value difference;

[0140] The third determination module is used to determine the potential bleeding area according to the possibility to judge the digital printing quality of the high weather-resistant printed fabric.

[0141] In the embodiment of the present application, the functions of each module in the system can be found in the corresponding description in the above method and will not be repeated here.

[0142] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0143] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A visual inspection method for digital printing quality of highly weather-resistant printed fabrics, characterized in that: The method comprises: Acquire a surface image and a template image of a highly weather-resistant printed fabric, and process the surface image by an image algorithm to obtain color features of different channels; Determine the total value of the difference between each pixel value of the pixel points at the same position of the surface image and the template image according to the color features of different channels, the surface image and the template image; Determine, based on the preset local window and the total value of the pixel value difference, the possibility of color bleeding at the color level within the preset local window for each pixel point; According to the possibility, determining the potential bleeding area to judge the digital printing quality of the highly weather-resistant printed fabric; The surface image is processed by an image algorithm to obtain color features of different channels, including: Filtering and down-sampling the surface image by a Gaussian filter to obtain multi-channel images at different scales; wherein the multi-channel image includes a three-channel image; According to the multi-channel image, a color Gaussian pyramid is constructed, and color features of different channels at different scales are determined from the color Gaussian pyramid; wherein the color features of different channels include color features of the R channel image, color features of the G channel image, color features of the B channel image, and color features of the Y channel image; Determining the total value of the difference between each pixel value of the pixel points at the same position of the surface image and the template image according to the color features of different channels, the surface image and the template image comprises: According to the color features of different channels, converting the surface image into a first four-channel image and converting the template image into a second four-channel image; Determine, according to the first four-channel image and the second four-channel image, a difference value between each pixel value of a pixel point at the same position of the surface image and the template image in each channel; The sum of the difference values ​​of all the channels corresponding to each pixel point at the same position is calculated respectively to obtain the total difference value of each pixel value of the pixel point at the same position between the surface image and the template image; Determining the possibility of color bleeding at the color level of each pixel point within the preset local window according to the preset local window and the total value of the pixel value difference includes: Determine a first pixel point from the surface image, and determine each second pixel point in the preset local window, except the first pixel point, with the first pixel point as the center of the preset local window; respectively determining a difference between the total pixel value difference value of the first pixel point and the total pixel value difference value of each of the second pixel points, and determining an absolute value of each of the differences; Determine a first sum of all the differences and a second sum of all the absolute values, and determine a specific trend of the first pixel point within the preset local window according to a ratio of the first sum to the second sum; Determining, according to a specific trend of the first pixel point in the preset local window, a possibility that the first pixel point has color bleeding in the color layer in the preset local window; Return to the step of determining the first pixel point from the surface image until a possibility of color bleeding at the color level within the preset local window is obtained for each pixel point.

2. The visual inspection method for digital printing quality of highly weather-resistant printed fabric according to claim 1, characterized in that: Determining the possibility that the first pixel point has bleeding in the color level in the preset local window according to the specific trend of the first pixel point in the preset local window includes: The possibility of color bleeding of the first pixel point in the preset local window at the color level is determined based on the product of the specific trend of the first pixel point in the preset local window and the total value of the pixel value difference of the first pixel point.

3. The visual inspection method for digital printing quality of highly weather-resistant printed fabric according to claim 1 is characterized in that: Determining the potential bleeding area according to the possibility includes: Determine an edge image of the template image, and determine the directional authenticity of the color gradient of each pixel in the surface image according to the edge image, the surface image and the possibility; According to the authenticity, determining the weight of each pixel point in the surface image on the color bleeding problem in terms of color features; Determining a target saliency map according to the color features of different channels and the performance weights; The target saliency map is subjected to threshold segmentation to determine potential bleeding areas.

4. The method for visually inspecting the digital printing quality of highly weather-resistant printed fabric according to claim 3, characterized in that: Determining the authenticity of the direction of the color gradient of each pixel in the surface image according to the edge image, the surface image and the possibility includes: Determine a third pixel point from the surface image, and determine each fourth pixel point in the preset local window except the first pixel point, with the third pixel point as the center of the preset local window; Obtaining respective possibility differences according to the difference between the possibility of the third pixel point and the possibility of the fourth pixel point, and using the absolute value of the possibility difference as the vector modulus between the third pixel point and each of the fourth pixel points to construct a plurality of initial vectors; Accumulating all the initial vectors to obtain a first target vector passing through the third pixel point, selecting a first target fourth pixel point located in the edge image and having the shortest distance from the third pixel point along the reverse direction of the first target vector, selecting a preset number of second target fourth pixel points between the third pixel point and the first target fourth pixel point along the reverse direction of the first target vector, and arranging the third pixel point, the first target fourth pixel point, and a preset number of the second target fourth pixel points in sequence along the reverse direction of the first target vector to obtain a pixel point set; Determine a second target vector between the first target fourth pixel and each of the second target fourth pixel in the pixel set, and determine a vector direction difference between two adjacent pixels according to the first target vector and each of the second target vectors; According to the vector direction difference, the directional authenticity of the color gradient of the third pixel point in the surface image is determined, and the step of determining the third pixel point from the surface image is returned until the directional authenticity of the color gradient of each pixel point in the surface image is obtained.

5. The method for visually inspecting the digital printing quality of highly weather-resistant printed fabric according to claim 4, characterized in that: Determining the weight of each pixel in the surface image on the color feature of the color bleeding problem according to the authenticity includes: Respectively determine the range of the vector modulus length corresponding to each of the pixel point sets, determine the first arrangement sequence number of the third pixel point in each of the pixel point sets, and determine the second arrangement sequence number of the fifth pixel point with the largest vector modulus length in each of the pixel point sets; Determine the sequence number difference between the first arrangement sequence number and the second arrangement sequence number respectively, and obtain the authenticity of the module length of each third pixel point in the color gradient according to the extreme difference of the vector module length corresponding to each pixel point set and the ratio of the sequence number difference; Obtaining a weight of the third pixel in the surface image on the color bleeding problem in terms of color features according to the product of the direction authenticity of the third pixel and the modulus authenticity and a normalization function; Return to the step of determining the third pixel point from the surface image until the weight of each pixel point in the surface image on the color bleeding problem in terms of color features is obtained.

6. The method for visually inspecting the digital printing quality of highly weather-resistant printed fabric according to claim 3, characterized in that: The color feature includes a pixel value; and determining the target saliency map according to the color features of different channels and the representation weights includes: In the multi-channel images at different scales, a difference operation is performed on the pixel value of the R channel and the pixel value of the G channel respectively to obtain a first difference feature of each pixel point corresponding to the multi-channel images at different scales, and a difference operation is performed on the pixel value of the B channel and the pixel value of the Y channel respectively to obtain a second difference feature of each pixel point corresponding to the multi-channel images at different scales; Determine the product of the first differential feature and the performance weight of the pixel corresponding to the first differential feature, determine the color bleeding differential feature value of each pixel according to the sum of the product and the second differential feature, and determine the first saliency map of color bleeding at different scales according to the color bleeding differential feature value of the pixel; According to the multi-channel images at different scales, construct the brightness Gaussian pyramid and direction pyramid; Determine a first brightness channel image of the brightness Gaussian pyramid and a first directional characteristic image of the direction pyramid at a first scale, and determine a second brightness channel image of the brightness Gaussian pyramid and a second directional characteristic image of the direction pyramid at a second scale, wherein the first scale has a higher degree of fineness than the second scale; interpolating the second brightness channel image to the first brightness channel image and performing matrix subtraction to obtain a second saliency map, and interpolating the second directional feature image to the first directional feature image and performing matrix subtraction to obtain a third saliency map; A target saliency map is obtained according to the first saliency map, the second saliency map and the third saliency map.

7. A visual inspection system for digital printing quality of highly weather-resistant printed fabrics, characterized in that: The system is used to execute the method according to any one of claims 1 to 6, comprising: An acquisition module is used to acquire a surface image and a template image of the high-weather-resistant printed fabric, and process the surface image through an image algorithm to obtain color features of different channels; A first determination module, configured to determine a total value of differences between pixel values ​​of pixels at the same position of the surface image and the template image according to the color features of different channels, the surface image and the template image; A second determination module is used to determine the possibility of color bleeding at the color level of each pixel point within the preset local window according to the preset local window and the total value of the pixel value difference; The third determination module is used to determine the potential bleeding area according to the possibility to judge the digital printing quality of the high weather-resistant printed cloth.

Images