A printed pattern image recognition method and system

Abstract

The application discloses a printed pattern image recognition method and system, and relates to the technical field of image recognition, and is used for recognizing the printing defects of a film containing a metal particle coating or a microstructure grating. The method comprises the following steps: acquiring a continuous image sequence of the film; comparing each frame of image in the continuous image sequence with a standard pattern image stored in advance to identify a potential abnormal area in which visual features are different in the continuous image sequence; based on analysis of the continuity of the potential abnormal area in continuous multiple frames of image or based on analysis of response features of the potential abnormal area under a preset physical or optical disturbance, the potential abnormal area is distinguished into a continuously existing printing defect or a transient optical artifact; and in response to distinguishing the potential abnormal area into the printing defect, an alarm signal of the existing printing defect is output. The application can effectively distinguish the real printing defect from the transient optical artifact, so that the false positive rate and the false negative rate are reduced.

Description

Technical Field

[0001] This application relates to the field of image recognition technology, and in particular to a method and system for recognizing printed patterns. Background Technology

[0002] In automated online inspection of products such as automotive stickers, capturing images of high-speed moving films using industrial cameras and comparing them with preset standard patterns is a conventional method for identifying printing defects. However, when the film contains a coating of metallic particles or a microstructured grating, it produces unique optical effects, posing a significant challenge to existing image recognition methods.

[0003] These materials are extremely sensitive to light reflection, and the microscopic physical vibrations of the thin film during high-speed operation on the printing production line are unavoidable. The combined effect of these two factors generates numerous non-defective, transient optical artifacts in the acquired images (such as randomly flickering bright spots or momentary color shifts). Traditional systems relying on static comparison struggle to distinguish these artifacts from genuine printing defects (such as omissions or smudges).

[0004] To reduce false alarms, engineers often need to lower the detection sensitivity, but this can lead to missed detections of subtle, real defects. Therefore, existing technologies face a dilemma when dealing with such special materials: high false alarm rates and high missed detection rates, severely limiting the reliability and efficiency of automated inspection. Summary of the Invention

[0005] This application proposes a printed pattern image recognition method and system, aiming to solve the technical problem that the existing technology has difficulty in accurately distinguishing between real printing defects and transient optical artifacts when identifying printing defects in thin films containing metal particle coatings or microstructure gratings, resulting in a high false alarm rate or high false alarm rate.

[0006] In a first aspect, this application provides a method for recognizing printed pattern images, used to identify printing defects in thin films containing metal particle coatings or microstructure gratings, the method comprising the following steps:

[0007] Acquire a continuous image sequence of the film moving at a preset linear speed on a printing production line;

[0008] Each frame of the continuous image sequence is compared with a pre-stored standard pattern image to identify potential abnormal regions in the continuous image sequence where visual features differ; wherein, the standard pattern image is an image generated by design, obtained by photographing a defect-free sample, or obtained by differential calculation after photographing a defect-free sample;

[0009] Based on the analysis of the continuity of the potential anomalous region in consecutive frames of images, or based on the analysis of the response characteristics of the potential anomalous region under the applied preset physical or optical perturbation, the potential anomalous region is distinguished as a persistent printing defect or a transient optical artifact.

[0010] In response to classifying the potential abnormal area as the printing defect, an alarm signal indicating the presence of the printing defect is output.

[0011] As some embodiments of this application, the step of acquiring a continuous image sequence of the film moving at a preset linear speed on a printing production line includes:

[0012] The thin film is illuminated based on a linear light source, and images are acquired at a frequency matching the linear velocity using a linear array industrial camera, thereby continuously imaging the same physical area of ​​the thin film to obtain multiple consecutive image frames of the thin film.

[0013] The multiple consecutive image frames are used to form a continuous image sequence of the thin film.

[0014] As some embodiments of this application, the step of illuminating the thin film based on a linear light source, acquiring images at a frequency matching the linear velocity using a linear array industrial camera, thereby continuously imaging the same physical area of ​​the thin film and obtaining multiple consecutive image frames of the thin film includes:

[0015] The moving thin film is illuminated using a time-division pulse method based on at least two sets of linear light sources with different illumination directions.

[0016] Based on a linear array industrial camera, images are acquired at a frequency matching the linear velocity of the film and in a time-division pulse mode synchronized with the linear light source, thereby continuously imaging the same physical area of ​​the film within a preset time interval, obtaining at least two sets of continuous multiple image frames of the film, each set of continuous multiple image frames corresponding to the illumination direction of a set of linear light sources.

[0017] As some embodiments of this application, the step of comparing each frame of the continuous image sequence with a pre-stored standard pattern image to identify potential abnormal regions in the continuous image sequence where visual features differ includes:

[0018] For the same physical region of the thin film, each corresponding frame image is extracted from at least two sets of continuous image sequences and differential operation is performed to generate a real-time differential image;

[0019] The real-time differential image is compared with a pre-stored standard pattern image to identify potential abnormal regions in the continuous image sequence where visual features differ. The standard pattern image is pre-generated and stored by obtaining the corresponding image from a defect-free standard sample using the same illumination and acquisition method as the continuous multiple image frames and performing the same differential operation.

[0020] As some embodiments of this application, the step of distinguishing the potential anomalous region as a persistent printing defect or a transient optical artifact based on the analysis of the continuity of the potential anomalous region in consecutive frames of images, or based on the analysis of the response characteristics of the potential anomalous region under an applied preset physical or optical perturbation, includes:

[0021] In a first preset number of consecutive frames:

[0022] The proportion of images where color channels are saturated in the potentially abnormal regions is statistically analyzed.

[0023] The actual center position of the potential anomaly region in the first frame image is recorded as a reference position, and the predicted position of the potential anomaly region is calculated based on the image acquisition frequency of the thin film, the reference position, and the linear velocity.

[0024] Calculate the deviation between the actual center location and the predicted location of the potential anomaly region;

[0025] If the ratio value is greater than a preset ratio threshold and the deviation value is less than a preset deviation threshold, then based on the analysis of the continuity of the potential abnormal region in consecutive frames of images, the potential abnormal region is classified as a persistent printing defect or a transient optical artifact.

[0026] Otherwise, based on the analysis of the response characteristics of the potential anomalous region under the applied preset physical or optical perturbation, the potential anomalous region is distinguished as a persistent printing defect or a transient optical artifact.

[0027] As some embodiments of this application, the step of distinguishing the potential anomalous region from a persistent printing defect or a transient optical artifact based on the analysis of the continuity of the potential anomalous region in consecutive multi-frame images includes:

[0028] In a second preset number of consecutive frames that is greater than the first preset number:

[0029] The actual center position of the potential anomaly region in the first frame image is recorded as a reference position, and the predicted position of the potential anomaly region is calculated based on the image acquisition frequency of the thin film, the reference position, and the linear velocity.

[0030] The deviation between the actual center location and the predicted location of the potential anomaly region is calculated to obtain the location persistence score;

[0031] Calculate the rate of change of the area or perimeter of the potential abnormal region to obtain the shape persistence score;

[0032] Calculate the change in the average brightness or color value of the potential abnormal region to obtain the visual feature persistence score;

[0033] The position persistence score, shape persistence score, and visual feature persistence score are compared with preset position persistence thresholds, shape persistence thresholds, and visual feature persistence thresholds, respectively.

[0034] If the position persistence score is greater than the position persistence threshold, the shape persistence score is greater than the shape persistence threshold, and the visual feature persistence score is greater than the visual feature persistence threshold, then the potential abnormal region is classified as a persistent printing defect; otherwise, the potential abnormal region is classified as a transient optical artifact.

[0035] As some embodiments of this application, the step of distinguishing the potential anomalous region from a persistent printing defect or a transient optical artifact based on the analysis of the response characteristics of the potential anomalous region under an applied preset physical or optical perturbation includes:

[0036] A preset physical or optical perturbation is applied to the potential anomaly region, and a continuous image sequence of the potential anomaly region under the physical or optical perturbation is acquired as a response image sequence.

[0037] In the response image sequence:

[0038] Calculate the positional deviation of the actual position of the potential anomaly region relative to the reference position before the disturbance to obtain the positional change value;

[0039] The visual feature value of the potential abnormal region is calculated relative to the feature value of the reference value before the disturbance to obtain the visual feature change value;

[0040] The position change value and / or visual feature change value are compared with a preset position change threshold and / or visual feature change threshold.

[0041] If the position change value is greater than the position change threshold, and / or if the visual feature change value is greater than the visual feature change threshold, then the potential abnormal region is classified as a transient optical artifact; otherwise, the potential abnormal region is classified as a persistent printing defect.

[0042] As some embodiments of this application, the preset physical or optical disturbance is one of the following:

[0043] Based on the controlled air valve, an airflow pulse of a preset duration is applied to the surface of the thin film where the potential abnormal area is located;

[0044] At the instant the response image sequence is acquired, the illumination angle or polarization direction of the linear light source of the thin film is changed.

[0045] As some embodiments of this application, the step of outputting an alarm signal indicating the presence of a printing defect in response to classifying the potential abnormal area as a printing defect includes:

[0046] In response to classifying the potential abnormal area as the printing defect, an alarm signal for the presence of the printing defect is output to the printing production line of the film to trigger at least one subsequent processing method, including audible and visual alarm, production line deceleration or shutdown, and automatic sorting.

[0047] Secondly, this application also provides a printed pattern image recognition system for identifying printing defects in thin films containing metal particle coatings or microstructure gratings, the system comprising:

[0048] The image sequence acquisition module is used to acquire a continuous image sequence of the film moving at a preset linear speed on the printing production line.

[0049] The image difference recognition module is used to compare each frame of the continuous image sequence with a pre-stored standard pattern image to identify potential abnormal regions in the continuous image sequence where visual features differ; wherein, the standard pattern image is an image generated by design, obtained by photographing a defect-free sample, or obtained by differential calculation after photographing a defect-free sample.

[0050] The differentiation module is used to differentiate the potential anomalous region into a persistent printing defect or a transient optical artifact based on the analysis of the continuity of the potential anomalous region in consecutive frames of images, or based on the analysis of the response characteristics of the potential anomalous region under a preset physical or optical disturbance.

[0051] An alarm module is used to output an alarm signal indicating the presence of the printing defect in response to classifying the potential abnormal area as the printing defect.

[0052] The technical solution according to the embodiments of this application has at least the following beneficial effects:

[0053] This application, by introducing continuous analysis of potential abnormal regions or analysis of disturbance response characteristics, can effectively distinguish between real printing defects and transient optical artifacts, thereby reducing false alarm rate and false negative rate, and improving the accuracy and reliability of special material thin film printing quality detection.

[0054] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0055] The accompanying drawings are used to provide a further understanding of the technical solutions of this application and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of this application and do not constitute a limitation on the technical solutions of this application.

[0056] Figure 1 This is a flowchart illustrating a printed pattern image recognition method provided in an embodiment of this application.

[0057] Figure 2 This is a schematic diagram of the architecture of a printed pattern image recognition system provided in an embodiment of this application. Detailed Implementation

[0058] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0059] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0060] Traditional automated online inspection systems often produce non-defective, transient optical artifacts in images when identifying printing defects in thin films containing metal particle coatings or microstructured gratings. This is because the thin film material is sensitive to light reflection and there are minute physical vibrations during high-speed movement. As a result, the system has difficulty in accurately distinguishing between real printing defects and these visual interferences, leading to a high false alarm rate or a high missed detection rate.

[0061] In this regard, such as Figure 1 As shown, this application discloses a method for recognizing printed patterns, used to identify printing defects in thin films containing metal particle coatings or microstructure gratings. The method includes the following steps:

[0062] S110, acquire a continuous image sequence of the film moving at a preset linear speed on the printing production line;

[0063] S120, each frame of the continuous image sequence is compared with a pre-stored standard pattern image to identify potential abnormal regions in the continuous image sequence where visual features differ; wherein, the standard pattern image is an image generated by design, obtained by photographing a defect-free sample, or obtained by differential calculation after photographing a defect-free sample.

[0064] S130, based on the analysis of the continuity of the potential anomalous region in consecutive frames of images, or based on the analysis of the response characteristics of the potential anomalous region under the applied preset physical or optical perturbation, the potential anomalous region is distinguished as a persistent printing defect or a transient optical artifact.

[0065] S140, in response to classifying the potential abnormal area as the printing defect, an alarm signal indicating the presence of the printing defect is output.

[0066] To facilitate a clearer understanding of the technical solutions in this application, some key terms and implementation environments are first explained. The thin films identified in this application specifically refer to thin films containing metal particle coatings or microstructured gratings. Due to their unique optical structure, these thin films exhibit high directionality and sensitivity in light reflection. Printing defects refer to persistent physical or chemical defects on the surface of the thin film that deviate from the standard pattern, such as omissions, scratches, or stains. Transient optical artifacts refer to non-persistent visual interference similar to defects that appear in the image due to minute vibrations of the thin film during high-speed movement or instantaneous changes in ambient light.

[0067] The implementation environment of this application is typically a printing production line, in which the film moves continuously at a preset linear speed. The image recognition method acquires a continuous sequence of images of the film using an industrial camera system and analyzes these images using image processing algorithms. The standard pattern image is a pre-stored reference image representing a defect-free film, which can be a computer-aided design (CAD) drawing, an image obtained by photographing a defect-free sample, or an image obtained by photographing a defect-free sample and then performing differential processing.

[0068] First, a continuous image sequence of the thin film is acquired. For example, a high-speed linear scan industrial camera can be used, along with a linear light source, to illuminate the thin film and acquire images at a frequency matching the film's linear velocity. In this way, the same physical region of the thin film can be continuously imaged, resulting in a series of consecutive image frames that together constitute a continuous image sequence of the thin film. Another approach is to use multiple area scan cameras to capture images of the thin film with overlapping fields of view, and then generate a continuous image sequence using image stitching technology.

[0069] Secondly, potential anomalous regions with discrepancies in visual features are identified. For example, the difference between a real-time image and a standard pattern image can be calculated by comparing pixel-level grayscale or color values. When the difference exceeds a preset threshold, the region is marked as a potential anomalous region. Another comparison method is to use a feature-matching algorithm to extract local features (such as edges and corners) from the real-time image and the standard pattern image, and calculate the distance between feature descriptors to identify mismatched regions.

[0070] Furthermore, based on the analysis of the continuity of potential anomalous regions in consecutive multi-frame images, or based on the analysis of the response characteristics of potential anomalous regions under applied preset physical or optical perturbations, potential anomalous regions are distinguished as persistent printing defects or transient optical artifacts.

[0071] Finally, in response to classifying potentially abnormal areas as printing defects, the system will output an alarm signal indicating the presence of a printing defect. For example, the alarm signal could be a simple indicator light illuminating or an audible alarm via a buzzer. Alternatively, the alarm signal could be a data packet containing information such as the location and type of the defect, sent via network to the production line control system to trigger subsequent automated processing.

[0072] The printed pattern image recognition method of this application works by accurately distinguishing between real defects and optical artifacts through multi-dimensional analysis. First, a continuous image sequence of the thin film is acquired using a high-speed image acquisition system to ensure that all visual information of the film during its movement is captured. Then, each frame of the image is compared with a standard pattern image to quickly identify potential abnormal areas where all visual features differ. This preliminary identification step can comprehensively identify all potentially problematic areas, including both real defects and optical artifacts.

[0073] To address the problem that traditional methods cannot distinguish between these two types of anomalies, this application introduces two key analysis strategies.

[0074] The first strategy is based on the continuity analysis of potential anomalous regions across multiple consecutive image frames. Since genuine printing defects are physically present and persistent on the film, their position, shape, and visual features exhibit high continuity and consistency across consecutive image frames. Conversely, transient optical artifacts are caused by minute vibrations of the film or instantaneous changes in light; they often exhibit non-continuous characteristics such as positional drift, irregular shape changes, or flickering visual features in consecutive images. By quantifying these characteristics, the system can determine whether an anomalous region is a persistent defect or a transient artifact.

[0075] The second strategy is based on analyzing the response characteristics of potential anomalous areas under applied pre-defined physical or optical perturbations. This method leverages the different ways in which real defects and optical artifacts respond to external perturbations. Real printing defects are physical flaws, and their location and visual characteristics typically do not change significantly under these slight perturbations. However, optical artifacts, due to their nature as instantaneous changes in light reflection characteristics, will show significant changes in their location or visual characteristics in the image when external light or physical conditions are disturbed. By analyzing the changes in the location and / or visual characteristics of the anomalous area before and after the perturbation, the system can further confirm the nature of the anomalous area.

[0076] This application, by introducing continuous analysis of potential abnormal regions or analysis of disturbance response characteristics, can effectively distinguish between real printing defects and transient optical artifacts, thereby reducing false alarm rate and false negative rate, and improving the accuracy and reliability of special material thin film printing quality detection.

[0077] In a specific embodiment of this application, the step of acquiring a continuous image sequence of the film moving at a preset linear speed on a printing production line preferably includes:

[0078] The thin film is illuminated based on a linear light source, and images are acquired at a frequency matching the linear velocity using a linear array industrial camera, thereby continuously imaging the same physical area of ​​the thin film to obtain multiple consecutive image frames of the thin film.

[0079] The multiple consecutive image frames are used to form a continuous image sequence of the thin film.

[0080] A linear light source is a device that provides narrow-band, high-intensity, uniform linear illumination. Its light typically illuminates a linear region perpendicular to the direction of film movement, ensuring consistent and sufficient illumination for each scan line of the film during image acquisition. A line-scan industrial camera is specifically designed for high-speed, high-resolution continuous imaging. Its photosensitive elements are arranged linearly, enabling it to scan the surface of a moving object line by line. By matching the image acquisition frequency of the line-scan industrial camera to the linear velocity of the film, it is ensured that the camera can continuously capture image data of the film surface during film movement. Thus, when a specific physical area on the film passes through the camera's field of view, that area is continuously scanned, generating multiple image frames. These image frames together constitute a continuous image sequence of the film, providing the foundational data for subsequent defect identification.

[0081] The proposed solution utilizes a linear light source and a linear array industrial camera to precisely match the acquisition frequency with the linear velocity, ensuring continuous and seamless imaging of the same physical region of the thin film, thereby obtaining high-quality, continuous multiple image frames. These image frames not only clearly reflect the surface features of the thin film but also provide a reliable data foundation for subsequent defect differentiation based on continuity analysis.

[0082] The following is a specific example to illustrate this.

[0083] Assume a thin film moves at a linear velocity of 10 meters per second (m / s) on a printing production line. To acquire a high-quality continuous image sequence, a high-brightness LED linear light source with an illumination area 5 millimeters wide, perpendicular to the film's direction of movement, can be configured. Simultaneously, a line-scan industrial camera with a resolution of 2048 pixels (µm) is deployed. To achieve image acquisition matching the linear velocity, the camera's line rate is set to 50,000 lines per second (lines / s). This means 50,000 lines of image data can be acquired per second, each line corresponding to a 5-millimeter-wide area on the film. In this way, the camera can continuously and seamlessly capture image data of the film surface as it moves at 10 m / s. For example, when a specific physical area on the film passes through the field of view of the linear light source and the line-scan industrial camera, that area is continuously scanned, generating multiple image frames. These continuous image frames are then stitched together in real time or arranged chronologically to form a complete, high-resolution continuous image sequence for subsequent defect identification.

[0084] In a further embodiment of this application, the step of illuminating the thin film based on a linear light source, acquiring images at a frequency matching the linear velocity using a linear array industrial camera, thereby continuously imaging the same physical area of ​​the thin film to obtain multiple consecutive image frames of the thin film preferably includes:

[0085] The moving thin film is illuminated using a time-division pulse method based on at least two sets of linear light sources with different illumination directions.

[0086] Based on a linear array industrial camera, images are acquired at a frequency matching the linear velocity of the film and in a time-division pulse mode synchronized with the linear light source, thereby continuously imaging the same physical area of ​​the film within a preset time interval, obtaining at least two sets of continuous multiple image frames of the film, each set of continuous multiple image frames corresponding to the illumination direction of a set of linear light sources.

[0087] "At least two sets of linear light sources with different illumination directions" refers to setting at least two independent sets of linear light sources along the movement path of the thin film, with each set having a different illumination angle or direction relative to the film surface. For example, one set of linear light sources can illuminate the film from the left side at a fixed angle, while another set illuminates the film from the right side at a different angle; or one set of light sources can use a direct illumination method, while the other uses an oblique illumination method. The purpose of this multi-angle illumination is to capture the reflection and scattering characteristics of the film surface from different directions, thereby more comprehensively revealing various types of printing defects, such as dents, bumps, scratches, foreign objects, or uneven coating.

[0088] "Time-division pulse illumination" of thin films can be understood as the sequential switching on and off of each group of linear light sources in a pre-defined order within extremely short time intervals. For example, when the first group of linear light sources is on, the other groups are off; when the first group is off, the second group is on, and so on. This time-division pulse illumination method ensures that only one group of linear light sources illuminates the thin film at any given time, avoiding light interference between different linear light sources, thus guaranteeing that each image frame clearly reflects the state of the thin film under the corresponding illumination conditions.

[0089] "A linear scan industrial camera acquires images at a frequency matched to the linear velocity of the thin film and synchronized with the time-division pulse mode of the linear light source." This means that when acquiring images, the acquisition frequency of the linear scan industrial camera is precisely matched with the movement speed of the thin film to ensure continuous and uninterrupted imaging of the same physical area of ​​the thin film. Simultaneously, the image acquisition shutter or exposure time of the linear scan industrial camera is strictly synchronized with the pulse opening time of the linear light source. For example, when the first set of linear light sources emits pulsed light, the linear scan industrial camera acquires the first frame image; when the second set of linear light sources emits pulsed light, the linear scan industrial camera acquires the second frame image. Through this synchronous acquisition method, it can be ensured that each frame image corresponds to the thin film state under different illumination directions, thereby obtaining multiple image sequences with different illumination characteristics.

[0090] Therefore, by using the above method, at least two sets of consecutive image frames of the thin film can be obtained, with each set of image frames corresponding to the illumination direction of a linear light source. These multiple sets of image frames together constitute a more comprehensive and richer description of the thin film surface, providing a more solid data foundation for subsequent defect identification.

[0091] Through the above technical solution, this application can obtain richer and more comprehensive visual information about the film surface. This enables the system to capture subtle defects that are easily overlooked under traditional single lighting conditions, such as tiny scratches, dents, or foreign objects that appear due to changes in the angle of light reflection, as well as coating unevenness with complex optical properties. Therefore, this application effectively improves the accuracy and comprehensiveness of defect identification, reduces the missed detection rate, and thus enhances the quality control level of the printing production line.

[0092] The following is a specific example to illustrate this.

[0093] Imagine a film moving at a linear velocity of 1 meter per second on a printing production line. To acquire a continuous sequence of images of the film, two sets of linear light sources can be configured. The first set of linear light sources is configured to illuminate the film at a 45-degree angle from above, while the second set is configured to illuminate the film at a 45-degree angle from below. These two sets of linear light sources are alternately activated by a controller using time-division pulses; for example, the first set of light sources is activated for 10 microseconds and then deactivated, followed immediately by the second set for 10 microseconds and deactivated, and so on. A linear scan industrial camera acquires images at a frequency matched to the film's linear velocity (e.g., 10,000 lines per second) and is strictly synchronized with the pulse activation of the linear light sources. When the first set of light sources is activated, the camera acquires the first frame; when the second set of light sources is activated, the camera acquires the second frame. In this way, two sets of image frames can be continuously acquired for the same physical area of ​​the film within a very short time interval. These multi-angle image data provide more comprehensive information for subsequent defect identification. For example, some dents that are not obvious under overhead lighting may become clearly visible under under overhead lighting due to the shadow effect.

[0094] In a further embodiment of this application, the step of comparing each frame of the continuous image sequence with a pre-stored standard pattern image to identify potential abnormal regions in the continuous image sequence where visual features differ preferably includes:

[0095] For the same physical region of the thin film, each corresponding frame image is extracted from at least two sets of continuous image sequences and differential operation is performed to generate a real-time differential image;

[0096] The real-time differential image is compared with a pre-stored standard pattern image to identify potential abnormal regions in the continuous image sequence where visual features differ. The standard pattern image is pre-generated and stored by obtaining the corresponding image from a defect-free standard sample using the same illumination and acquisition method as the continuous multiple image frames and performing the same differential operation.

[0097] Differential operations refer to pixel-level subtraction between two images to highlight their differences. By performing differential operations on images of the same physical area acquired under different lighting conditions, it is possible to effectively suppress inherent texture or pattern features of the film surface that appear consistent under different lighting conditions, while highlighting features that exhibit significant differences under different lighting conditions due to defects. Real-time differential images are images acquired in real-time on the printing production line and processed using differential operations. The standard pattern image is pre-generated and stored using the same lighting and acquisition methods as the real-time image acquisition, and performing the same differential operations on a defect-free standard sample. This means that the standard pattern image itself contains the expected differential response of the defect-free film under different lighting conditions, thus providing a more accurate and robust benchmark for subsequent real-time comparison.

[0098] Through the above technical solution, this application effectively overcomes the limitations of traditional single-illumination or simple image comparison methods, which are susceptible to interference from background noise and inherent optical characteristics when processing complex optical surfaces. Through differential operations, defect features are enhanced while normal background is suppressed, enabling even minute or inconspicuous defects to be clearly detected. Furthermore, using a standard pattern image generated through the same differential operation for comparison further ensures the accuracy of the comparison, thereby effectively reducing the false alarm rate and false negative rate, and improving the quality control level of the printing production line.

[0099] It is worth mentioning that the two analysis strategies in this application can be selectively applied according to the actual situation. In the preferred embodiment of this application, the step of distinguishing potential anomalous regions from persistent printing defects or transient optical artifacts based on the analysis of the continuity of potential anomalous regions in consecutive multi-frame images, or based on the analysis of the response characteristics of potential anomalous regions under applied preset physical or optical perturbations, includes:

[0100] In a first preset number of consecutive frames:

[0101] The proportion of images where color channels are saturated in the potentially abnormal regions is statistically analyzed.

[0102] The actual center position of the potential anomaly region in the first frame image is recorded as a reference position, and the predicted position of the potential anomaly region is calculated based on the image acquisition frequency of the thin film, the reference position, and the linear velocity.

[0103] Calculate the deviation between the actual center location and the predicted location of the potential anomaly region;

[0104] If the ratio value is greater than a preset ratio threshold and the deviation value is less than a preset deviation threshold, then based on the analysis of the continuity of the potential abnormal region in consecutive frames of images, the potential abnormal region is classified as a persistent printing defect or a transient optical artifact.

[0105] Otherwise, based on the analysis of the response characteristics of the potential anomalous region under the applied preset physical or optical perturbation, the potential anomalous region is distinguished as a persistent printing defect or a transient optical artifact.

[0106] "A first preset number of consecutive multi-frame images" refers to a set of consecutive image frames acquired by the system for initial evaluation after a potential anomalous region is initially detected. This number is usually small, designed to quickly acquire the basic dynamic and visual characteristics of the anomalous region; for example, it can be set to 3-10 frames, depending on the linear velocity of the film and the image acquisition frequency.

[0107] "The proportion of images in which the color channels of the potential anomalous region are saturated" refers to the calculation of the proportion of image frames in which the color channels (e.g., any one or more channels in the RGB channels) of the potential anomalous region reach their maximum brightness value (i.e., saturation) in these consecutive image frames. Color channel saturation is usually associated with strong reflections or glare, which are often characteristic of optical artifacts.

[0108] "Recording the actual center position of the potential abnormal region in the first frame image as a reference position" means using the geometric center point of the abnormal region in the first image frame as the benchmark for subsequent position tracking.

[0109] "Calculating the predicted position of the potential anomalous region based on the image acquisition frequency of the thin film, the reference position, and the linear velocity" refers to using the known motion parameters of the thin film (linear velocity) and the parameters of the image acquisition system (acquisition frequency) to deduce the theoretically expected position of the anomalous region in subsequent image frames. This helps determine whether the anomalous region is stably attached to the moving thin film.

[0110] "Calculating the deviation between the actual center position and the predicted position of the potential anomalous region" refers to comparing the detected center position of the anomalous region in the actual image frame with the predicted position obtained through the above calculation, quantifying the difference in position. A smaller deviation value usually indicates that the anomalous region is consistent with the motion of the film and may be a physical defect; while a larger deviation value may indicate that it is a transient optical artifact and its position is not fixed on the film.

[0111] The "proportion threshold" and "deviation threshold" are preset parameters used to define the critical points for the aforementioned statistical proportions and deviation values. These thresholds can be determined through extensive experiments and data analysis on samples of known defects and optical artifacts to optimize classification accuracy.

[0112] This application's solution introduces a preliminary judgment mechanism based on color channel saturation and positional deviation, enabling intelligent selection of subsequent analysis strategies. When a potential anomalous region exhibits high color channel saturation (potentially an optical artifact) for a short period but simultaneously has a small positional deviation (consistent with film movement, potentially a physical defect), this constitutes a somewhat ambiguous scenario. In this case, the system is configured to prioritize "continuous analysis," observing its persistence and feature stability over a longer image sequence to more accurately determine whether it is a persistent printing defect or a specific, long-lasting optical artifact. Conversely, if the potential anomalous region does not meet the above conditions (e.g., low saturation, or a large positional deviation indicating it is not fixed to the film), the system selects "analysis of response characteristics under applied preset physical or optical perturbations." This analysis method is particularly effective in distinguishing transient optical artifacts, as optical artifacts are typically more sensitive to illumination angles, polarization, or physical perturbations (such as airflow), while physical defects are relatively stable. This conditional strategy selection avoids performing the most complex analysis on all anomalous regions, thereby improving detection efficiency.

[0113] The following is a specific example to illustrate this.

[0114] Suppose that the system detects a potentially abnormal area on the printing production line.

[0115] First, the system acquires data for that region across multiple consecutive frames of images within a "first preset number". For example, the "first preset number" is set to 5 frames.

[0116] In these 5 frames, the system will make the following judgments:

[0117] Color channel saturation statistics: Count the number of images in these 5 frames where color channel saturation occurs in this potentially abnormal region. Assuming 4 frames show color channel saturation in this region, the calculated ratio is 0.8. If the preset ratio threshold is 0.7, then 0.8 is greater than 0.7, satisfying the first condition.

[0118] Position deviation calculation: The actual center position of the potential anomaly region in the first frame image is recorded as a reference position. Based on the image acquisition frequency and linear velocity of the thin film, the predicted position of the region in subsequent frames is calculated. Then, the deviation value between the actual center position and the predicted position of the region in subsequent frames is calculated. It is assumed that the calculated average deviation value is 2 pixels. If the preset deviation threshold is 5 pixels, then 2 pixels is less than 5 pixels, satisfying the second condition.

[0119] Since both of the above conditions (the ratio value is greater than the preset ratio threshold, and the deviation value is less than the preset deviation threshold) are met, the system will determine that the potential abnormal area belongs to the type that requires "continuity analysis". This means that the area may be a printing defect with high reflectivity but stable adhesion to the film, and its persistence needs to be confirmed through continuous observation over a longer period of time.

[0120] Conversely, if any of the above conditions are not met, for example, if the color channel saturation ratio is less than 0.7, or the position deviation is greater than 5 pixels (indicating that the abnormal area does not move stably with the film), the system will choose to perform "analysis of response characteristics under applied preset physical or optical perturbations".

[0121] In some embodiments of this application described above, the step of distinguishing potential anomalous regions from persistent printing defects or transient optical artifacts based on the analysis of the continuity of potential anomalous regions in consecutive multi-frame images preferably includes:

[0122] In a second preset number of consecutive frames that is greater than the first preset number:

[0123] The actual center position of the potential anomaly region in the first frame image is recorded as a reference position, and the predicted position of the potential anomaly region is calculated based on the image acquisition frequency of the thin film, the reference position, and the linear velocity.

[0124] The deviation between the actual center location and the predicted location of the potential anomaly region is calculated to obtain the location persistence score;

[0125] Calculate the rate of change of the area or perimeter of the potential abnormal region to obtain the shape persistence score;

[0126] Calculate the change in the average brightness or color value of the potential abnormal region to obtain the visual feature persistence score;

[0127] The position persistence score, shape persistence score, and visual feature persistence score are compared with preset position persistence thresholds, shape persistence thresholds, and visual feature persistence thresholds, respectively.

[0128] If the position persistence score is greater than the position persistence threshold, the shape persistence score is greater than the shape persistence threshold, and the visual feature persistence score is greater than the visual feature persistence threshold, then the potential abnormal region is classified as a persistent printing defect; otherwise, the potential abnormal region is classified as a transient optical artifact.

[0129] To more accurately assess the persistence of potential anomalous regions, this application employs analysis across multiple consecutive frames, exceeding a first preset number and a second preset number. The second preset number is typically greater than the first preset number; for example, the first preset number could be 5 frames, while the second preset number could be 10 frames or more, to provide observation over a longer time span, thereby enhancing the reliability of the judgment.

[0130] The location persistence score is obtained as follows: First, the actual center position of the potential anomalous region in the first frame is recorded as a reference position. Then, based on the film's image acquisition frequency, this reference position, and the film's linear velocity, the predicted position of the potential anomalous region in subsequent frames can be calculated. The stability of its location is quantified by comparing the deviation between the actual center position of the potential anomalous region in the actual frame and the calculated predicted position. A smaller deviation indicates better location persistence, and vice versa. This score aims to assess whether the anomalous region remains stable despite film movement, rather than appearing randomly or drifting.

[0131] Shape persistence score is obtained by calculating the rate of change of area or perimeter of a potential anomalous region across multiple consecutive frames. A persistent printing defect typically maintains relatively stable shape characteristics, while transient optical artifacts may exhibit significant shape variations or irregularities across different frames. By quantifying these rates of change, the two types of anomalousness can be effectively distinguished.

[0132] The visual feature persistence score is obtained by calculating the amount of change in the average brightness or color value of potentially anomalous areas. Genuine printing defects, such as flaws in metallic particle coatings or microstructured gratings, should maintain relatively consistent visual features (such as brightness and color) across consecutive frames. Transient optical artifacts, such as reflections or shadows, may exhibit drastic fluctuations in brightness or color values ​​due to minute changes in lighting conditions or viewing angle. Analyzing these variations in visual features can further enhance discriminative power.

[0133] The obtained positional persistence score, shape persistence score, and visual feature persistence score are compared with preset positional persistence thresholds, shape persistence thresholds, and visual feature persistence thresholds, respectively. These thresholds, obtained based on empirical data or trained through machine learning, are used to define the boundaries between persistent printing defects and transient optical artifacts. If all three persistence scores are greater than their corresponding preset thresholds, the potential anomalous area is considered to have high persistence and is thus classified as a persistent printing defect. Conversely, if any score fails to meet the threshold conditions, it is classified as a transient optical artifact.

[0134] This application's solution, by comprehensively considering the persistence of location, shape, and visual features, and utilizing longer image sequences for analysis, significantly improves the accuracy of distinguishing between persistent printing defects and transient optical artifacts. This helps reduce false alarms and missed alarms, avoids unnecessary production line downtime or slowdowns caused by transient interference, and ensures that genuine printing defects can be detected and addressed in a timely manner, thereby improving product quality and production efficiency.

[0135] The following is a specific example to illustrate this.

[0136] Suppose that on a printing production line, the system initially identifies a potential anomaly area. Following the method described above, this area is further tracked and analyzed in detail within a second preset number (e.g., 20 frames) of consecutive images.

[0137] Specifically, the system first records the center position of the potential anomaly region in the first frame of the image as (x0, y0). Based on the linear velocity of the thin film and the image acquisition frequency, the system calculates the predicted center position of the region in each subsequent frame. For example, in the 10th frame, the predicted position is (xp10, yp10). If the actual center position of the region in the 10th frame is (xa10, ya10), the deviation between this position and the predicted position is calculated. By statistically analyzing the deviations over these 20 frames, a position persistence score is obtained, for example, 0.95 (out of 1.0, representing a perfect match).

[0138] Simultaneously, the system calculates the area change rate and perimeter change rate of the potential anomaly region across 20 frames of images. If the area change rate is within 5% and the perimeter change rate is within 8%, a shape persistence score is calculated, for example, 0.92.

[0139] In addition, the system calculates the change in the average brightness or color value of the region across 20 frames of images. If the change in average brightness is within 10%, a visual feature persistence score is obtained, for example, 0.90.

[0140] Assuming the preset persistence thresholds are 0.85 for position, 0.80 for shape, and 0.82 for visual feature, since the calculated persistence scores of 0.95 (0.95) and 0.80 (0.92) are both greater than the calculated scores of 0.80 and 0.82 (0.90), this potential abnormal area is ultimately classified as a persistent printing defect, triggering the corresponding alarm signal.

[0141] Conversely, if another potentially anomalous region has a location persistence score of only 0.70 during tracking, failing to reach the threshold of 0.85, the region will still be classified as a transient optical artifact, even if the shape and visual feature scores are high, thus avoiding false alarms.

[0142] In some embodiments of this application described above, the step of distinguishing the potential anomalous region from a persistent printing defect or a transient optical artifact based on the analysis of the response characteristics of the potential anomalous region under an applied preset physical or optical perturbation preferably includes:

[0143] A preset physical or optical perturbation is applied to the potential anomaly region, and a continuous image sequence of the potential anomaly region under the physical or optical perturbation is acquired as a response image sequence.

[0144] In the response image sequence:

[0145] Calculate the positional deviation of the actual position of the potential anomaly region relative to the reference position before the disturbance to obtain the positional change value;

[0146] The visual feature value of the potential abnormal region is calculated relative to the feature value of the reference value before the disturbance to obtain the visual feature change value;

[0147] The position change value and / or visual feature change value are compared with a preset position change threshold and / or visual feature change threshold.

[0148] If the position change value is greater than the position change threshold, and / or if the visual feature change value is greater than the visual feature change threshold, then the potential abnormal region is classified as a transient optical artifact; otherwise, the potential abnormal region is classified as a persistent printing defect.

[0149] Applying a pre-set physical or optical perturbation refers to an external intervention on a detected potential anomalous region, designed to elicit a response from that region. The purpose is to determine the nature of the anomalous region by observing its changes before and after the perturbation. After applying the perturbation, a continuous image sequence of the potential anomalous region under the physical or optical perturbation is acquired as a response image sequence. This response image sequence is used to capture the dynamic changes of the potential anomalous region under the perturbation. Through continuous imaging, its response characteristics can be analyzed more comprehensively, such as the trajectory of movement, gradual or abrupt changes in visual features, etc.

[0150] Calculating the positional deviation of the potential anomaly region relative to its pre-disturbance reference position to obtain the positional change value involves first determining the actual center position of the potential anomaly region in each frame of the response image sequence. Simultaneously, the reference position of the potential anomaly region before the disturbance needs to be obtained. The positional change value is the spatial distance or displacement between the actual position of the potential anomaly region after the disturbance and the pre-disturbance reference position. This value reflects whether the potential anomaly region has undergone significant physical displacement under the disturbance. Further, calculating the difference in visual feature values ​​of the potential anomaly region relative to the pre-disturbance reference value to obtain the visual feature change value involves a quantitative analysis of the visual attributes of the potential anomaly region. Visual feature values ​​may include, but are not limited to, brightness, color (such as RGB values, HSV values), texture, contrast, etc. Before the disturbance, the visual feature reference value of the potential anomaly region needs to be obtained. The feature value difference is the quantitative difference between the visual feature value of the potential anomaly region after the disturbance and the pre-disturbance reference value. This value reflects whether the appearance characteristics of the potential anomaly region have changed significantly under the disturbance.

[0151] Comparing the positional change value and / or visual feature change value with preset positional change thresholds and / or visual feature change thresholds is to establish a judgment criterion. These thresholds are preset based on empirical data, experimental results, or specific application requirements to define what degree of change is considered "significant." Therefore, if the positional change value is greater than the positional change threshold, and / or if the visual feature change value is greater than the visual feature change threshold, the potential anomalous area is classified as a transient optical artifact. This means that when the position or visual feature of a potential anomalous area changes significantly beyond the preset threshold after being disturbed, it is more likely to be a transient optical phenomenon caused by specific lighting or observation conditions, rather than an inherent physical defect of the film itself. Conversely, if the positional change value is not greater than the positional change threshold, and the visual feature change value is not greater than the visual feature change threshold, the potential anomalous area is classified as a persistent printing defect. This indicates that when the position and visual features of a potential anomalous area remain relatively stable after being disturbed, and the change amount does not reach the preset threshold, it is more likely to be a real, physical printing defect.

[0152] The solution proposed in this application avoids misjudgments that may result from relying solely on static image features or simple continuity analysis, and reduces invalid alarms triggered by optical artifacts, thereby improving the inspection efficiency and automation level of the production line. Simultaneously, by accurately identifying persistent printing defects, it ensures product quality, reduces the defect rate, and provides a more robust technical guarantee for the quality control of high-precision printed materials.

[0153] The following is a specific example to illustrate this.

[0154] Suppose that during the detection of printing defects on a thin film containing a metal particle coating, the image difference recognition module identifies a potential anomalous region. In the initial continuity analysis, this region does not fully meet the criteria for a persistent printing defect, therefore further perturbation response analysis is required. In this case, the following method can be used for differentiation:

[0155] First, the system records the precise center location and average brightness value of the potential abnormal area before the disturbance as a reference.

[0156] Next, a weak airflow pulse of a preset duration (e.g., 50 milliseconds) is applied to the thin film surface where the potentially anomalous area is located through a controlled air valve. Simultaneously with the application of the airflow pulse, a linear industrial camera continuously acquires a sequence of images of the area at a high frame rate (e.g., 1000 frames per second), forming a response image sequence.

[0157] In the acquired response image sequence, the system will track the actual center location and average brightness value of the potential abnormal area in real time.

[0158] If the actual center position of the potential abnormal area is observed to have shifted by more than 2 pixels relative to the reference position before the disturbance (i.e., the position change value is greater than the preset position change threshold), and its average brightness value has changed by more than 10% relative to the reference value before the disturbance (i.e., the visual feature change value is greater than the preset visual feature change threshold), the system will determine that the potential abnormal area is a transient optical artifact, such as caused by the movement of tiny dust particles on the film surface under the action of airflow or changes in their reflective properties.

[0159] Conversely, if the positional displacement of the potential abnormal area is less than 2 pixels and the average brightness change is less than 10%, the system will classify it as a persistent printing defect, such as a physical scratch or coating defect on the film.

[0160] In this way, even subtle differences that are difficult to distinguish with the naked eye can be accurately classified through quantitative response feature analysis, greatly improving the level of intelligence in detection.

[0161] It should be noted that the preset physical or optical disturbance is one of the following:

[0162] Based on the controlled air valve, an airflow pulse of a preset duration is applied to the surface of the thin film where the potential abnormal area is located;

[0163] At the instant the response image sequence is acquired, the illumination angle or polarization direction of the linear light source of the thin film is changed.

[0164] The first disturbance method is based on a controlled air valve, which applies a pre-set airflow pulse to the surface of the thin film containing the potential anomalous region. The controlled air valve can be understood as a device capable of controlling the intensity, direction, and duration of the airflow, such as a solenoid valve or a piezoelectric valve. The pre-set duration of the airflow pulse can be adjusted according to factors such as the material of the thin film, its movement speed, and the size of the potential anomalous region, ensuring that the airflow effectively acts on the target area without affecting other areas. Its purpose is to detect the nature of the potential anomalous region through physical action.

[0165] The second perturbation method involves changing the illumination angle or polarization direction of the linear light source of the thin film at the instant of acquiring the response image sequence. Changing the illumination angle can be achieved by adjusting the physical position of the linear light source or using a controllable array of mirrors. Changing the polarization direction can be achieved by placing a rotatable polarizer in front of the light source or using an LED light source with switchable polarization. "Instantaneously" refers to completing the switching of light source parameters within a very short time during image acquisition to ensure that the response image sequence can capture the immediate changes before and after the perturbation. Its purpose is to detect the properties of potential anomalous regions through optical effects.

[0166] The introduction of these two perturbation methods enables the system to verify anomalies from both physical and optical dimensions, thereby significantly improving the accuracy and robustness of printing defect identification, effectively reducing the false alarm rate, avoiding unnecessary downtime or slowdown of the production line due to optical artifacts, and thus improving production efficiency and product quality.

[0167] In a specific embodiment of this application, the step of outputting an alarm signal indicating the presence of the printing defect in response to classifying the potential abnormal area as the printing defect preferably includes:

[0168] In response to classifying the potential abnormal area as the printing defect, an alarm signal for the presence of the printing defect is output to the printing production line of the film to trigger at least one subsequent processing method, including audible and visual alarm, production line deceleration or shutdown, and automatic sorting.

[0169] When the system confirms the existence of a persistent printing defect, the generated alarm signal is not merely displayed on the operating interface, but is directly sent to the control system of the film printing production line. "Output to the film printing production line" can be understood as transmitting alarm data packets or control commands to the relevant execution units or central control units on the production line via wired or wireless communication interfaces. The purpose is to achieve real-time sharing of defect information and automated response.

[0170] Audible and visual alarms refer to the use of equipment such as warning lights, buzzers, or loudspeakers on the production line to alert operators to abnormal situations through visual and auditory means, prompting them to take manual intervention or conduct inspections.

[0171] Production line slowdown or shutdown refers to the automatic reduction of the film conveyor speed or complete cessation of production line operation after an alarm signal is received by the production line control system, based on preset logic or the severity of the defect. The purpose is to prevent the generation of more defective products and to provide operators with time to inspect and troubleshoot.

[0172] Automated sorting refers to the process at the end of the production line or a specific location where robotic arms, pneumatic devices, or other automated mechanisms separate film products with printing defects from the qualified product stream. Its purpose is to ensure that only products meeting quality standards enter subsequent stages, reducing the labor intensity and error rate of manual sorting.

[0173] In practical applications, these follow-up processing methods can be triggered individually or in combination according to actual needs and defect types. For example, minor defects may only trigger audible and visual alarms and deceleration, while severe defects may directly trigger shutdown and automatic sorting.

[0174] This application combines alarm signals with specific post-processing methods on the printing production line, enabling the instant conversion and efficient utilization of defect information. This not only effectively avoids the continuous production of defective products, reducing production costs and scrap rates, but also alleviates the burden on operators and improves the automation level and response speed of the production line. Therefore, the solution proposed in this application ensures product quality while also enhancing the intelligent management level of the entire printing production process.

[0175] like Figure 2 As shown, this application also discloses a printed pattern image recognition system 200 for identifying printing defects in thin films containing metal particle coatings or microstructure gratings. The system includes:

[0176] Image sequence acquisition module 210 is used to acquire a continuous image sequence of the film moving at a preset linear speed on the printing production line;

[0177] The image difference recognition module 220 is used to compare each frame of the continuous image sequence with a pre-stored standard pattern image to identify potential abnormal regions in the continuous image sequence where visual features differ; wherein, the standard pattern image is an image generated by design, obtained by photographing a defect-free sample, or obtained by differential calculation after photographing a defect-free sample.

[0178] The differentiation module 230 is used to differentiate the potential anomalous region into a persistent printing defect or a transient optical artifact based on the analysis of the continuity of the potential anomalous region in consecutive multi-frame images, or based on the analysis of the response characteristics of the potential anomalous region under an applied preset physical or optical disturbance.

[0179] Alarm module 240 is configured to output an alarm signal indicating the presence of the printing defect in response to classifying the potential abnormal area as the printing defect.

[0180] The image sequence acquisition module 210 may include a high-speed linear scan industrial camera and a linear light source. The linear light source is used to illuminate the thin film, while the linear scan industrial camera acquires images at a frequency matching the linear velocity of the thin film. With this configuration, the same physical area of ​​the thin film can be continuously imaged, resulting in a series of consecutive image frames that together constitute a continuous image sequence of the thin film. In another embodiment, the image sequence acquisition module may employ multiple area scan cameras to capture images of the thin film in an overlapping field of view, and generate a continuous image sequence using image stitching technology.

[0181] The image difference recognition module 220 can use an algorithm based on pixel-level grayscale or color value comparison to calculate the difference between the real-time image and the standard pattern image. When the difference exceeds a preset threshold, the region is marked as a potential abnormal region. As another implementation, the image difference recognition module can use a feature matching-based algorithm to extract local features, such as edges and corners, from the real-time image and the standard pattern image, and calculate the distance between feature descriptors to identify mismatched regions.

[0182] The differentiation module 230 can be structurally implemented as a dedicated processing unit integrating data processing, decision logic, and control interfaces. The data processing unit comprises three core computing components: a short-series analyzer, a long-series tracker, and a perturbation response analyzer, responsible for performing preliminary screening, complete time-series stability analysis, and response characteristic analysis under active perturbation, respectively. The decision logic unit has built-in preset thresholds and judgment rules, including screening logic for intelligent triage, a scorer for generating quantified persistence scores, and a perturbation decision-maker for direct verification. The differentiation module 230 coordinates the workflow of each component through an internal trigger interface and ultimately sends the judgment result, containing information such as defect type and location, to the alarm module or the upper-level control system through an output interface.

[0183] Alarm module block 240 can be configured to output an alarm signal by simply illuminating an indicator light or emitting an audible alarm via a buzzer. Alternatively, the alarm module can generate a data packet containing information such as the location and type of the defect, and send it to an external control system via a data interface for further processing.

[0184] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0185] The foregoing has provided a detailed description of the preferred embodiments of this application. However, this application is not limited to the above-described embodiments. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application. All such equivalent modifications or substitutions are included within the scope defined in this application.

Claims

1. A method for recognizing printed pattern images, characterized in that, The method for identifying printing defects in thin films containing metal particle coatings or microstructured gratings includes the following steps: Acquire a continuous image sequence of the film moving at a preset linear speed on a printing production line; Each frame of the continuous image sequence is compared with a pre-stored standard pattern image to identify potential abnormal regions in the continuous image sequence where visual features differ; wherein, the standard pattern image is an image generated by design, obtained by photographing a defect-free sample, or obtained by differential calculation after photographing a defect-free sample; Based on the analysis of the continuity of the potential anomalous region in consecutive frames of images, or based on the analysis of the response characteristics of the potential anomalous region under the applied preset physical or optical perturbation, the potential anomalous region is distinguished as a persistent printing defect or a transient optical artifact. In response to classifying the potential abnormal area as the printing defect, an alarm signal indicating the presence of the printing defect is output.

2. The method for recognizing printed pattern images according to claim 1, characterized in that, The step of acquiring a continuous image sequence of the film moving at a preset linear speed on the printing production line includes: The thin film is illuminated based on a linear light source, and images are acquired at a frequency matching the linear velocity using a linear array industrial camera, thereby continuously imaging the same physical area of ​​the thin film to obtain multiple consecutive image frames of the thin film. The multiple consecutive image frames are used to form a continuous image sequence of the thin film.

3. The method for recognizing printed pattern images according to claim 2, characterized in that, The steps of illuminating the thin film based on a linear light source, acquiring images at a frequency matching the linear velocity using a linear array industrial camera, and thereby continuously imaging the same physical area of ​​the thin film to obtain multiple consecutive image frames of the thin film include: The moving thin film is illuminated using a time-division pulse method based on at least two sets of linear light sources with different illumination directions. Based on a linear array industrial camera, images are acquired at a frequency matching the linear velocity of the film and in a time-division pulse mode synchronized with the linear light source, thereby continuously imaging the same physical area of ​​the film within a preset time interval, obtaining at least two sets of continuous multiple image frames of the film, each set of continuous multiple image frames corresponding to the illumination direction of a set of linear light sources.

4. The method for recognizing printed pattern images according to claim 3, characterized in that, The step of comparing each frame of the continuous image sequence with a pre-stored standard pattern image to identify potential anomalous regions in the continuous image sequence that have differences in visual features includes: For the same physical region of the thin film, each corresponding frame image is extracted from at least two sets of the continuous image sequences and differential operation is performed to generate a real-time differential image; The real-time differential image is compared with a pre-stored standard pattern image to identify potential abnormal regions in the continuous image sequence where visual features differ. The standard pattern image is pre-generated and stored by obtaining the corresponding image from a defect-free standard sample using the same illumination and acquisition method as the continuous multiple image frames and performing the same differential operation.

5. The method for recognizing printed pattern images according to claim 1, characterized in that, The step of distinguishing the potential anomalous region from a persistent printing defect or a transient optical artifact based on the analysis of the continuity of the potential anomalous region in consecutive frames of images, or based on the analysis of the response characteristics of the potential anomalous region under an applied preset physical or optical perturbation, includes: In a first preset number of consecutive frames: The proportion of images where color channels are saturated in the potentially abnormal regions is statistically analyzed. The actual center position of the potential anomaly region in the first frame image is recorded as a reference position, and the predicted position of the potential anomaly region is calculated based on the image acquisition frequency of the thin film, the reference position, and the linear velocity. Calculate the deviation between the actual center location and the predicted location of the potential anomaly region; If the ratio value is greater than a preset ratio threshold and the deviation value is less than a preset deviation threshold, then based on the analysis of the continuity of the potential abnormal region in consecutive frames of images, the potential abnormal region is classified as a persistent printing defect or a transient optical artifact. Otherwise, based on the analysis of the response characteristics of the potential anomalous region under the applied preset physical or optical perturbation, the potential anomalous region is distinguished as a persistent printing defect or a transient optical artifact.

6. The method for recognizing printed pattern images according to claim 5, characterized in that, The step of distinguishing the potential anomalous region from persistent printing defects or transient optical artifacts based on the analysis of the continuity of the potential anomalous region in consecutive frames of images includes: In a second preset number of consecutive frames that is greater than the first preset number: The actual center position of the potential anomaly region in the first frame image is recorded as a reference position, and the predicted position of the potential anomaly region is calculated based on the image acquisition frequency of the thin film, the reference position, and the linear velocity. The deviation between the actual center location and the predicted location of the potential anomaly region is calculated to obtain the location persistence score; Calculate the rate of change of the area or perimeter of the potential abnormal region to obtain the shape persistence score; Calculate the change in the average brightness or color value of the potential abnormal region to obtain the visual feature persistence score; The position persistence score, shape persistence score, and visual feature persistence score are compared with preset position persistence thresholds, shape persistence thresholds, and visual feature persistence thresholds, respectively. If the position persistence score is greater than the position persistence threshold, the shape persistence score is greater than the shape persistence threshold, and the visual feature persistence score is greater than the visual feature persistence threshold, then the potential abnormal region is classified as a persistent printing defect; otherwise, the potential abnormal region is classified as a transient optical artifact.

7. The method for recognizing printed pattern images according to claim 5, characterized in that, The step of distinguishing the potential anomalous regions from persistent printing defects or transient optical artifacts based on the analysis of their response characteristics under applied preset physical or optical perturbations includes: A preset physical or optical perturbation is applied to the potential anomaly region, and a continuous image sequence of the potential anomaly region under the physical or optical perturbation is acquired as a response image sequence. In the response image sequence: Calculate the positional deviation of the actual position of the potential anomaly region relative to the reference position before the disturbance to obtain the positional change value; The visual feature value of the potential abnormal region is calculated relative to the feature value of the reference value before the disturbance to obtain the visual feature change value; The position change value and / or visual feature change value are compared with a preset position change threshold and / or visual feature change threshold. If the position change value is greater than the position change threshold, and / or if the visual feature change value is greater than the visual feature change threshold, then the potential abnormal region is classified as a transient optical artifact; otherwise, the potential abnormal region is classified as a persistent printing defect.

8. The method for recognizing printed pattern images according to claim 7, characterized in that, The preset physical or optical disturbance is one of the following: Based on the controlled air valve, an airflow pulse of a preset duration is applied to the surface of the thin film where the potential abnormal area is located; At the instant the response image sequence is acquired, the illumination angle or polarization direction of the linear light source of the thin film is changed.

9. The method for recognizing printed pattern images according to claim 1, characterized in that, The step of outputting an alarm signal indicating the presence of a printing defect in response to classifying the potential abnormal area as the printing defect includes: In response to classifying the potential abnormal area as the printing defect, an alarm signal for the presence of the printing defect is output to the printing production line of the film to trigger at least one subsequent processing method, including audible and visual alarm, production line deceleration or shutdown, and automatic sorting.

10. A printed pattern image recognition system, characterized in that, The system for identifying printing defects in thin films containing metallic particle coatings or microstructured gratings includes: The image sequence acquisition module is used to acquire a continuous image sequence of the film moving at a preset linear speed on the printing production line. The image difference recognition module is used to compare each frame of the continuous image sequence with a pre-stored standard pattern image to identify potential abnormal regions in the continuous image sequence where visual features differ; wherein, the standard pattern image is an image generated by design, obtained by photographing a defect-free sample, or obtained by differential calculation after photographing a defect-free sample. The differentiation module is used to differentiate the potential anomalous region into a persistent printing defect or a transient optical artifact based on the analysis of the continuity of the potential anomalous region in consecutive frames of images, or based on the analysis of the response characteristics of the potential anomalous region under a preset physical or optical disturbance. An alarm module is used to output an alarm signal indicating the presence of the printing defect in response to classifying the potential abnormal area as the printing defect.

Images