WO2024160201A1 - 印刷缺陷检测方法、装置、设备及存储介质 - Google Patents

印刷缺陷检测方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2024160201A1
WO2024160201A1 PCT/CN2024/074707 CN2024074707W WO2024160201A1 WO 2024160201 A1 WO2024160201 A1 WO 2024160201A1 CN 2024074707 W CN2024074707 W CN 2024074707W WO 2024160201 A1 WO2024160201 A1 WO 2024160201A1
Authority
WO
WIPO (PCT)
Prior art keywords
target image
image
reference image
initial
transformation
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2024/074707
Other languages
English (en)
French (fr)
Inventor
李嘉敏
钟国通
钟穗盈
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Laboratory For Artificial Intelligence In Design Ltd
Royal College of Art
Hong Kong Polytechnic University HKPU
Original Assignee
Laboratory For Artificial Intelligence In Design Ltd
Royal College of Art
Hong Kong Polytechnic University HKPU
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Laboratory For Artificial Intelligence In Design Ltd, Royal College of Art, Hong Kong Polytechnic University HKPU filed Critical Laboratory For Artificial Intelligence In Design Ltd
Publication of WO2024160201A1 publication Critical patent/WO2024160201A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G06T7/001—Industrial image inspection using an image reference approach
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00—Geometric image transformations in the plane of the image
    • G06T3/02—Affine transformations
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T5/00—Image enhancement or restoration
    • G06T5/70—Denoising; Smoothing
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30144—Printing quality
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Definitions

  • the present disclosure is based on the Chinese patent application with application number 202310106900.3, application date February 2, 2023, and invention name “Printing defect detection method, device, equipment and storage medium”, and claims the priority of the Chinese patent application.
  • the entire content of the Chinese patent application is hereby introduced into the present disclosure as a reference.
  • the present disclosure relates to the field of computer technology, and in particular to a printing defect detection method, device, electronic device and readable storage medium.
  • the purpose of the present disclosure is to provide a printing defect detection method, device, electronic device and readable storage medium, which can automatically detect printing defects.
  • a printing defect detection method comprising: obtaining an initial target image of a printed product to be detected; obtaining a reference image corresponding to the printed product to be detected; aligning the initial target image to the reference image in a plurality of ways, and obtaining a plurality of candidate transformation target images; determining the transformation target image based on the degree of similarity between each candidate transformation target image among the multiple candidate transformation target images and the reference image; performing color distribution feature normalization processing on the transformation target image based on the reference image to obtain a standardized transformation target image; obtaining a final differential result based on the standardized transformation target image and the reference image to detect the printing defects of the printed product to be detected according to the final differential result.
  • the multiple candidate transformation target images include a first candidate transformation target image; the initial target image is aligned to the reference image in a variety of ways to obtain multiple candidate transformation target images respectively, including: using an iterative nearest neighbor algorithm to calculate the affine transformation from the initial target image to the reference image to obtain an initial registration result of the initial target image and the reference image; using a mask method to subtract pixels other than the overlapping points of the initial target image and the reference image in the initial registration result to obtain a contour point sequence of the initial target image aligned to the reference image; transforming the initial target image according to the contour point sequence of the initial target image aligned to the reference image to obtain the first candidate transformation target image.
  • the multiple candidate transformation target images include a second candidate transformation target image; the initial target image is aligned to the reference image in a variety of ways to obtain multiple candidate transformation target images, including: obtaining feature points in the initial target image and feature points in the reference image respectively; matching the feature points in the initial target image with the feature points in the reference image to obtain a first homography matrix between the initial target image and the reference image; obtaining a first initial transformation target image and a first mask image according to the initial target image and the first homography matrix; removing blank pixels in the first initial transformation target image according to the first mask image to obtain the second candidate transformation target image.
  • a first initial transformed target image and a first mask image are obtained respectively according to the initial target image and the first homography matrix, including: multiplying the first homography matrix with the initial target image to obtain the first initial transformed target image; multiplying the first homography matrix with the generated image to obtain the first mask image, wherein the generated image is an image filled with white pixels and of the same size as the initial target image.
  • the plurality of candidate transformation target images include a third candidate transformation target image; the initial target image is aligned to the reference image in a plurality of ways, A plurality of candidate transformation target images are obtained respectively, including: obtaining marks in an initial target image and marks in a reference image respectively; making marks in the initial target image correspond to marks in the reference image to obtain a second homography matrix between the initial target image and the reference image; obtaining a second initial transformation target image and a second mask image respectively according to the initial target image and the second homography matrix; removing blank pixels in the second initial transformation target image according to the second mask image to obtain the third candidate transformation target image.
  • color distribution feature normalization processing is performed on the transformation target image based on the reference image to obtain a standardized transformation target image, including: obtaining a reference image after mask processing; converting the transformation target image and the reference image after mask processing from RGB color space to CIELAB color space respectively to obtain a transformation target image in the CIELAB color space and a reference image in the CIELAB color space; color distribution feature normalization processing is performed on the transformation target image in the CIELAB color space based on the reference image in the CIELAB color space to obtain the standardized transformation target image.
  • color distribution feature normalization processing is performed on a transformation target image in the CIELAB color space based on a reference image in the CIELAB color space to obtain a standardized transformation target image, including: denoising the transformation target image in the CIELAB color space using a Gaussian convolution kernel; obtaining a pixel mean and a standard deviation of the transformation target image in the CIELAB color space after the denoising processing; obtaining a pixel mean and a standard deviation of the reference image in the CIELAB color space; color distribution feature normalization processing is performed on the transformation target image in the CIELAB color space according to the pixel mean and the standard deviation of the transformation target image in the CIELAB color space after the denoising processing and the pixel mean and the standard deviation of the reference image in the CIELAB color space to obtain the standardized transformation target image.
  • a transformation target image is determined based on the degree of similarity between each candidate transformation target image among the multiple candidate transformation target images and the reference image, including: masking the reference image to obtain a reference image after masking; for each candidate transformation target image, a root mean square error measurement method is used to obtain a similarity score between the candidate transformation target image and the reference image after masking, wherein a local search method is used to iteratively update the candidate transformation target image during the root mean square error measurement process; and an updated image of the candidate transformation target image with the highest similarity score to the reference image after masking among the multiple candidate transformation target images is used as the transformation target image.
  • the standardized transformation target image is an image in the CIELAB color space
  • a final differential result is obtained based on the standardized transformation target image and the reference image to detect printing defects of the printed product to be inspected according to the final differential result, including: obtaining a reference image in the CIELAB color space; performing differential processing on the standardized transformation target image and the reference image in the CIELAB color space to obtain an initial differential result; performing threshold segmentation processing on the initial differential result to obtain a binary differential result; performing morphological operation processing on the binary differential result to obtain the final differential result, so as to detect printing defects of the printed product to be inspected according to the final differential result.
  • a printing defect detection device comprising: an acquisition module, for acquiring an initial target image of a printed product to be detected; the acquisition module, further for acquiring a reference image corresponding to the printed product to be detected; an image alignment module, for aligning the initial target image to the reference image in a plurality of ways, and obtaining a plurality of candidate transformation target images respectively; an image calibration module, for determining a transformation target image based on a degree of similarity between each candidate transformation target image in the plurality of candidate transformation target images and the reference image; a color standardization module, for performing color distribution feature standardization processing on the transformation target image based on the reference image, and obtaining a standardized transformation target image; and a difference module, for obtaining a final difference result based on the standardized transformation target image and the reference image, so as to detect printing defects of the printed product to be detected according to the final difference result.
  • an electronic device comprising: a memory, a processor, and executable instructions stored in the memory and executable in the processor, wherein the processor implements any of the above methods when executing the executable instructions.
  • a computer-readable storage medium on which computer-executable instructions are stored, and when the executable instructions are executed by a processor, any of the above methods is implemented.
  • FIG1 is a schematic diagram showing a structure of a printing defect detection system in an embodiment of the present disclosure.
  • FIG. 2 shows a flow chart of a printing defect detection method in an embodiment of the present disclosure.
  • FIG. 3 is a schematic diagram showing the processing process of step S206 shown in FIG. 2 in one embodiment.
  • FIG. 4 is a schematic diagram of an image alignment process according to FIG. 3 .
  • FIG. 5 is a schematic diagram showing the processing procedure of step S206 shown in FIG. 2 in another embodiment.
  • FIG. 6 is a schematic diagram showing the processing process of step S506 shown in FIG. 5 in one embodiment.
  • FIG. 7 is a schematic diagram of another image alignment process shown in FIG. 5 and FIG. 6 .
  • FIG. 8 is a schematic diagram showing the processing procedure of step S206 shown in FIG. 2 in yet another embodiment.
  • FIG. 9 is a schematic diagram of another image alignment process shown in FIG. 8 and FIG. 6 .
  • FIG. 10 is a schematic diagram showing the processing process of step S208 shown in FIG. 2 in one embodiment.
  • FIG. 11 is a schematic diagram showing the processing process of step S210 shown in FIG. 2 in one embodiment.
  • FIG. 12 is a schematic diagram showing the processing procedure of step S1106 shown in FIG. 11 in one embodiment.
  • FIG. 13 is a schematic diagram showing the processing process of step S212 shown in FIG. 2 in one embodiment.
  • FIG. 14 is a schematic diagram of a printing defect detection process according to FIG. 2 to FIG. 13 .
  • FIG. 15 shows a block diagram of a printing defect detection device in an embodiment of the present disclosure.
  • FIG. 16 shows a block diagram of another printing defect detection device in an embodiment of the present disclosure.
  • FIG. 17 is a schematic diagram showing the structure of an electronic device in an embodiment of the present disclosure.
  • first, second, etc. are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as “first” or “second” may explicitly or implicitly include one or more of the features.
  • the meaning of “plurality” is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.
  • the symbol “/” generally indicates that the objects associated with each other are in an "or” relationship.
  • connection should be understood in a broad sense, for example, it can be an electrical connection or can communicate with each other; it can be directly connected or indirectly connected through an intermediate medium.
  • connection should be understood in a broad sense, for example, it can be an electrical connection or can communicate with each other; it can be directly connected or indirectly connected through an intermediate medium.
  • Common printing defects include dirty spots, excessive ink absorption, ghosting, gear marks, etc. After these printing defects are manually checked, the defective products will be thrown away, and the machine will be recalibrated and put into operation. This process needs to be repeated until the product that meets the standards is produced, which undoubtedly wastes a lot of time and economic costs.
  • monitoring system Some solutions to printing defects have been developed in the related art. For example, industrial printing press manufacturers such as Heidelberg Drucker maschine and Koenig & Bauer provide online inspection and monitoring systems equipped with high-resolution color cameras and LED lighting technology for their own offset printing presses. However, this system (hereinafter referred to as the monitoring system) also has some problems.
  • the monitoring system is a special function that comes with the factory printing system. If the printing system is modified or updated after sales, the monitoring system may not accept it, and it may not be compatible with older models of printing presses. If the modification is made by a third-party supplier, the original manufacturer cannot provide lifetime after-sales service and warranty.
  • the monitoring system usually consists of multiple cameras and LED lighting, which is installed after the last printing unit or coater of a large printing press (such as an offset press). If one of the cameras or LED lighting does not work, the overall detection performance will be greatly reduced.
  • the equipment cost and maintenance cost of the printing press with the monitoring system are high.
  • the high-resolution color camera used for inspection is unique and expensive, and it is difficult to find a substitute in the market.
  • image processing is the core of computer vision, which usually involves different steps: image acquisition, image enhancement, image representation, image description, image recognition, etc. The specific steps involved depend on the purpose and application of image processing.
  • Image acquisition is one of the basic steps of the image processing system, that is, to obtain an optical image through a suitable device, and use an analog-to-digital converter to convert it into an image for computer processing, and then use the computer to further process it, such as enhancement, segmentation, representation, description and recognition.
  • Image enhancement refers to the process of transforming and adjusting the image to make the resulting image more suitable for display and further analysis or operation.
  • Image segmentation is to divide the image into multiple parts and reduce its complexity so that each part can be further analyzed to identify the target in the image.
  • the goal of image representation is to represent and describe the fragments in an appropriate form, such as boundaries, or complete areas, for further processing.
  • Image description can extract structured and comparable attributes from the image for distinction and analysis.
  • the image recognition step can detect and analyze the image to achieve automated processing, such as classifying places, people, and objects.
  • the present disclosure provides a printing defect detection method, which uses multiple methods to align an initial target image of a printed product to be detected with a corresponding reference image to obtain multiple candidate transformation target images, and then determines the transformation target image based on the similarity between each candidate transformation target image in the multiple candidate transformation target images and the reference image, performs color distribution feature standardization processing on the transformation target image based on the reference image to obtain a standardized transformation target image, and then obtains a final differential result based on the standardized transformation target image and the reference image, so as to detect the printing defects of the printed product to be detected according to the final differential result, thereby using the intelligent image processing technology in computer vision to automatically detect various defects on printed products, and then identify them, which can reduce the image acquisition cost.
  • Set hardware requirements uses multiple methods to align an initial target image of a printed product to be detected with a corresponding reference image to obtain multiple candidate transformation target images, and then determines the transformation target image based on the similarity between each candidate transformation target image in the multiple candidate transformation target images and the reference image, performs color
  • FIG. 1 shows an exemplary system architecture 10 to which the printing defect detection method or printing defect detection apparatus of the present disclosure can be applied.
  • the system architecture 10 may include a terminal device 102, a network 104, and a server 106.
  • the terminal device 102 may be various devices with a shooting function, including but not limited to various devices such as smart phones, digital cameras, industrial cameras, scanners, etc.
  • the network 104 is used to provide a medium for a communication link between the terminal device 102 and the server 106.
  • the network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.
  • the server 106 may be a server or server cluster that provides various services, such as a server for data processing, a database server for storing data, etc.
  • the terminal device 102 can interact with the server 106 through the network 104 to receive or send data, etc.
  • the terminal device 102 photographs the printed product 100 to be inspected to obtain an initial target image, and then sends the initial target image to the server 106 through the network 104 for processing.
  • the server 106 executes the printing defect detection method in the embodiment of the present disclosure to obtain a final differential result, and then sends the final differential result to the terminal device 102 such as a smart phone through the network 104 for the user to view.
  • the terminal device 102 executes the printing defect detection method in the embodiment of the present disclosure to obtain a final differential result, and then sends the final differential result to the server 106 through the network 104 for storage.
  • terminal devices, networks and servers in Figure 1 is only illustrative. Any number of terminal devices, networks and servers may be provided according to implementation requirements.
  • Fig. 2 is a flow chart of a printing defect detection method according to an exemplary embodiment. The method shown in Fig. 2 can be applied to the server side of the system shown in Fig. 1, or to the terminal device shown in Fig. 1.
  • the method 20 provided in the embodiment of the present disclosure may include the following steps.
  • step S202 an initial target image of the printed product to be inspected is acquired.
  • the terminal device 102 shown in FIG. 1 may be used to photograph the printed product to be inspected to obtain an initial target image.
  • step S204 a reference image corresponding to the printed product to be inspected is obtained.
  • the terminal device 102 shown in FIG. 1 may be used to photograph a reference (standard product without defects) corresponding to the printed product to be inspected to obtain a reference image.
  • the terminal device 102 shown in FIG. 1 may be used to respectively capture a reference image and an initial target image under the same lighting conditions to minimize changes for better comparison and defect detection.
  • step S206 the initial target image is aligned to the reference image in a variety of ways to obtain a plurality of candidate transformation target images.
  • the reference image and the initial target image may not be placed at the same position when the product is captured, resulting in incomplete alignment between the reference image and the initial target image.
  • a variety of methods can be used to align the initial target image to the reference image, and corresponding candidate transformed target images are obtained respectively, and then the alignment result that is most similar to the reference image is selected.
  • the plurality of candidate transform target images may include a first candidate transform target image, a second candidate transform target image, and a third candidate transform target image.
  • a contouring method may be used to align the initial target image with the reference image to obtain a first candidate transformed target image.
  • a contouring method may be used to align the initial target image with the reference image to obtain a first candidate transformed target image.
  • a feature point matching method may be used to align the initial target image with the reference image to obtain a second candidate transformed target image.
  • a feature point matching method may be used to align the initial target image with the reference image to obtain a second candidate transformed target image.
  • a marker matching method may be used to align the initial target image with the reference image to obtain a third candidate transformed target image.
  • a marker matching method may be used to align the initial target image with the reference image to obtain a third candidate transformed target image.
  • a transform target image is determined based on the degree of similarity between each of the plurality of transform target image candidates and the reference image.
  • similarity scores between the first candidate transformation target image, the second candidate transformation target image, the third candidate transformation target image and the reference image may be calculated respectively, and the image with the highest similarity score may be selected as the transformation target image.
  • the image with the highest similarity score may be selected as the transformation target image.
  • step S210 a color distribution feature normalization process is performed on the transformation target image based on the reference image to obtain a normalized transformation target image.
  • the reference image (after mask processing) and the transformation target image may be first converted from the RGB color space to the CIELAB color space, and then the color distribution feature normalization process may be performed.
  • the color distribution feature normalization process may be performed.
  • step S212 a final differential result is obtained based on the normalized transformation target image and the reference image, so as to detect printing defects of the printed product to be inspected according to the final differential result.
  • the normalized transformation target image may be compared to the mask image in the CIELAB color space.
  • the reference image after coding is differentiated and then threshold segmentation and morphological operation are performed to obtain the final differential result.
  • FIG. 13 please refer to FIG. 13 .
  • a database of final differential results of a large number of printed products can be established, and the final differential results can be classified using algorithms such as deep learning models to detect different types of defects.
  • the classification algorithm can be implemented in the form of software or an application for a mobile device.
  • the initial target image of the printed product to be detected is aligned to the corresponding reference image by using multiple methods, and then the transformed target image with the highest similarity to the reference image is selected for color distribution feature standardization processing to obtain the final differential result, thereby realizing automatic detection of factor defects and solving the technical problems of poor adaptability and low accuracy of a single image alignment method when processing various types of defects of a large number of products.
  • the method adopted by the embodiment of the present disclosure relies on the processing of the initial target image and the reference image, does not require a high-resolution camera to collect images, has low hardware requirements, low installation costs, and is more flexible than online detection equipment.
  • Fig. 3 is a schematic diagram showing the processing process of step S206 in one embodiment shown in Fig. 2. As shown in Fig. 3, in the embodiment of the present disclosure, the above step S206 may further include the following steps.
  • Step S302 using an iterative nearest neighbor algorithm to calculate the affine transformation from the initial target image to the reference image, and obtaining an initial registration result of the initial target image and the reference image.
  • Step S304 Subtract pixels other than the overlapped points of the initial target image and the reference image in the initial registration result by using a mask method, so as to obtain a contour point sequence of the initial target image aligned with the reference image.
  • Step S306 transforming the initial target image according to the contour point sequence aligned with the reference image to obtain a first candidate transformed target image.
  • FIG4 is a schematic diagram of an image alignment process according to FIG3 .
  • the initial target image 4024 (with printing defects 400) is aligned with the reference image 4022 to obtain an initial alignment result 404 of the initial target image 4024 and the reference image 4022.
  • a binary mask is applied to mask out pixels other than the intersection of the two images to obtain a contour point sequence 406 of the initial target image aligned with the reference image.
  • the initial target image is transformed according to the contour point sequence 406 of the initial target image aligned with the reference image to obtain a transformed initial target image 408.
  • the transformed initial target image 408 can then be masked to remove blank pixels in the image to obtain a first candidate transformation.
  • the target image 4102, 4104 in the figure is a mask image obtained based on the contour point sequence 406 of the initial target image aligned to the reference image and an image filled with white pixels and of the same size as the initial target image.
  • an iterative nearest neighbor algorithm is used to calculate the affine transformation of the initial target image to the reference image, a binary mask is applied to mask out the pixels other than the intersection of the two images, and a contour point sequence is obtained to align the initial target image to the reference image. This can provide an aligned target image and reference image for subsequent differential processing, thereby improving the accuracy of defect detection.
  • Fig. 5 is a schematic diagram showing the processing process of step S206 shown in Fig. 2 in another embodiment. As shown in Fig. 5, in the embodiment of the present disclosure, the above step S206 may further include the following steps.
  • Step S502 respectively obtaining feature points in the initial target image and feature points in the reference image.
  • Step S504 correspond the feature points in the initial target image to the feature points in the reference image to obtain a first homography matrix between the initial target image and the reference image.
  • the homography matrix can represent a unique corresponding transformation relationship between two planes.
  • Step S506 obtaining a first initial transformed target image and a first mask image according to the initial target image and the first homography matrix respectively.
  • the first homography matrix may be multiplied by the initial target image and the blank image respectively to obtain a first initial transformed target image and a first mask image respectively.
  • the first homography matrix may be multiplied by the initial target image and the blank image respectively to obtain a first initial transformed target image and a first mask image respectively.
  • Step S508 removing blank pixels in the first initial transformation target image according to the first mask image to obtain a second candidate transformation target image.
  • the first initial transformation target image may be represented as T
  • the first mask image may be represented as M
  • Fig. 6 is a schematic diagram showing the processing process of step S506 in one embodiment shown in Fig. 5. As shown in Fig. 6, in the embodiment of the present disclosure, the above step S506 may further include the following steps.
  • Step S602 multiply the first homography matrix by the initial target image to obtain a first initial transformed target image.
  • Step S604 multiply the first homography matrix by the generated image to obtain a first mask image.
  • the generated image is an image filled with white pixels and has the same size as the initial target image.
  • FIG7 is a schematic diagram of another image alignment process according to FIG5 and FIG6.
  • step S502 feature points in the initial target image 7024 (with printing defects 700) and feature points in the reference image 7022 are obtained respectively.
  • step S504 the feature points in the initial target image 7024 are matched with the feature points in the reference image 7022 to obtain a first homography matrix 704 between the initial target image 7024 and the reference image 7022.
  • step S602 the first homography matrix 704 is multiplied with the initial target image 7024 to obtain a first initial transformed target image 7062, and in step S604, the first homography matrix 704 is multiplied with a generated image 7026 filled with white pixels and having the same size as the initial target image 7024 to obtain a first mask image 7064.
  • the black triangle area of the first mask image 7064 represents blank pixels.
  • the initial target image 7024 and the reference image 7022 share the same blank pixel area, and which black pixels in the initial target image 7024 are blank pixels can be determined by the first mask image 7064.
  • the blank pixels in the first initial transformation target image 7062 can be removed according to the first mask image 7064 to obtain the second candidate transformation target image 708.
  • a first homography matrix representing the transformation relationship between the initial target image and the reference image is obtained, and then the first homography matrix is multiplied by the initial target image to obtain a first initial transformed target image, and a mask is used to remove background blank pixels in the first initial transformed target image to obtain a second candidate transformed target image.
  • This can provide an aligned target image and a reference image for subsequent differential processing, thereby improving the accuracy of defect detection.
  • Fig. 8 is a schematic diagram showing the processing process of step S206 in another embodiment shown in Fig. 2. As shown in Fig. 8, in the embodiment of the present disclosure, the above step S206 may further include the following steps.
  • Step S802 respectively obtaining the markers in the initial target image and the markers in the reference image.
  • reference marks may be used to align, sort, position, and orient printed products during the printing process.
  • Step S804 Match the markers in the initial target image with the markers in the reference image to obtain a second homography matrix between the initial target image and the reference image.
  • Step S806 obtaining the second initial target image and the second homography matrix respectively. Transform the target image and the second mask image.
  • the second homography matrix may be multiplied by the initial target image and the blank image respectively to obtain a second initial transformed target image and a second mask image respectively.
  • the specific implementation is similar to FIG. 6 .
  • Step S808 removing blank pixels in the second initial transformation target image according to the second mask image to obtain a third candidate transformation target image.
  • FIG. 9 is a schematic diagram of another image alignment process according to FIG. 8 and FIG. 6 .
  • step S802 feature points in the initial target image 9024 (with printing defects 900) and feature points in the reference image 9022 are obtained respectively.
  • step S804 the feature points in the initial target image 9024 are matched with the feature points in the reference image 9022 to obtain a second homography matrix 904 between the initial target image 9024 and the reference image 9022.
  • step S806 the second homography matrix 904 is multiplied with the initial target image 9024 to obtain a second initial transformed target image 9062 , and the second homography matrix 904 is multiplied with a generated image 9026 filled with white pixels and having the same size as the initial target image 9024 to obtain a second mask image 9064.
  • step S808 the blank pixels in the second initial transformed target image 9062 can be removed according to the second mask image 9064 to obtain a third candidate transformed target image 908.
  • a second homography matrix representing the transformation relationship between the initial target image and the reference image is obtained, and then the second homography matrix is multiplied by the initial target image to obtain a second initial transformed target image, and a mask is used to remove background blank pixels in the second initial transformed target image to obtain a third candidate transformed target image, which can provide an aligned target image and a reference image for subsequent differential processing, thereby improving the accuracy of defect detection.
  • Fig. 10 is a schematic diagram showing the processing process of step S208 in one embodiment shown in Fig. 2. As shown in Fig. 10, in the embodiment of the present disclosure, the above step S208 may further include the following steps.
  • Step S1002 performing mask processing on the reference image to obtain a reference image after mask processing.
  • Step S1004 for each candidate transformation target image, a root mean square error measurement method is used to obtain a similarity score between the candidate transformation target image and the reference image after mask processing, wherein a local search method is used to iteratively update the candidate transformation target image during the root mean square error measurement process.
  • the similarity score SM (Similarity Metric) between the candidate transformation target image and the reference image after mask processing can be expressed as follows:
  • R′ (i,j) represents the pixel of the i-th row and j-th column of the reference image R′ after mask processing
  • T′ (i,j) represents the pixel of the i-th row and j-th column of the candidate transformation target image T′ (after mask processing)
  • M (i,j) represents the pixel of the i-th row and j-th column of the corresponding mask image (for example, the first mask image and/or the second mask image mentioned above), that is, ⁇ ij M (i,j) represents the sum of the pixel values after removing the blank pixels in the candidate transformation target image
  • T′′ (i,j) represents the pixel after T′ (i,j) is iteratively updated.
  • Step S1006 taking the updated image of the candidate transformation target image having the highest similarity score with the reference image after masking among the plurality of candidate transformation target images as the transformation target image.
  • the method provided by the embodiment of the present disclosure adopts a local search method to perform similarity measurement (i.e., calculate the SM value) in each iteration, obtains candidate objects through operations such as translation, rotation, and scaling, and selects the best candidate object for affine transformation, and then enters the next iteration to obtain a finer search space scale, thereby achieving more accurate and uniform sub-pixel alignment.
  • similarity measurement i.e., calculate the SM value
  • the root mean square error (RMSE) indicator is used to measure the similarity between the candidate transformed target image and the reference image after masking, replacing the mean absolute error (MAE) indicator in related technologies.
  • RMSE has ideal sensitivity to outliers, so any offset in image alignment will produce a significant RMSE error value. Dividing it by the total pixel value after removing blank pixels provides a weighted error that can be used to compare with product images of different sizes.
  • RMSE takes a short time to calculate while measuring accurately, so it is suitable for multiple calculations during the iterative calibration process, which improves the efficiency of the calibration process.
  • FIG11 is a schematic diagram showing the processing process of step S210 shown in FIG2 in an embodiment. As shown in FIG11 , in the embodiment of the present disclosure, the above step S210 may further include: Next step.
  • Step S1102 obtaining a reference image after mask processing.
  • the masked reference image may be denoted as R'.
  • Step S1104 respectively converting the transformation target image and the reference image after the mask processing from the RGB color space to the CIELAB color space to obtain the transformation target image in the CIELAB color space and the reference image in the CIELAB color space.
  • the transformation target image in the CIELAB color space may be represented as TL
  • the reference image in the CIELAB color space may be represented as RL .
  • Step S1106 performing color distribution feature normalization processing on the transformation target image in the CIELAB color space based on the reference image in the CIELAB color space to obtain a normalized transformation target image.
  • Fig. 12 is a schematic diagram showing the processing process of step S1106 shown in Fig. 11 in one embodiment. As shown in Fig. 12, in the embodiment of the present disclosure, the above step S1106 may further include the following steps.
  • Step S1202 Use a Gaussian convolution kernel to perform denoising on the transformed target image in the CIELAB color space.
  • the transformed target image in the CIELAB color space after denoising can be represented as T L' .
  • Step S1204 obtaining the pixel mean and standard deviation of the transformed target image in the CIELAB color space after denoising.
  • Step S1206 obtaining the pixel mean and standard deviation of the reference image in the CIELAB color space.
  • Step S1208 performing color distribution feature normalization processing on the transform target image in the CIELAB color space according to the pixel mean and standard deviation of the transform target image in the CIELAB color space after denoising and the pixel mean and standard deviation of the reference image in the CIELAB color space to obtain a standardized transform target image.
  • the normalized transformation target image TS may be expressed as follows:
  • the pixel mean of the transformed target image in the CIELAB color space after processing represents the pixel standard deviation of the reference image in CIELAB color space, Indicates the pixel standard deviation of the transformed target image in the CIELAB color space after denoising.
  • the RGB format image captured in the image acquisition stage is converted to the CIELAB color space, and pixel standardization is performed to transfer the color distribution characteristics of the target image to the color distribution characteristics of the reference image.
  • the reference image and the initial target image are taken under the same lighting conditions, no matter what image acquisition device is used, the conversion between the physical print and the digital image will enhance the difference between the reference image and the initial target image.
  • the CIELAB color space can offset the impact of the conversion of the print (CMYK) to the RGB space digital image, offset the image color difference caused by the camera's automatic exposure or color correction and other self-adjustment settings, can ensure that it does not affect the comparison of the reference image and the target image, and can improve the accuracy of subsequent differential results.
  • Fig. 13 is a schematic diagram showing the processing process of step S212 shown in Fig. 2 in one embodiment.
  • the standardized transformation target image may be an image in the CIELAB color space.
  • the above step S212 may further include the following steps.
  • Step S1302 Obtain a reference image in the CIELAB color space.
  • Step S1304 performing a difference process on the standardized transformation target image and the reference image in the CIELAB color space to obtain an initial difference result.
  • Step S1306 performing threshold segmentation processing on the initial difference result to obtain a binary difference result.
  • a thresholding method is used to detect abnormal values in the initial differential result, and the binary differential result output by the threshold segmentation is a binary mask (for example, if the pixel value is below a preset threshold, it is 0, and if it is greater than or equal to the threshold, it is 1).
  • Step S1308 performing morphological operation processing on the binarized difference result to obtain a final difference result, so as to detect printing defects of the printed product to be inspected according to the final difference result.
  • the morphological operation process may include erosion, dilation, and the like.
  • the noise in the binary mask during threshold segmentation can be eliminated and the accuracy of the final differential result can be improved.
  • Fig. 14 is a schematic diagram of a printing defect detection process according to Fig. 2 to Fig. 13.
  • the printing defect detection process shown in Fig. 14 may include the following steps S1402 to S1414.
  • Step S1402 image acquisition: using a photographing device to acquire images of the printed product to be inspected and its corresponding reference product.
  • Step S1404 pre-processing the image captured by the shooting device, such as enhancement, segmentation, representation, etc., to obtain a reference image 14002 and an initial target image 14004 .
  • Step S1406 image alignment, can be performed in three ways.
  • contouring S14062
  • the specific implementation can be referred to Figures 3 and 4.
  • feature point alignment S14064
  • the specific implementation can be referred to Figures 5 to 7.
  • marker alignment S14066
  • the specific implementation can be referred to Figures 8 and 9.
  • Step S1408 calibrating the result of the initial alignment.
  • Step S1410 color correction is performed on the calibration result.
  • FIG. 11 and FIG. 12 please refer to FIG. 11 and FIG. 12 .
  • Step S1412 threshold segmentation processing is performed on the initial difference result 14006 (refer to step S1304) between the color-corrected transformation target image and the reference image to obtain a binary difference result 14008.
  • the specific implementation method can refer to step S1306.
  • Step S1414 perform morphological operation processing on the threshold result to obtain a final difference result 14010.
  • the specific implementation method can refer to step S1308.
  • Fig. 15 is a block diagram of a printing defect detection device according to an exemplary embodiment.
  • the device shown in Fig. 15 can be applied to the server side of the system shown in Fig. 1, or to the terminal device shown in Fig. 1.
  • the apparatus 150 may include an acquisition module 1502 , an image alignment module 1504 , an image calibration module 1506 , a color standardization module 1508 and a difference module 1510 .
  • the acquisition module 1502 may be used to acquire an initial target image of the printed product to be inspected.
  • the acquisition module 1502 may also be used to acquire a reference image corresponding to the printed product to be inspected.
  • the image alignment module 1504 may be used to align the initial target image to the reference image in a variety of ways to obtain a plurality of candidate transformed target images.
  • the image calibration module 1506 may be used to determine a transformation target image based on a degree of similarity between each candidate transformation target image among a plurality of candidate transformation target images and a reference image.
  • the color standardization module 1508 may be used to perform color distribution feature standardization processing on the transformation target image based on the reference image to obtain a standardized transformation target image.
  • the difference module 1510 may be used to obtain a final difference result based on the normalized transformed target image and the reference image, so as to detect printing defects of the printed product to be inspected according to the final difference result.
  • Fig. 16 is a block diagram of a printing defect detection device according to an exemplary embodiment.
  • the device shown in Fig. 16 can be applied to the server side of the system shown in Fig. 1, or to the terminal device shown in Fig. 1.
  • the device 160 may include an acquisition module 1602, an image alignment module 1604, an image calibration module 1606, a color standardization module 1608 and a difference module 1610, wherein the image alignment module 1604 may include a contouring module 16042, a feature point module 16044, and a marking module 16046, and the difference module 1610 may include a thresholding module 16102 and a morphological operation module 16104.
  • the image alignment module 1604 may include a contouring module 16042, a feature point module 16044, and a marking module 16046
  • the difference module 1610 may include a thresholding module 16102 and a morphological operation module 16104.
  • the acquisition module 1602 may be used to acquire an initial target image of the printed product to be inspected.
  • the acquisition module 1602 may also be used to acquire a reference image corresponding to the printed product to be inspected.
  • the image alignment module 1604 may be used to align the initial target image with the reference image in a variety of ways to obtain a plurality of candidate transformed target images.
  • the plurality of candidate transform target images may include a first candidate transform target image, a second candidate transform target image, and a third candidate transform target image.
  • the contouring module 16042 can be used to calculate the affine transformation of the initial target image to the reference image using an iterative nearest neighbor algorithm to obtain an initial alignment result of the initial target image and the reference image; use a mask method to subtract pixels outside the overlapping points of the initial target image and the reference image in the initial alignment result to obtain a sequence of contour points that align the initial target image with the reference image; transform the initial target image according to the sequence of contour points that align the initial target image with the reference image to obtain a first candidate transformed target image.
  • the feature point module 16044 can be used to obtain feature points in the initial target image and feature points in the reference image respectively; correspond the feature points in the initial target image with the feature points in the reference image to obtain a first homography matrix between the initial target image and the reference image; obtain a first initial transformed target image and a first mask image according to the initial target image and the first homography matrix respectively; remove blank pixels in the first initial transformed target image according to the first mask image to obtain a second candidate transformed target image.
  • the feature point module 16044 can also be used to multiply the first homography matrix with the initial target image to obtain a first initial transformed target image; multiply the first homography matrix with the generated image to obtain a first mask image, wherein the generated image is an image filled with white pixels and of the same size as the initial target image.
  • the marking module 16046 can be used to obtain the marks in the initial target image and the marks in the reference image respectively; correspond the marks in the initial target image with the marks in the reference image to obtain a second homography matrix between the initial target image and the reference image; obtain a second initial transformed target image and a second mask image according to the initial target image and the second homography matrix respectively; remove blank pixels in the second initial transformed target image according to the second mask image to obtain a third candidate transformed target image.
  • the image calibration module 1606 may be used to determine a transformation target image based on a degree of similarity between each candidate transformation target image among a plurality of candidate transformation target images and a reference image.
  • the image calibration module 1606 can also be used to perform mask processing on the reference image to obtain the reference image after mask processing; for each candidate transformation target image, a root mean square error measurement method is used to obtain a similarity score between the candidate transformation target image and the reference image after mask processing, wherein a local search method is used to iteratively update the candidate transformation target image during the root mean square error measurement process; and the updated image of the candidate transformation target image with the highest similarity score to the reference image after mask processing among multiple candidate transformation target images is used as the transformation target image.
  • the color standardization module 1608 may be used to perform color distribution feature standardization processing on the transformation target image based on the reference image to obtain a standardized transformation target image.
  • the color standardization module 1608 can also be used to obtain a reference image after mask processing; convert the transformation target image and the reference image after mask processing from the RGB color space to the CIELAB color space respectively to obtain a transformation target image in the CIELAB color space and a reference image in the CIELAB color space; perform color distribution feature standardization processing on the transformation target image in the CIELAB color space based on the reference image in the CIELAB color space to obtain a standardized transformation target image.
  • the color standardization module 1608 can also be used to perform denoising on the transformed target image in the CIELAB color space using a Gaussian convolution kernel; obtain the pixel mean and standard deviation of the transformed target image in the CIELAB color space after the denoising process; obtain the pixel mean and standard deviation of the reference image in the CIELAB color space; and obtain the pixel mean and standard deviation of the transformed target image in the CIELAB color space according to the pixel mean and standard deviation of the transformed target image in the CIELAB color space after the denoising process and the pixel mean and standard deviation of the reference image in the CIELAB color space.
  • the standard deviation performs color distribution characteristic standardization on the transformation target image in the CIELAB color space to obtain a standardized transformation target image.
  • the difference module 1610 may be used to obtain a final difference result based on the normalized transformed target image and the reference image, so as to detect printing defects of the printed product to be inspected according to the final difference result.
  • the difference module 1610 can also be used to obtain a reference image in the CIELAB color space; perform difference processing on the standardized transformation target image and the reference image in the CIELAB color space to obtain an initial difference result.
  • the thresholding module 16102 may be used to perform threshold segmentation processing on the initial differential result to obtain a binary differential result.
  • the morphological operation module 16104 may be used to perform morphological operation processing on the binary difference result to obtain a final difference result, so as to detect printing defects of the printed product to be inspected according to the final difference result.
  • Fig. 17 is a schematic diagram showing the structure of an electronic device in an embodiment of the present disclosure. It should be noted that the device shown in Fig. 17 is only an example of a computer system and should not bring any limitation to the functions and scope of use of the embodiment of the present disclosure.
  • the device 1700 includes a central processing unit (CPU) 1701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1702 or a program loaded from a storage portion 1708 into a random access memory (RAM) 1703.
  • ROM read-only memory
  • RAM random access memory
  • various programs and data required for the operation of the device 1700 are also stored.
  • the CPU 1701, the ROM 1702, and the RAM 1703 are connected to each other via a bus 1704.
  • An input/output (I/O) interface 1705 is also connected to the bus 1704.
  • the following components are connected to the I/O interface 1705: an input section 1706 including a keyboard, a mouse, etc.; an output section 1707 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 1708 including a hard disk, etc.; and a communication section 1709 including a network interface card such as a LAN card, a modem, etc.
  • the communication section 1709 performs communication processing via a network such as the Internet.
  • a drive 1710 is also connected to the I/O interface 1705 as needed.
  • a removable medium 1711 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1710 as needed, so that a computer program read therefrom is installed into the storage section 1708 as needed.
  • an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains a program code for executing the method shown in the flowchart.
  • the computer program can be downloaded and installed from a network through a communication part 1709, and/or installed from a removable medium 1711.
  • CPU central processing unit
  • the computer-readable medium shown in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two.
  • the computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above.
  • Computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.
  • a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device.
  • a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried.
  • This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above.
  • Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device.
  • the program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.
  • each box in the flowchart or block diagram may represent a module, a program segment, or a portion of code, which contains one or more executable instructions for implementing the specified logical functions.
  • the box marked The functions of the drawings may also occur in a different order than that indicated in the drawings. For example, two blocks shown in succession may actually be executed substantially in parallel, or they may sometimes be executed in the opposite order, depending on the functions involved.
  • each block in a block diagram or flow chart, and combinations of blocks in a block diagram or flow chart may be implemented by a dedicated hardware-based system that performs the specified functions or operations, or may be implemented by a combination of dedicated hardware and computer instructions.
  • the modules involved in the embodiments described in the present disclosure may be implemented by software or by hardware.
  • the modules described may also be set in a processor, for example, it may be described as: a processor includes an acquisition module, an image alignment module, an image calibration module, a color standardization module and a difference module.
  • a processor includes an acquisition module, an image alignment module, an image calibration module, a color standardization module and a difference module.
  • the names of these modules do not constitute a limitation on the modules themselves in some cases, for example, the acquisition module may also be described as a "module for acquiring a captured image from a connected shooting device".
  • the present disclosure further provides a computer-readable medium, which may be included in the device described in the above embodiment; or may exist independently without being assembled into the device.
  • the above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device implements:

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Quality & Reliability (AREA)
  • Image Processing (AREA)

Abstract

提供一种印刷缺陷检测方法、装置、设备及存储介质，涉及计算机技术领域。方法包括：获取待检测印刷产品的初始目标图像；获取与待检测印刷产品对应的参考图像；采用多种方式将初始目标图像向参考图像对齐，分别获得多个候选变换目标图像；基于多个候选变换目标图像中各个候选变换目标图像与参考图像的相似程度确定变换目标图像；基于参考图像对变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像；基于标准化变换目标图像和参考图像获得最终差分结果，以根据最终差分结果检测待检测印刷产品的印刷缺陷。

Description

印刷缺陷检测方法、装置、设备及存储介质
本公开基于申请号为202310106900.3、申请日为2023年02月02日、发明名称为《印刷缺陷检测方法、装置、设备及存储介质》的中国专利申请提出，并要求该中国专利申请的优先权，该中国专利申请的全部内容在此引入本公开作为参考。
技术领域
本公开涉及计算机技术领域，具体而言，涉及一种印刷缺陷检测方法、装置、电子设备及可读存储介质。
背景技术
如今，纸制品制造和印刷行业的机械设备已经相当齐全，但在印前、印刷、印后、装配等各个环节，仍严重依赖人工检查印刷缺陷。由于机器通常处于高速运转状态，因此在人工检查之前，大量有缺陷的产品已经生产出来了，则需要扔掉有缺陷的产品，再重新校准机器。这个过程无疑造成了时间成本和经济成本的浪费。因此，如何自动化地检测印刷缺陷成为亟待解决的问题。
在所述背景技术部分公开的上述信息仅用于加强对本公开的背景的理解，因此它可以包括不构成对本领域普通技术人员已知的现有技术的信息。
发明内容
本公开的目的在于提供一种印刷缺陷检测方法、装置、电子设备及可读存储介质，可自动化地检测印刷缺陷。
本公开的其他特性和优点将通过下面的详细描述变得显然，或部分地通过本公开的实践而习得。
根据本公开的一方面，提供一种印刷缺陷检测方法，包括：获取待检测印刷产品的初始目标图像；获取与所述待检测印刷产品对应的参考图像；采用多种方式将所述初始目标图像向所述参考图像对齐，分别获得多个 候选变换目标图像；基于所述多个候选变换目标图像中各个候选变换目标图像与所述参考图像的相似程度确定变换目标图像；基于所述参考图像对所述变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像；基于所述标准化变换目标图像和所述参考图像获得最终差分结果，以根据所述最终差分结果检测所述待检测印刷产品的印刷缺陷。
根据本公开的一实施例，所述多个候选变换目标图像包括第一候选变换目标图像；采用多种方式将所述初始目标图像向所述参考图像对齐，分别获得多个候选变换目标图像，包括：采用迭代最近邻点算法计算所述初始目标图像到所述参考图像的仿射变换，获得初始目标图像与参考图像的初始配准结果；采用掩码方式将所述初始配准结果中初始目标图像与参考图像重合点之外的像素进行减除，获得初始目标图像向参考图像对齐的轮廓点序列；根据所述初始目标图像向参考图像对齐的轮廓点序列对所述初始目标图像进行变换，获得所述第一候选变换目标图像。
根据本公开的一实施例，所述多个候选变换目标图像包括第二候选变换目标图像；采用多种方式将所述初始目标图像向所述参考图像对齐，分别获得多个候选变换目标图像，包括：分别获得初始目标图像中的特征点和参考图像中的特征点；将所述初始目标图像中的特征点与所述参考图像中的特征点对应，获得所述初始目标图像与所述参考图像之间的第一单应性矩阵；根据所述初始目标图像与所述第一单应性矩阵分别获得第一初始变换目标图像和第一掩码图像；根据所述第一掩码图像去除所述第一初始变换目标图像中的空白像素，获得所述第二候选变换目标图像。
根据本公开的一实施例，根据所述初始目标图像与所述第一单应性矩阵分别获得第一初始变换目标图像和第一掩码图像，包括：将所述第一单应性矩阵与所述初始目标图像相乘，获得所述第一初始变换目标图像；将所述第一单应性矩阵与生成图像相乘，获得所述第一掩码图像，其中，所述生成图像为填充白色像素的、与所述初始目标图像相同大小的图像。
根据本公开的一实施例，所述多个候选变换目标图像包括第三候选变换目标图像；采用多种方式将所述初始目标图像向所述参考图像对齐， 分别获得多个候选变换目标图像，包括：分别获得初始目标图像中的标记和参考图像中的标记；将所述初始目标图像中的标记与所述参考图像中的标记对应，获得所述初始目标图像与所述参考图像之间的第二单应性矩阵；根据所述初始目标图像与所述第二单应性矩阵分别获得第二初始变换目标图像和第二掩码图像；根据所述第二掩码图像去除所述第二初始变换目标图像中的空白像素，获得所述第三候选变换目标图像。
根据本公开的一实施例，基于所述参考图像对所述变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像，包括：获得掩码处理后的参考图像；分别将所述变换目标图像和所述掩码处理后的参考图像从RGB颜色空间转换至CIELAB颜色空间，获得CIELAB颜色空间的变换目标图像和CIELAB颜色空间的参考图像；基于所述CIELAB颜色空间的参考图像对所述CIELAB颜色空间的变换目标图像进行颜色分布特征标准化处理，获得所述标准化变换目标图像。
根据本公开的一实施例，基于所述CIELAB颜色空间的参考图像对所述CIELAB颜色空间的变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像，包括：利用高斯卷积核对所述CIELAB颜色空间的变换目标图像进行去噪处理；获得去噪处理后的CIELAB颜色空间的变换目标图像的像素均值和标准差；获得CIELAB颜色空间的参考图像的像素均值和标准差；根据所述去噪处理后的CIELAB颜色空间的变换目标图像的像素均值和标准差、所述CIELAB颜色空间的参考图像的像素均值和标准差对所述CIELAB颜色空间的变换目标图像进行颜色分布特征标准化处理，获得所述标准化变换目标图像。
根据本公开的一实施例，基于所述多个候选变换目标图像中各个候选变换目标图像与所述参考图像的相似程度确定变换目标图像，包括：对所述参考图像进行掩码处理，获得掩码处理后的参考图像；对于所述各个候选变换目标图像，采用均方根误差度量方法获得所述候选变换目标图像与掩码处理后的参考图像的相似性得分，其中，在均方根误差度量过程中采用局部搜索方法对所述候选变换目标图像进行迭代更新；将所述多个候选变换目标图像中与掩码处理后的参考图像的相似性得分最高的候选变换目标图像的更新的图像作为所述变换目标图像。
根据本公开的一实施例，所述标准化变换目标图像为CIELAB颜色空间的图像；基于所述标准化变换目标图像和所述参考图像获得最终差分结果，以根据所述最终差分结果检测所述待检测印刷产品的印刷缺陷，包括：获得CIELAB颜色空间的参考图像；对所述标准化变换目标图像与所述CIELAB颜色空间的参考图像进行差分处理，获得初始差分结果；对所述初始差分结果进行阈值分割处理，获得二值化差分结果；对所述二值化差分结果进行形态运算处理，获得所述最终差分结果，以根据所述最终差分结果检测所述待检测印刷产品的印刷缺陷。
根据本公开的再一方面，提供一种印刷缺陷检测装置，包括：获取模块，用于获取待检测印刷产品的初始目标图像；所述获取模块，还用于获取与所述待检测印刷产品对应的参考图像；图像对齐模块，用于采用多种方式将所述初始目标图像向所述参考图像对齐，分别获得多个候选变换目标图像；图像校准模块，用于基于所述多个候选变换目标图像中各个候选变换目标图像与所述参考图像的相似程度确定变换目标图像；颜色标准化模块，用于基于所述参考图像对所述变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像；差分模块，用于基于所述标准化变换目标图像和所述参考图像获得最终差分结果，以根据所述最终差分结果检测所述待检测印刷产品的印刷缺陷。
根据本公开的再一方面，提供一种电子设备，包括：存储器、处理器及存储在所述存储器中并可在所述处理器中运行的可执行指令，所述处理器执行所述可执行指令时实现如上述任一种方法。
根据本公开的再一方面，提供一种计算机可读存储介质，其上存储有计算机可执行指令，所述可执行指令被处理器执行时实现如上述任一种方法。
应当理解的是，以上的一般描述和后文的细节描述仅是示例性的，并不能限制本公开。
附图说明
通过参照附图详细描述其示例实施例，本公开的上述和其它目标、特征及优点将变得更加显而易见。
图1示出本公开实施例中一种印刷缺陷检测系统结构的示意图。
图2示出本公开实施例中一种印刷缺陷检测方法的流程图。
图3示出了图2中所示的步骤S206在一实施例中的处理过程示意图。
图4是根据图3示出的一种图像对齐的过程示意图。
图5示出了图2中所示的步骤S206在另一实施例中的处理过程示意图。
图6示出了图5中所示的步骤S506在一实施例中的处理过程示意图。
图7是根据图5和图6示出的另一种图像对齐的过程示意图。
图8示出了图2中所示的步骤S206在又一实施例中的处理过程示意图。
图9是根据图8和图6示出的另一种图像对齐的过程示意图。
图10示出了图2中所示的步骤S208在一实施例中的处理过程示意图。
图11示出了图2中所示的步骤S210在一实施例中的处理过程示意图。
图12示出了图11中所示的步骤S1106在一实施例中的处理过程示意图。
图13示出了图2中所示的步骤S212在一实施例中的处理过程示意图。
图14是根据图2至图13示出的一种印刷缺陷检测流程示意图。
图15示出本公开实施例中一种印刷缺陷检测装置的框图。
图16示出本公开实施例中另一种印刷缺陷检测装置的框图。
图17示出本公开实施例中一种电子设备的结构示意图。
具体实施方式
现在将参考附图更全面地描述示例实施例。然而，示例实施例能够以多种形式实施，且不应被理解为限于在此阐述的范例；相反，提供这些实施例使得本公开将更加全面和完整，并将示例实施例的构思全面地传达给本领域的技术人员。附图仅为本公开的示意性图解，并非一定是按比例绘制。图中相同的附图标记表示相同或类似的部分，因而将省略 对它们的重复描述。
此外，所描述的特征、结构或特性可以以任何合适的方式结合在一个或更多实施例中。在下面的描述中，提供许多具体细节从而给出对本公开的实施例的充分理解。然而，本领域技术人员将意识到，可以实践本公开的技术方案而省略所述特定细节中的一个或更多，或者可以采用其它的方法、装置、步骤等。在其它情况下，不详细示出或描述公知结构、方法、装置、实现或者操作以避免喧宾夺主而使得本公开的各方面变得模糊。
此外，术语“第一”、“第二”等仅用于描述目的，而不能理解为指示或暗示相对重要性或者隐含指明所指示的技术特征的数量。由此，限定有“第一”、“第二”的特征可以明示或者隐含地包括一个或者更多个该特征。在本公开的描述中，“多个”的含义是至少两个，例如两个，三个等，除非另有明确具体的限定。符号“/”一般表示前后关联对象是一种“或”的关系。
在本公开中，除非另有明确的规定和限定，“连接”等术语应做广义理解，例如，可以是电连接或可以互相通讯；可以是直接相连，也可以通过中间媒介间接相连。对于本领域的普通技术人员而言，可以根据具体情况理解上述术语在本公开中的具体含义。
常见的印刷缺陷有脏斑、吸墨过度、重影、齿轮印等。在人工检查出这些印刷缺陷后，会扔掉有缺陷的产品，再重新校准机器并投入运行。需要重复这个过程，直到生产出符合标准的产品，无疑大量浪费了时间成本和经济成本。
相关技术中开发了一些对于印刷缺陷的解决方案。例如，工业印刷机制造商如海德堡德鲁克马斯钦公司和柯尼格鲍尔公司，为它们自己的胶印机提供配备了高分辨率彩色照相机和LED照明技术的联机检查和监测系统。然而，这个系统(以下简称为该监测系统)也存在着一些问题。
首先，该监测系统为随出厂印刷机系统附带的一项特殊功能，若售后对印刷机系统进行修改、更新，该监测系统可能不会接受，而且也可能不兼容旧型号的印刷机。如果由第三方供应商进行修改，原厂商则不能提供终身售后服务和保修。
其次，该监测系统通常由多个摄像头和LED照明组成，安装在一台大型印刷机(如胶印机)的最后一个印刷单元或涂布机之后。如果其中一个摄像头或LED照明不工作，整体检测性能将大大降低。
此外，附加该监测系统的印刷机设备成本和维护费用较高。用于检查的高分辨率彩色相机是独特的和昂贵的，在市场上难以找到替代品。
最后，并不是所有的工序都能被该监测系统覆盖。如丝印、模切、铺箔和装配工序等等，这些过程通常依赖于手工检查，需要有经验的工人发现缺陷，并评估它是关键缺陷还是在可接受水平内。
在深度学习时代，计算机视觉是最热门的领域之一，其可应用于自动驾驶、面部识别等各个领域。图像处理是计算机视觉的核心，通常涉及到不同的步骤：图像采集、图像增强、图像表示、图像描述、图像识别等，具体包括哪些步骤取决于图像处理的目的和应用。
图像采集是图像处理系统的基本步骤之一，即通过合适的器件获取光学图像，并利用模数转换器将其转换为用于计算机处理的图像，然后再利用计算机对其进行进一步的处理，如增强、分割、表示、描述和识别等。图像增强是指对图像进行变换和调整，使结果图像更适合显示并进行进一步分析或操作的过程。图像分割是将图像分割成多个部分，降低其复杂度，从而进一步分析每个部分，识别图像中的目标。图像表示的目标是用适当的形式表示和描述片段，如边界、或完整的区域等，以供进一步处理。图像描述可从图像中提取结构化的、可比较的属性进行区分和分析。在增强、分割、表示、描述步骤完成后，图像识别步骤可检测和分析图像，从而实现自动化处理，例如对地点、人物和物体进行分类等。
本公开提供了一种印刷缺陷检测方法，通过采用多种方式将待检测印刷产品的初始目标图像向对应的参考图像对齐，分别获得多个候选变换目标图像，然后基于多个候选变换目标图像中各个候选变换目标图像与参考图像的相似程度确定变换目标图像，基于参考图像对变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像，再基于标准化变换目标图像和参考图像获得最终差分结果，以根据最终差分结果检测待检测印刷产品的印刷缺陷，从而利用计算机视觉中的智能图像处理技术，可自动化对印刷品上的各种缺陷进行检测，进而对其进行识别，可降低对图像采 集硬件的要求。
图1示出了可以应用本公开的印刷缺陷检测方法或印刷缺陷检测装置的示例性系统架构10。
如图1所示，系统架构10可以包括终端设备102、网络104和服务器106。终端设备102可以是具有拍摄功能的各种设备，包括但不限于智能手机、数码相机、工业相机、扫描仪等各种设备等等。网络104用以在终端设备102和服务器106之间提供通信链路的介质。网络104可以包括各种连接类型，例如有线、无线通信链路或者光纤电缆等等。服务器106可以是提供各种服务的服务器或服务器集群等，例如用于数据处理的服务器、用于存储数据的数据库服务器等等。
终端设备102可以通过网络104与服务器106交互，以接收或发送数据等。例如终端设备102拍摄待检测印刷产品100获得初始目标图像，然后将初始目标图像通过网络104发送到服务器106进行处理。又例如服务器106执行本公开实施例中的印刷缺陷检测方法获得最终差分结果，然后将最终差分结果通过网络104发送到如智能手机等终端设备102以便用户查看。再例如终端设备102执行本公开实施例中的印刷缺陷检测方法获得最终差分结果，然后将最终差分结果通过网络104发送到服务器106进行存储。
应该理解，图1中的终端设备、网络和服务器的数目仅仅是示意性的。根据实现需要，可以具有任意数目的终端设备、网络和服务器。
图2是根据一示例性实施例示出的一种印刷缺陷检测方法的流程图。如图2所示的方法例如可以应用于图1所示的系统的服务器端，也可以应用于图1所示的终端设备。
参考图2，本公开实施例提供的方法20可以包括以下步骤。
在步骤S202中，获取待检测印刷产品的初始目标图像。
在一些实施例中，可利用图1中所示的终端设备102拍摄待检测印刷产品，获得初始目标图像。
在步骤S204中，获取与待检测印刷产品对应的参考图像。
在一些实施例中，可利用图1中所示的终端设备102拍摄与待检测印刷产品对应的参考(无缺陷的标准)产品，获得参考图像。
在一些实施例中，可利用图1中所示的终端设备102在相同的光照条件下分别采集参考图像和初始目标图像，以尽量减少变化，以便更好地比较和检测缺陷。
在步骤S206中，采用多种方式将初始目标图像向参考图像对齐，分别获得多个候选变换目标图像。
在一些实施例中，参考图像和初始目标图像在采集时被拍摄的产品可能不是放在统一的位置，导致参考图像和初始目标图像之间没有完全对齐。可采用多种方式将初始目标图像向参考图像对齐，分别获得对应的候选变换目标图像，再从中选择与参考图像最为相似的对齐结果。
在一些实施例中，多个候选变换目标图像可以包括第一候选变换目标图像、第二候选变换目标图像、第三候选变换目标图像。
例如，可采用轮廓化的方法将初始目标图像向参考图像对齐，获得第一候选变换目标图像，具体实施方式可参照图3和图4。
再例如，可采用特征点匹配的方法，将初始目标图像向参考图像对齐，获得第二候选变换目标图像，具体实施方式可参照图5至图7。
又例如，可采用标记匹配的方法，将初始目标图像向参考图像对齐，获得第三候选变换目标图像，具体实施方式可参照，图8和图9。
在步骤S208中，基于多个候选变换目标图像中各个候选变换目标图像与参考图像的相似程度确定变换目标图像。
在一些实施例中，可分别计算第一候选变换目标图像、第二候选变换目标图像、第三候选变换目标图像与参考图像相似性得分，选择其中相似性得分最高作为变换目标图像，具体实施方式可参照图10。
在步骤S210中，基于参考图像对变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像。
在一些实施例中，可先将(掩码处理后的)参考图像和变换目标图像从RGB颜色空间转换到CIELAB颜色空间，再进行颜色分布特征标准化处理，具体实施方式可参照图11和图12。
在步骤S212中，基于标准化变换目标图像和参考图像获得最终差分结果，以根据最终差分结果检测待检测印刷产品的印刷缺陷。
在一些实施例中，可将标准化变换目标图像与CIELAB颜色空间的掩 码处理后的参考图像进行差分后进行阈值分割及形态运算处理，获得最终差分结果，具体实施方式可参照图13。
在一些实施例中，可建立大量印刷产品的最终差分结果的数据库，利用深度学习模型等算法对最终差分结果进行分类，以检测不同类型的缺陷。分类算法可以软件或移动设备的应用程序等形式实现。
根据本公开实施例提供的印刷缺陷检测方法，通过采用多种方式将待检测印刷产品的初始目标图像向对应的参考图像对齐，再从中选择与参考图像的相似程度最高的变换目标图像进行颜色分布特征标准化处理，以获得最终差分结果，实现了自动化检测因素缺陷，解决了单一图像对齐方法处理大量产品各种类型的缺陷时适应性较差、准确率较低的技术问题。本公开实施例采用的方法依赖于对初始目标图像和参考图像的处理，不需要高分辨率相机采集图像，硬件要求低，安装成本低，与在线检测设备相比更灵活。
图3示出了图2中所示的步骤S206在一实施例中的处理过程示意图。如图3所示，本公开实施例中，上述步骤S206可以进一步包括以下步骤。
步骤S302，采用迭代最近邻点算法计算初始目标图像到参考图像的仿射变换，获得初始目标图像与参考图像的初始配准结果。
步骤S304，采用掩码方式将初始配准结果中初始目标图像与参考图像重合点之外的像素进行减除，获得初始目标图像向参考图像对齐的轮廓点序列。
步骤S306，根据初始目标图像向参考图像对齐的轮廓点序列对初始目标图像进行变换，获得第一候选变换目标图像。
图4是根据图3示出的一种图像对齐的过程示意图。如图4所示，参照图3，步骤S302中将(有印刷缺陷400的)初始目标图像4024与参考图像4022进行配准，获得初始目标图像4024与参考图像4022的初始配准结果404。然后步骤S304中应用二进制掩码来掩除两个图像交点以外的像素，获得初始目标图像向参考图像对齐的轮廓点序列406。然后步骤S306中根据初始目标图像向参考图像对齐的轮廓点序列406对初始目标图像进行变换，获得变换后的初始目标图像408。然后可对变换后的初始目标图像408进行掩码处理以去除图像中的空白像素，获得第一候选变换 目标图像4102，图中4104为根据初始目标图像向参考图像对齐的轮廓点序列406和填充白色像素的、与初始目标图像相同大小的图像获得的掩码图像。
根据本公开实施例提供的图像对齐方法，采用迭代最近邻点算法计算初始目标图像到参考图像的仿射变换，应用二进制掩码来掩除两个图像交点以外的像素，获得轮廓点序列使初始目标图像向参考图像对齐，可为后续差分处理提供对齐的目标图像和参考图像，从而提高了缺陷检测的准确性。
图5示出了图2中所示的步骤S206在另一实施例中的处理过程示意图。如图5所示，本公开实施例中，上述步骤S206可以进一步包括以下步骤。
步骤S502，分别获得初始目标图像中的特征点和参考图像中的特征点。
步骤S504，将初始目标图像中的特征点与参考图像中的特征点对应，获得初始目标图像与参考图像之间的第一单应性矩阵。单应性矩阵可以表示两个平面之间的唯一的对应变换关系。
步骤S506，根据初始目标图像与第一单应性矩阵分别获得第一初始变换目标图像和第一掩码图像。
在一些实施例中，可将第一单应性矩阵与初始目标图像和空白图像分别相乘，分别获得第一初始变换目标图像和第一掩码图像，具体实施方式可参照图6。
步骤S508，根据第一掩码图像去除第一初始变换目标图像中的空白像素，获得第二候选变换目标图像。
在一些实施例中，可将第一初始变换目标图像表示为T，第一掩码图像表示为M，则掩码处理后的参考图像T’＝T×M。
图6示出了图5中所示的步骤S506在一实施例中的处理过程示意图。如图6所示，本公开实施例中，上述步骤S506可以进一步包括以下步骤。
步骤S602，将第一单应性矩阵与初始目标图像相乘，获得第一初始变换目标图像。
步骤S604，将第一单应性矩阵与生成图像相乘，获得第一掩码图像， 其中，生成图像为填充白色像素的、与初始目标图像相同大小的图像。
图7是根据图5和图6示出的另一种图像对齐的过程示意图。如图7所示，参照图5，首先在步骤S502中分别获得(有印刷缺陷700的)初始目标图像7024中的特征点和参考图像7022中的特征点。然后在步骤S504中将初始目标图像7024中的特征点与参考图像7022中的特征点对应，获得初始目标图像7024与参考图像7022之间的第一单应性矩阵704。参照图6，在步骤S602中将第一单应性矩阵704与初始目标图像7024相乘，获得第一初始变换目标图像7062，并在步骤S604中将第一单应性矩阵704与填充白色像素的、与初始目标图像7024相同大小的生成图像7026相乘，获得第一掩码图像7064。
在一些实施例中，第一掩码图像7064的黑色三角区域代表空白像素。初始目标图像7024与参考图像7022共享相同的空白像素区域，可通过第一掩码图像7064确定初始目标图像7024中的哪些黑色像素是空白像素。在步骤S508中可根据第一掩码图像7064去除第一初始变换目标图像7062中的空白像素，获得第二候选变换目标图像708。
根据本公开实施例提供的图像对齐方法，通过匹配初始目标图像与参考图像对应的特征点，获得表示初始目标图像与参考图像之间的变换关系的第一单应性矩阵，再将第一单应性矩阵与初始目标图像相乘获得第一初始变换目标图像，并使用掩码去除第一初始变换目标图像中的背景空白像素，获得第二候选变换目标图像，可为后续差分处理提供对齐的目标图像和参考图像，从而提高了缺陷检测的准确性。
图8示出了图2中所示的步骤S206在又一实施例中的处理过程示意图。如图8所示，本公开实施例中，上述步骤S206可以进一步包括以下步骤。
步骤S802，分别获得初始目标图像中的标记和参考图像中的标记。
在一些实施例中，可利用印刷品在印刷过程中的对齐、排序、定位和方向等参考标记。
步骤S804，将初始目标图像中的标记与参考图像中的标记对应，获得初始目标图像与参考图像之间的第二单应性矩阵。
步骤S806，根据初始目标图像与第二单应性矩阵分别获得第二初始 变换目标图像和第二掩码图像。
在一些实施例中，可将第二单应性矩阵与初始目标图像和空白图像分别相乘，分别获得第二初始变换目标图像和第二掩码图像，具体实施方式类似图6。
步骤S808，根据第二掩码图像去除第二初始变换目标图像中的空白像素，获得第三候选变换目标图像。
图9是根据图8和图6示出的另一种图像对齐的过程示意图。如图9所示，参照图8，首先在步骤S802中分别获得(有印刷缺陷900的)初始目标图像9024中的特征点和参考图像9022中的特征点。然后在步骤S804中将初始目标图像9024中的特征点与参考图像9022中的特征点对应，获得初始目标图像9024与参考图像9022之间的第二单应性矩阵904。然后在步骤S806中将第二单应性矩阵904与初始目标图像9024相乘，获得第二初始变换目标图像9062，并将第二单应性矩阵904与填充白色像素的、与初始目标图像9024相同大小的生成图像9026相乘，获得第二掩码图像9064。然后在步骤S808中可根据第二掩码图像9064去除第二初始变换目标图像9062中的空白像素，获得第三候选变换目标图像908。
根据本公开实施例提供的图像对齐方法，通过匹配初始目标图像与参考图像对应的标记，获得表示初始目标图像与参考图像之间的变换关系的第二单应性矩阵，再将第二单应性矩阵与初始目标图像相乘获得第二初始变换目标图像，并使用掩码去除第二初始变换目标图像中的背景空白像素，获得第三候选变换目标图像，可为后续差分处理提供对齐的目标图像和参考图像，从而提高了缺陷检测的准确性。
图10示出了图2中所示的步骤S208在一实施例中的处理过程示意图。如图10所示，本公开实施例中，上述步骤S208可以进一步包括以下步骤。
步骤S1002，对参考图像进行掩码处理，获得掩码处理后的参考图像。
在一些实施例中，可将参考图像表示为R，对应的掩码图像(例如上述的第一掩码图像和/或第二掩码图像)表示为M，则掩码处理后的参考 图像R’＝R×M。
步骤S1004，对于各个候选变换目标图像，采用均方根误差度量方法获得候选变换目标图像与掩码处理后的参考图像的相似性得分，其中，在均方根误差度量过程中采用局部搜索方法对候选变换目标图像进行迭代更新。
在一些实施例中，可将候选变换目标图像与掩码处理后的参考图像的相似性得分SM(Similarity Metric)表示如下：
式中，R′(i,j)表示掩码处理后的参考图像R’的第i行、第j列的像素，T′(i,j)表示(掩码处理后的)候选变换目标图像T’的第i行、第j列的像素，M(i,j)表示对应的掩码图像(例如上述的第一掩码图像和/或第二掩码图像)的第i行、第j列的像素，即∑ijM(i,j)表示候选变换目标图像中去除了空白像素后的各个像素值之和则T″(i,j)表示T′(i,j)迭代更新后的像素。
步骤S1006，将多个候选变换目标图像中与掩码处理后的参考图像的相似性得分最高的候选变换目标图像的更新的图像作为变换目标图像。
本公开实施例提供的方法，采用局部搜索方法在每次迭代中进行相似性度量(即计算SM值)，通过平移、旋转和缩放等操作获得候选对象并从中选择最佳候选对象进行仿射变换，然后进入下次迭代以获得更精细的搜索空间尺度，从而实现更精确和更均匀的亚像素对齐。
采用均方根误差(Root Mean Squared Error，RMSE)指标度量候选变换目标图像与掩码处理后的参考图像的相似性，代替相关技术中的平均绝对误差指标(Mean Absolute Error，MAE)，在纯色背景上的小图案的情况下，RMSE对离群值具有理想的敏感性，因此图像对齐中的任何偏移都会产生显著的RMSE误差值。再除以去除了空白像素后总像素值提供了一个加权误差，可用于与不同尺寸的产品图像进行比较。而且RMSE在度量准确的前提下计算时间也较短，因此适用于在迭代校准过程中进行多次计算，提高了校准过程的效率。
图11示出了图2中所示的步骤S210在一实施例中的处理过程示意图。如图11所示，本公开实施例中，上述步骤S210可以进一步包括以 下步骤。
步骤S1102，获得掩码处理后的参考图像。
在一些实施例中，掩码处理后的参考图像可表示为R’。
步骤S1104，分别将变换目标图像和掩码处理后的参考图像从RGB颜色空间转换至CIELAB颜色空间，获得CIELAB颜色空间的变换目标图像和CIELAB颜色空间的参考图像。
在一些实施例中，CIELAB颜色空间的变换目标图像可表示为TL，CIELAB颜色空间的参考图像可表示为RL。
步骤S1106，基于CIELAB颜色空间的参考图像对CIELAB颜色空间的变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像。
图12示出了图11中所示的步骤S1106在一实施例中的处理过程示意图。如图12所示，本公开实施例中，上述步骤S1106可以进一步包括以下步骤。
步骤S1202，利用高斯卷积核对CIELAB颜色空间的变换目标图像进行去噪处理。
在一些实施例中，去噪处理后的CIELAB颜色空间的变换目标图像可表示为TL’。
步骤S1204，获得去噪处理后的CIELAB颜色空间的变换目标图像的像素均值和标准差。
步骤S1206，获得CIELAB颜色空间的参考图像的像素均值和标准差。
步骤S1208，根据去噪处理后的CIELAB颜色空间的变换目标图像的像素均值和标准差、CIELAB颜色空间的参考图像的像素均值和标准差对CIELAB颜色空间的变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像。
在一些实施例中，可将标准化变换目标图像TS表示如下：
式中，表示CIELAB颜色空间的参考图像的像素均值，表示去噪处 理后的CIELAB颜色空间的变换目标图像的像素均值，表示CIELAB颜色空间的参考图像的像素标准差，表示去噪处理后的CIELAB颜色空间的变换目标图像的像素标准差。
根据本公开实施例提供的方法，将在图像采集阶段捕获的RGB格式图像转换到CIELAB颜色空间，并进行像素标准化，将目标图像的颜色分布特征转移到参考图像的颜色分布特征上。虽然参考图像和初始目标图像都是在相同的光照条件下拍摄的，但无论采用何种图像采集设备，实体印刷品与数字图像之间的转换都会强化参考图像与初始目标图像之间的差异。CIELAB色彩空间可以抵消印刷品(CMYK)向RGB空间数字图像转换造成的影响，抵消相机的自动曝光或色彩校正等自调节设置导致的图像色差，可保证不影响参考图像和目标图像的比较，可提高后续差分结果的准确性。
图13示出了图2中所示的步骤S212在一实施例中的处理过程示意图。参照图12，标准化变换目标图像可以为CIELAB颜色空间的图像。如图13所示，本公开实施例中，上述步骤S212可以进一步包括以下步骤。
步骤S1302，获得CIELAB颜色空间的参考图像。
步骤S1304，对标准化变换目标图像与CIELAB颜色空间的参考图像进行差分处理，获得初始差分结果。
步骤S1306，对初始差分结果进行阈值分割处理，获得二值化差分结果。
在一些实施例中，采用阈值化方法检测初始差分结果中的异常值，阈值分割输出的二值化差分结果是一个二进制掩码(例如若像素值在预设阈值以下为0，如果大于或等于该阈值则为1)。
步骤S1308，对二值化差分结果进行形态运算处理，获得最终差分结果，以根据最终差分结果检测待检测印刷产品的印刷缺陷。
在一些实施例中，形态运算处理可以包括侵蚀、膨胀等处理。
根据本公开实施例提供的方法，通过对标准化变换目标图像与参考图像的初始差分结果阈值分割后进行形态运算处理，可以消除阈值分割时二进制掩码中的噪声，提高最终差分结果的准确性。
图14是根据图2至图13示出的一种印刷缺陷检测流程示意图。如图14所示的印刷缺陷检测流程可以包括以下步骤S1402至步骤S1414。
步骤S1402，图像采集，利用拍摄设备采集待检测印刷产品及其对应参考产品的图像。
步骤S1404，对拍摄设备采集图像进行预处理，例如可进行增强、分割、表示等处理，获得参考图像14002和初始目标图像14004。
步骤S1406，图像对齐，可分别采用三种方式进行图像对齐。例如轮廓化(S14062)，具体实施方式可参照图3和图4。例如特征点对齐(S14064)，具体实施方式可参照图5至图7。例如标记对齐(S14066)，具体实施方式可参照图8和图9。
步骤S1408，对初对齐的结果进行校准。具体实施方式可参照图10。
步骤S1410，对校准的结果进行颜色校正。具体实施方式可参照图11和图12。
步骤S1412，对颜色校正后的变换目标图像与参考图像的初始差分结果14006(参照步骤S1304)进行阈值分割处理，获得二值化差分结果14008。具体实施方式可参照步骤S1306。
步骤S1414，对于阈值化结果进行形态运算处理，获得最终差分结果14010。具体实施方式可参照步骤S1308。
图15是根据一示例性实施例示出的一种印刷缺陷检测装置的框图。如图15所示的装置例如可以应用于图1所示的系统的服务器端，也可以应用于图1所示的终端设备。
参考图15，本公开实施例提供的装置150可以包括获取模块1502、图像对齐模块1504、图像校准模块1506、颜色标准化模块1508和差分模块1510。
获取模块1502可用于获取待检测印刷产品的初始目标图像。
获取模块1502还可用于获取与待检测印刷产品对应的参考图像。
图像对齐模块1504可用于采用多种方式将初始目标图像向参考图像对齐，分别获得多个候选变换目标图像。
图像校准模块1506可用于基于多个候选变换目标图像中各个候选变换目标图像与参考图像的相似程度确定变换目标图像。
颜色标准化模块1508可用于基于参考图像对变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像。
差分模块1510可用于基于标准化变换目标图像和参考图像获得最终差分结果，以根据最终差分结果检测待检测印刷产品的印刷缺陷。
图16是根据一示例性实施例示出的一种印刷缺陷检测装置的框图。如图16所示的装置例如可以应用于图1所示的系统的服务器端，也可以应用于图1所示的终端设备。
参考图16，本公开实施例提供的装置160可以包括获取模块1602、图像对齐模块1604、图像校准模块1606、颜色标准化模块1608和差分模块1610，其中图像对齐模块1604可以包括轮廓化模块16042、特征点模块16044、标记模块16046，差分模块1610可以包括阈值化模块16102和形态运算模块16104。
获取模块1602可用于获取待检测印刷产品的初始目标图像。
获取模块1602还可用于获取与待检测印刷产品对应的参考图像。
图像对齐模块1604可用于采用多种方式将初始目标图像向参考图像对齐，分别获得多个候选变换目标图像。
多个候选变换目标图像可以包括第一候选变换目标图像、第二候选变换目标图像和第三候选变换目标图像。
轮廓化模块16042可用于采用迭代最近邻点算法计算初始目标图像到参考图像的仿射变换，获得初始目标图像与参考图像的初始配准结果；采用掩码方式将初始配准结果中初始目标图像与参考图像重合点之外的像素进行减除，获得初始目标图像向参考图像对齐的轮廓点序列；根据初始目标图像向参考图像对齐的轮廓点序列对初始目标图像进行变换，获得第一候选变换目标图像。
特征点模块16044可用于分别获得初始目标图像中的特征点和参考图像中的特征点；将初始目标图像中的特征点与参考图像中的特征点对应，获得初始目标图像与参考图像之间的第一单应性矩阵；根据初始目标图像与第一单应性矩阵分别获得第一初始变换目标图像和第一掩码图像；根据第一掩码图像去除第一初始变换目标图像中的空白像素，获得第二候选变换目标图像。
特征点模块16044还可用于将第一单应性矩阵与初始目标图像相乘，获得第一初始变换目标图像；将第一单应性矩阵与生成图像相乘，获得第一掩码图像，其中，生成图像为填充白色像素的、与初始目标图像相同大小的图像。
标记模块16046可用于分别获得初始目标图像中的标记和参考图像中的标记；将初始目标图像中的标记与参考图像中的标记对应，获得初始目标图像与参考图像之间的第二单应性矩阵；根据初始目标图像与第二单应性矩阵分别获得第二初始变换目标图像和第二掩码图像；根据第二掩码图像去除第二初始变换目标图像中的空白像素，获得第三候选变换目标图像。
图像校准模块1606可用于基于多个候选变换目标图像中各个候选变换目标图像与参考图像的相似程度确定变换目标图像。
图像校准模块1606还可用于对参考图像进行掩码处理，获得掩码处理后的参考图像；对于各个候选变换目标图像，采用均方根误差度量方法获得候选变换目标图像与掩码处理后的参考图像的相似性得分，其中，在均方根误差度量过程中采用局部搜索方法对候选变换目标图像进行迭代更新；将多个候选变换目标图像中与掩码处理后的参考图像的相似性得分最高的候选变换目标图像的更新的图像作为变换目标图像。
颜色标准化模块1608可用于基于参考图像对变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像。
颜色标准化模块1608还可用于获得掩码处理后的参考图像；分别将变换目标图像和掩码处理后的参考图像从RGB颜色空间转换至CIELAB颜色空间，获得CIELAB颜色空间的变换目标图像和CIELAB颜色空间的参考图像；基于CIELAB颜色空间的参考图像对CIELAB颜色空间的变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像。
颜色标准化模块1608还可用于利用高斯卷积核对CIELAB颜色空间的变换目标图像进行去噪处理；获得去噪处理后的CIELAB颜色空间的变换目标图像的像素均值和标准差；获得CIELAB颜色空间的参考图像的像素均值和标准差；根据去噪处理后的CIELAB颜色空间的变换目标图像的像素均值和标准差、CIELAB颜色空间的参考图像的像素均值和 标准差对CIELAB颜色空间的变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像。
差分模块1610可用于基于标准化变换目标图像和参考图像获得最终差分结果，以根据最终差分结果检测待检测印刷产品的印刷缺陷。
差分模块1610还可用于获得CIELAB颜色空间的参考图像；对标准化变换目标图像与CIELAB颜色空间的参考图像进行差分处理，获得初始差分结果。
阈值化模块16102可用于对初始差分结果进行阈值分割处理，获得二值化差分结果。
形态运算模块16104可用于对二值化差分结果进行形态运算处理，获得最终差分结果，以根据最终差分结果检测待检测印刷产品的印刷缺陷。
本公开实施例提供的装置中的各个模块的具体实现可以参照上述方法中的内容，此处不再赘述。
图17示出本公开实施例中一种电子设备的结构示意图。需要说明的是，图17示出的设备仅以计算机系统为示例，不应对本公开实施例的功能和使用范围带来任何限制。
如图17所示，设备1700包括中央处理单元(CPU)1701，其可以根据存储在只读存储器(ROM)1702中的程序或者从存储部分1708加载到随机访问存储器(RAM)1703中的程序而执行各种适当的动作和处理。在RAM 1703中，还存储有设备1700操作所需的各种程序和数据。CPU1701、ROM 1702以及RAM 1703通过总线1704彼此相连。输入/输出(I/O)接口1705也连接至总线1704。
以下部件连接至I/O接口1705：包括键盘、鼠标等的输入部分1706；包括诸如阴极射线管(CRT)、液晶显示器(LCD)等以及扬声器等的输出部分1707；包括硬盘等的存储部分1708；以及包括诸如LAN卡、调制解调器等的网络接口卡的通信部分1709。通信部分1709经由诸如因特网的网络执行通信处理。驱动器1710也根据需要连接至I/O接口1705。可拆卸介质1711，诸如磁盘、光盘、磁光盘、半导体存储器等等，根据需要安装在驱动器1710上，以便于从其上读出的计算机程序根据需要被安装入存储部分1708。
特别地，根据本公开的实施例，上文参考流程图描述的过程可以被实现为计算机软件程序。例如，本公开的实施例包括一种计算机程序产品，其包括承载在计算机可读介质上的计算机程序，该计算机程序包含用于执行流程图所示的方法的程序代码。在这样的实施例中，该计算机程序可以通过通信部分1709从网络上被下载和安装，和/或从可拆卸介质1711被安装。在该计算机程序被中央处理单元(CPU)1701执行时，执行本公开的系统中限定的上述功能。
需要说明的是，本公开所示的计算机可读介质可以是计算机可读信号介质或者计算机可读存储介质或者是上述两者的任意组合。计算机可读存储介质例如可以是——但不限于——电、磁、光、电磁、红外线、或半导体的系统、装置或器件，或者任意以上的组合。计算机可读存储介质的更具体的例子可以包括但不限于：具有一个或多个导线的电连接、便携式计算机磁盘、硬盘、随机访问存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、光纤、便携式紧凑磁盘只读存储器(CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。在本公开中，计算机可读存储介质可以是任何包含或存储程序的有形介质，该程序可以被指令执行系统、装置或者器件使用或者与其结合使用。而在本公开中，计算机可读的信号介质可以包括在基带中或者作为载波一部分传播的数据信号，其中承载了计算机可读的程序代码。这种传播的数据信号可以采用多种形式，包括但不限于电磁信号、光信号或上述的任意合适的组合。计算机可读的信号介质还可以是计算机可读存储介质以外的任何计算机可读介质，该计算机可读介质可以发送、传播或者传输用于由指令执行系统、装置或者器件使用或者与其结合使用的程序。计算机可读介质上包含的程序代码可以用任何适当的介质传输，包括但不限于：无线、电线、光缆、RF等等，或者上述的任意合适的组合。
附图中的流程图和框图，图示了按照本公开各种实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段、或代码的一部分，上述模块、程序段、或代码的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。也应当注意，在有些作为替换的实现中，方框中所标注 的功能也可以以不同于附图中所标注的顺序发生。例如，两个接连地表示的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图或流程图中的每个方框、以及框图或流程图中的方框的组合，可以用执行规定的功能或操作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
描述于本公开实施例中所涉及到的模块可以通过软件的方式实现，也可以通过硬件的方式来实现。所描述的模块也可以设置在处理器中，例如，可以描述为：一种处理器包括获取模块、图像对齐模块、图像校准模块、颜色标准化模块和差分模块。其中，这些模块的名称在某种情况下并不构成对该模块本身的限定，例如，获取模块还可以被描述为“从所连接的拍摄设备获取拍摄的图像的模块”。
作为另一方面，本公开还提供了一种计算机可读介质，该计算机可读介质可以是上述实施例中描述的设备中所包含的；也可以是单独存在，而未装配入该设备中。上述计算机可读介质承载有一个或者多个程序，当上述一个或者多个程序被一个该设备执行时，使得该设备实现：
获取待检测印刷产品的初始目标图像；获取与待检测印刷产品对应的参考图像；采用多种方式将初始目标图像向参考图像对齐，分别获得多个候选变换目标图像；基于多个候选变换目标图像中各个候选变换目标图像与参考图像的相似程度确定变换目标图像；基于参考图像对变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像；基于标准化变换目标图像和参考图像获得最终差分结果，以根据最终差分结果检测待检测印刷产品的印刷缺陷。
以上具体地示出和描述了本公开的示例性实施例。应可理解的是，本公开不限于这里描述的详细结构、设置方式或实现方法；相反，本公开意图涵盖包含在所附权利要求的精神和范围内的各种修改和等效设置。

Claims (12)

  1. 一种印刷缺陷检测方法，其中，包括：
    获取待检测印刷产品的初始目标图像；
    获取与所述待检测印刷产品对应的参考图像；
    采用多种方式将所述初始目标图像向所述参考图像对齐，分别获得多个候选变换目标图像；
    基于所述多个候选变换目标图像中各个候选变换目标图像与所述参考图像的相似程度确定变换目标图像；
    基于所述参考图像对所述变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像；
    基于所述标准化变换目标图像和所述参考图像获得最终差分结果，以根据所述最终差分结果检测所述待检测印刷产品的印刷缺陷。
  2. 根据权利要求1所述的方法，其中，所述多个候选变换目标图像包括第一候选变换目标图像；
    采用多种方式将所述初始目标图像向所述参考图像对齐，分别获得多个候选变换目标图像，包括：
    采用迭代最近邻点算法计算所述初始目标图像到所述参考图像的仿射变换，获得初始目标图像与参考图像的初始配准结果；
    采用掩码方式将所述初始配准结果中初始目标图像与参考图像重合点之外的像素进行减除，获得初始目标图像向参考图像对齐的轮廓点序列；
    根据所述初始目标图像向参考图像对齐的轮廓点序列对所述初始目标图像进行变换，获得所述第一候选变换目标图像。
  3. 根据权利要求1所述的方法，其中，所述多个候选变换目标图像包括第二候选变换目标图像；
    采用多种方式将所述初始目标图像向所述参考图像对齐，分别获得多个候选变换目标图像，包括：
    分别获得初始目标图像中的特征点和参考图像中的特征点；
    将所述初始目标图像中的特征点与所述参考图像中的特征点对应，获得所述初始目标图像与所述参考图像之间的第一单应性矩阵；
    根据所述初始目标图像与所述第一单应性矩阵分别获得第一初始变 换目标图像和第一掩码图像；
    根据所述第一掩码图像去除所述第一初始变换目标图像中的空白像素，获得所述第二候选变换目标图像。
  4. 根据权利要求3所述的方法，其中，根据所述初始目标图像与所述第一单应性矩阵分别获得第一初始变换目标图像和第一掩码图像，包括：
    将所述第一单应性矩阵与所述初始目标图像相乘，获得所述第一初始变换目标图像；
    将所述第一单应性矩阵与生成图像相乘，获得所述第一掩码图像，其中，所述生成图像为填充白色像素的、与所述初始目标图像相同大小的图像。
  5. 根据权利要求1所述的方法，其中，所述多个候选变换目标图像包括第三候选变换目标图像；
    采用多种方式将所述初始目标图像向所述参考图像对齐，分别获得多个候选变换目标图像，包括：
    分别获得初始目标图像中的标记和参考图像中的标记；
    将所述初始目标图像中的标记与所述参考图像中的标记对应，获得所述初始目标图像与所述参考图像之间的第二单应性矩阵；
    根据所述初始目标图像与所述第二单应性矩阵分别获得第二初始变换目标图像和第二掩码图像；
    根据所述第二掩码图像去除所述第二初始变换目标图像中的空白像素，获得所述第三候选变换目标图像。
  6. 根据权利要求1所述的方法，其中，基于所述参考图像对所述变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像，包括：
    获得掩码处理后的参考图像；
    分别将所述变换目标图像和所述掩码处理后的参考图像从RGB颜色空间转换至CIELAB颜色空间，获得CIELAB颜色空间的变换目标图像和CIELAB颜色空间的参考图像；
    基于所述CIELAB颜色空间的参考图像对所述CIELAB颜色空间的变换目标图像进行颜色分布特征标准化处理，获得所述标准化变换目标图 像。
  7. 根据权利要求6所述的方法，其中，基于所述CIELAB颜色空间的参考图像对所述CIELAB颜色空间的变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像，包括：
    利用高斯卷积核对所述CIELAB颜色空间的变换目标图像进行去噪处理；
    获得去噪处理后的CIELAB颜色空间的变换目标图像的像素均值和标准差；
    获得CIELAB颜色空间的参考图像的像素均值和标准差；
    根据所述去噪处理后的CIELAB颜色空间的变换目标图像的像素均值和标准差、所述CIELAB颜色空间的参考图像的像素均值和标准差对所述CIELAB颜色空间的变换目标图像进行颜色分布特征标准化处理，获得所述标准化变换目标图像。
  8. 根据权利要求1所述的方法，其中，基于所述多个候选变换目标图像中各个候选变换目标图像与所述参考图像的相似程度确定变换目标图像，包括：
    对所述参考图像进行掩码处理，获得掩码处理后的参考图像；
    对于所述各个候选变换目标图像，采用均方根误差度量方法获得所述候选变换目标图像与掩码处理后的参考图像的相似性得分，其中，在均方根误差度量过程中采用局部搜索方法对所述候选变换目标图像进行迭代更新；
    将所述多个候选变换目标图像中与掩码处理后的参考图像的相似性得分最高的候选变换目标图像的更新的图像作为所述变换目标图像。
  9. 根据权利要求1所述的方法，其中，所述标准化变换目标图像为CIELAB颜色空间的图像；
    基于所述标准化变换目标图像和所述参考图像获得最终差分结果，以根据所述最终差分结果检测所述待检测印刷产品的印刷缺陷，包括：
    获得CIELAB颜色空间的参考图像；
    对所述标准化变换目标图像与所述CIELAB颜色空间的参考图像进行差分处理，获得初始差分结果；
    对所述初始差分结果进行阈值分割处理，获得二值化差分结果；
    对所述二值化差分结果进行形态运算处理，获得所述最终差分结果，以根据所述最终差分结果检测所述待检测印刷产品的印刷缺陷。
  10. 一种印刷缺陷检测装置，其中，包括：
    获取模块，用于获取待检测印刷产品的初始目标图像；
    所述获取模块，还用于获取与所述待检测印刷产品对应的参考图像；
    图像对齐模块，用于采用多种方式将所述初始目标图像向所述参考图像对齐，分别获得多个候选变换目标图像；
    图像校准模块，用于基于所述多个候选变换目标图像中各个候选变换目标图像与所述参考图像的相似程度确定变换目标图像；
    颜色标准化模块，用于基于所述参考图像对所述变换目标图像进行颜色分布特征标准化处理，获得标准化变换目标图像；
    差分模块，用于基于所述标准化变换目标图像和所述参考图像获得最终差分结果，以根据所述最终差分结果检测所述待检测印刷产品的印刷缺陷。
  11. 一种电子设备，包括：存储器、处理器及存储在所述存储器中并可在所述处理器中运行的可执行指令，其中，所述处理器执行所述可执行指令时实现如权利要求1-9任一项所述的方法。
  12. 一种计算机可读存储介质，其上存储有计算机可执行指令，其中，所述可执行指令被处理器执行时实现如权利要求1-9任一项所述的方法。
PCT/CN2024/074707 2023-02-02 2024-01-30 印刷缺陷检测方法、装置、设备及存储介质 Ceased WO2024160201A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202310106900.3A CN118446950B (zh) 2023-02-02 2023-02-02 印刷缺陷检测方法、装置、设备及存储介质
CN202310106900.3 2023-02-02

Publications (1)

Publication Number Publication Date
WO2024160201A1 true WO2024160201A1 (zh) 2024-08-08

Family

ID=92045665

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2024/074707 Ceased WO2024160201A1 (zh) 2023-02-02 2024-01-30 印刷缺陷检测方法、装置、设备及存储介质

Country Status (2)

Country Link
CN (1) CN118446950B (zh)
WO (1) WO2024160201A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119360432A (zh) * 2024-12-23 2025-01-24 湖北微模式科技发展有限公司 一种卡证人像打印缺失检测方法及装置
CN120031881A (zh) * 2025-04-23 2025-05-23 宁德思客琦智能装备有限公司 一种基于图像处理和深度学习的缺陷检测方法和系统
CN120411099A (zh) * 2025-07-02 2025-08-01 湖北微模式科技发展有限公司 一种印刷缺陷数据生成方法、印刷缺陷检测方法及装置

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060158703A1 (en) * 2005-01-14 2006-07-20 Pavel Kisilev Print defect detection
KR20130068243A (ko) * 2011-12-15 2013-06-26 한국전자통신연구원 영상 정합 장치 및 방법
CN110473184A (zh) * 2019-08-06 2019-11-19 福建工程学院 一种pcb板缺陷检测方法
CN110956630A (zh) * 2019-12-18 2020-04-03 浙江大学 平面印刷缺陷检测方法、装置以及系统
CN111028213A (zh) * 2019-12-04 2020-04-17 北大方正集团有限公司 图像缺陷检测方法、装置、电子设备及存储介质
CN112508826A (zh) * 2020-11-16 2021-03-16 哈尔滨工业大学（深圳） 一种基于特征配准及梯度形状匹配融合的印刷品缺陷检测方法
CN115100158A (zh) * 2022-07-01 2022-09-23 北京大恒图像视觉有限公司 一种实现图像精准匹配的印刷检测方法

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2008292398A (ja) * 2007-05-28 2008-12-04 Omron Corp 部品電極の浮き不良の検査方法および基板検査装置
CN110335262A (zh) * 2019-06-28 2019-10-15 西安理工大学 一种机器视觉中印刷品缺陷自动分类方法
JP7456267B2 (ja) * 2020-04-28 2024-03-27 コニカミノルタ株式会社 画像検査装置、画像形成システム、画像検査方法および画像検査プログラム
CN113658176B (zh) * 2021-09-07 2023-11-07 重庆科技学院 基于交互注意力与卷积神经网络的瓷砖表面缺陷检测方法

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20060158703A1 (en) * 2005-01-14 2006-07-20 Pavel Kisilev Print defect detection
KR20130068243A (ko) * 2011-12-15 2013-06-26 한국전자통신연구원 영상 정합 장치 및 방법
CN110473184A (zh) * 2019-08-06 2019-11-19 福建工程学院 一种pcb板缺陷检测方法
CN111028213A (zh) * 2019-12-04 2020-04-17 北大方正集团有限公司 图像缺陷检测方法、装置、电子设备及存储介质
CN110956630A (zh) * 2019-12-18 2020-04-03 浙江大学 平面印刷缺陷检测方法、装置以及系统
CN112508826A (zh) * 2020-11-16 2021-03-16 哈尔滨工业大学（深圳） 一种基于特征配准及梯度形状匹配融合的印刷品缺陷检测方法
CN115100158A (zh) * 2022-07-01 2022-09-23 北京大恒图像视觉有限公司 一种实现图像精准匹配的印刷检测方法

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN119360432A (zh) * 2024-12-23 2025-01-24 湖北微模式科技发展有限公司 一种卡证人像打印缺失检测方法及装置
CN120031881A (zh) * 2025-04-23 2025-05-23 宁德思客琦智能装备有限公司 一种基于图像处理和深度学习的缺陷检测方法和系统
CN120411099A (zh) * 2025-07-02 2025-08-01 湖北微模式科技发展有限公司 一种印刷缺陷数据生成方法、印刷缺陷检测方法及装置

Also Published As

Publication number Publication date
CN118446950A (zh) 2024-08-06
CN118446950B (zh) 2025-11-14

Similar Documents

Publication Publication Date Title
CN118446950B (zh) 印刷缺陷检测方法、装置、设备及存储介质
CN110705405B (zh) 目标标注的方法及装置
CN111985465B (zh) 文本识别方法、装置、设备及存储介质
CN112861861B (zh) 识别数码管文本的方法、装置及电子设备
CN111178359A (zh) 车牌号码识别方法、装置和设备及计算机存储介质
KR20220101373A (ko) 영상 기반 pcb 결함 검출 장치 및 방법
CN113785181A (zh) Oled屏幕点缺陷判定方法、装置、存储介质及电子设备
CN118275449A (zh) 铜带表面缺陷检测方法、装置及设备
CN119438210A (zh) 玻璃酒瓶表面缺陷检测方法及系统
CN112348126A (zh) 一种印刷物品中目标对象的识别方法和装置
CN117011216A (zh) 一种缺陷检测方法、装置、电子设备及存储介质
CN115511821A (zh) 管道缺陷识别方法、装置、电子设备及存储介质
WO2025055548A1 (zh) 图像处理方法、装置及相关设备
Mukhopadhyay et al. PCB inspection in the context of smart manufacturing
CN115512381A (zh) 文本识别方法、装置、设备、存储介质及作业机械
CN120954020A (zh) 印章区域文字识别方法、装置及非易失性存储介质
CN118865219A (zh) 电力场景图像目标识别方法、多模态数据生成方法及系统
CN112287905A (zh) 车辆损伤识别方法、装置、设备及存储介质
CN110310341B (zh) 颜色算法中默认参数的生成方法、装置、设备和存储介质
CN112784830B (zh) 一种字符识别方法及装置
CN115393855A (zh) 一种车牌产品质量检测方法、系统及设备
CN114581933A (zh) 基于ocr的多类型快递单识别转换方法
CN116993654B (zh) 摄像头模组缺陷检测方法、装置、设备、存储介质及产品
HK40110784A (zh) 印刷缺陷检测方法、装置、设备及存储介质
CN117095417A (zh) 一种屏摄表单图像文本识别方法、装置、设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 24749680

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 24749680

Country of ref document: EP

Kind code of ref document: A1