Packaging bag printing quality detection method and detection system based on image transmission

Abstract

The application discloses a packaging bag printing quality detection method and system based on image transmission, belongs to the technical field of fault diagnosis, collects multi-source operation data such as vibration, temperature, electromagnetism and acoustics of a permanent magnet electric motor roller, and generates a multi-dimensional characteristic tensor through time-frequency calibration fusion; intrinsic modal decomposition is performed on the multi-dimensional characteristic tensor, and combined with physical constraints such as mechanical vibration, heat conduction and electromagnetic induction, a mixed characteristic is decoupled into independent physical process representation, and a characteristic primitive set with physical meaning is reconstructed; then, a dynamic evolution model is established based on the characteristic primitive, an operation state trajectory is predicted, and an abnormal state is identified; finally, the abnormal state is back-propagated and traced, a fault source area is located, and a diagnosis decision is generated; the application can improve fault recognition accuracy, diagnosis interpretability and fault source positioning reliability.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology, and specifically to a method and system for detecting the printing quality of packaging bags based on image transmission. Background Technology

[0002] In the online printing inspection scenario of pre-cooked food packaging bags undergoing high-temperature steaming, the packaging bag surface simultaneously exhibits matte ink, localized glossy anti-counterfeiting patterns, and light-colored batch numbers on the heat-sealed edge. Cameras typically transmit images to the inspection server via the workshop's wireless network. Due to the combined effects of steam from the sealing machine, metal frame obstruction, and high-speed traction vibration, transmitted images are prone to localized line band misalignment, brief ghosting, or frame drops. Existing inspection methods often directly compare the received image with a standard image, easily misjudging transmission-induced breaks as misprints, and potentially mistaking missing strokes in the genuine heat-sealed batch number as transmission noise. This results in a small number of incorrectly printed bags entering the sterilization process, subsequent invalidation of the entire box's traceability code, high rework costs, and difficulties in reproducing the defect. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for detecting the printing quality of packaging bags based on image transmission, so as to overcome the shortcomings of the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for detecting the printing quality of packaging bags based on image transmission, comprising: Acquire the received image of the packaging bag to be inspected at the printing inspection station and sent through the image transmission link, as well as the standard printed image corresponding to the packaging bag of the same version; Based on the arrival order of the received image segments, the edge continuity between adjacent image bands, and the printed positioning marks, a transmission disturbance region corresponding to the received image is generated. Based on the transmission disturbance area, the received image is divided into a reliable transmission area and a transmission area to be corrected. The overprint boundary, text baseline and color block outline in the reliable transmission area are used to reconnect the position of the transmission area to be corrected, so as to obtain the printed image to be compared. Based on the correspondence between the printed image to be compared and the standard printed image, the printing difference features of batch number characters, traceability code, heat sealing edge fine line and anti-counterfeiting glossy oil texture are extracted. Differences in printing difference features that are located in the transmission disturbance region and disappear after position reconnection are identified as transmission pseudo-defects, while differences that are located in the transmission reliable region or remain after position reconnection are identified as real printing defects. Based on the location, continuity, and correlation of actual printing defects with traceability codes and batch number characters, the packaging bag printing quality inspection results are output.

[0005] Preferably, acquiring the received image and the standard printed image includes the following steps: When the printed positioning mark of the packaging bag to be inspected enters the field of view, a pre-reading image containing the heat-sealed blank strip and the printed positioning mark is acquired, and the orientation of the plate is determined according to the relative position of the heat-sealed blank strip and the printed positioning mark. The effective detection boundary is determined based on the area that recedes from the outer boundary of the printed area in the pre-read image towards the blank area of ​​the heat-sealing edge. The image within the effective detection boundary is divided into multiple image bands according to the traction direction of the packaging bag, and the acquisition sequence identifier that does not participate in the comparison of printing defects is written in the blank band of the heat-sealed edge corresponding to each image band. Based on the orientation of the page and the effective detection boundary, retrieve and crop the standard printed image corresponding to the same packaging bag, so that the received image and the standard printed image have the same comparison range.

[0006] Preferably, the formation of the received image includes the following steps: Each image band is sent to the detection server via the image transmission link; The image bands are stitched together on the detection server according to their actual arrival order, instead of immediately restoring the original order according to the acquisition sequence identifier. Read the acquisition sequence identifier of each image band in the stitched received image located within the heat-sealed blank band, and save the original acquisition number expressed by the acquisition sequence identifier and the actual arrangement position of the image band in the received image.

[0007] Preferably, generating the transmission disturbance region includes the following steps: Check the acquisition sequence identifier against the actual arrangement position for each band, and identify the intersection of adjacent image bands where the acquisition sequence identifier is skipped, reversed or repeated as an abnormal sequence intersection. At the junctions of sequential anomalies and their adjacent junctions, the grayscale texture direction of the color block outline endpoints, text baseline endpoints, and heat-sealed blank bands is extracted respectively. When there are color block outline endpoints that can be connected on both sides of the same boundary but the grayscale texture direction is broken, or the grayscale texture direction is continuous but the text baseline is misaligned, the boundary is determined as a pseudo-continuous boundary. The sequential anomaly boundary and the pseudo-continuous boundary are used as the initial boundary locations of the transmission disturbance region.

[0008] Preferably, the expansion of the transmission disturbance region includes the following steps: Select printed positioning marks that do not pass through the intersection of sequence abnormalities as anchoring objects, and form a line connecting the centers of the printed positioning marks along the traction direction of the packaging bag; Extend the line connecting the centers of the printed positioning marks to the pseudo-continuum boundary, and compare the extended position with the actual center position of the positioning mark edge at the pseudo-continuum boundary. When the comparison results indicate that there is a traction offset at the pseudo-continuous boundary, the edge of the adjacent image band corresponding to the pseudo-continuous boundary is incorporated into the transmission disturbance region. The edge of the incorporated image band is then extended laterally to the nearest heat-sealed blank band to form the transmission disturbance region corresponding to the received image.

[0009] Preferably, dividing the transmission trusted region and the transmission to be corrected region includes the following steps: Tracing adjacent image bands back and forth along the outer edge of the transmission disturbance area towards the direction of the packaging bag's traction; Image content with continuous acquisition sequence identification, no traction offset of the center line of printed positioning mark, and continuous grayscale texture direction of heat-sealed blank band is identified as a reliable transmission area. Image content that is connected to the transmission disturbance area and does not simultaneously meet the requirements of continuous acquisition sequence identification, no traction offset of the center line, and continuous grayscale texture direction is identified as the transmission correction area.

[0010] Preferably, the location reconnection for the transmission area to be corrected includes the following steps: In the trusted transmission region, the overprint boundary, text baseline and color block outline that span the front and back sides of the transmission area to be corrected are selected as the reconnection reference. A reference chain is formed based on the front-to-back attachment relationship of the overprint boundary, text baseline and color block outline in the standard printed image, and the reference chain is extended to the area to be corrected during transmission. Extract the boundary ends of the same color, the baseline ends of the same character height, and the turning ends of the same contour that correspond to the reference chain within the transmission area to be corrected; Following the order of first aligning with the text baseline, then with the overprint boundary, and finally with the color block outline, the area to be corrected is segmented and locally rotated to obtain the printed image to be compared.

[0011] Preferably, extracting printing difference features includes the following steps: Based on the overprint boundary, text baseline and color block outline after the reference chain is closed, the printed image to be compared and the standard printed image are divided into corresponding local comparison segments. The image content that still cannot be closed with the reference chain after reconnection is attached to the end of the adjacent local comparison segment. The starting points for reading differences are the end of the strokes of the batch number characters, the corner points of the traceability code, the continuous edges of the heat-sealing line, and the reflective direction of the anti-counterfeiting glossy oil. Read the disconnection, offset, overlap and missing states along the extension direction of each printed object, and merge the abnormal states that are connected in the same extension direction into printing difference segments to form printing difference features.

[0012] Preferably, determining the transmission of false defects and the actual printing defects and outputting the detection results includes the following steps: The original position of each printing difference segment in the received image is traced back, and the original position is correlated with the transmission disturbance area, the transmission reliable area, and the attachment direction of the non-closable content. Read the segmented translation direction and local rotation direction of the area to be corrected during the position reconnection process, and mark the printing difference segment whose abnormal state direction is opposite to the segmented translation direction or local rotation direction as a difference that can be canceled by reconnection. If the difference that can be offset by reconnection is read again along the extension direction of the printed object itself, and if it restores a continuous correspondence with the standard printed image, it is determined to be a transmission pseudo-defect. Differences located in the reliable transmission area, or differences located in the transmission disturbance area but which still do not correspond continuously with the standard printed image after position reconnection, are determined to be real printing defects.

[0013] The present invention also provides a packaging bag printing quality inspection system based on image transmission, comprising: The image acquisition module acquires the received image of the packaging bag to be inspected after it is captured at the printing inspection station and sent through the image transmission link, as well as the standard printed image corresponding to the same packaging bag. The transmission disturbance region generation module generates the transmission disturbance region corresponding to the received image based on the arrival order of the received image segments, the edge continuity between adjacent image bands, and the printed positioning marks. The region segmentation module divides the received image into a reliable transmission region and a transmission correction region based on the transmission disturbance region. It then uses the overprinting boundary, text baseline, and color block outline in the reliable transmission region to perform position reconnection on the transmission correction region, thus obtaining the printed image to be compared. The printing difference feature extraction module extracts printing difference features such as batch number characters, traceability code, heat sealing edge fine lines, and anti-counterfeiting glossy oil texture based on the correspondence between the printed image to be compared and the standard printed image. The defect type determination module identifies differences in printing difference features that are located in the transmission disturbance area and disappear after repositioning as transmission pseudo-defects, and differences that are located in the transmission reliable area or remain after repositioning as real printing defects. The quality inspection result output module outputs the packaging bag printing quality inspection results based on the location, continuity, and correlation of the actual printing defects with the traceability code and batch number characters.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention differs from existing technologies that perform mode decomposition on only a single raw signal or spectral signal. Instead, it first performs time synchronization, time-frequency calibration, and fusion of heterogeneous data from multiple sources, such as vibration, temperature, electromagnetic, and acoustic data, to construct a multidimensional feature tensor containing time, frequency, and multi-source channel dimensions. This approach solves the problem that single signals are susceptible to noise and operating condition fluctuations and fail to reflect the overall state of the equipment. It allows anomaly information from different physical sources to corroborate each other within a unified time-frequency framework, thereby improving the completeness and reliability of the operational status representation.

[0015] This invention does not directly use the modal decomposition results as physical states. Instead, it applies physical constraints such as mechanical vibration, heat conduction, and electromagnetic induction to the candidate components obtained from multidimensional feature tensor decomposition, and maps and reconstructs them into independent physical process representations and feature primitive sets based on cross-channel correlation and physical laws. This invention solves the problems of unclear physical meaning of existing spectral components and difficulty in locating fault causes, enabling diagnostic features to correspond to specific operating mechanisms, directly improving the interpretability of fault identification, the accuracy of source tracing, and the stability of anomaly prediction. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a flowchart of the method of the present invention.

[0018] Figure 2 This is a flowchart of the system modules of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, 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] Example 1, please refer to Figure 1 As shown in this embodiment, the packaging bag printing quality inspection method based on image transmission includes: Acquire the received image of the packaging bag to be inspected at the printing inspection station and sent through the image transmission link, as well as the standard printed image corresponding to the packaging bag of the same version; This embodiment relates to the process of acquiring received images and standard printed images, which is used to ensure that the subsequent transmission disturbance area has a traceable data source under the conditions of high-speed traction of packaging bags, transmission of images via transmission links and out-of-order arrival.

[0021] In this embodiment, the packaging bag to be inspected is a pre-cooked food high-temperature cooking packaging bag. The surface of the packaging bag includes a heat-sealed blank strip, printed positioning marks, batch number characters, traceability code, and patterned color blocks. A line scan camera or area scan camera is set up at the printing inspection station along the traction direction of the packaging bag to capture a field of view that covers the entire width of the packaging bag laterally and the length of at least one printed positioning mark in the traction direction.

[0022] When the printed positioning mark enters the acquisition field of view, a pre-reading acquisition is performed first. The pre-reading image is only used to determine the orientation of the printing plate and the effective detection boundary, and is not used for printing defect comparison. The pre-reading image includes the heat-sealing blank band and the positioning mark. The detection server first reads the grayscale changes column by column along the horizontal direction in the pre-reading image, and identifies the area where continuous ink boundaries appear and which is adjacent to the heat-sealing blank band as the printing area. The heat-sealing blank band is identified based on its low texture characteristics, that is, within a width of 20 consecutive pixels, the difference between the maximum and minimum grayscale values ​​does not exceed 8 grayscale levels, and this area is connected to the image edge.

[0023] The positioning markers are identified based on their preset shapes. The positioning markers are either black rectangles or combinations of dots. The detection server reads the position of the positioning markers relative to the heat-sealed edge blank band. When the positioning marker is to the left of the heat-sealed edge blank band, the page orientation is determined to be positive; when the positioning marker is to the right of the heat-sealed edge blank band, the page orientation is determined to be negative. The effective detection boundary is obtained by backing back 6 pixels from the outer rectangle of the printed area towards the heat-sealed edge blank band. This backed-back area is used to avoid the space occupied by subsequent acquisition sequence markers, ensuring that writing markers does not change the printing defect comparison range.

[0024] After confirming the orientation of the image and the effective detection boundary, the image within the effective detection boundary is split according to the traction direction of the packaging bag. During splitting, the traction direction is taken as the vertical axis, and the horizontal axis of the packaging bag as the horizontal axis. Starting from the vertical starting point of the effective detection boundary, one image band is extracted every 96 rows of pixels. If less than 96 rows of pixels remain, the remaining rows are used as the last image band. Each image band retains its full horizontal pixel width, ensuring that the heat-sealing blank band, the positioning mark extension area, and the printed content are all within the same band. The image bands are numbered sequentially according to the acquisition order as Band 1, Band 2, and so on until Band 1. This number is not directly written into the printed pattern area but is written into the corresponding heat-sealing blank band for each image band. The acquisition sequence identifier is formed by a combination of four light gray short lines and one check short line. The gray value of the light gray short lines is set to the average gray value of the heat-sealing blank band minus 12 gray levels. The short line length is 12 pixels, the width is 2 pixels, and the interval between adjacent short lines is 4 pixels. Four light gray short lines represent the lower four bits of the image band number in binary format. The check line is used to indicate the parity result of the sum of the bits of the image band number. Because the identifier is located in the back-off area outside the effective detection boundary, this position is not read when comparing batch number characters, traceability codes, heat seal edge lines, and anti-counterfeiting glossy patterns. Therefore, the acquisition sequence identifier is not included in the comparison of printing defects.

[0025] It should be noted that when image strips are sent via the image transmission link, each image strip carries image data, a strip number corresponding to the acquisition sequence identifier, and the acquisition completion time. After receiving the image strips, the detection server does not immediately restore the original order according to the strip number. Instead, it first splices them according to their actual arrival order to form the received image. The actual arrival order is based on the order in which the detection server completes reception. The first complete image strip that arrives is placed at the vertical starting point of the received image, the second complete image strip is placed next to it, and so on until the last complete image strip that arrives. After splicing, the detection server reads the acquisition sequence identifier within the heat-sealed blank strip of each image strip and compares the identified strip number with the current position number in the received image.

[0026] For example, if an image band is located at position 5 in the received image, but its original acquisition number is indicated by the acquisition sequence identifier as band 7, then this image band is recorded as having an arrival advance difference. If the current position number matches the original acquisition number, it is recorded as having no arrival difference. This corresponding difference, along with the continuity of the edges of adjacent image bands, is used by subsequent transmission disturbance areas to distinguish between local misconnections caused by transmission and actual printing defects.

[0027] The acquisition of standard printed images is carried out simultaneously based on the orientation of the printing plate and the effective detection boundary. The detection server retrieves a standard printed image that matches the current printing batch and orientation of the packaging bag from the standard image library of the same type of packaging bag.

[0028] Printing batches can be obtained from the plate number block near the printing positioning mark. The plate number block contains 3 digits, and the same plate information is determined by matching each digit using a character template. After retrieving the standard printed image, the same boundary is cropped according to the valid detection boundary obtained from the pre-read image; if the current plate orientation is reversed, the standard printed image is rotated 180 degrees first, and then the boundary cropping is performed.

[0029] The cropped standard printed image and the received image are consistent in terms of horizontal pixel count, vertical splicing range, and effective detection position. The area occupied by the heat-sealed blank band in the received image is marked as a non-comparison area.

[0030] It should be noted that the received image retains the original splicing state caused by the transmission arrival, and also carries the corresponding difference between the acquisition sequence identifier and the actual arrival order; the standard printed image has the same comparison range, providing a data basis for subsequent printing quality judgment based on the transmission disturbance area, the transmission reliable area, and the transmission correction area.

[0031] Based on the arrival order of the received image segments, the edge continuity between adjacent image bands, and the printed positioning marks, a transmission disturbance region corresponding to the received image is generated. The detection server sequentially reads the acquisition sequence identifiers of each image band within the heat-sealed blank area along the traction direction of the received image. During reading, a rectangular reading area containing the acquisition sequence identifier is first extracted from the heat-sealed blank area of ​​each image band. The upper boundary of this rectangular reading area is 8 pixels from the upper edge of the image band, and the left boundary is 10 pixels from the inner edge of the heat-sealed blank area. The reading area size is 40 pixels by 12 pixels. Within the reading area, the detection server identifies four light gray short lines and one check short line. The four light gray short lines are converted into the original acquisition number of the image band using the aforementioned binary method, and the check short line is used to verify the number identification result. After the number identification is completed, the actual arrangement positions of the first to the last image band in the received image are sequentially numbered, and then the actual arrangement position numbers are compared with the original acquisition numbers band by band. If the original acquisition numbers of two adjacent image bands are not consecutively increasing by 1, it is determined that there is a jump at the boundary of the adjacent image bands; if the original acquisition number of the later image band is less than the original acquisition number of the earlier image band, it is determined that there is an inversion at the boundary of the adjacent image bands; if the same original acquisition number appears in adjacent or spaced image bands, it is determined that there is a repetition at the boundary on both sides of the image band with the repeated number.

[0032] The intersection of adjacent image bands corresponding to the above jumps, inversions, and repetitions is recorded as an abnormal sequence intersection.

[0033] For each sequentially aberrant boundary, as well as the one preceding and one following adjacent boundary, the grayscale texture direction of the color block outline endpoints, text baseline endpoints, and heat-sealed blank bands on both sides of the boundary is extracted. The color block outline endpoints are obtained by the color block edges within 6 rows of pixels above and below the boundary. The detection server scans the color change positions horizontally and records the pixel points where the same color block edge terminates at the boundary as the color block outline endpoints.

[0034] The text baseline endpoints are obtained from the lower edges of the batch number and traceability code characters. The detection server calculates the lower edge straight line along the lowest row of pixels of each character's connected strokes, and extends this line to the intersection of the image bands to obtain the text baseline endpoints. The grayscale texture direction is obtained from a continuous 24-pixel wide area in the heat-sealed blank band. First, the average grayscale value of each row in this area is calculated, and then the increasing or decreasing order of the average grayscale values ​​of adjacent rows is compared to form a texture direction sequence along the traction direction.

[0035] If the horizontal distance between the endpoints of the color blocks on both sides of the same boundary does not exceed 2 pixels and the vertical distance does not exceed 3 pixels, the endpoints of the outlines are considered to be able to connect; at the same time, if the order of increase and decrease of 5 consecutive positions in the grayscale texture direction sequence on both sides of the boundary is reversed, the texture direction is considered to be broken.

[0036] If the order of increase or decrease of 10 consecutive positions in the grayscale texture direction sequence on both sides of the same boundary is consistent, the texture direction is considered continuous; at the same time, if the lateral distance between the endpoints of the text baselines on both sides of the boundary reaches 4 pixels and continues to appear on more than 2 text baselines, the text baselines are considered misaligned.

[0037] A boundary where the contour endpoints can be aligned but the texture direction is broken, or where the texture direction is continuous but the text baseline is misaligned, is denoted as a pseudo-continuous boundary. This processing can identify splicing positions that appear smooth but are actually caused by transmission misordering.

[0038] The printed positioning mark that has not passed through the intersection of abnormal sequences is selected as the anchor object. The detection server first reads the four boundary points of the bounding rectangle of the anchor object, and then takes the intersection of the middle column of the left and right boundary pixel columns and the middle row of the top and bottom boundary pixel rows as the center of the anchor object. After connecting the centers of two adjacent anchor objects, a center line is formed, which extends along the traction direction of the packaging bag to each pseudo-continuous intersection. The extension process is recursively calculated according to the horizontal displacement corresponding to each row of pixels that the center line advances: first, the vertical row number between the two anchor object centers is counted, then the horizontal pixel difference between the two centers is counted, and the horizontal pixel difference is evenly distributed to each row according to the vertical row number; then, starting from the anchor object center closest to the pseudo-continuous intersection, the horizontal displacement is accumulated row by row until the row where the pseudo-continuous intersection is located is reached. Upon arrival, the horizontal position of the extended center line is compared with the actual positioning mark edge center position at the pseudo-continuous boundary. When the horizontal difference is 0 to 2 pixels and the vertical difference is 0 to 2 pixels, it is recorded as no traction offset. When the horizontal difference exceeds 2 pixels or the vertical difference exceeds 2 pixels, it is recorded as having a traction offset state, and the offset direction and the number of offset pixels are saved.

[0039] The edges of sequentially aberrant boundaries, pseudo-continuous boundaries, and adjacent image bands exhibiting traction offset are collectively considered as objects to be delineated. For each object to be delineated, the boundary row is extended forward and backward by 48 rows of pixels in the traction direction; if the boundary is both a sequentially aberrant boundary and a pseudo-continuous boundary, it is extended forward and backward by 72 rows of pixels. The lateral extent begins from the edge of the printed content closest to the boundary in the image band and extends laterally to the inner edge of the nearest heat-sealed blank band; the edge of the printed content is jointly determined by the outer boundary of color blocks, text, traceability codes, or anti-counterfeiting glossy textures.

[0040] When the longitudinal extension ranges of two adjacent objects to be delineated overlap, the two ranges are merged into one continuous range. The merged region is the transmission disturbance region corresponding to the received image. This region includes both locations where transmission disorder can be directly proven and pseudo-smooth misconnection locations obtained by inferring from edge continuity and printed positioning marks, enabling subsequent reliable transmission regions and transmission regions to be corrected to be delineated based on the same received image.

[0041] Based on the transmission disturbance area, the received image is divided into a reliable transmission area and a transmission area to be corrected. The overprint boundary, text baseline and color block outline in the reliable transmission area are used to reconnect the position of the transmission area to be corrected, so as to obtain the printed image to be compared.

[0042] This embodiment is used to separate the image content that can be directly used for comparison from the image content that needs to be reconnected after the transmission disturbance area has been obtained. This avoids directly treating the inter-band misconnection caused by image transmission as a printing defect, while retaining the real printing differences that still exist after reconnection.

[0043] The detection server starts from the outer edge of each transmission disturbance area and reads adjacent image bands forward and backward along the traction direction of the packaging bag. During reading, the acquisition sequence identifiers of adjacent image bands are first checked. If the original acquisition numbers of two consecutive image bands differ by 1 and the numbering direction is consistent with the traction direction, the acquisition sequence identifiers are considered continuous. Then, the traction offset status of the center line connecting the aforementioned printing positioning marks is retrieved. Image band edges without recorded lateral or longitudinal offsets are considered as having no traction offset at the center line. Next, the grayscale texture direction of the heat-sealed blank band is read. If more than 18 out of 20 consecutive adjacent rows of grayscale increment / decrease relationships match the corresponding positions of adjacent image bands, the grayscale texture direction is considered continuous. Image content that simultaneously satisfies the conditions of continuous acquisition sequence identifiers, no traction offset at the center line connecting the printing positioning marks, and continuous grayscale texture direction of the heat-sealed blank band is identified as a reliable transmission area. Image content connected to a transmission disturbance area but lacking any of these relationships is identified as a transmission area requiring correction. This division ensures that subsequent reconnection only affects image content within the transmission link's influence range, without altering the actual printing state within the reliable transmission area.

[0044] Within the trusted transmission area, overprinting boundaries, text baselines, and color block outlines that can cross the front and rear sides of the area to be corrected are selected as reconnection references. During selection, the standard printed image is used as the sequential reference. First, it is determined which segment of the color block outline the text baseline is attached to in the traction direction, and then the connection order between the overprinting boundary and that color block outline is confirmed. If the text baseline below the same batch number character in the standard printed image first passes the lower edge of the blue color block and then is adjacent to the red overprinting boundary, then the reconnection references in the received image are also arranged in the order of text baseline, blue color block outline, and red overprinting boundary. The detection server sequentially connects the reference endpoints in the front trusted transmission area and the reference endpoints in the rear trusted transmission area to form a reference chain, and extends the reference chain to the area to be corrected according to the attachment relationship in the standard printed image, obtaining the positions that each reference object in the area to be corrected should reach.

[0045] Within the area to be corrected, extract the same-color boundary ends, same-character height baseline ends, and same-contour transition ends corresponding to the reference chain. Same-color boundary ends are obtained through overprinting boundaries with consistent color categories; same-character height baseline ends are obtained through the row containing the lower edge of the character; and same-contour transition ends are obtained by the position where the color block contour direction changes from horizontal to vertical or vice versa. Position reconnection first aligns with the text baseline, i.e., reads the row difference between the same-character height baseline end and the extended end of the text baseline in the reference chain, and moves the entire image segment containing the same character row in the area to be corrected along the traction direction by the same number of rows. After completing the text baseline alignment, align the overprinting boundary, i.e., reads the column difference between the same-color boundary end and the same-color overprinting boundary end in the reference chain, and moves the image segment that has already been moved in the row direction along the horizontal direction by the same number of columns.

[0046] Finally, by aligning the color block outline and reading the column and row differences between the two turning points of the same outline, the image content between the two turning points is divided into continuous short segments. Each segment is then allocated horizontal and vertical compensation amounts based on its proportion of the total number of rows to the previous turning point, thus completing the local rotation. This sequence first ensures that the text recognition area is not bent, then ensures consistent overprinting boundaries, and finally eliminates angular offsets in the color block outline at local misalignments.

[0047] After segmented translation and local rotation, the transmission area to be corrected is reattached between the front and rear transmission reliable areas. During reattachment, the pixel positions of the transmission reliable areas remain unchanged; only the pixel content of the transmission area to be corrected is filled into the empty spaces after the reference chain is closed. If, after reattachment, there are still image contents with row differences exceeding 2 pixels or column differences exceeding 2 pixels between the two ends of the reference chain, their original pixels are retained and marked as unclosable content. The resulting comparative printed image corrects the positional anomalies caused by image band disorder, misalignment, and local tilt, while retaining the broken strokes, missing prints, reprints, and incomplete traceability codes that still exist after reattachment, providing a stable basis for subsequently distinguishing between transmission pseudo-defects and real printing defects.

[0048] Based on the correspondence between the printed image to be compared and the standard printed image, the printing difference features of batch number characters, traceability code, heat sealing edge fine line and anti-counterfeiting glossy oil texture are extracted.

[0049] First, read the overprint boundary, text baseline, and color block outline after the reference chain is closed, and use these three to determine the boundary of the local comparison segment. The local comparison segment is divided as follows: the text line where the text baseline is located is the first priority boundary, and the area where the batch number character is located is extended by 12 rows of pixels before and after along the traction direction; the rectangle outside the corner point of the traceability code is the second priority boundary, and the area where the traceability code is located is extended by 10 pixels in all directions; the continuous edge of the heat-sealing edge is the third priority boundary, and the area on both sides of the edge of the edge is extended by 6 pixels; the reflective direction of the anti-counterfeiting glossy varnish is the fourth priority boundary, and the outer edge of the glossy varnish is extended outward by 8 pixels.

[0050] The printed image to be compared and the standard printed image are cropped using the same segment boundaries, ensuring that both have the same pixel range within each local comparison segment. For image content that cannot be closed with the reference chain after reconnection, it is not merged into the middle of the segment, but rather assigned according to its edge distance from the previous or subsequent local comparison segment; when the distances on both sides are not equal, it is attached to the side with the shorter distance; when the distances on both sides are equal, it is attached to the end of the segment behind the traction direction, and the attachment direction is recorded.

[0051] Instead of using the top left corner of the segment as a unified starting point for reading, the starting point for reading differences is determined according to the traceable ends of the printed object. For batch number characters, the end of the stroke is used as the starting point for reading differences. First, the character stroke skeleton is extracted from the standard printed image, and then the endpoint of the stroke skeleton that is connected by only one adjacent stroke line is selected as the stroke end. When there are multiple stroke ends for the same character, the endpoint that is closer to the text baseline and located in front of the traction direction is selected first.

[0052] The traceability code uses three corner points as the starting point for difference reading. First, the circumscribed squares of the three corner points in the standard printed image are determined, and then the alternating black and white edges of the corresponding corner points are read in the printed image to be compared. The heat-sealing edge fine line uses continuous edges as the starting point for difference reading, selecting the position where the first 20 consecutive pixels from the starting end of the fine line in the standard printed image maintain a consistent line width. The anti-counterfeiting glossy varnish uses the direction of reflection as the starting point for difference reading, reading the first complete change segment of the glossy varnish's grayscale from bright to dark and then from dark to bright in the standard printed image, and taking the bright end of this change segment as the starting point.

[0053] Starting from the point of difference reading, the difference between the printed image to be compared and the standard printed image is read along the extension direction of the corresponding printed object. The batch number character is read pixel by pixel along the stroke skeleton direction. If the standard printed image contains a stroke but the corresponding ink is absent for three consecutive pixels in the printed image to be compared, it is recorded as a break. If the stroke pixels of the printed image to be compared deviate from the standard printed image by more than two pixels and the continuous length reaches five pixels, it is recorded as an offset. The traceability code is read line by line from the positioning corner point into the code grid. The black and white status of each code grid is determined by its center nine pixels. A black grid turning into a white grid is recorded as a missing grid, and a white grid turning into a black grid is recorded as an overlap. The heat-sealing edge fine line is read along the line direction. If the line width changes from 2 pixels in the standard printed image to 0 pixels and remains for 4 pixels, it is recorded as a break. If the line width increases to 4 pixels and remains for 4 pixels, it is recorded as an overlap. The anti-counterfeiting glossy texture is read along the reflective direction. If a complete variation segment in the compared printed image lacks a bright or dark end, it is recorded as missing. If the position of the variation segment differs from the standard printed image by more than 3 pixels, it is recorded as offset. Adjacent abnormal states in the same extension direction with an interval of no more than 2 pixels are merged into the same printing difference segment.

[0054] A record is created for each printing difference segment. The record includes the type of printed object, the local comparison segment number, the start position within the segment, the end position within the segment, the type of abnormal state, and its connection to non-closable content. The start and end positions are both based on the top-left corner of the local comparison segment, and the number of pixels horizontally and the number of rows in the traction direction are counted respectively; for example, if the start point is 15 pixels to the right of the left boundary of the segment and 9 rows downwards from the top boundary, it is recorded as 15 pixels horizontally and 9 rows vertically. For printing difference segments connected to non-closable content, the contact edge is also recorded as the front, back, left, or right side.

[0055] After the above records are made, the printing differences in batch number characters, traceability codes, heat-sealing edge lines, and anti-counterfeiting glossy patterns can simultaneously reflect the defect morphology, defect location, and its relationship with the re-received residual content, providing a verifiable basis for distinguishing between transmission fake defects and real printing defects.

[0056] Differences in printing difference features that are located in the transmission disturbance region and disappear after repositioning are identified as transmission pseudo-defects, while differences located in the transmission reliable region or that remain after repositioning are identified as real printing defects.

[0057] The start and end positions within each printing difference segment are read, and the original position of the printing difference segment in the received image is deduced based on the cropped position of the local comparison segment in the printed image to be compared. The deduction process is as follows: First, the number of horizontal pixels and vertical rows of the upper left corner of the local comparison segment in the printed image to be compared is obtained, and then the number of horizontal pixels and vertical rows of the printing difference segment recorded within the segment is added to obtain the position of the printing difference segment in the printed image to be compared; then, the segmented translation amount and local rotation amount experienced by this position during the position reconnection process are read, and the horizontal translation amount, vertical translation amount, and rotation compensation amount are deducted in the opposite direction to obtain the original position in the received image.

[0058] After tracing back to the original location, the original location is overlapped and verified with the transmission disturbance area and the transmission reliable area. If the printed difference segment is connected to non-closable content, the attachment direction of the non-closable content relative to the front, back, left, or right side of the printed difference segment is also recorded. The above content together constitutes the difference source record.

[0059] The segmented translation direction and local rotation direction of the area to be corrected during position reconnection are read. The segmented translation direction is determined by the number of columns moved laterally and the number of rows moved vertically. For example, if the image segment moves 3 pixels to the right and 2 rows down during reconnection, the segmented translation direction is recorded as lower right. The local rotation direction is determined by the row and column changes of the two contour transition ends before and after rotation. If the transition end closer to the front of the traction direction remains stationary, and the transition end closer to the rear of the traction direction compensates 2 pixels to the left, the local rotation direction is recorded as leftward convergence. Subsequently, the abnormal state direction of the printing difference segment is read. For the disconnected state, the direction in which the two ends of the break move away from each other is the abnormal state direction. For the offset state, the direction in which the printed image to be compared deviates from the standard printed image is the abnormal state direction. For the overlapping state, the direction in which the excess ink extends relative to the standard edge is the abnormal state direction. If the direction of the abnormal state is opposite to the direction of segmented translation or the direction of local rotation, it indicates that the difference may be canceled by the reconnection action, and the printed difference segment is marked as a difference that can be canceled by reconnection; if the directions are the same or do not correspond to each other, no such mark is made.

[0060] To address discrepancies that can be offset by reconnection, the print image to be compared is read again along the extension direction of the print object itself. The batch number character is read according to the stroke skeleton direction. If the strokes that are continuous in the standard print image are also continuous in the corresponding positions in the print image to be compared, and the horizontal and vertical differences between the centers of the corresponding strokes do not exceed 1 pixel, then the strokes of the batch number character are considered to have been restored to continuous correspondence.

[0061] The traceability code is read sequentially by grid, with the center 9 pixels of each grid taken. When the black and white state of the same central area in the compared printed image and the standard printed image matches, the grid is considered a correspondence. When all grids involved in the reconnection that can cancel out differences correspond, the traceability code grids are considered to have restored continuous correspondence. The heat-sealing edge line is read along the line direction. When the line width in the compared printed image is within one pixel of the standard printed image's line width plus or minus one pixel, and its continuous length covers the start and end range of the original printing difference segment, the heat-sealing edge line is considered to have restored continuous correspondence. The anti-counterfeiting glossy varnish is read along the reflective direction. When the order of the bright end, dark end, and the next bright end matches the standard printed image, and the positional difference of the three does not exceed two pixels, the anti-counterfeiting glossy varnish is considered to have restored continuous correspondence. Printing difference segments that satisfy the continuous correspondence relationship of the corresponding objects are identified as differences that disappear after position reconnection.

[0062] The differences are categorized based on the source records and the results of subsequent readings. If the original position of a printing difference falls within a transmission disturbance area and continuous correspondence has been restored, the difference is identified as a false transmission defect. If the original position of a printing difference falls within a reliable transmission area, regardless of whether it is adjacent to a transmission disturbance area, it is identified as a genuine printing defect. If the original position of a printing difference falls within a transmission disturbance area, but after position reconciliation, there are still issues such as discontinuous strokes in the batch number characters, mismatched traceability code grids, incomplete heat-sealing edge lines, or incomplete anti-counterfeiting glossy varnish variations, it is also identified as a genuine printing defect.

[0063] It should be noted that transmission pseudo-defects only correspond to differences caused by misconnection of image bands that can be eliminated by reconnection. Real printing defects correspond to abnormalities in the actual printing quality of the packaging bag. This can reduce the number of erroneous rejections and prevent defective packaging bags from entering the subsequent sterilization and packing processes.

[0064] Based on the location, continuity, and correlation of actual printing defects with traceability codes and batch number characters, the packaging bag printing quality inspection results are output.

[0065] First, the start and end positions, the corresponding local comparison segment number, and the abnormal state type of each actual printing defect are read. Position statistics are based on the top-left corner of the local comparison segment as the origin, recorded horizontally by the number of pixel columns and vertically by the number of rows in the traction direction. Then, based on the cropped position of the local comparison segment in the printed image to be compared, the actual printing defect is converted into its position in the entire printed image. During the conversion, the horizontal pixel count at the top-left corner of the local comparison segment is added to the horizontal pixel count at the defect's starting point to obtain the horizontal starting position of the entire image; the vertical row count at the top-left corner of the local comparison segment is added to the vertical row count at the defect's starting point to obtain the vertical starting position of the entire image. The end position is converted in the same way. This yields the precise position of the actual printing defect within the entire packaging bag layout, which is used for subsequent association with traceability codes and batch number characters.

[0066] The degree of continuity is determined based on the coverage length of the actual printing defect along the extension direction of the printed object itself. For batch number character defects, statistics are compiled along the stroke skeleton direction. If the length of consecutive pixels covering the same stroke is more than 6 pixels, it is recorded as a stroke continuity defect; if the coverage length is less than 6 pixels but located at a stroke intersection, a closed loop of a number, or the end of a character, it is recorded as a critical stroke defect. For traceability code defects, statistics are compiled by code grid. When the number of consecutive adjacent abnormal code grids reaches 2, it is recorded as a continuous code grid defect; when a single abnormal code grid is located within the outer square of 3 positioning corner points, it is recorded as a positioning corner point defect. For heat-sealing edge fine line defects, statistics are compiled along the line direction. When the length of consecutive breaks reaches 8 pixels, it is recorded as a continuity impairment defect. For anti-counterfeiting glossy oil pattern defects, statistics are compiled along the reflective direction. When any one of the bright end, dark end, or the next bright end is missing and continuously covers a complete changing segment, it is recorded as a verification pattern defect.

[0067] The association is determined based on the spatial connection and reading order between the actual printing defect and the traceability code / batch number characters. If the full image location of the actual printing defect falls within the outer area of ​​the batch number characters, or is within 3 pixels of the edge of the outer area of ​​the batch number characters, and the defect direction is consistent with the stroke direction of the character, then the actual printing defect is considered to be associated with the batch number characters.

[0068] If a genuine printing defect falls within the area bounded by the traceability code, or intersects with the square circumscribed by the corner point of the traceability code, the genuine printing defect is considered to be associated with the traceability code. If the same genuine printing defect is simultaneously connected to both non-closable content and the corner point of the traceability code, it is processed primarily based on the traceability code association; if the same genuine printing defect is simultaneously connected to both batch number characters and the heat seal edge line, it is processed primarily based on the batch number characters. This order ensures that traceability and batch number information are prioritized in the non-conformance determination.

[0069] The test results are categorized as qualified, pending re-inspection, and unqualified. If no actual printing defects are detected, or if only isolated ink dots that are not connected to the traceability code, batch number characters, heat seal edge lines, or anti-counterfeiting glossy texture and have a continuous length of less than 4 pixels exist, the output is qualified.

[0070] If defects exist in the anti-counterfeiting glossy verification pattern or the continuity of the heat-sealed edge fine lines, and these defects are not associated with traceability codes or batch number characters, output "Pending Re-inspection" and indicate the defect object, its location in the entire image, and its continuous length in the result. If defects exist in batch number character association, traceability code association, positioning corner point defects, continuous code grid defects, or if there are more than three defects requiring re-inspection on the same packaging bag, output "Unqualified".

[0071] When outputting the packaging bag printing quality inspection results, the packaging bag orientation, actual printing defect category, full image location of the defect, continuity, related objects, and processing conclusion should be written simultaneously.

[0072] For unqualified results, the corresponding transmission disturbance area and location reconnection record are retained in the received image to demonstrate that the defect is not a transmission pseudo-defect. For results to be re-inspected, the corresponding local comparison segment in the printed image to be compared is retained to facilitate manual verification of glossy varnish lines and heat-sealing edge fine lines.

[0073] It should be noted that the detection results no longer rely solely on the area of ​​image differences, but are output based on the actual impact of real printing defects on traceability codes and batch number characters. This can reduce false rejections caused by transmitting fake defects and prevent packaging bags that affect traceability information from flowing into subsequent processes.

[0074] Example 2, please refer to Figure 2 As shown, the image transmission-based packaging bag printing quality inspection system described in this embodiment includes: The image acquisition module acquires the received image of the packaging bag to be inspected after it is captured at the printing inspection station and sent through the image transmission link, as well as the standard printed image corresponding to the same packaging bag. The transmission disturbance region generation module generates the transmission disturbance region corresponding to the received image based on the arrival order of the received image segments, the edge continuity between adjacent image bands, and the printed positioning marks. The region segmentation module divides the received image into a reliable transmission region and a transmission correction region based on the transmission disturbance region. It then uses the overprinting boundary, text baseline, and color block outline in the reliable transmission region to perform position reconnection on the transmission correction region, thus obtaining the printed image to be compared. The printing difference feature extraction module extracts printing difference features such as batch number characters, traceability code, heat sealing edge fine lines, and anti-counterfeiting glossy oil texture based on the correspondence between the printed image to be compared and the standard printed image. The defect type determination module identifies differences in printing difference features that are located in the transmission disturbance area and disappear after repositioning as transmission pseudo-defects, and differences that are located in the transmission reliable area or remain after repositioning as real printing defects. The quality inspection result output module outputs the packaging bag printing quality inspection results based on the location, continuity, and correlation of the actual printing defects with the traceability code and batch number characters.

[0075] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for detecting the printing quality of packaging bags based on image transmission, characterized in that, include: Acquire the received image of the packaging bag to be inspected at the printing inspection station and sent through the image transmission link, as well as the standard printed image corresponding to the packaging bag of the same version; Based on the arrival order of the received image segments, the edge continuity between adjacent image bands, and the printed positioning marks, a transmission disturbance region corresponding to the received image is generated. Based on the transmission disturbance area, the received image is divided into a reliable transmission area and a transmission area to be corrected. The overprint boundary, text baseline and color block outline in the reliable transmission area are used to reconnect the position of the transmission area to be corrected, so as to obtain the printed image to be compared. Based on the correspondence between the printed image to be compared and the standard printed image, the printing difference features of batch number characters, traceability code, heat sealing edge fine line and anti-counterfeiting glossy oil texture are extracted. Differences in printing difference features that are located in the transmission disturbance region and disappear after position reconnection are identified as transmission pseudo-defects, while differences that are located in the transmission reliable region or remain after position reconnection are identified as real printing defects. Based on the location, continuity, and correlation of actual printing defects with traceability codes and batch number characters, the packaging bag printing quality inspection results are output.

2. The packaging bag printing quality inspection method based on image transmission according to claim 1, characterized in that, Obtaining the received image and the standard printed image includes the following steps: When the printed positioning mark of the packaging bag to be inspected enters the field of view, a pre-reading image containing the heat-sealed blank strip and the printed positioning mark is acquired, and the orientation of the plate is determined according to the relative position of the heat-sealed blank strip and the printed positioning mark. The effective detection boundary is determined based on the area that recedes from the outer boundary of the printed area in the pre-read image towards the blank area of ​​the heat-sealing edge. The image within the effective detection boundary is divided into multiple image bands according to the traction direction of the packaging bag, and the acquisition sequence identifier that does not participate in the comparison of printing defects is written in the blank band of the heat-sealed edge corresponding to each image band. Based on the orientation of the page and the effective detection boundary, retrieve and crop the standard printed image corresponding to the same packaging bag, so that the received image and the standard printed image have the same comparison range.

3. The method for detecting the printing quality of packaging bags based on image transmission according to claim 2, characterized in that, The formation of a received image includes the following steps: Each image band is sent to the detection server via the image transmission link; The image bands are stitched together on the detection server according to their actual arrival order, instead of immediately restoring the original order according to the acquisition sequence identifier. Read the acquisition sequence identifier of each image band in the stitched received image located within the heat-sealed blank band, and save the original acquisition number expressed by the acquisition sequence identifier and the actual arrangement position of the image band in the received image.

4. The packaging bag printing quality inspection method based on image transmission according to claim 1, characterized in that, Generating a transmission disturbance region includes the following steps: Check the acquisition sequence identifier against the actual arrangement position for each band, and identify the intersection of adjacent image bands where the acquisition sequence identifier is skipped, reversed or repeated as an abnormal sequence intersection. At the junctions of sequential anomalies and their adjacent junctions, the grayscale texture direction of the color block outline endpoints, text baseline endpoints, and heat-sealed blank bands is extracted respectively. When there are color block outline endpoints that can be connected on both sides of the same boundary but the grayscale texture direction is broken, or the grayscale texture direction is continuous but the text baseline is misaligned, the boundary is determined as a pseudo-continuous boundary. The sequential anomaly boundary and the pseudo-continuous boundary are used as the initial boundary locations of the transmission disturbance region.

5. The method for detecting the printing quality of packaging bags based on image transmission according to claim 4, characterized in that, The expansion of the transmission disturbance area includes the following steps: Select printed positioning marks that do not pass through the intersection of sequence abnormalities as anchoring objects, and form a line connecting the centers of the printed positioning marks along the traction direction of the packaging bag; Extend the line connecting the centers of the printed positioning marks to the pseudo-continuum boundary, and compare the extended position with the actual center position of the positioning mark edge at the pseudo-continuum boundary. When the comparison results indicate that there is a traction offset at the pseudo-continuous boundary, the edge of the adjacent image band corresponding to the pseudo-continuous boundary is incorporated into the transmission disturbance region. The edge of the incorporated image band is then extended laterally to the nearest heat-sealed blank band to form the transmission disturbance region corresponding to the received image.

6. The method for detecting the printing quality of packaging bags based on image transmission according to claim 5, characterized in that, The process of dividing the transmission into trusted regions and transmission regions to be corrected includes the following steps: Tracing adjacent image bands back and forth along the outer edge of the transmission disturbance area towards the direction of the packaging bag's traction; Image content with continuous acquisition sequence identification, no traction offset of the center line of printed positioning mark, and continuous grayscale texture direction of heat-sealed blank band is identified as a reliable transmission area. Image content that is connected to the transmission disturbance area and does not simultaneously meet the requirements of continuous acquisition sequence identification, no traction offset of the center line, and continuous grayscale texture direction is identified as the transmission correction area.

7. The method for detecting the printing quality of packaging bags based on image transmission according to claim 6, characterized in that, The location reconnection for the transmission area to be corrected includes the following steps: In the trusted transmission region, the overprint boundary, text baseline and color block outline that span the front and back sides of the transmission area to be corrected are selected as the reconnection reference. A reference chain is formed based on the front-to-back attachment relationship of the overprint boundary, text baseline and color block outline in the standard printed image, and the reference chain is extended to the area to be corrected during transmission. Extract the boundary ends of the same color, the baseline ends of the same character height, and the turning ends of the same contour that correspond to the reference chain within the transmission area to be corrected; Following the order of first aligning with the text baseline, then with the overprint boundary, and finally with the color block outline, the area to be corrected is segmented and locally rotated to obtain the printed image to be compared.

8. The method for detecting the printing quality of packaging bags based on image transmission according to claim 7, characterized in that, Extracting printing difference features includes the following steps: Based on the overprint boundary, text baseline and color block outline after the reference chain is closed, the printed image to be compared and the standard printed image are divided into corresponding local comparison segments. The image content that still cannot be closed with the reference chain after reconnection is attached to the end of the adjacent local comparison segment. The starting points for reading differences are the end of the strokes of the batch number characters, the corner points of the traceability code, the continuous edges of the heat-sealing line, and the reflective direction of the anti-counterfeiting glossy oil. Read the disconnection, offset, overlap and missing states along the extension direction of each printed object, and merge the abnormal states that are connected in the same extension direction into printing difference segments to form printing difference features.

9. The method for detecting the printing quality of packaging bags based on image transmission according to claim 8, characterized in that, The process of identifying spurious defects and genuine printing defects and outputting the inspection results includes the following steps: The original position of each printing difference segment in the received image is traced back, and the original position is correlated with the transmission disturbance area, the transmission reliable area, and the attachment direction of the non-closable content. Read the segmented translation direction and local rotation direction of the area to be corrected during the position reconnection process, and mark the printing difference segment whose abnormal state direction is opposite to the segmented translation direction or local rotation direction as a difference that can be canceled by reconnection. If the difference that can be offset by reconnection is read again along the extension direction of the printed object itself, and if it restores a continuous correspondence with the standard printed image, it is determined to be a transmission pseudo-defect. Differences located in the reliable transmission area, or differences located in the transmission disturbance area but which still do not correspond continuously with the standard printed image after position reconnection, are determined to be real printing defects.

10. A packaging bag printing quality inspection system based on image transmission, used to implement the packaging bag printing quality inspection method based on image transmission as described in any one of claims 1-9, characterized in that, include: The image acquisition module acquires the received image of the packaging bag to be inspected after it is captured at the printing inspection station and sent through the image transmission link, as well as the standard printed image corresponding to the same packaging bag. The transmission disturbance region generation module generates the transmission disturbance region corresponding to the received image based on the arrival order of the received image segments, the edge continuity between adjacent image bands, and the printed positioning marks. The region segmentation module divides the received image into a reliable transmission region and a transmission correction region based on the transmission disturbance region. It then uses the overprinting boundary, text baseline, and color block outline in the reliable transmission region to perform position reconnection on the transmission correction region, thus obtaining the printed image to be compared. The printing difference feature extraction module extracts printing difference features such as batch number characters, traceability code, heat sealing edge fine lines, and anti-counterfeiting glossy oil texture based on the correspondence between the printed image to be compared and the standard printed image. The defect type determination module identifies differences in printing difference features that are located in the transmission disturbance area and disappear after repositioning as transmission pseudo-defects, and differences that are located in the transmission reliable area or remain after repositioning as real printing defects. The quality inspection result output module outputs the packaging bag printing quality inspection results based on the location, continuity, and correlation of the actual printing defects with the traceability code and batch number characters.

Images