A method for detecting imaging defects in a laser printed image

Abstract

The application provides a laser printing image defect detection method, comprising: obtaining a laser image after laser interference is weakened as a template image, a training image and a detection image; aligning the training image through a positioning template to generate an aligned training image; stacking multiple aligned training images to generate a non-laser standard image; processing the aligned detection image and the non-laser standard image to obtain a non-light column detection image in which laser interference is eliminated; taking the non-light column detection image in which laser interference is eliminated as a new positioning template to generate a template to be detected; aligning the detection image and the template to be detected to generate an aligned detection image; further processing the aligned detection image and the non-laser standard image to obtain a detection image non-light column detection image; performing difference on the detection image non-light column detection image and the template image to obtain a difference image; performing binaryzation processing on the obtained difference image, screening defects through defect feature analysis, and performing waste kicking according to a judgment result.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of laser printing technology, in particular to a laser printed product imaging defect detection method. BACKGROUND

[0002] Laser printing is a printing technology that uses laser ink (metallic optical color-changing anti-counterfeiting ink) and advanced laser printing technology to print laser ink on various flat and smooth transparent materials. The printed product can show different colors with the change of viewing angle, and has obvious dynamic color-changing effect, and can produce rainbow ring effect under spotlight.

[0003] In the prior art, the laser printing technology such as CN202110379065.1 "Light column laser printing detection device and method based on column lens array" cooperates with a ring light source to eliminate the uneven bright and dark effect caused by laser on imaging, and at the same time realizes the compatibility of laser products and ordinary products, and improves the detection effect of laser products under the premise of ensuring that the detection accuracy of ordinary products does not decrease. This technology can reduce the influence of laser light column on imaging gray scale and imaging color change to a certain extent, but it cannot completely eliminate the uneven distribution of light caused by laser, and it cannot well adapt to laser light columns of different angles for different product laser grating lines.

[0004] And CN202110630454.7 "Double light source columnar laser paper color detection device based on rhombic diffraction pattern" mentions that a pair of symmetrically arranged lampshades and corresponding rod-shaped light sources arranged in the lampshades are used to irradiate laser products, and a bright and dark alternating period changing image is obtained. By positioning the dark light column position, only the dark light column area is detected for color quality detection, and the color quality detection of the dark light source area represents the overall quality of the product, which solves the problem that laser stripes cause laser products to be unable to accurately detect quality to a certain extent. However, the shortcomings are also obvious, and the quality of the product under the bright light column cannot be guaranteed. SUMMARY

[0005] The present application provides a laser printed product imaging defect detection method, which comprises the following steps:

[0006] Step 1, adjust the illumination angle of the laser printed product imaging system and the shooting angle of the image acquisition module, and obtain the laser image after weakening the laser interference;

[0007] Step 2, maintain the same angle and parameter to collect a plurality of laser products, and divide the collected images into template images, training images and detection images;

[0008] Step 3: create a positioning template using the template image;

[0009] Step 4: align the training images with the positioning template to generate aligned training images;

[0010] Step 5: stack multiple aligned training images to calculate the average value of the gray scale dimension of each pixel to generate a standard image without laser;

[0011] Step 6: difference the aligned training images and the standard image without laser to obtain a laser light column energy distribution map, pre-process the laser light column energy distribution map, and difference the aligned training images and the pre-processed laser light column energy distribution map again to obtain a training image without light column after eliminating laser interference;

[0012] Step 7: use the training image without light column after eliminating laser interference as a new positioning template to generate a template to be detected;

[0013] Step 8: align the detection image with the template to be detected to generate an aligned detection image;

[0014] Step 9: difference the aligned detection image and the standard image without laser to obtain a detection image laser light column energy distribution map, pre-process the detection image laser light column energy distribution map, and difference the aligned detection image and the pre-processed detection image laser light column energy distribution map again to obtain a detection image without light column;

[0015] Step 10: difference the detection image without light column and the template image to obtain a difference image;

[0016] Step 11: perform binaryzation processing on the obtained difference image, filter defects through defect feature analysis, and perform waste kicking according to the judgment result.

[0017] Further, in step 1, the illumination light source is a dome light source, and the illumination light source generates uniform diffusion;

[0018] The image acquisition module acquires the laser image through an angle between 5-15° with the normal.

[0019] Further, in step 3, the positioning template is generated by creating a pattern matching model on the preset multiple positioning points, and matching the shapes on the pattern matching model and the corresponding patterns in the detection image.

[0020] Further, in step 4, find the points matched with the preset multiple positioning points on the training image to generate an affine transformation matrix, and perform rigid transformation such as translation and rotation to align the training image with the positioning template.

[0021] Further, in step 5, the following steps are further included:

[0022] Step 51: perform mean filtering on the aligned training image;

[0023] Step 52: Stack multiple mean filtered aligned training images to calculate the global mean of the image gray value dimension, and obtain a standard image without laser;

[0024] Step 53: Determine whether the training image meets the requirements; the requirements include that the number of training images for stacking is greater than 10, and the laser column position is distributed in different positions of the image, and the output image of the training is free of laser column interference.

[0025] Further, in step 6, the following steps are further included:

[0026] Step 61: Divide the aligned training image into small blocks according to the size of n x n, perform n x n mean filtering on each block area, subtract the filtered mean value from the image median, multiply the subtraction result by a decay factor, add the original image gray value, and finally round the result to generate an equalization filtered training image;

[0027] Step 62: Calculate the local mean of the equalization filtered training image according to the size of n x n, subtract the local mean from the global mean, and obtain a laser column energy distribution map;

[0028] Step 63: Perform non-maximum suppression on the laser column energy distribution map according to the size of n x n area;

[0029] Step 64: Map the non-maximum suppressed laser column energy distribution map to the training image according to the position, and smooth the laser column energy distribution using bilinear interpolation;

[0030] Step 65: Perform difference between the aligned training image and the bilinear interpolation smoothed laser column energy distribution map to obtain a laser interference free light column training image.

[0031] Further, in step 61, the equalization filtered training image is generated as:

[0032] new = round((val - mean) * factor + orig)

[0033] Where new is the processed pixel value, val is the image gray level median, round represents the rounding operation, factor is the decay factor, and orig is the pixel value before processing.

[0034] Further, in step 62, the laser column energy distribution map processing process is:

[0035]

[0036] Where A(x, y)0is the laser column energy distribution map, I (x, y) is the gray value of the aligned training image, and I0 (x, y) is the gray value of the non-laser standard image.

[0037] Further, in step 63, the non-maximum suppression processing procedure is:

[0038]

[0039] wherein A (x, y) is the non-maximum suppression laser column energy distribution map, A (x, y) 0 is the laser column energy distribution map before processing, and max nxn (A (x, y) 0) represents the maximum value of the laser column energy obtained in the nxn range.

[0040] Further, in step 65, the non-laser training image processing procedure is:

[0041] I0 (x, y) = I (x, y) - A' (x, y)

[0042] wherein I0 (x, y) is the gray value of the non-laser detection image after removing the laser interference, I (x, y) is the gray value of the aligned training image, and A' (x, y) is the bilinear interpolation smoothed laser column energy distribution map.

[0043] The present application has the following beneficial effects:

[0044] The present application aims to provide a laser printed product imaging system and a defect detection method, and the conventional printed product imaging system is greatly disturbed by the laser area, which can form a light column with obvious light and dark contrast on the image, the image in the light column area is in an overexposed state, the image details are lost, and the image state of the product cannot be completely restored by using an algorithm processing, which causes great difficulty for defect detection.

[0045] The present application is based on a laser printed product imaging system, adopts a laser product special light source, collects images at a specific shooting angle, makes the light column caused by the laser narrow and weak, maximally eliminates the laser interference, and obtains an image after weakening the laser interference. On this basis, a non-laser interference standard image is calculated, a laser column equalization algorithm processing is performed on the laser image, the influence of uneven laser column energy distribution is removed, and the original defect information of the image can be maximally retained, the accuracy of image quality detection is ensured, and the industry problems of low defect detection precision, poor effect and easy to miss detection in the laser area are solved.

[0046] For laser light column printing detection, a method is proposed to extract the laser light column energy distribution on the image by a standard image without laser interference, and then the laser light column energy distribution is used to eliminate the influence of the laser light column on the image. The image without light column template and the detection image can be established better, the original information of the image is retained to the greatest extent, the influence of the laser light column on detection is effectively suppressed, and the defect information is better recognized. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 A system schematic diagram of a laser printing imaging system is provided for the present application;

[0048] Figure 2 A detection flowchart of a laser printing defect detection method is provided for the present application;

[0049] Figure 3 A standard image without laser interference calculation flowchart in a laser printing defect detection method is provided for the present application;

[0050] Figure 4 A laser light column equalization algorithm flowchart in a laser printing defect detection method is provided for the present application;

[0051] Figure 5 Image comparison of a common image acquisition platform (left) and a laser image acquisition platform (right);

[0052] Figure 6 A standard image without laser training flowchart in a laser printing defect detection method is provided for the present application;

[0053] Figure 7 A brightness equalization algorithm schematic diagram in a laser printing defect detection method is provided for the present application;

[0054] Figure 8 Laser image (left), standard image without laser (middle) and magnified laser energy distribution graph (right);

[0055] Figure 9 Laser image (left), transformed laser energy distribution graph (middle) and laser light column equalization algorithm processed image without laser light column interference; DETAILED DESCRIPTION

[0056] The advantages and characteristics of the present application will become more apparent with the description. However, these examples are only exemplary and do not constitute any limitation on the scope of the present application. Those skilled in the art should understand that the details and forms of the technical solutions of the present application can be modified or replaced without departing from the spirit and scope of the present application, and such modifications and replacements all fall within the protection scope of the present application.

[0057] As the drawingsFigure 1 As shown, the present application provides a laser printing image system, which uses a special light source for laser products to collect images at a specific shooting angle, so that the light column caused by laser is narrowed and weakened, that is, the width and intensity of the bright light column are reduced, which can greatly reduce the influence of laser light column on imaging, and obtain images after weakening the interference of laser.

[0058] The imaging system comprises an image collection module 1, an illumination light source 2, a collection control module 3 and a data processing system 4; the collection control module 3 is connected with the image collection module 1 and the illumination light source 2 respectively, and the collection control module 3 is connected with the data processing system 4.

[0059] The image collection module 1 is arranged above the to-be-tested printing product 5, and is used to collect laser printing product images at a specific angle (the collection angle is between 5-15° with the normal line), reduce the reflection interference caused by the laser process, enhance the details and contrast of the laser collected images, and transmit the collected images to the data processing system for analysis and processing.

[0060] The illumination light source 2 is a dome light source, which uses a uniform diffuse light source, and the uniform diffuse light source is fixedly installed between the to-be-tested printing product 5 and the image collection module 1. The top of the dome light source has an opening, so that the image collection module 1 can collect images of the to-be-tested printing product 5 through the opening.

[0061] The printing product is subjected to line scanning.

[0062] As Figure 2 The present application also provides a laser printing image defect detection method, which comprises the following steps:

[0063] Step 1, place the laser product on the conveying belt of the laser image collection platform, adjust the light source irradiation angle and the camera shooting angle, start the conveying belt to perform line array camera scanning and collect images, and obtain laser images after weakening the laser interference;

[0064] Since the illumination light source 2 is a uniform diffuse light source, the consistency of the brightness of the light source in each direction is good, so the laser elimination has a good effect, and belongs to a large degree of interference weakening;

[0065] The camera shooting angle is designed to be 5-15° with the normal line, which is an accurate weakening degree, and the collection position is in the most uniform position of the light, so as to avoid the shadow influence of the center opening position of the light source.

[0066] Step 2, maintain the same angle and parameter acquisition of a plurality of laser products, the collected images are divided into template images, training images and detection images, the template images need to be selected to be clear and defect-free images, the training images need to be selected to be clear and defect-free images, and the laser light column position is distributed at different positions of the images, and the number of training images needs to be greater than 10;

[0067] Step 3: create a positioning template using the selected template image;

[0068] The positioning template is created by selecting several (2-4) positioning points, and a pattern matching model is created on the several positioning points. The shape on the pattern matching model is matched with the corresponding pattern in the detection image.

[0069] Step 4: align the training images in step 2 with the positioning template in step 3 to generate aligned training images;

[0070] Find the matching points on the training images, generate an affine transformation matrix, and align the training images with the positioning template through translation, rotation and other rigid transformations.

[0071] Step 5: input the aligned training images into the laser light column energy distribution model for training to obtain a no-laser standard image, as shown in Figure 6

[0072] As shown in Figure 2 , the no-laser interference standard image generation includes the following steps:

[0073] Step 51: mean filtering of the aligned training images;

[0074] Step 52: stack the images filtered in step 51 to obtain the mean value of the image gray value dimension to obtain a no-laser standard image;

[0075]

[0076] Wherein, is the no-laser standard image; N is the number of training images; f(x, y) is the training image, and x, y represent the coordinates of the pixel points on the image;

[0077] Step 53: determine whether the training images meet the requirements; the requirements include that the number of training images used for stacking is greater than 10, and the laser light column position is distributed at different positions of the images, and the output image of the training is free of laser light column interference.

[0078] ​Step 6: Difference the aligned training image of step 4 and the output non-laser standard image of step 5 to obtain the laser light column energy distribution image, and perform non-maximum suppression and original image position space mapping on the laser light column energy distribution image to generate a preprocessed laser light column energy distribution image; then difference the aligned training image of step 4 and the preprocessed laser light column energy distribution image to obtain a non-column detection image eliminating laser interference, as shown in Figure 6

[0079] As Figure 3 , the laser light column equalization algorithm processing step includes:

[0080] Step 61: Divide the aligned training image into small blocks according to the size of n x n, perform n x n mean filtering on each block, subtract the image median from the filtered mean value, multiply the subtraction result by a decay factor, add the original image gray value, and finally take the integer of the result by rounding method. This operation realizes the preliminary equalization of the image brightness, so that the bright places of the image become dark and the dark places become bright, and the processing process is as shown in Figure 7 .

[0081] new = round((val - mean) * factor + orig)

[0082] Where new is the processed pixel value; val is the image gray level median, such as the median of a 255 gray level image is 128; mean is the value of the pixel point after mean filtering; round represents the rounding operation; factor is the decay factor, which is determined according to the usual block size of n x n, and the value is between 0.55 and 0.8; orig is the pixel value before processing.

[0083] Step 62: Calculate the local mean value of the image filtered by the brightness equalization of step 61 according to the size of n x n, subtract the global mean value obtained in step 52 from the local mean value to obtain the laser light column energy distribution image, and the processing process is as shown in the following formula:

[0084]

[0085] Where A(x, y)0is the laser light column energy distribution image, is the non-laser standard image.

[0086] Step 63: Perform non-maximum suppression on the laser light column energy distribution image according to the size of n x n region, and the processing process is as shown in the following formula:

[0087]

[0088] ​Wherein, A(x, y) is the non-maximum suppression laser column energy distribution map; A(x, y) 0 is the laser column energy distribution map before processing; max nxn (A(x, y) 0) represents the maximum value of the laser column energy obtained in the nxn range.

[0089] Step 64: the non-maximum suppression laser column energy distribution map is mapped to the position of the alignment training image, and the laser column energy distribution is smoothed by using the bilinear interpolation.

[0090] Step 65: difference is made with the alignment training image and the bilinear interpolation smoothed laser column energy distribution map, so that the laser interference free light column detection map is obtained, as shown in the following formula. Figure 9 The red box in the figure is a defect, and the obtained light column detection map has obvious defect contrast, and the detection effect is better than that of the untreated laser map, and the processing process is as shown in the following formula.

[0091] I0(x, y) = I(x, y)-A'(x, y)

[0092] Wherein, I0(x, y) represents the gray value of the laser interference free light column detection image, I(x, y) represents the gray value of the alignment training image, and A'(x, y) represents the bilinear interpolation smoothed laser column energy distribution map.

[0093] Step 7: the laser interference free light column detection map obtained by eliminating the laser interference in step 6 is used as a new positioning template to generate a to-be-detected template;

[0094] Step 8: the detection image collected in step 2 is aligned with the to-be-detected template in step 7 to generate an alignment detection image.

[0095] Step 9: the alignment detection image is processed by the same laser column algorithm as step 6 to obtain a light column detection map.

[0096] Step 10: the light column detection map obtained in step 9 is differentiated with the template image to obtain a difference map.

[0097] Step 11: the obtained difference map is binarized, and defects are screened through defect feature analysis, and the judgment result is output to a kick-out module, the kick-out module judges the kick-out opportunity, and the kick-out action is completed.

[0098] The present application is not limited to the above specific embodiments, and those skilled in the art can use other various specific embodiments to implement the present application according to the content disclosed in the embodiments and the drawings, therefore, any design structure and idea using the present application, and some simple changes or changes, fall within the scope of the present application.

Claims

1. A method for detecting imaging defects in laser-printed materials, characterized in that, The laser printing defect detection method includes the following steps: Step 1: Adjust the illumination angle of the laser printing imaging system (2) and the shooting angle of the image acquisition module (1) to obtain the laser image after the laser interference is weakened; Step 2: Acquire multiple laser products while maintaining the same angle and parameters, and divide the acquired images into template images, training images, and detection images; Step 3: Create a positioning template using the template image; Step 4: Align the training images using the positioning template to generate aligned training images; Step 5: Stack multiple aligned training images, calculate the mean grayscale value of each pixel, and generate a laser-free standard image; Step 6: Difference the aligned training image with the laser-free standard image to obtain the laser beam energy distribution map. Preprocess the laser beam energy distribution map, and then difference the aligned training image with the preprocessed laser beam energy distribution map again to obtain a laser-free detection map without laser interference. Step 7: Use the laser-interference-free beam detection map as a new positioning template to generate the test template; Step 8: Align the detection image with the template to be tested to generate an aligned detection image; Step 9: Difference the aligned detection image with the standard image without laser beams to obtain the laser beam energy distribution map of the detection image. Preprocess the laser beam energy distribution map of the detection image. Difference the laser beam energy distribution map of the aligned detection image with the preprocessed detection image again to obtain the laser beam detection image without laser beams. Step 10: Difference the detection image without light pillars between the detection image and the template image to obtain a difference image; Step 11: Binarize the obtained difference map, filter defects through defect feature analysis, and discard defective products based on the judgment results.

2. The laser printing defect detection method according to claim 1, characterized in that, In step 1, the lighting source (2) is a dome light source, and the lighting source (2) produces uniform diffusion; The image acquisition module (1) acquires laser images at an angle between 5-15° and the normal.

3. The laser printing defect detection method according to claim 1, characterized in that, In step 3, the positioning template is created by setting multiple positioning points and creating a pattern matching model on the positioning points. The shape on the pattern matching model is matched with the corresponding pattern in the image to be detected.

4. The laser printing defect detection method according to claim 3, characterized in that, In step 4, points matching multiple preset positioning points are found on the training image, an affine transformation matrix is ​​generated, and the training image is aligned with the positioning template through rigid transformation of translation and rotation.

5. The laser printing defect detection method according to claim 1, characterized in that, Step 5 also includes the following steps: Step 51: Apply mean filtering to the aligned training images; Step 52: Stack multiple mean-filtered aligned training images together and calculate the global mean of the image grayscale dimension to obtain a laser-free standard image; Step 53: Determine whether the training images meet the requirements; the requirements include that the number of training images used for stacking is greater than 10, the laser beam positions are distributed in different positions of the images, and the training output images are free from laser beam interference.

6. The laser printing defect detection method according to claim 5, characterized in that, Step 6 also includes the following steps: Step 61: Divide the aligned training image into small blocks of size n×n, perform n×n mean filtering on each block, subtract the mean of the filtered image from the median of the image, multiply the result by the attenuation coefficient, add the gray value of the original image, and finally round the result to the nearest integer to generate the equalized filtered training image. Step 62: Calculate the local mean of the equalization filter training image in an n×n size, and subtract the local mean from the global mean to obtain the laser beam energy distribution map; Step 63: Perform non-maximum suppression on the laser beam energy distribution map according to an n×n region; Step 64: Map the laser beam energy distribution map after non-maximum suppression to the training image according to the location, and smooth the laser beam energy distribution using bilinear interpolation; Step 65: Perform differential analysis using the aligned training image and the laser beam energy distribution map smoothed by bilinear interpolation to obtain a laser beam-free detection map without laser interference.

7. The laser printing defect detection method according to claim 6, characterized in that, In step 61, the generation of the equalization filter training image is as follows: new=round((val-mean)*factor+orig) Where new is the processed pixel value, val is the median gray level of the image, mean is the value of the pixel after mean filtering, round indicates rounding to the nearest integer, factor is the attenuation coefficient, and orig is the pixel value before processing.

8. The laser printing defect detection method according to claim 7, characterized in that, In step 62, the laser beam energy distribution map processing procedure is as follows: Where A(x,y)0 is the energy distribution diagram of the laser beam. It is the laser-free standard image obtained in step 53, where I(x,y) represents the grayscale value of the aligned training image.

9. The laser printing defect detection method according to claim 8, characterized in that, In step 63, the nonmaximum suppression process is as follows: Where A(x,y) is the energy distribution of the laser beam after nonmaximum suppression; A(x,y)0 is the energy distribution of the laser beam before processing; max nxn (A(x,y)0) represents the maximum energy value of the laser beam obtained in the range of nxn.

10. The laser printing defect detection method according to claim 9, characterized in that, In step 65, the process for processing the image without light pillars is as follows: I0(x,y)=I(x,y)-A′(x,y) Where I0(x,y) represents the grayscale value of the laser beam detection image after removing laser interference, I(x,y) represents the grayscale value of the aligned training image, and A′(x,y) represents the laser beam energy distribution map after bilinear interpolation smoothing.

Images