A positioning overprint quality intelligent detection method of a holographic anti-counterfeiting label

Abstract

The present application relates to the technical field of anti-fake label quality detection, and discloses a positioning overprint quality intelligent detection method for holographic anti-fake labels, which comprises the following steps: collecting a reference image and establishing a label reference coordinate; controlling a multi-direction light source to irradiate in sequence, collecting a directional response image and recording the light source direction; extracting a posture reference area reflection feature, matching and interpolating to obtain local posture data; combining the light source direction and the local posture to calculate an effective incidence state; extracting a holographic response position, matching a standard response position and calculating a state position deviation; screening multiple state position deviations to obtain a real overprint offset and a state residual deviation, and generating an overprint state marker and a holographic response state marker. The effective incidence state is corrected by the local posture data, and the standard holographic response position is matched accordingly. The real overprint offset is extracted from the multiple state position deviations to distinguish the local posture change, the holographic angle response change and the real overprint offset.

Description

Technical Field

[0001] This invention relates to the field of anti-counterfeiting label quality inspection technology, specifically to an intelligent inspection method for the positioning and overprinting quality of holographic anti-counterfeiting labels. Background Technology

[0002] Holographic anti-counterfeiting labels utilize holographic gratings, diffraction structures, or other optical microstructures to form anti-counterfeiting marks with angular variation characteristics. They can exhibit different brightness, color, or pattern changes under different viewing angles or lighting directions, and are widely used in the packaging of alcoholic beverages, cosmetics, pharmaceuticals, and other products. To ensure that the positional relationship between the holographic area and non-holographic printing areas such as QR codes, text, and borders meets design requirements, the label's registration quality is typically inspected during production. This is achieved by using machine vision to capture label images, obtain the positions of the holographic area and non-holographic positioning features, and determine whether there has been any lateral or vertical registration misalignment of the holographic layer.

[0003] Current methods for detecting overprinting typically involve acquiring label images under preset illumination conditions and determining overprinting based on the deviation between the detected holographic response region position and a preset standard position. However, in actual detection, the imaging position of the holographic grating changes with the incident direction. Simultaneously, local tilting during label application, lamination, or processing alters the actual effective incident state of the detection light relative to the holographic region. This means the holographic response position obtained by the camera is simultaneously affected by both the label's local orientation and the holographic angle response. Therefore, the apparent positional deviation obtained is easily mixed with optical positional changes caused by variations in the effective incident state, making it difficult to accurately distinguish this positional change from the actual geometric overprinting offset of the holographic layer relative to a non-holographic reference region. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an intelligent detection method for the positioning and overprinting quality of holographic anti-counterfeiting labels, solving the problem in existing detection methods that make it difficult to distinguish between holographic angle response, local label posture changes, and actual overprinting offset.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for intelligent detection of the positioning and overprinting quality of holographic anti-counterfeiting labels, comprising: S1. Acquire the reference image of the label to be detected, extract non-holographic positioning features, and establish the label reference coordinate data; S2. Control the detection light sources from different directions to illuminate the tag in sequence, collect the directional response image and record the light source direction data; S3. Locate the attitude reference area according to the label reference coordinate data, extract the reflection features in the orientation response image to form an attitude response feature group, compare the attitude response feature group with the pre-stored attitude response feature-surface attitude correspondence data and interpolate to obtain local attitude data. S4. Convert the light source direction data to the label reference coordinate system, and calculate the corresponding effective incident state data by combining the local attitude data. S5. Extract the holographic response position data of the holographic detection region in the orientation response image, match the current state standard response position from the pre-stored effective incident state-standard holographic response position corresponding data according to the effective incident state data, and calculate the state position deviation between the two. S6. Sort the horizontal and vertical deviations of multiple state position deviations, remove deviations that exceed the preset discrete thresholds, and then calculate the average to obtain the true overprint offset data; subtract the true overprint offset data from each state position deviation to obtain the state residual deviation, and compare the true overprint offset data and the state residual deviation with the corresponding thresholds to generate overprint state markers and holographic response state markers.

[0006] Preferably, in step S2, the process of acquiring the directional response image and recording the light source direction data includes: The detection light sources in different directions are turned on in sequence according to the preset illumination order, and the other detection light sources are turned off while the current detection light source is turned on, while keeping the imaging position of the camera unchanged. During the period when the label is illuminated by the current detection light source, the corresponding directional response image is acquired. After the acquisition is completed, the current detection light source is turned off and the next detection light source is turned on, until the directional response image acquisition for each detection light source is completed. The illumination direction of each detection light source in the preset device coordinate system is obtained, the corresponding light source direction data is generated, and the direction response images are associated and stored with the corresponding light source direction data according to the acquisition order of the direction response images.

[0007] Preferably, in step S3, the process of obtaining the attitude response feature set includes: Based on the label reference coordinate data, the attitude reference region is located in the response image in each direction, and the corresponding attitude reference region image is cropped. The pixel grayscale values ​​in each pose reference region image are compared with a preset grayscale threshold, and response pixels whose grayscale values ​​reach the preset grayscale threshold are selected. The number of response pixels is counted and the average grayscale value of the response pixels is calculated. The proportion of the number of response pixels to the total number of pixels in the pose reference region is calculated respectively, and the average gray value is normalized according to a preset gray range. The obtained proportion is combined with the normalized average gray value to obtain the directional reflection feature value corresponding to each directional response image. The directional reflection feature values ​​are arranged according to the light source direction data corresponding to each directional response image to form the attitude response feature group.

[0008] Preferably, in step S3, the process of determining the local attitude data includes: Retrieve pre-stored attitude response features—surface attitude corresponding data, which includes multiple sets of calibrated attitude response feature sets and corresponding surface attitude data; Calculate the absolute value of the difference between the attitude response feature group and the corresponding direction reflection feature value in each calibrated attitude response feature group. Accumulate the absolute values ​​of each difference to obtain the feature difference value. Sort the feature difference values ​​from smallest to largest and filter the calibrated attitude response feature groups whose feature difference values ​​are not greater than a preset difference threshold to form a candidate attitude set. The difference between the preset difference threshold and the difference values ​​of each feature in the candidate pose set is calculated. The obtained difference is normalized to obtain the interpolation weight. The corresponding surface pose data is weighted and summed according to the interpolation weight to obtain the local pose data.

[0009] Preferably, in step S4, the calculation process of the effective incident state data includes: The light source direction data is transformed from the preset device coordinate system to the coordinate system corresponding to the label reference coordinate data to obtain the light source direction data in the label coordinate system. Based on the local attitude data, the local surface normal of the holographic detection area is determined. The angle between each light source direction and the local surface normal is calculated. Each light source direction is projected onto the tangent plane corresponding to the local surface normal. The azimuth difference between the projection direction and the preset reference direction is calculated to obtain the incident tilt parameter and the incident azimuth parameter. By combining the incident tilt parameter and incident azimuth parameter corresponding to the same detection light source, the corresponding effective incident state data can be obtained.

[0010] Preferably, in step S5, the process of extracting the holographic response position data includes: Based on the label reference coordinate data, the holographic detection area is located and cropped in the response image in each direction. The pixel grayscale value in the holographic detection area is compared with the preset response threshold, and response pixels that meet the preset response threshold are selected. The response pixels are divided into connected regions, the number of pixels in each connected region is calculated, and connected regions with a number of pixels less than a preset threshold are removed to obtain the effective response region. The average position of each response pixel in the effective response region is calculated to obtain the holographic response position data corresponding to the response image in each direction.

[0011] Preferably, in step S5, the process of establishing the data corresponding to the effective incident state and the standard holographic response position includes: Select standard labels whose overprinting state meets the preset standard conditions, set multiple preset surface postures, and illuminate the standard labels with detection light sources in different directions under each preset surface posture, and collect the corresponding calibration direction response images and light source direction data. Calculate the calibrated effective incident state data based on the preset surface posture and the corresponding light source direction data, and extract the standard holographic response position data of the holographic detection area from the corresponding calibration direction response image; The calibrated effective incident state data is associated and stored with the corresponding standard holographic response position data to form the effective incident state-standard holographic response position corresponding data.

[0012] Preferably, in step S5, the matching process for the current state standard response position includes: Calculate the state difference value between the current effective incident state data and each calibrated effective incident state data in the effective incident state-standard holographic response position corresponding data, and sort them in ascending order of the state difference value; Valid incident state data with state difference values ​​not greater than a preset state difference threshold are selected. The corresponding interpolation weight is determined according to the state difference value. The corresponding standard holographic response position data is weighted and summed according to the interpolation weight to obtain the current state standard response position. The horizontal and vertical position differences between the holographic response position data and the corresponding current state standard response position are calculated respectively to form the state position deviations corresponding to the response images in each direction.

[0013] Preferably, in step S6, the process of obtaining the actual overprint offset data includes: The lateral deviation value and the longitudinal deviation value are extracted from multiple state position deviations respectively. The lateral deviation value and the longitudinal deviation value are sorted respectively, and the median value of the corresponding sorted sequence is determined as the lateral reference deviation value and the longitudinal reference deviation value. Calculate the absolute value of the difference between each lateral deviation value and the lateral reference deviation value, and the absolute value of the difference between each longitudinal deviation value and the longitudinal reference deviation value, and remove the corresponding deviation values ​​whose absolute value of the difference exceeds a preset discrete threshold. The average values ​​of the retained lateral and longitudinal deviations are calculated separately to obtain the lateral true overprint offset value and the longitudinal true overprint offset value, and the lateral true overprint offset value and the longitudinal true overprint offset value are combined to form the true overprint offset data.

[0014] Preferably, in step S6, the generation process of the overlay state marker and the holographic response state marker includes: Subtract the corresponding lateral and longitudinal true overprint offset values ​​from the lateral and longitudinal position deviation values ​​of each state position deviation to obtain the state residual deviation corresponding to each state position deviation. The absolute values ​​of the horizontal true overprint offset value and the vertical true overprint offset value are compared with the corresponding overprint offset thresholds. When both do not exceed the corresponding overprint offset thresholds, a normal overprint mark is generated. When at least one exceeds the corresponding overprint offset threshold, an abnormal overprint mark is generated. The absolute values ​​of the horizontal and vertical components of the residual deviation of each state are compared with the corresponding residual deviation thresholds, and the number of states in which at least one component exceeds the corresponding residual deviation threshold is counted. When the number of states reaches a preset threshold, a holographic response anomaly marker is generated, and when the number of states does not reach the preset threshold, a holographic response normal marker is generated.

[0015] This invention provides an intelligent detection method for the positioning and overprinting quality of holographic anti-counterfeiting labels. It has the following beneficial effects: 1. This invention corrects the effective incident state of the detection light source by combining local attitude data, matches the standard holographic response position according to the effective incident state, and extracts the true overprint offset from the multi-state position deviation, which can distinguish between the local attitude change of the label, the holographic angle response change and the true geometric overprint offset.

[0016] 2. This invention sorts, discretely filters, and averages the positional deviations in response images from multiple directions, eliminating positional deviations that deviate from most states. This allows the true overprint offset data to be determined by multiple effective lighting states, reducing the impact of single lighting state anomalies on the overprint judgment results.

[0017] 3. This invention obtains the residual state deviation by subtracting the actual overprinting offset data from the position deviation of each state, and performs threshold judgment on the actual overprinting offset data and the residual state deviation respectively, which can generate overprinting state markers and holographic response state markers respectively, thereby realizing the classification and detection of geometric overprinting anomalies and holographic response anomalies. Attached Figure Description

[0018] Figure 1 This is a flowchart of an intelligent detection method for the positioning and overprinting quality of a holographic anti-counterfeiting label according to the present invention; Figure 2 This is a flowchart illustrating the actual overprint offset data acquisition process of the present invention; Figure 3 This is a schematic diagram of the planar structure of the holographic anti-counterfeiting label of the present invention. Detailed Implementation

[0019] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see the appendix Figure 1 and attached Figure 3 This invention provides an intelligent detection method for the positioning and overprinting quality of holographic anti-counterfeiting labels, comprising: S1. Acquire the reference image of the label to be detected, extract non-holographic positioning features, and establish the label reference coordinate data; Specifically, after the tag to be detected enters the camera imaging area, the reference illumination source is turned on and a reference image is acquired. The reference image undergoes grayscale processing and region segmentation, extracting QR code location detection patterns, printed positioning patterns, or other non-holographic positioning features. The center position of each positioning feature is calculated, and the origin, coordinate direction, and scale parameters of the tag reference coordinates are determined based on the positional relationships between multiple positioning features, forming tag reference coordinate data. Subsequently, the pose reference region and holographic detection region in the orientation response image are both mapped according to the tag reference coordinate data, ensuring that image data obtained under different illumination conditions uses a unified tag position reference.

[0021] S2. Control the detection light sources from different directions to illuminate the tag in sequence, collect the directional response image and record the light source direction data; Furthermore, in S2, the process of acquiring the directional response image and recording the light source direction data includes: The detection light sources in different directions are turned on in sequence according to the preset illumination order, and the other detection light sources are turned off while the current detection light source is turned on, while keeping the imaging position of the camera unchanged. During the period when the label is illuminated by the current detection light source, the corresponding directional response image is acquired. After the acquisition is completed, the current detection light source is turned off and the next detection light source is turned on, until the directional response image acquisition for each detection light source is completed. The illumination direction of each detection light source in the preset device coordinate system is obtained, the corresponding light source direction data is generated, and the direction response images are associated and stored with the corresponding light source direction data according to the acquisition order of the direction response images.

[0022] Specifically, in one embodiment, multiple detection light sources can be positioned in different spatial directions within the label detection area. The installation position and illumination direction of each detection light source are calibrated during the equipment debugging phase, and a preset equipment coordinate system is established using the detection device where the camera is located. Based on the spatial positional relationship of the detection light sources relative to the coordinate axes of the preset equipment coordinate system, the illumination direction of each detection light source is determined, and the corresponding direction parameters are pre-stored in the parameter storage area of ​​the detection device. After detection begins, the camera remains in a fixed installation state. Once the detection label enters the preset imaging area, the camera's focal length, shooting position, and imaging direction remain unchanged throughout the detection process, ensuring that images acquired under different lighting conditions have a unified imaging benchmark.

[0023] Each detection light source operates sequentially according to a pre-set illumination order. Once the current detection light source is turned on, the others remain off, ensuring that the optical response in the currently acquired image primarily corresponds to the illumination direction of the current detection light source. After a detection light source is turned on, a preset stabilization time is allowed before the camera exposure is triggered, allowing the light emission state of the current detection light source to enter a stable range. After the camera completes image acquisition, it saves the obtained image as the directional response image corresponding to the current illumination state, then turns off the current detection light source and turns on the next detection light source according to the illumination order. This process continues until all detection light sources in the preset illumination order have completed the acquisition of their corresponding directional response images. During the acquisition of the same tag, the tag's position within the imaging area remains unchanged to reduce the impact of tag displacement on the spatial correspondence between different directional response images.

[0024] For each image acquisition, the detection device reads the illumination direction of the currently activated detection light source in the preset device coordinate system based on the pre-stored installation parameters, and generates corresponding light source direction data. The light source direction data characterizes the spatial direction of light rays pointing from the detection light source to the label detection area, and can be recorded using direction vector parameters in the device coordinate system. After the direction response image is acquired, it is associated with the light source direction data corresponding to the current detection light source according to the actual acquisition sequence, and written into the data record corresponding to the same label to be detected.

[0025] For example, in one feasible implementation, four detection light sources are set up, which sequentially illuminate the label detection area from different directions. After the device completes four illuminations and four image acquisitions in a preset order, it forms four directional response images and simultaneously obtains four sets of corresponding light source direction data. In subsequent processing, each directional response image calls its associated light source direction data, so that subsequent local attitude calculations and effective incident state calculations can use the illumination direction information corresponding to the current directional response image. Through this associated storage method, the spatial illumination conditions corresponding to each directional response image can be retained without changing the camera imaging position, and a continuous data foundation can be provided for subsequent attitude response feature extraction and holographic response position processing.

[0026] S3. Locate the attitude reference area according to the label reference coordinate data, extract the reflection features in the orientation response image to form an attitude response feature group, compare the attitude response feature group with the pre-stored attitude response feature-surface attitude correspondence data and interpolate to obtain local attitude data. Furthermore, in S3, the process of obtaining the attitude response feature set includes: Based on the label reference coordinate data, the attitude reference region is located in the response images of each direction, and the corresponding attitude reference region image is cropped. The pixel grayscale values ​​in the reference region images of each pose are compared with the preset grayscale threshold. Response pixels whose grayscale values ​​reach the preset grayscale threshold are selected, the number of response pixels is counted, and the average grayscale value of the response pixels is calculated. The proportion of the number of response pixels to the total number of pixels in the pose reference area is calculated, and the average gray value is normalized according to the preset gray range. The obtained proportion is combined with the normalized average gray value to obtain the directional reflection feature value corresponding to the response image in each direction. Arrange the directional reflection feature values ​​according to the light source direction data corresponding to each directional response image to form an attitude response feature group.

[0027] Specifically, after associating and storing the response images in each direction with the light source direction data, the previously obtained label reference coordinate data is used as the basis for spatial positioning. The attitude reference region can be pre-determined in the label design coordinates, and its position and size relative to the origin of the label reference coordinates are recorded. When detecting the current label, based on the origin position, coordinate axis direction, and scale parameters in the label reference coordinate data, the pre-recorded boundary of the attitude reference region is mapped to the response images in each direction, and the attitude reference region image corresponding to the same label position is extracted from each response image. Since the response images in each direction are acquired sequentially under the condition that the camera imaging position remains unchanged, the spatial position of the attitude reference region in different response images is determined using the same label reference coordinates.

[0028] For the cropped attitude reference region image, the image is converted to grayscale data, and the grayscale value of each pixel within the attitude reference region is read. A preset grayscale threshold can be determined during the device calibration stage based on the reflectance grayscale distribution of a standard label under various detection light source illumination conditions. The grayscale value of each pixel is compared with the preset grayscale threshold; pixels whose grayscale values ​​reach the preset threshold are identified as response pixels, while other pixels are not included in the current direction reflection feature statistics. After filtering, the number of response pixels is counted, and the grayscale values ​​of all response pixels are summed. The summed values ​​are then averaged using the number of response pixels to obtain the average grayscale value of the response pixels corresponding to the current attitude reference region image.

[0029] The proportion of responding pixels is further calculated based on the total number of pixels in the attitude reference region. This proportion records the area where a significant reflection response occurs under the current illumination direction, and the average grayscale value of the responding pixels records the reflection intensity of that area. To ensure a uniform data scale for grayscale data from different images, the average grayscale value of the responding pixels is normalized according to a preset grayscale range. This preset grayscale range corresponds to the grayscale bit depth of the orientation response image; for example, when using an eight-bit grayscale image, normalization can be performed based on a grayscale range of 0 to 255.

[0030] The proportion of response pixels and the normalized average gray value corresponding to each directional response image are used as two reflection feature components, which are combined in a fixed data arrangement order to form the directional reflection feature value of that directional response image. For example, if the attitude reference region in a certain directional response image has 10,000 pixels, and 4,000 pixels reach a preset gray value threshold, then the proportion of response pixels is 0.4; if the normalized average gray value of the response pixels is 0.68, then the directional reflection feature value is composed of two feature components: 0.4 and 0.68. This data only characterizes the reflection state of the attitude reference region under the illumination direction corresponding to the current detected light source and is not directly used as local attitude data.

[0031] After extracting the reflection features of the response images in each direction, the light source direction data associated with and stored during the preceding acquisition process is retrieved, and the corresponding directional reflection feature values ​​are arranged according to a preset light source direction order. When detecting light sources from four directions, four sets of directional reflection feature values ​​can be arranged sequentially. Each set of directional reflection feature values ​​retains two feature components: the response pixel ratio and the normalized average gray value, thus forming the attitude response feature group of the current label to be detected. The attitude response feature group also retains the correlation between different illumination directions and their corresponding reflection features, and serves as the data basis for subsequent comparisons with pre-stored attitude response features and surface attitude corresponding data.

[0032] Furthermore, in S3, the process of determining local pose data includes: Retrieve pre-stored attitude response features—surface attitude corresponding data. The attitude response features—surface attitude corresponding data includes multiple sets of calibrated attitude response feature sets and corresponding surface attitude data. Calculate the absolute value of the difference between the attitude response feature group and the corresponding direction reflection feature value in each calibrated attitude response feature group. Accumulate the absolute values ​​of each difference to obtain the feature difference value. Sort the feature difference values ​​from smallest to largest and select the calibrated attitude response feature groups whose feature difference values ​​are not greater than the preset difference threshold to form a candidate attitude set. Calculate the difference between the preset difference threshold and the difference values ​​of each feature in the candidate pose set, normalize the obtained difference to obtain the interpolation weight, and perform weighted summation on the corresponding surface pose data according to the interpolation weight to obtain the local pose data.

[0033] Specifically, after obtaining the attitude response feature set of the tag to be detected, the attitude response features—surface attitude corresponding data—are retrieved from the device's calibration data storage area. This corresponding data can be pre-established during the device calibration stage. Specifically, the standard tag can be adjusted to multiple known surface attitudes, and response images in each direction are acquired in the same order as the actual detection light source under each surface attitude. The reflection feature values ​​in each direction are extracted in the aforementioned manner, and the reflection feature values ​​in multiple directions under the same surface attitude form a calibration attitude response feature set. At the same time, the surface normal data corresponding to this calibration attitude is recorded. Thus, each set of calibration attitude response feature sets has corresponding surface attitude data.

[0034] For the current attitude response feature group, it is compared item by item with each calibrated attitude response feature group according to the direction of the detection light source and the arrangement of the feature components. The feature difference value corresponding to the j-th calibrated attitude response feature group is calculated according to the following formula: ; In the formula, Indicates the current attitude response feature group and the first Feature difference values ​​between each calibrated attitude response feature group Indicates the number of light sources detected. This indicates the number of feature components contained in the reflection feature value in each direction. Indicates the current tag is in the [number]th position. The first detection light source under the first detection light source One reflection characteristic component, Indicates the first The calibration posture in the first The first detection light source under the first detection light source One calibrated reflection feature component. When the reflection feature value in the current direction consists of the proportion of response pixels and the normalized average gray value, P is 2.

[0035] After calculating all feature difference values, they are arranged in ascending order of feature difference values, and feature difference values ​​not exceeding a preset difference threshold are selected. The calibrated attitude response feature set is retained as the candidate attitude set. The interpolation weights for each calibrated pose in the candidate pose set are determined according to the following formula: ; In the formula, Indicates the first The interpolation weights corresponding to each candidate calibration pose. This indicates the preset difference threshold. This represents the corresponding feature difference value. Represents the set of candidate poses. This represents a data index within the candidate pose set. This represents a preset non-zero correction value. The smaller the feature difference value, the larger the corresponding interpolation weight, ensuring that the current attitude response features maintain a greater degree of data correlation with the calibration attitude with a higher degree of matching.

[0036] Then, the surface normal data corresponding to the candidate poses are weighted and summed according to each interpolation weight: ; The weighted results are then normalized. ; In the formula, Indicates the first The surface normal vectors corresponding to each candidate calibration pose. This represents the intermediate attitude vector obtained by weighted summation. This represents the normalized local surface normal vector. This represents the magnitude of the intermediate attitude vector. The local surface normal vector serves as the local attitude data for the current label and continues to participate in the calculation of subsequent valid incident state data.

[0037] For example, with a preset difference threshold of 0.20, three candidate poses are selected, with feature difference values ​​of 0.08, 0.12, and 0.18, respectively. The non-zero correction is set to 0.001, resulting in interpolation weights of approximately 0.543, 0.363, and 0.094. By weighting and normalizing the surface normal vectors corresponding to the three candidate poses according to these weights, the local surface normal vector corresponding to the current pose reference region can be obtained. This vector serves as the pose basis for correcting the actual incident state of the detection light source in the next processing stage.

[0038] S4. Convert the light source direction data to the label reference coordinate system, and calculate the corresponding effective incident state data by combining the local attitude data. Furthermore, in S4, the calculation process for the effective incident state data includes: Transform the light source direction data from the preset device coordinate system to the coordinate system corresponding to the label reference coordinate data to obtain the light source direction data in the label coordinate system; Based on the local attitude data, the local surface normal of the holographic detection area is determined. The angle between each light source direction and the local surface normal is calculated. Each light source direction is projected onto the tangent plane corresponding to the local surface normal. The azimuth difference between the projection direction and the preset reference direction is calculated to obtain the incident tilt parameter and the incident azimuth parameter. By combining the incident tilt parameters and incident azimuth parameters corresponding to the same detection light source, the corresponding effective incident state data can be obtained.

[0039] Specifically, after obtaining the light source direction data corresponding to each detection light source and the local attitude data of the label to be detected, the two types of data are first unified to the label reference coordinate system. During the equipment calibration stage, the direction parameters of each detection light source in the preset equipment coordinate system are recorded, and the coordinate transformation relationship between the preset equipment coordinate system and the label reference coordinate system is saved. During detection, the orientation relationship of the current label relative to the preset equipment coordinate system is determined based on the previously obtained label reference coordinate data. The direction vectors of each detection light source are rotated and transformed according to the coordinate transformation relationship to obtain the light source direction vector in the label reference coordinate system. This transformation only changes the coordinate expression of the light source direction vector and does not change the actual spatial illumination direction of the detection light source.

[0040] The aforementioned local attitude data is represented using normalized local surface normal vectors. For the first... For each detection light source, the dot product operation is performed between the light source direction vector in the label's reference coordinate system and the local surface normal vector. The incident tilt parameter is calculated based on the relationship between the resulting dot product and the magnitudes of the two vectors. ; In the formula, Indicates the first The incident tilt parameters corresponding to each detection light source This indicates the first [number] after transformation to the label reference coordinate system. Each light source direction vector This represents the local surface normal vector corresponding to the aforementioned local attitude data. Represents the dot product of two vectors. This represents the magnitude of the light source direction vector. This represents the magnitude of the local surface normal vector.

[0041] To further determine the orientational relationship of the light rays relative to the local surface, the first... The projection vector of a light source direction vector is obtained by projecting it onto the tangent plane corresponding to the local surface normal vector, as follows: ; in, Indicates the first The projection vector of each light source direction onto the local tangent plane. and The meaning is the same as described above. Since the local surface normal vector has been normalized, the component of the light source direction vector in the local surface normal direction is subtracted from the original light source direction vector to obtain the direction component located in the local tangent plane.

[0042] Within the local tangent plane, a first reference direction and a second reference direction perpendicular to the first reference direction are determined based on the label reference coordinate data, and the incident azimuth parameters are calculated based on the components of the projection vector in the two reference directions. ; In the formula, Indicates the first The incident azimuth parameters corresponding to each detection light source This represents the first reference direction vector within the local tangent plane. This represents the second reference direction vector within the local tangent plane. This represents the arctangent operation that determines the orientation of the projection vector based on the two directional components.

[0043] For example, the direction vector of a certain detection light source after coordinate transformation is: The current local surface normal vector is Then, according to the above calculation, the incident tilt parameters of the detection light source relative to the local surface of the current label can be obtained, and the incident azimuth parameters can be determined based on its tangent plane projection. The parameters calculated using the same detection light source... and The data is combined in a fixed order to form the valid incident state data corresponding to the directional response image. Each directional response image generates corresponding valid incident state data in the same way, and maintains its association with the preceding light source direction data and directional response image, for use in subsequent standard holographic response position matching processing.

[0044] S5. Extract the holographic response position data of the holographic detection region in the orientation response image, match the current state standard response position from the pre-stored effective incident state-standard holographic response position corresponding data according to the effective incident state data, and calculate the state position deviation between the two. Furthermore, in S5, the process of extracting holographic response position data includes: Based on the label reference coordinate data, the holographic detection area is located and cropped in the response image in each direction. The pixel grayscale value in the holographic detection area is compared with the preset response threshold, and the response pixels that meet the preset response threshold are selected. The response pixels are divided into connected regions, the number of pixels in each connected region is calculated, and connected regions with a number of pixels less than a preset threshold are removed to obtain the effective response region. Calculate the average position of each response pixel in the effective response region to obtain the holographic response position data corresponding to the response image in each direction.

[0045] Specifically, after obtaining the response images in each direction and the corresponding valid incident state data, the position of the holographic detection region in each direction's response image is determined according to the previously established label reference coordinate data. The boundary of the holographic detection region can be pre-recorded in the label reference coordinate system during the label design stage. During detection, the preset region boundary is mapped to the corresponding direction's response image based on the current label's reference coordinate origin, coordinate axis direction, and scale parameters. Since the response images in each direction are continuously acquired using a fixed camera, the same region boundary mapping relationship is used to extract the holographic detection region from each direction's response image, ensuring that the subsequently obtained position data all use the label reference coordinates as the position reference.

[0046] For the captured holographic detection area, the grayscale values ​​of each pixel within the area are read. When the directional response image is a color image, the corresponding grayscale image can be obtained first according to the device's preset grayscale conversion parameters, and then pixel filtering can be performed. The preset response threshold can be determined during the device calibration stage based on the holographic response grayscale distribution of the standard label under different lighting directions. The grayscale values ​​of each pixel within the holographic detection area are compared one by one with the preset response threshold. Pixels whose grayscale values ​​reach the preset response threshold are retained as response pixels, and the horizontal and vertical coordinates of each response pixel in the label's reference coordinate system are recorded.

[0047] After filtering the response pixels, connected regions are divided based on their spatial adjacency, grouping consecutively located response pixels into the same region. For each region, the number of response pixels it contains is counted and compared to a preset threshold. Connected regions with fewer than the threshold are removed from the current processing results, while those with the threshold are retained. The preset threshold can be determined based on the area of ​​the holographic response region of the standard label and the image resolution, ensuring that small, isolated bright spots are not included in subsequent position calculations. If multiple connected regions satisfying the threshold exist within the holographic detection region, the retained regions are collectively considered as valid response regions.

[0048] For the effective response area, the average horizontal position is obtained by summing the horizontal coordinates of all responding pixels and dividing by the total number of responding pixels. The average vertical position is obtained by processing the vertical coordinates of all responding pixels in the same way. The average horizontal and vertical positions together constitute the holographic response position data corresponding to the response image in the current direction. This position data is represented using the label reference coordinates, so the holographic response position data obtained under different illumination directions can be directly compared.

[0049] For example, a holographic detection region in a response image of a certain direction contains 12,000 pixels. After grayscale thresholding, 1,800 response pixels are obtained. Connected regions are then divided into three regions, corresponding to 1,600, 150, and 50 pixels respectively. When the preset threshold is set to 100 pixels, the first two connected regions are retained, and the third is discarded. Subsequently, the average horizontal and vertical coordinates of the 1,750 response pixels within the retained region are calculated, and these coordinates are used as the holographic response position data for that direction's response image.

[0050] Each directional response image undergoes the same data processing procedure to obtain corresponding holographic response position data, while retaining the association with the preceding valid incident state data. Subsequent processing calls the valid incident state data corresponding to the current directional response image, obtains the current state standard response position from the pre-stored valid incident state and standard holographic response position correspondence data, and then calculates the position difference between this standard position and the holographic response position data obtained in this stage.

[0051] Furthermore, in S5, the process of establishing the data corresponding to the effective incident state and the standard holographic response position includes: Select standard labels whose overprinting state meets the preset standard conditions, set multiple preset surface postures, and illuminate the standard labels with detection light sources in different directions under each preset surface posture, and collect the corresponding calibration direction response images and light source direction data. Calculate the calibrated effective incident state data based on the preset surface posture and the corresponding light source direction data, and extract the standard holographic response position data of the holographic detection area from the corresponding calibration direction response image; The calibrated effective incident state data is associated and stored with the corresponding standard holographic response position data to form the effective incident state-standard holographic response position correspondence data.

[0052] Specifically, before establishing standard response data, labels that meet preset standard conditions after positioning and overprinting quality inspection are selected as standard labels, and these standard labels are installed in the same imaging position as the actual testing station. The label reference coordinates of the standard labels are determined using the same method as those of the labels to be tested, ensuring that the position data obtained in the calibration stage and the actual testing stage use a unified coordinate reference. The preset standard conditions can be set based on the product's allowable lateral and longitudinal overprinting deviations; only labels whose corresponding deviations are all within the allowable range participate in subsequent calibration.

[0053] During calibration, the spatial orientation of the standard label relative to the camera and the detection light source is adjusted so that the standard label is sequentially positioned in multiple known surface orientations. The local surface orientation corresponding to each preset surface orientation is determined before calibration and saved as surface orientation data. For each preset surface orientation, detection light sources in different directions are turned on sequentially according to the illumination sequence used in the actual detection stage. The camera maintains its imaging position and acquires calibration orientation response images corresponding to each detection light source. Each calibration orientation response image retains the light source orientation data corresponding to the current detection light source, thus establishing a correlation between the preset surface orientation, light source orientation data, and calibration orientation response images.

[0054] Subsequently, using the same data processing method as the aforementioned effective incident state data calculation process, the local surface normal corresponding to the current preset surface attitude and the light source direction data of the detection light source are unified to the tag reference coordinate system. The incident tilt parameter and incident azimuth parameter of the light source direction relative to the local surface are calculated, and the two are combined to form the calibrated effective incident state data. Since the local surface direction corresponding to the same detection light source is different under different preset surface attitudes, different calibrated effective incident state data are generated separately, and the detection light source number is not directly used as the index basis for the standard response data.

[0055] For each calibration orientation response image, the holographic detection region is determined based on the label reference coordinates of the standard label. Then, following the aforementioned holographic response position extraction method, response pixels are filtered, connected regions are divided, and connected regions with fewer than a preset threshold of pixels are removed. Finally, the positions of the response pixels within the retained response regions are statistically analyzed to obtain the standard holographic response position data corresponding to that calibration orientation response image. The standard holographic response position data is represented using label reference coordinates and is associated with the calibration effective incident state data corresponding to the current calibration orientation response image.

[0056] For example, five preset surface postures can be set, and four detection light sources can be used to illuminate each surface posture sequentially, thereby obtaining twenty sets of calibrated valid incident state data and corresponding twenty sets of standard holographic response position data. If different calibration states produce the same or repeated valid incident states, the valid incident states can be merged according to the preset data precision, and the corresponding standard holographic response position data can be retained.

[0057] After calibrating all preset surface poses and detection light source combinations, each set of calibrated effective incident state data is used as a retrieval condition, and the corresponding standard holographic response position data is stored as associated data, forming a correspondence between the effective incident state and the standard holographic response position. During actual detection, the current effective incident state data obtained in S4 can be used as the basis for querying, compared with the calibrated effective incident state in the corresponding data, and the standard response position of the current state is subsequently determined accordingly.

[0058] Furthermore, in S5, the matching process for the current state standard response position includes: Calculate the state difference value between the current effective incident state data and the corresponding data of the effective incident state-standard holographic response position, and sort them in ascending order of state difference value; Select calibrated valid incident state data whose state difference value is not greater than the preset state difference threshold, determine the corresponding interpolation weight according to the state difference value, and perform weighted summation on the corresponding standard holographic response position data according to the interpolation weight to obtain the current state standard response position; The horizontal and vertical position differences between the holographic response position data and the corresponding current state standard response position are calculated to form the state position deviations corresponding to the response images in each direction.

[0059] Specifically, after establishing the corresponding data of the effective incident states and standard holographic response positions, standard response position matching is performed on each direction response image of the current label to be detected. The current direction response image has already been associated with the corresponding effective incident state data in the preprocessing, which includes the incident tilt parameter and the incident azimuth parameter. During processing, each calibrated effective incident state is sequentially read from the pre-stored corresponding data, and the state difference value between the current effective incident state and each calibrated effective incident state is calculated. To ensure that both the incident tilt direction and the incident azimuth direction participate in the matching, the state difference value is calculated according to the following formula: ; In the formula, Indicates the first The current effective incident state and the first The state difference value between each calibrated valid incident state Indicates the first The incident tilt parameters for the current valid incident state. Indicates the first The incident azimuth parameters of the current valid incident state. Indicates the first The incident tilt parameters for calibrating the effective incident state. Indicates the first The incident azimuth parameters for calibrating the effective incident state. and These represent the preset difference weights corresponding to the incident tilt parameter and the incident azimuth parameter, respectively, and the sum of the two preset difference weights is 1.

[0060] After calculating all state difference values, sort them from smallest to largest, and select those that are not greater than a preset state difference threshold. The calibrated valid incident state is preserved, and the corresponding records are used to form the first... A candidate standard set of current valid incident states The interpolation weights for each calibration record in the candidate standard set are determined based on the state difference value: ; In the formula, Indicates the first The first one under the current effective incident state The interpolation weights of the candidate standard records, This indicates a preset threshold for state differences. This represents the corresponding state difference value. Represents the set of candidate criteria. This represents the record index in the candidate criteria set. This indicates a preset non-zero correction amount. The sum of the normalized interpolation weights is 1, and candidate records with smaller state difference values ​​receive larger interpolation weights.

[0061] The standard holographic response position data associated with each record in the candidate standard set is retrieved and weighted summed according to the corresponding interpolation weights: ; in, Indicates the first The current state standard response position corresponding to the response image in each direction; Indicates the first The candidate standard record is associated with the standard holographic response location data; This indicates the corresponding interpolation weight. The current state standard response position obtained in this way is not directly adopted from the standard position corresponding to a fixed detection light source, but is determined by multiple calibration data corresponding to the actual effective incident state of the current tag.

[0062] For example, with a preset state difference threshold of 0.15, three candidate standard records are obtained from filtering a certain current valid incident state, with corresponding state difference values ​​of 0.06, 0.10, and 0.14, respectively. When the non-zero correction is 0.001, the interpolation weights of the three records are approximately 0.595, 0.333, and 0.072, respectively. If the standard holographic response positions corresponding to the three records are 12.30 mm, 12.34 mm, and 12.39 mm in the horizontal coordinates and 8.40 mm, 8.38 mm, and 8.35 mm in the vertical coordinates, then the weighted standard response position of the current state is approximately 12.320 mm in the horizontal coordinates and 8.390 mm in the vertical coordinates.

[0063] The coordinate difference between the holographic response position data extracted from the current direction response image and the current state standard response position is then calculated: ; In the formula, Indicates the first The state position deviation corresponding to the response image in each direction Indicates the first Holographic response location data in each direction response image This indicates the corresponding current state standard response position. Indicates the difference in lateral position. This represents the longitudinal position difference. The state position deviations corresponding to each effective incident state are obtained sequentially using the same processing method, and the obtained state position deviations are saved according to the correlation of the direction response images for use in the subsequent extraction process of the actual overlay offset data.

[0064] Please see the appendix Figure 2 S6. Sort the horizontal and vertical deviations of multiple state position deviations, remove deviations that exceed the preset discrete threshold, and then calculate the average to obtain the true overlay offset data; subtract the true overlay offset data from each state position deviation to obtain the state residual deviation, and compare the true overlay offset data and the state residual deviation with the corresponding thresholds to generate overlay state markers and holographic response state markers.

[0065] Furthermore, in S6, the process of obtaining the actual overlay offset data includes: The lateral and longitudinal deviation values ​​are extracted from multiple state position deviations, sorted, and the median value of the corresponding sorted sequence is determined as the lateral and longitudinal reference deviation values. Calculate the absolute value of the difference between each lateral deviation value and the lateral reference deviation value, and the absolute value of the difference between each longitudinal deviation value and the longitudinal reference deviation value, and remove the corresponding deviation values ​​whose absolute value of the difference exceeds the preset discrete threshold. Calculate the average of the retained horizontal and vertical deviation values ​​to obtain the horizontal true overprint offset value and the vertical true overprint offset value, and combine the horizontal true overprint offset value and the vertical true overprint offset value to form the true overprint offset data.

[0066] Specifically, after obtaining the state position deviations corresponding to the response images in each direction, the state position deviations of the same label to be detected are sorted according to the horizontal and vertical components, forming a horizontal deviation sequence and a vertical deviation sequence. Since different effective incident states may have situations such as excessive local reflection, incomplete response areas, or deviations in feature positions, some state position deviations may significantly deviate from most detection results. Therefore, the median value after sorting is first used to establish a deviation screening benchmark, and then the true overlay offset is calculated on the retained data.

[0067] For those involved in the calculation For each state position deviation, sort the lateral and longitudinal deviation values ​​from smallest to largest, and determine the lateral and longitudinal reference deviation values ​​according to the following formula: ; ; In the formula, Indicates the lateral reference deviation value. Indicates the longitudinal reference deviation value. Indicates the first The lateral deviation value of each state position deviation. Indicates the first The longitudinal deviation value of each state position deviation. This indicates the number of state position deviations for the currently detected label in the calculation, and `median` represents the median value of the sorted sequence. When the number of state position deviations is even, the average of the two deviation values ​​at the middle position after sorting is taken as the corresponding baseline deviation value.

[0068] After obtaining the two reference deviation values, calculate the dispersion of each lateral deviation value and longitudinal deviation value relative to the corresponding reference deviation value: ; ; In the formula, This represents the lateral discrete value corresponding to the k-th lateral deviation value. Indicates the first Each longitudinal deviation value corresponds to a longitudinal discrete value. The lateral discrete value is compared with a preset lateral discrete threshold; lateral deviation values ​​exceeding the preset lateral discrete threshold are removed. Similarly, the longitudinal discrete value is compared with a preset longitudinal discrete threshold; longitudinal deviation values ​​exceeding the preset longitudinal discrete threshold are removed. After filtering, lateral and longitudinal retention sets are formed respectively.

[0069] Calculate the average for both the horizontal and vertical retention sets: ; ; In the formula, This represents the horizontal actual overprint offset value. This represents the true vertical overprinting offset value. This represents the set of state indices retained after filtering by horizontal discrete conditions. This represents the set of state indices retained after filtering based on vertical discrete conditions. and These represent the number of states retained in the corresponding set.

[0070] Taking five state position deviations as an example, the lateral deviation values ​​are 0.10 mm, 0.11 mm, 0.09 mm, 0.12 mm, and 0.31 mm, respectively. After sorting, the lateral baseline deviation value is 0.11 mm. When the preset lateral dispersion threshold is 0.05 mm, the lateral dispersion value corresponding to 0.31 mm is 0.20 mm, so this lateral deviation value is discarded. The average value of the remaining four lateral deviation values ​​is 0.105 mm, which is used as the lateral true overprint offset value. The longitudinal deviation value is processed in the same way. Finally, the lateral true overprint offset value and the longitudinal true overprint offset value constitute the true overprint offset data of the label to be detected, and the difference is calculated with each state position deviation to generate the subsequent state residual deviation.

[0071] Furthermore, in S6, the generation process of the overlay state marker and the holographic response state marker includes: Subtract the corresponding lateral and longitudinal true overprint offset values ​​from the lateral and longitudinal position deviation values ​​of each state position deviation to obtain the state residual deviation corresponding to each state position deviation. The absolute values ​​of the horizontal and vertical true overprint offsets are compared with the corresponding overprint offset thresholds. When neither of them exceeds the corresponding overprint offset threshold, a normal overprint mark is generated. When at least one of them exceeds the corresponding overprint offset threshold, an abnormal overprint mark is generated. The absolute values ​​of the horizontal and vertical components of the residual deviation of each state are compared with the corresponding residual deviation thresholds, and the number of states in which at least one component exceeds the corresponding residual deviation threshold is counted. When the number of states reaches the preset threshold, a holographic response anomaly marker is generated, and when the number of states does not reach the preset threshold, a holographic response normal marker is generated.

[0072] Specifically, after obtaining the lateral and vertical true overprint offset values ​​of the label to be detected, the state position deviations corresponding to the aforementioned response images in each direction are then called. For each valid illumination state participating in the detection, the lateral true overprint offset value is subtracted from the lateral deviation value of that state, and the vertical true overprint offset value is subtracted from the vertical deviation value, to obtain the lateral and vertical residuals after deducting the common geometric displacement under that illumination state. The state residual deviation is calculated according to the following formula: ; ; In the formula, Indicates the first The lateral state residual deviation corresponding to each effective lighting state Indicates the first Longitudinal state residual deviation corresponding to each effective lighting state Indicates the first The lateral deviation value in each state position deviation. Indicates the first The longitudinal deviation value in each state position deviation This represents the horizontal actual overprint offset value. This represents the true longitudinal overprint offset value. The lateral state residual deviation and the longitudinal state residual deviation corresponding to each effective lighting state together constitute the state residual deviation of that lighting state.

[0073] To determine the overprinting status, the absolute values ​​of the horizontal and vertical true overprinting offsets are obtained and compared with pre-set horizontal and vertical overprinting offset thresholds. A normal overprinting mark is generated when the absolute value of both the horizontal and vertical true overprinting offsets does not exceed the horizontal and vertical overprinting offset thresholds. An abnormal overprinting mark is generated when the absolute value of any true overprinting offset exceeds the corresponding overprinting offset threshold. The overprinting offset thresholds can be pre-stored based on the allowable overprinting tolerances for the current specification label during production and retrieved when inspecting the corresponding specification label.

[0074] For determining the holographic response state, the absolute value of the lateral residual deviation of each effective illumination state is compared with the lateral residual deviation threshold, and the absolute value of the longitudinal residual deviation is compared with the longitudinal residual deviation threshold. For the [specific case / state]... Each valid lighting state sets a residual anomaly flag value: ; Further statistics on the number of residual abnormal states: ; In the formula, Indicates the first The residual anomaly flag value corresponds to each valid lighting state. A value of 1 indicates that the state has a component that exceeds the residual deviation threshold, while a value of 0 indicates that neither of the two residual deviation components exceeds the corresponding threshold. This represents the threshold for lateral residual bias. Indicates the longitudinal residual deviation threshold. Indicates the number of residual abnormal states. This indicates the number of valid lighting states for the tag to be detected to participate in the residual deviation judgment. The serial number indicating the effective lighting status.

[0075] The number of residual abnormal states is compared with a preset threshold. When the number of residual abnormal states reaches the preset threshold, a holographic response abnormality marker is generated; when the number of residual abnormal states does not reach the preset threshold, a holographic response normal marker is generated. For example, if a tag has five valid illumination states and the preset threshold is set to 2, and at least one residual deviation component in three illumination states exceeds the corresponding residual deviation threshold, then the number of residual abnormal states is 3, and a holographic response abnormality marker is generated. The overprinting state marker and the holographic response state marker are retained separately and used together as the positioning overprinting detection result of the current tag to be detected.

[0076] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent detection of the positioning and overprinting quality of holographic anti-counterfeiting labels, characterized in that, include: S1. Acquire the reference image of the label to be detected, extract non-holographic positioning features, and establish the label reference coordinate data; S2. Control the detection light sources from different directions to illuminate the tag in sequence, collect the directional response image and record the light source direction data; S3. Locate the attitude reference area according to the label reference coordinate data, extract the reflection features in the orientation response image to form an attitude response feature group, compare the attitude response feature group with the pre-stored attitude response feature-surface attitude correspondence data and interpolate to obtain local attitude data. S4. Convert the light source direction data to the label reference coordinate system, and calculate the corresponding effective incident state data by combining the local attitude data. S5. Extract the holographic response position data of the holographic detection region in the orientation response image, match the current state standard response position from the pre-stored effective incident state-standard holographic response position corresponding data according to the effective incident state data, and calculate the state position deviation between the two. S6. Sort the horizontal and vertical deviations of multiple state position deviations, remove deviations that exceed the preset discrete thresholds, and then calculate the average to obtain the true overprint offset data; subtract the true overprint offset data from each state position deviation to obtain the state residual deviation, and compare the true overprint offset data and the state residual deviation with the corresponding thresholds to generate overprint state markers and holographic response state markers.

2. The intelligent detection method for the positioning and overprinting quality of a holographic anti-counterfeiting label according to claim 1, characterized in that, In step S2, the process of acquiring the orientation response image and recording the light source orientation data includes: The detection light sources in different directions are turned on in sequence according to the preset illumination order, and the other detection light sources are turned off while the current detection light source is turned on, while keeping the imaging position of the camera unchanged. During the period when the label is illuminated by the current detection light source, the corresponding directional response image is acquired. After the acquisition is completed, the current detection light source is turned off and the next detection light source is turned on, until the directional response image acquisition for each detection light source is completed. The illumination direction of each detection light source in the preset device coordinate system is obtained, the corresponding light source direction data is generated, and the direction response images are associated and stored with the corresponding light source direction data according to the acquisition order of the direction response images.

3. The intelligent detection method for the positioning and overprinting quality of a holographic anti-counterfeiting label according to claim 1, characterized in that, In step S3, the process of obtaining the attitude response feature set includes: Based on the label reference coordinate data, the attitude reference region is located in the response image in each direction, and the corresponding attitude reference region image is cropped. The pixel grayscale values ​​in each pose reference region image are compared with a preset grayscale threshold, and response pixels whose grayscale values ​​reach the preset grayscale threshold are selected. The number of response pixels is counted and the average grayscale value of the response pixels is calculated. The proportion of the number of response pixels to the total number of pixels in the pose reference region is calculated respectively, and the average gray value is normalized according to a preset gray range. The obtained proportion is combined with the normalized average gray value to obtain the directional reflection feature value corresponding to each directional response image. The directional reflection feature values ​​are arranged according to the light source direction data corresponding to each directional response image to form the attitude response feature group.

4. The intelligent detection method for the positioning and overprinting quality of a holographic anti-counterfeiting label according to claim 3, characterized in that, In step S3, the process of determining the local pose data includes: Retrieve pre-stored attitude response features—surface attitude corresponding data, which includes multiple sets of calibrated attitude response feature sets and corresponding surface attitude data; Calculate the absolute value of the difference between the attitude response feature group and the corresponding direction reflection feature value in each calibrated attitude response feature group. Accumulate the absolute values ​​of each difference to obtain the feature difference value. Sort the feature difference values ​​from smallest to largest and filter the calibrated attitude response feature groups whose feature difference values ​​are not greater than a preset difference threshold to form a candidate attitude set. The difference between the preset difference threshold and the difference values ​​of each feature in the candidate pose set is calculated. The obtained difference is normalized to obtain the interpolation weight. The corresponding surface pose data is weighted and summed according to the interpolation weight to obtain the local pose data.

5. The intelligent detection method for the positioning and overprinting quality of a holographic anti-counterfeiting label according to claim 4, characterized in that, In step S4, the calculation process for the effective incident state data includes: The light source direction data is transformed from the preset device coordinate system to the coordinate system corresponding to the label reference coordinate data to obtain the light source direction data in the label coordinate system. Based on the local attitude data, the local surface normal of the holographic detection area is determined. The angle between each light source direction and the local surface normal is calculated. Each light source direction is projected onto the tangent plane corresponding to the local surface normal. The azimuth difference between the projection direction and the preset reference direction is calculated to obtain the incident tilt parameter and the incident azimuth parameter. By combining the incident tilt parameter and incident azimuth parameter corresponding to the same detection light source, the corresponding effective incident state data can be obtained.

6. The intelligent detection method for the positioning and overprinting quality of a holographic anti-counterfeiting label according to claim 1, characterized in that, In step S5, the process of extracting holographic response position data includes: Based on the label reference coordinate data, the holographic detection area is located and cropped in the response image in each direction. The pixel grayscale value in the holographic detection area is compared with the preset response threshold, and response pixels that meet the preset response threshold are selected. The response pixels are divided into connected regions, the number of pixels in each connected region is calculated, and connected regions with a number of pixels less than a preset threshold are removed to obtain the effective response region. The average position of each response pixel in the effective response region is calculated to obtain the holographic response position data corresponding to the response image in each direction.

7. The intelligent detection method for the positioning and overprinting quality of a holographic anti-counterfeiting label according to claim 1, characterized in that, In step S5, the process of establishing the data corresponding to the effective incident state and the standard holographic response position includes: Select standard labels whose overprinting state meets the preset standard conditions, set multiple preset surface postures, and illuminate the standard labels with detection light sources in different directions under each preset surface posture, and collect the corresponding calibration direction response images and light source direction data. Calculate the calibrated effective incident state data based on the preset surface posture and the corresponding light source direction data, and extract the standard holographic response position data of the holographic detection area from the corresponding calibration direction response image; The calibrated effective incident state data is associated and stored with the corresponding standard holographic response position data to form the effective incident state-standard holographic response position corresponding data.

8. The intelligent detection method for the positioning and overprinting quality of a holographic anti-counterfeiting label according to claim 7, characterized in that, In step S5, the matching process for the current state standard response position includes: Calculate the state difference value between the current effective incident state data and each calibrated effective incident state data in the effective incident state-standard holographic response position corresponding data, and sort them in ascending order of the state difference value; Valid incident state data with state difference values ​​not greater than a preset state difference threshold are selected. The corresponding interpolation weight is determined according to the state difference value. The corresponding standard holographic response position data is weighted and summed according to the interpolation weight to obtain the current state standard response position. The horizontal and vertical position differences between the holographic response position data and the corresponding current state standard response position are calculated respectively to form the state position deviations corresponding to the response images in each direction.

9. The intelligent detection method for the positioning and overprinting quality of a holographic anti-counterfeiting label according to claim 8, characterized in that, In step S6, the process of obtaining the actual overprint offset data includes: The lateral deviation value and the longitudinal deviation value are extracted from multiple state position deviations respectively. The lateral deviation value and the longitudinal deviation value are sorted respectively, and the median value of the corresponding sorted sequence is determined as the lateral reference deviation value and the longitudinal reference deviation value. Calculate the absolute value of the difference between each lateral deviation value and the lateral reference deviation value, and the absolute value of the difference between each longitudinal deviation value and the longitudinal reference deviation value, and remove the corresponding deviation values ​​whose absolute value of the difference exceeds a preset discrete threshold. The average values ​​of the retained lateral and longitudinal deviations are calculated separately to obtain the lateral true overprint offset value and the longitudinal true overprint offset value, and the lateral true overprint offset value and the longitudinal true overprint offset value are combined to form the true overprint offset data.

10. The intelligent detection method for the positioning and overprinting quality of a holographic anti-counterfeiting label according to claim 9, characterized in that, In step S6, the generation process of the overlay state marker and the holographic response state marker includes: Subtract the corresponding lateral and longitudinal true overprint offset values ​​from the lateral and longitudinal position deviation values ​​of each state position deviation to obtain the state residual deviation corresponding to each state position deviation. The absolute values ​​of the horizontal true overprint offset value and the vertical true overprint offset value are compared with the corresponding overprint offset thresholds. When both do not exceed the corresponding overprint offset thresholds, a normal overprint mark is generated. When at least one exceeds the corresponding overprint offset threshold, an abnormal overprint mark is generated. The absolute values ​​of the horizontal and vertical components of the residual deviation of each state are compared with the corresponding residual deviation thresholds, and the number of states in which at least one component exceeds the corresponding residual deviation threshold is counted. When the number of states reaches a preset threshold, a holographic response anomaly marker is generated, and when the number of states does not reach the preset threshold, a holographic response normal marker is generated.

Images