Packing box defect detection system and method based on image analysis

Abstract

The invention relates to a packaging box defect detection system based on image analysis, belongs to digital culture product manufacturing, and particularly relates to the field of general image data processing or generation. And the intelligent defect positioning module is used for intelligently analyzing the printing number of the printing object with the printing defect of the latest printing product and the corresponding defect type identifier based on each piece of basic information including the customized multiple pieces of image data by using the intelligent defect positioning model. The invention further relates to a packaging box defect detection method based on image analysis. According to the invention, aiming at the technical problem that the printing defect area of the packaging box is difficult to quickly position in the prior art, various basic information screened in a targeted manner and an intelligent defect positioning model designed by a customized structure are introduced; the printing character string or the printing pattern and the corresponding defect type of the defect of each newest printing product of the packaging box are intelligently analyzed and set, and therefore the technical problem is solved.

Description

Technical Field

[0001] The present invention relates to the production of digital cultural products, and more specifically to the field of general image data processing or generation, and in particular to a packaging box defect detection system and method based on image analysis. Background Art

[0002] After completing the design of the printing scheme for various packaging boxes, digital printing of these boxes using digital printing machines is a major branch of digital cultural product production. Designed and digitally printed packaging boxes can be used in a variety of specific segments, including medicine boxes, cartons, brochures, posters, handbags, and instruction manuals. In pursuit of efficiency, digital printing technology utilizes a high-speed, piece-by-piece printing process. With increasingly rapid printing operations, printed boxes inevitably exhibit defects such as color difference, misalignment of printed objects, distortion of printed patterns, and missing text. These defects require timely on-site detection and rapid adjustment of printing strategies to ensure the quality of the finished boxes produced by the digital printing production line.

[0003] For example, Chinese invention patent publication CN114923920A proposes a device for detecting defects in packaging boxes. The device comprises: a conveying mechanism mounted on a frame for conveying packaging boxes; a rotating assembly disposed between the two conveying mechanisms for transferring the packaging boxes on the two conveying mechanisms; a bottom detection assembly mounted on the frame for detecting the bottoms of the packaging boxes on the rotating assembly; a symmetrical curved plate mounted on the frame for half-wrapping the packaging boxes on the rotating assembly at predetermined intervals; and a detection mechanism disposed on the curved plate for detecting surfaces other than the bottom of the packaging boxes. When used, the device can automatically perform multi-angle inspections on packaging boxes instead of manual labor, thereby effectively improving work efficiency and enhancing product inspection quality.

[0004] For example, Chinese invention patent publication CN117274689A proposes a method and system for detecting packaging defects. Based on machine vision-based artificial intelligence (AI) detection technology, this method extracts and fuses multi-scale features from packaging images to determine whether defects exist. This effectively addresses the inefficiency of manual inspection while also improving detection accuracy.

[0005] Obviously, the above-mentioned existing technologies involve either mechanical detection of packaging box defects or visual detection of packaging box defects, but both are for detecting whether there are defects in the packaging box as a whole, and are unable to locate defects in each newly produced packaging box and analyze the defect types of located defects at the same time. In fact, there are multiple printed character strings and multiple printed patterns on the packaging box at the same time. It is necessary to perform on-site positioning of defects of a certain object including a printed character string or a printed pattern on each newly produced packaging box, which facilitates the subsequent rapid adjustment of the printing strategy of a certain object on the box body, making the printing strategy adjustment of the packaging box more targeted, thereby ensuring the quality of the finished products of the packaging boxes under the digital printing production line. Summary of the Invention

[0006] In order to solve the technical problems in the prior art, the present invention provides a packaging box defect detection system and method based on image analysis. By introducing targeted screened basic information and a customized structurally designed intelligent defect positioning model, the system intelligently analyzes and sets the printed character strings or printed patterns with defects in each latest printed product of the packaging box and intelligently analyzes the corresponding defect types. The targeted screened basic information includes customized multiple image data, and the customized structurally designed intelligent defect positioning model is based on the neural network architecture of a convolutional neural network, thereby completing the intelligent detection of printing defects of each packaging box product based on image analysis, facilitating and quickly detecting printing defects and formulating corresponding printing adjustment strategies, and improving the quality and efficiency of digital cultural product production.

[0007] According to one aspect of the present invention, a packaging box defect detection system based on image analysis is provided, the system comprising: The first extraction mechanism is used to obtain a plurality of ASCII code values corresponding to a plurality of printed character strings set on the packaging box and a plurality of JPEG format data corresponding to a plurality of printed patterns; The second extraction mechanism is configured to receive a high-definition image of the box body of the latest printed product of the set packaging box, identify a sub-image of the surface printing area of the latest printed product in the high-definition image of the box body as a target sub-image corresponding to the latest printed product; a third extraction mechanism, configured to execute each training action on the convolutional neural network to obtain a convolutional neural network after executing each training action and output the convolutional neural network as an intelligent defect location model, and to accumulate the number of multiple printed character strings and the number of multiple printed patterns of the set packaging box to obtain the total number of printed objects corresponding to the set packaging box, wherein the number of training actions executed by the convolutional neural network is positively correlated with the total number of printed objects corresponding to the set packaging box; The defect positioning mechanism is respectively connected to the first extraction mechanism, the second extraction mechanism and the third extraction mechanism, and is used to use an intelligent defect positioning model to intelligently analyze the printing number of the printing object with printing defects in the latest printing product and the defect type identification of the printing object with printing defects in the latest printing product based on the total number of printing objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printing character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printing patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printing product.

[0008] According to another aspect of the present invention, a method for detecting packaging box defects based on image analysis is provided, the method comprising: Obtain multiple ASCII code values corresponding to multiple printed character strings and multiple JPEG format data corresponding to multiple printed patterns of the set packaging box; Receiving a high-definition image of a box body of a latest printed product of a set packaging box, identifying a sub-image of a surface printing area of the latest printed product in the high-definition image of the box body as a target sub-image corresponding to the latest printed product; Performing each training action on the convolutional neural network to obtain a convolutional neural network after each training action is performed and outputting it as an intelligent defect location model, accumulating the number of multiple printed character strings and the number of multiple printed patterns of the set packaging box to obtain the total number of printed objects corresponding to the set packaging box, wherein the number of training actions performed by the convolutional neural network is positively correlated with the total number of printed objects corresponding to the set packaging box; An intelligent defect location model is used to intelligently analyze the print numbers of the print objects with printing defects in the latest printed products and the defect type identification of the print objects with printing defects in the latest printed products based on the total number of printed objects corresponding to the set packaging box, multiple ASCII code values corresponding to multiple printed strings of the set packaging box, multiple JPEG format data corresponding to multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printed product.

[0009] It can be seen that the present invention has at least the following five outstanding substantive features: First, by introducing targeted screening of basic information and a customized structurally designed intelligent defect location model, we intelligently analyze the defective printed strings or patterns for each newly printed product on the packaging box, as well as the corresponding defect types. The targeted screening of basic information includes customized multiple image data sets. The customized structurally designed intelligent defect location model is based on the neural network architecture of a convolutional neural network, thus completing intelligent detection of printing defects for each packaging product based on image analysis. This facilitates the rapid detection of printing defects and the formulation of corresponding printing adjustment strategies, thereby improving the quality and efficiency of digital cultural product production. Secondly, an intelligent defect localization model is introduced to intelligently locate defective objects on the latest printed products of a set packaging box. The intelligent defect localization model is a convolutional neural network that has performed various training actions. The number of training actions performed by the convolutional neural network is positively correlated with the total number of printed objects corresponding to the set packaging box. The printed objects corresponding to the set packaging box are the printed strings or patterns on the set packaging box. The customized structural design of the above intelligent defect localization model ensures the effectiveness and stability of the intelligent detection results of printing defects on the packaging box products. Thirdly: To complete the intelligent positioning of defective objects of the latest printed products on the set packaging boxes, various basic information that is specifically screened includes the total number of printed objects corresponding to the set packaging boxes, multiple ASCII code values corresponding to multiple printed strings on the set packaging boxes, multiple JPEG format data corresponding to multiple printed patterns, and the cyan component value, magenta component value, yellow component value, and black component value of each pixel point of the target sub-image corresponding to the latest printed products. The above-mentioned targeted screening of various basic information including customized multiple image data further ensures the effectiveness and stability of the intelligent detection results of printing defects on the packaging boxes; Furthermore: specifically, the ASCII code value corresponding to each printed character string is a combination of ASCII code values obtained by concatenating the ASCII code values corresponding to the respective printed characters of the printed character string end to end; the JPEG format data corresponding to each printed pattern is a digital representation of a JPEG format file corresponding to a reference pattern of the printed pattern at an ultra-high-definition resolution; and an ultra-high-definition image of a box body of a latest printed product of a set packaging box is received, and a sub-image of a surface printed area of the latest printed product in the ultra-high-definition image of the box body is identified as a target sub-image corresponding to the latest printed product; Finally: in each training action performed on the convolutional neural network, the printing number of the printed object with printing defects of a certain historical printing product of the packaging box and the defect type identification of the printed object with printing defects of the said historical printing product are set as the output content of the convolutional neural network item by item, and the total number of printed objects corresponding to the packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the said historical printing product are set as the input content of the convolutional neural network item by item to complete this training action of the convolutional neural network, thereby ensuring the training effect of each training action of the convolutional neural network. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The embodiments of the present invention will be described below with reference to the accompanying drawings, in which: Figure 1 Schematic diagram of the working scenario of a packaging box defect detection system and method based on image analysis according to the present invention.

[0011] Figure 2 FIG1 is an internal structure diagram of a packaging box defect detection system based on image analysis according to the first embodiment of the present invention.

[0012] Figure 3 FIG2 is a diagram showing the internal structure of a packaging box defect detection system based on image analysis according to a second embodiment of the present invention.

[0013] Figure 4 FIG1 is an internal structure diagram of a packaging box defect detection system based on image analysis according to the third embodiment of the present invention.

[0014] Figure 5 FIG4 is an internal structure diagram of a packaging box defect detection system based on image analysis according to the fourth embodiment of the present invention.

[0015] Figure 6 FIG1 is an internal structure diagram of a packaging box defect detection system based on image analysis according to the fifth embodiment of the present invention.

[0016] Figure 7 The figure is a flowchart showing the steps of a packaging box defect detection method based on image analysis according to the sixth embodiment of the present invention. DETAILED DESCRIPTION

[0017] like Figure 1As shown, a schematic diagram of the working scenario of a packaging box defect detection system and method based on image analysis according to the present invention is given. The general image data processing or generation involved in the present invention belongs to the production of digital cultural products.

[0018] The specific technical process of the present invention is as follows: Technical process A: To complete the intelligent positioning of defective objects of the latest printed products of the packaging box, an intelligent defect positioning model is introduced. Different packaging boxes correspond to intelligent defect positioning models with different customized structures, such as Figure 1 The set packaging box may be a set packaging box for medicines; Specifically, the intelligent defect location model has the following customized structural designs: First, the intelligent defect location model is a convolutional neural network after executing each training action; Second, the number of training actions performed by the convolutional neural network is positively correlated with the total number of printed objects corresponding to the set packaging box, where the printed objects corresponding to the set packaging box are the printed strings or printed patterns on the set packaging box; Third, in each training action performed on the convolutional neural network, the printing number of the printed object of a certain historical printing product of the packaging box with a printing defect and the defect type identifier of the printed object of the certain historical printing product with a printing defect are set as the output content of the convolutional neural network item by item, and the total number of printed objects corresponding to the packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-image corresponding to the certain historical printing product are set as the input content of the convolutional neural network item by item to complete this training action of the convolutional neural network, thereby ensuring the training effect of each training action of the convolutional neural network; In this way, the customized structural design of the above intelligent defect localization model ensures the effectiveness and stability of the intelligent detection results of printing defects of packaging products; Technical process B: To complete the intelligent positioning of defective objects of the latest printed products of the packaging box, various basic information are screened in a targeted manner; For example, the basic information includes setting the total number of printing objects corresponding to the packaging box, setting multiple ASCII code values corresponding to multiple printed character strings of the packaging box, setting multiple JPEG format data corresponding to multiple printed patterns, and setting the cyan component value, magenta component value, yellow component value, and black component value of each pixel point of the target sub-image corresponding to the latest printed product; like Figure 1As shown, the basic information for targeted screening includes multiple ASCII code values, multiple JPEG format data, and multiple image data. The multiple image data are cyan component values, magenta component values, yellow component values, and black component values of each pixel point of the target sub-image corresponding to the latest printed product; As well as Figure 1 As shown, a picture acquisition mechanism can be used to acquire an ultra-high-definition picture of the box body of the latest printed product of the set packaging box, and identify a target sub-picture corresponding to the latest printed product from the ultra-high-definition picture of the box body of the latest printed product of the set packaging box. When the picture acquisition mechanism acquires the ultra-high-definition picture of the box body of the latest printed product of the set packaging box, the latest printed product of the set packaging box needs to be in a state of unfolding into a plane. For further example, the ASCII code value corresponding to each printed character string is a combination of ASCII code values obtained by concatenating the ASCII code values corresponding to the respective printed characters of the printed character string end to end, and the JPEG format data corresponding to each printed pattern is a digital representation of a JPEG format file corresponding to a reference pattern of the printed pattern at ultra-high definition resolution; As a further example, a high-definition image of a box body of a latest printed product of a set packaging box is received, and a sub-image of a surface printing area of the latest printed product in the high-definition image of the box body is identified as a target sub-image corresponding to the latest printed product; In this way, the targeted screening of various basic information including customized multiple image data further ensures the effectiveness and stability of the intelligent detection results of printing defects of packaging products; Technical Process C: Using the intelligent defect location model designed for customized packaging structures using Technical Process A, based on the basic information targeted by Technical Process B, intelligently analyzes and determines the defect objects of the latest printed products of the packaging boxes, completing the intelligent location of printing defects. Specifically, the result of the intelligent analysis includes the printing number of the printing object of the latest printing product that has printing defects and the defect type identification of the printing object of the latest printing product that has printing defects, such as Figure 1 As shown, that is, the printed number of the defect object and the defect type identification; Specifically, when the print number of the printed object with a printing defect in the latest printed product and the defect type identifier of the printed object with a printing defect in the latest printed product analyzed by the intelligent analysis are both empty characters, it is determined that the latest printed product of the packaging box does not have any printing defects; Technical process D: reporting the intelligent analysis results of technical process C to the remote product production monitoring server via the wireless communication network; For example, a data reporting mechanism may be used to enter the print number of the printed object with the printing defect in the latest printed product, the defect type identifier of the printed object with the printing defect in the latest printed product, and the print serial number of the latest printed product into a network data packet and wirelessly transmit the packet to a remote product production monitoring server; As a further example, the data reporting mechanism is also used to not perform network data packet packaging and wireless transmission when the print number of the printed object with printing defects in the latest printed product and the defect type identifier of the printed object with printing defects in the latest printed product received are both empty characters.

[0019] It can be seen that the present invention introduces various basic information that are targeted and an intelligent defect positioning model with customized structural design, intelligently analyzes and sets the printed character strings or printed patterns with defects in each latest printed product of the packaging box, and intelligently analyzes the corresponding defect types. The various basic information that are targeted and screened include customized multiple image data. The intelligent defect positioning model with customized structural design is based on the neural network architecture of the convolutional neural network, thereby completing the intelligent detection of printing defects of each packaging box product based on image analysis, facilitating and quickly detecting printing defects and formulating corresponding printing adjustment strategies, and improving the quality and efficiency of digital cultural product production.

[0020] The key points of the present invention are: targeted screening of various basic information of customized multiple image data, intelligent defect localization models for different customized structures of different packaging boxes, and targeted design of each training action performed by the convolutional neural network.

[0021] Hereinafter, a packaging box defect detection system and method based on image analysis of the present invention will be specifically described in the form of an embodiment.

[0022] First embodiment Figure 2 FIG1 is an internal structure diagram of a packaging box defect detection system based on image analysis according to the first embodiment of the present invention.

[0023] like Figure 2 As shown, the packaging box defect detection system based on image analysis includes the following components: The first extraction mechanism is used to obtain a plurality of ASCII code values corresponding to a plurality of printed character strings set on the packaging box and a plurality of JPEG format data corresponding to a plurality of printed patterns; For example, the value extraction unit may be used to obtain multiple ASCII code values corresponding to multiple printed character strings set on the packaging box, and the data extraction unit may be used to obtain multiple JPEG format data corresponding to multiple printed patterns set on the packaging box; The second extraction mechanism is configured to receive a high-definition image of the box body of the latest printed product of the set packaging box, identify a sub-image of the surface printing area of the latest printed product in the high-definition image of the box body as a target sub-image corresponding to the latest printed product; Specifically, receiving a high-definition image of a box body of a latest printed product of a set packaging box, identifying a sub-image of a surface printed area of the latest printed product in the high-definition image of the box body as a target sub-image corresponding to the latest printed product includes: the resolution of the high-definition image of the box body is 4 times the resolution of the high-definition resolution, that is, 3840×2160; a third extraction mechanism, configured to execute each training action on the convolutional neural network to obtain a convolutional neural network after executing each training action and output the convolutional neural network as an intelligent defect location model, and to accumulate the number of multiple printed character strings and the number of multiple printed patterns of the set packaging box to obtain the total number of printed objects corresponding to the set packaging box, wherein the number of training actions executed by the convolutional neural network is positively correlated with the total number of printed objects corresponding to the set packaging box; For example, the number of training actions executed by the convolutional neural network is positively correlated with the total number of printed objects corresponding to the package box, including: when the total number of printed objects corresponding to the package box is set to 8, the number of training actions executed by the corresponding convolutional neural network is 600 times; when the total number of printed objects corresponding to the package box is set to 10, the number of training actions executed by the corresponding convolutional neural network is 700 times; when the total number of printed objects corresponding to the package box is set to 12, the number of training actions executed by the corresponding convolutional neural network is 800 times; when the total number of printed objects corresponding to the package box is set to 14, the number of training actions executed by the corresponding convolutional neural network is 900 times, and so on; a defect locating mechanism connected to the first extraction mechanism, the second extraction mechanism, and the third extraction mechanism, respectively, for using an intelligent defect locating model to intelligently analyze the print number of the print object having a print defect in the latest printed product and the defect type identifier of the print object having a print defect in the latest printed product based on the total number of print objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple print character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple print patterns, and the cyan component value, magenta component value, yellow component value, and black component value of each pixel point of the target sub-image corresponding to the latest printed product; Specifically, the intelligent defect location model is used to intelligently analyze the print number of the print object with printing defects in the latest printed product and the defect type identification of the print object with printing defects in the latest printed product based on the total number of printed objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value, and black component value of each pixel point of the target sub-image corresponding to the latest printed product. The defect type identification of the print object with printing defects in the latest printed product includes: different values of the defect type identification of the print object can be used to represent different defect types of the print object; The ASCII code value corresponding to each printed character string is a combination of ASCII code values obtained by concatenating the ASCII code values corresponding to the printed characters of the printed character string, and the JPEG format data corresponding to each printed pattern is a digital representation of a JPEG format file corresponding to a reference pattern of the printed pattern at ultra-high-definition resolution. The printing object corresponding to the packaging box is set to be the printing string or printing pattern of the packaging box. Different printing objects have different printing numbers. The defect type identification of the printing object is used for resolution printing color difference defect, printing object misalignment defect, printing pattern distortion defect and printed text missing defect. In each training action performed on the convolutional neural network, the printing number of the printed object of a certain historical printing product of the packaging box with a printing defect and the defect type identifier of the printed object of the certain historical printing product with a printing defect are set as the output content of the convolutional neural network item by item, and the total number of printed objects corresponding to the packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-image corresponding to the certain historical printing product are set as the input content of the convolutional neural network item by item to complete this training action of the convolutional neural network; Wherein, using an intelligent defect location model to intelligently analyze the printing number of the printing object with printing defects in the latest printed product and the defect type identification of the printing object with printing defects in the latest printed product based on the total number of printed objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-image corresponding to the latest printed product, includes: when the printing number of the printing object with printing defects in the latest printed product and the defect type identification of the printing object with printing defects in the latest printed product obtained by analysis are both empty characters, determining the printing object without printing defects in the latest printed product; Accordingly, when the print number of the printed object with a printing defect in the latest printed product and the defect type identifier of the printed object with a printing defect in the latest printed product obtained by analysis are both not empty characters, it is determined that the printed object with a printing defect in the latest printed product has a printing defect; At the same time, when only one of the print number and defect type identifier of the latest printed product with printing defects is empty, it appears that a logical error has occurred in the intelligent analysis result, and the intelligent defect location model needs to be trained more times to perform intelligent analysis again; And wherein, an intelligent defect location model is used to intelligently analyze the print number of the print object with printing defects in the latest printed product and the defect type identification of the print object with printing defects in the latest printed product based on the total number of printed objects corresponding to the set packaging box, multiple ASCII code values corresponding to multiple printed character strings of the set packaging box and multiple JPEG format data corresponding to multiple printed patterns, as well as the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printed product, further including: the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printed product are the respective color component values under the CMYK color space.

[0024] Second embodiment Figure 3 FIG2 is a diagram showing the internal structure of a packaging box defect detection system based on image analysis according to a second embodiment of the present invention.

[0025] like Figure 3 As shown, compared with Figure 2 , the packaging box defect detection system based on image analysis also includes: The data reporting mechanism is connected to the defect locating mechanism and is used to enter the printing number of the printed object of the latest printed product with the printing defect, the defect type identifier of the printed object of the latest printed product with the printing defect, and the printing serial number of the latest printed product into a network data packet and wirelessly transmit it to a remote product production monitoring server; For example, the printing number of the printed object with printing defects in the latest printed product, the defect type identifier of the printed object with printing defects in the latest printed product, and the printing serial number of the latest printed product are entered into a network data packet and wirelessly sent to a remote product production monitoring server, including: the network data packet is an IP data packet; The data reporting mechanism is further configured not to package and wirelessly send the network data packet when the received print number of the print object with the printing defect in the latest printed product and the defect type identifier of the print object with the printing defect in the latest printed product are both empty characters.

[0026] Third embodiment Figure 4 FIG1 is an internal structure diagram of a packaging box defect detection system based on image analysis according to the third embodiment of the present invention.

[0027] like Figure 4 As shown, compared with Figure 3 , the packaging box defect detection system based on image analysis also includes: A parameter storage mechanism, connected to the third extraction mechanism, for completing model storage of the intelligent defect location model by storing various model parameters of the intelligent defect location model; Specifically, a TF memory chip or a FLASH memory chip can be selected to implement the parameter storage mechanism, which is connected to the third extraction mechanism and is used to complete the model storage of the intelligent defect location model by storing various model parameters of the intelligent defect location model.

[0028] Fourth embodiment Figure 5 FIG4 is an internal structure diagram of a packaging box defect detection system based on image analysis according to the fourth embodiment of the present invention.

[0029] like Figure 5 As shown, compared with Figure 4 , the packaging box defect detection system based on image analysis also includes: The instant display mechanism is connected to the defect positioning mechanism and is used to receive and display on site the printing number of the printed object with printing defects in the latest printed product and the defect type identification of the printed object with printing defects in the latest printed product Specifically, an LED display array or an LCD display array can be selected to implement the instant display mechanism, which is connected to the defect positioning mechanism to receive and display on-site the printing number of the printed object with printing defects in the latest printed product and the defect type identification of the printed object with printing defects in the latest printed product.

[0030] Fifth embodiment Figure 6 FIG1 is an internal structure diagram of a packaging box defect detection system based on image analysis according to the fifth embodiment of the present invention.

[0031] like Figure 6 As shown, compared with Figure 5 , the packaging box defect detection system based on image analysis also includes: The image acquisition mechanism is connected to the second extraction mechanism and is used to acquire an ultra-high-definition image of the box body of the latest printed product of the set packaging box, and send the ultra-high-definition image of the box body of the latest printed product of the set packaging box to the second extraction mechanism; The step of collecting the ultra-high-definition image of the latest printed product of the set packaging box and sending the ultra-high-definition image of the latest printed product of the set packaging box to the second extraction mechanism includes: using an ultra-high-definition resolution photoelectric sensor to complete the collection of the ultra-high-definition image of the latest printed product of the set packaging box; For example, using a photoelectric sensor with ultra-high resolution to complete the acquisition of ultra-high definition images of the box body of the latest printed product of the packaging box includes: the photoelectric sensor can be selected as a CCD sensor or a CMOS sensor; Among them, collecting the ultra-clear picture of the box body of the latest printed product of the set packaging box and sending the ultra-clear picture of the box body of the latest printed product of the set packaging box to the second extraction mechanism also includes: setting a filter on the top of the photoelectric sensor with ultra-clear resolution.

[0032] Next, various embodiments of the present invention will be further described.

[0033] In each of the above embodiments, optionally, in the packaging box defect detection system based on image analysis: An intelligent defect location model is used to intelligently analyze the print number of the print object with printing defects in the latest printed product and the defect type identification of the print object with printing defects in the latest printed product based on the total number of printed objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-image corresponding to the latest printed product. The value range of each color component value among the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-image corresponding to the latest printed product is also included: The receiving of the ultra-high-definition image of the box body of the latest printed product of the set packaging box and identifying a sub-image of the surface printing area of the latest printed product in the ultra-high-definition image of the box body as a target sub-image corresponding to the latest printed product includes: identifying the sub-image of the surface printing area of the latest printed product in the ultra-high-definition image of the box body based on a grayscale value distribution interval corresponding to the surface printing area of the latest printed product; Specifically, identifying the sub-image of the surface printing area of the latest printed product in the ultra-high-definition image of the box body based on the grayscale value distribution interval corresponding to the surface printing area of the latest printed product includes: using a pixel point in the ultra-high-definition image of the box body with a grayscale value within the grayscale value distribution interval corresponding to the surface printing area of the latest printed product as a constituent pixel point of the sub-image of the surface printing area of the latest printed product in the ultra-high-definition image of the box body; And wherein, the sub-screen of the surface printing area of the latest printed product in the ultra-high-definition screen of the box body is identified based on the grayscale value distribution interval corresponding to the surface printing area of the latest printed product, including: the grayscale value distribution interval corresponding to the surface printing area of the latest printed product is limited by the grayscale value upper limit threshold corresponding to the surface printing area of the latest printed product and the grayscale value lower limit threshold corresponding to the surface printing area of the latest printed product.

[0034] And in each of the above embodiments, optionally, in the packaging box defect detection system based on image analysis: Using an intelligent defect location model to intelligently analyze the print number of the print object with printing defects in the latest printed product and the defect type identification of the print object with printing defects in the latest printed product based on the total number of printed objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printed product also includes: synchronously inputting the total number of printing objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printed product into the intelligent defect location model to run the intelligent defect location model and obtain the print number of the print object with printing defects in the latest printed product and the defect type identification of the print object with printing defects in the latest printed product output by the intelligent defect location model; For example, the JPEG format data corresponding to each printed pattern is the digital representation data of the JPEG file of the printed pattern; Among them, the total number of printing objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printing character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printing patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-image corresponding to the latest printed product are synchronously input into the intelligent defect positioning model to run the intelligent defect positioning model, and obtain the print number of the printing object with printing defects in the latest printed product and the defect type identification of the printing object with printing defects in the latest printed product output by the intelligent defect positioning model, including: setting the total number of printing objects corresponding to the set packaging box, setting the multiple printing character strings of the set packaging box, and inputting the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-image corresponding to the latest printed product into the intelligent defect positioning model. Before the multiple ASCII code values corresponding to the character strings, the multiple JPEG format data corresponding to the multiple printing patterns, and the cyan component value, magenta component value, yellow component value, and black component value of each pixel point of the target sub-image corresponding to the latest printed product are synchronously input into the intelligent defect location model, binary value conversion processing is performed on the total number of printed objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printing character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printing patterns, and the cyan component value, magenta component value, yellow component value, and black component value of each pixel point of the target sub-image corresponding to the latest printed product; And wherein, the total number of printed objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printed product are synchronously input into the intelligent defect location model to run the intelligent defect location model, and obtain the printing number of the printing object with printing defects in the latest printed product and the defect type identification of the printing object with printing defects in the latest printed product output by the intelligent defect location model. It also includes: the printing number of the printing object with printing defects in the latest printed product and the defect type identification of the printing object with printing defects in the latest printed product output by the intelligent defect location model are both represented in the form of binary values.

[0035] Sixth embodiment Figure 7 The figure is a flowchart showing the steps of a packaging box defect detection method based on image analysis according to the sixth embodiment of the present invention.

[0036] like Figure 7 As shown, the packaging box defect detection method based on image analysis includes the following steps: Step 701: Acquire multiple ASCII code values corresponding to multiple printed character strings and multiple JPEG format data corresponding to multiple printed patterns of a set packaging box; For example, the value extraction unit may be used to obtain multiple ASCII code values corresponding to multiple printed character strings set on the packaging box, and the data extraction unit may be used to obtain multiple JPEG format data corresponding to multiple printed patterns set on the packaging box; Step 702: Receive a high-definition image of the box body of the latest printed product of the set packaging box, identify a sub-image of the surface printing area of the latest printed product in the high-definition image of the box body as a target sub-image corresponding to the latest printed product; Specifically, receiving a high-definition image of a box body of a latest printed product of a set packaging box, identifying a sub-image of a surface printed area of the latest printed product in the high-definition image of the box body as a target sub-image corresponding to the latest printed product includes: the resolution of the high-definition image of the box body is 4 times the resolution of the high-definition resolution, that is, 3840×2160; Step 703: Execute each training action on the convolutional neural network to obtain a convolutional neural network after executing each training action and output it as an intelligent defect location model. Accumulate the number of multiple printed character strings and the number of multiple printed patterns of the set package box to obtain the total number of printed objects corresponding to the set package box. The number of training actions executed by the convolutional neural network is positively correlated with the total number of printed objects corresponding to the set package box. For example, the number of training actions executed by the convolutional neural network is positively correlated with the total number of printed objects corresponding to the package box, including: when the total number of printed objects corresponding to the package box is set to 8, the number of training actions executed by the corresponding convolutional neural network is 600 times; when the total number of printed objects corresponding to the package box is set to 10, the number of training actions executed by the corresponding convolutional neural network is 700 times; when the total number of printed objects corresponding to the package box is set to 12, the number of training actions executed by the corresponding convolutional neural network is 800 times; when the total number of printed objects corresponding to the package box is set to 14, the number of training actions executed by the corresponding convolutional neural network is 900 times, and so on; Step 704: Using an intelligent defect location model, intelligently analyze the print number of the print object with a printing defect in the latest printed product and the defect type identifier of the print object with a printing defect in the latest printed product based on the total number of printed objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printed character strings and the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value, and black component value of each pixel point of the target sub-image corresponding to the latest printed product; Specifically, the intelligent defect location model is used to intelligently analyze the print number of the print object with printing defects in the latest printed product and the defect type identification of the print object with printing defects in the latest printed product based on the total number of printed objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value, and black component value of each pixel point of the target sub-image corresponding to the latest printed product. The defect type identification of the print object with printing defects in the latest printed product includes: different values of the defect type identification of the print object can be used to represent different defect types of the print object; The ASCII code value corresponding to each printed character string is a combination of ASCII code values obtained by concatenating the ASCII code values corresponding to the printed characters of the printed character string, and the JPEG format data corresponding to each printed pattern is a digital representation of a JPEG format file corresponding to a reference pattern of the printed pattern at ultra-high-definition resolution. The printing object corresponding to the packaging box is set to be the printing string or printing pattern of the packaging box. Different printing objects have different printing numbers. The defect type identification of the printing object is used for resolution printing color difference defect, printing object misalignment defect, printing pattern distortion defect and printed text missing defect. In each training action performed on the convolutional neural network, the printing number of the printed object of a certain historical printing product of the packaging box with a printing defect and the defect type identifier of the printed object of the certain historical printing product with a printing defect are set as the output content of the convolutional neural network item by item, and the total number of printed objects corresponding to the packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-image corresponding to the certain historical printing product are set as the input content of the convolutional neural network item by item to complete this training action of the convolutional neural network; Wherein, using an intelligent defect location model to intelligently analyze the printing number of the printing object with printing defects in the latest printed product and the defect type identification of the printing object with printing defects in the latest printed product based on the total number of printed objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-image corresponding to the latest printed product, includes: when the printing number of the printing object with printing defects in the latest printed product and the defect type identification of the printing object with printing defects in the latest printed product obtained by analysis are both empty characters, determining the printing object without printing defects in the latest printed product; Accordingly, when the print number of the printed object with a printing defect in the latest printed product and the defect type identifier of the printed object with a printing defect in the latest printed product obtained by analysis are both not empty characters, it is determined that the printed object with a printing defect in the latest printed product has a printing defect; At the same time, when only one of the print number and defect type identifier of the latest printed product with printing defects is empty, it appears that a logical error has occurred in the intelligent analysis result, and the intelligent defect location model needs to be trained more times to perform intelligent analysis again; And wherein, an intelligent defect location model is used to intelligently analyze the print number of the print object with printing defects in the latest printed product and the defect type identification of the print object with printing defects in the latest printed product based on the total number of printed objects corresponding to the set packaging box, multiple ASCII code values corresponding to multiple printed character strings of the set packaging box and multiple JPEG format data corresponding to multiple printed patterns, as well as the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printed product, further including: the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printed product are the respective color component values under the CMYK color space.

[0037] In addition, in a packaging box defect detection system and method based on image analysis according to the present invention: Accumulating the number of multiple printed character strings and the number of multiple printed patterns of the set packaging box to obtain the total number of printed objects corresponding to the set packaging box, wherein the number of training actions performed by the convolutional neural network is positively correlated with the total number of printed objects corresponding to the set packaging box, includes: using a dual-input single-output content mapping function to represent the content mapping relationship of the positive correlation between the number of training actions performed by the convolutional neural network and the total number of printed objects corresponding to the set packaging box; For example, the MATLAG toolbox can be selected to complete the testing and simulation of the data processing process of using a dual-input and single-output content mapping function to represent the content mapping relationship in which the number of training actions performed by the convolutional neural network is positively correlated with the total number of printed objects corresponding to a set packaging box; And wherein, a content mapping relationship in which a dual-input and single-output content mapping function is used to represent the positive correlation between the number of training actions performed by the convolutional neural network and the total number of printed objects corresponding to the set packaging box includes: in the content mapping function, the number of multiple printed character strings set for the packaging box and the number of multiple printed patterns set for the packaging box are two input contents, and the number of training actions performed by the convolutional neural network is a single output content.

[0038] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0039] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device / electronic device / computer-readable storage medium / computer program product embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The above is only a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are included in the scope of protection of the present invention.

Claims

1. A packaging box defect detection system based on image analysis, characterized in that: The system comprises: The first extraction mechanism is used to obtain a plurality of ASCII code values corresponding to a plurality of printed character strings set on the packaging box and a plurality of JPEG format data corresponding to a plurality of printed patterns; The second extraction mechanism is configured to receive an ultra-high-definition image of a box body of a latest printed product set as a packaging box, and identify a sub-image of a surface printing area of the latest printed product in the ultra-high-definition image of the box body as a target sub-image corresponding to the latest printed product; a third extraction mechanism, configured to execute each training action on the convolutional neural network to obtain a convolutional neural network after executing each training action and output the convolutional neural network as an intelligent defect location model, and to accumulate the number of multiple printed character strings and the number of multiple printed patterns of the set packaging box to obtain the total number of printed objects corresponding to the set packaging box, wherein the number of training actions executed by the convolutional neural network is positively correlated with the total number of printed objects corresponding to the set packaging box; The defect positioning mechanism is respectively connected to the first extraction mechanism, the second extraction mechanism and the third extraction mechanism, and is used to use an intelligent defect positioning model to intelligently analyze the printing number of the printing object with printing defects in the latest printing product and the defect type identification of the printing object with printing defects in the latest printing product based on the total number of printing objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printing character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printing patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printing product.

2. The packaging box defect detection system based on image analysis according to claim 1, characterized in that: The ASCII code value corresponding to each printed character string is a combination of ASCII code values obtained by concatenating the ASCII code values corresponding to the printed characters of the printed character string, and the JPEG format data corresponding to each printed pattern is a digital representation of a JPEG format file corresponding to a reference pattern of the printed pattern at ultra-high definition resolution; The printing object corresponding to the packaging box is set to be the printing string or printing pattern of the packaging box. Different printing objects have different printing numbers. The defect type identification of the printing object is used for resolution printing color difference defect, printing object misalignment defect, printing pattern distortion defect and printed text missing defect. Among them, in each training action performed on the convolutional neural network, the printing number of the printed object with printing defects in a certain historical printing product of the packaging box and the defect type identification of the printed object with printing defects in the said historical printing product are set as the output content of the convolutional neural network item by item, and the total number of printed objects corresponding to the packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the said historical printing product are set as the input content of the convolutional neural network item by item to complete this training action of the convolutional neural network.

3. The packaging box defect detection system based on image analysis according to claim 2, characterized in that: Using an intelligent defect location model to intelligently analyze the printing number of the printing object with printing defects in the latest printed product and the defect type identification of the printing object with printing defects in the latest printed product based on the total number of printing objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printing character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printing patterns, and the cyan component value, magenta component value, yellow component value, and black component value of each pixel point of the target sub-image corresponding to the latest printed product, including: when the printing number of the printing object with printing defects in the latest printed product and the defect type identification of the printing object with printing defects in the latest printed product obtained by analysis are both empty characters, determining the printing object without printing defects in the latest printed product; Among them, an intelligent defect positioning model is used to intelligently analyze the printing number of the printing object with printing defects in the latest printing product and the defect type identification of the printing object with printing defects in the latest printing product based on the total number of printing objects corresponding to the set packaging box, multiple ASCII code values corresponding to multiple printing strings of the set packaging box, multiple JPEG format data corresponding to multiple printing patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printing product. It also includes: the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printing product are the respective color component values under the CMYK color space.

4. The packaging box defect detection system based on image analysis according to claim 3, characterized in that: The system further comprises: The data reporting mechanism is connected to the defect locating mechanism and is used to enter the printing number of the printed object of the latest printed product with the printing defect, the defect type identifier of the printed object of the latest printed product with the printing defect, and the printing serial number of the latest printed product into a network data packet and wirelessly transmit it to a remote product production monitoring server; The data reporting mechanism is further configured not to package and wirelessly send the network data packet when the received print number of the print object with the printing defect in the latest printed product and the defect type identifier of the print object with the printing defect in the latest printed product are both empty characters.

5. The packaging box defect detection system based on image analysis according to claim 3, characterized in that: The system further comprises: The parameter storage mechanism is connected to the third extraction mechanism and is used to complete the model storage of the intelligent defect location model by storing various model parameters of the intelligent defect location model.

6. The packaging box defect detection system based on image analysis according to claim 3, characterized in that: The system further comprises: The instant display mechanism is connected to the defect positioning mechanism and is used to receive and display on site the printing number of the printing object with printing defects in the latest printed product and the defect type identification of the printing object with printing defects in the latest printed product.

7. The packaging box defect detection system based on image analysis according to claim 3, characterized in that: The system further comprises: The image acquisition mechanism is connected to the second extraction mechanism and is used to acquire an ultra-high-definition image of the box body of the latest printed product of the set packaging box, and send the ultra-high-definition image of the box body of the latest printed product of the set packaging box to the second extraction mechanism; The step of collecting the ultra-high-definition image of the latest printed product of the set packaging box and sending the ultra-high-definition image of the latest printed product of the set packaging box to the second extraction mechanism includes: using an ultra-high-definition resolution photoelectric sensor to complete the collection of the ultra-high-definition image of the latest printed product of the set packaging box; Among them, collecting the ultra-clear picture of the box body of the latest printed product of the set packaging box and sending the ultra-clear picture of the box body of the latest printed product of the set packaging box to the second extraction mechanism also includes: setting a filter on the top of the photoelectric sensor with ultra-clear resolution.

8. The packaging box defect detection system based on image analysis according to any one of claims 3 to 7, characterized in that: An intelligent defect location model is used to intelligently analyze the print number of the print object with printing defects in the latest printed product and the defect type identification of the print object with printing defects in the latest printed product based on the total number of printed objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-image corresponding to the latest printed product. The value range of each color component value among the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-image corresponding to the latest printed product is also included: The receiving of the ultra-high-definition image of the box body of the latest printed product of the set packaging box and identifying a sub-image of the surface printing area of the latest printed product in the ultra-high-definition image of the box body as a target sub-image corresponding to the latest printed product includes: identifying the sub-image of the surface printing area of the latest printed product in the ultra-high-definition image of the box body based on a grayscale value distribution interval corresponding to the surface printing area of the latest printed product; Among them, the sub-screen of the surface printing area of the latest printed product in the ultra-high-definition screen of the box body is identified based on the grayscale value distribution interval corresponding to the surface printing area of the latest printed product, including: the grayscale value distribution interval corresponding to the surface printing area of the latest printed product is limited by the grayscale value upper limit threshold corresponding to the surface printing area of the latest printed product and the grayscale value lower limit threshold corresponding to the surface printing area of the latest printed product.

9. The packaging box defect detection system based on image analysis according to any one of claims 3 to 7, characterized in that: Using an intelligent defect location model to intelligently analyze the print number of the print object with printing defects in the latest printed product and the defect type identification of the print object with printing defects in the latest printed product based on the total number of printed objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printed product also includes: synchronously inputting the total number of printing objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printed product into the intelligent defect location model to run the intelligent defect location model and obtain the print number of the print object with printing defects in the latest printed product and the defect type identification of the print object with printing defects in the latest printed product output by the intelligent defect location model; Among them, the total number of printing objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printing character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printing patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-image corresponding to the latest printed product are synchronously input into the intelligent defect positioning model to run the intelligent defect positioning model, and obtain the print number of the printing object with printing defects in the latest printed product and the defect type identification of the printing object with printing defects in the latest printed product output by the intelligent defect positioning model, including: setting the total number of printing objects corresponding to the set packaging box, setting the multiple printing character strings of the set packaging box, and inputting the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-image corresponding to the latest printed product into the intelligent defect positioning model. Before the multiple ASCII code values corresponding to the character strings, the multiple JPEG format data corresponding to the multiple printing patterns, and the cyan component value, magenta component value, yellow component value, and black component value of each pixel point of the target sub-image corresponding to the latest printed product are synchronously input into the intelligent defect location model, binary value conversion processing is performed on the total number of printed objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printing character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printing patterns, and the cyan component value, magenta component value, yellow component value, and black component value of each pixel point of the target sub-image corresponding to the latest printed product; Among them, the total number of printed objects corresponding to the set packaging box, the multiple ASCII code values corresponding to the multiple printed character strings of the set packaging box, the multiple JPEG format data corresponding to the multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printed product are synchronously input into the intelligent defect location model to run the intelligent defect location model, and obtain the printing number of the printing object with printing defects in the latest printed product and the defect type identification of the printing object with printing defects in the latest printed product output by the intelligent defect location model. It also includes: the printing number of the printing object with printing defects in the latest printed product and the defect type identification of the printing object with printing defects in the latest printed product output by the intelligent defect location model are both represented in the form of binary values.

10. A packaging box defect detection method based on image analysis, characterized in that: The method comprises: Obtain multiple ASCII code values corresponding to multiple printed character strings and multiple JPEG format data corresponding to multiple printed patterns of the set packaging box; Receiving a high-definition image of a box body of a latest printed product of a set packaging box, identifying a sub-image of a surface printing area of the latest printed product in the high-definition image of the box body as a target sub-image corresponding to the latest printed product; Performing each training action on the convolutional neural network to obtain a convolutional neural network after each training action is performed and outputting it as an intelligent defect location model, accumulating the number of multiple printed character strings and the number of multiple printed patterns of the set packaging box to obtain the total number of printed objects corresponding to the set packaging box, wherein the number of training actions performed by the convolutional neural network is positively correlated with the total number of printed objects corresponding to the set packaging box; An intelligent defect location model is used to intelligently analyze the print numbers of the print objects with printing defects in the latest printed products and the defect type identification of the print objects with printing defects in the latest printed products based on the total number of printed objects corresponding to the set packaging box, multiple ASCII code values corresponding to multiple printed strings of the set packaging box, multiple JPEG format data corresponding to multiple printed patterns, and the cyan component value, magenta component value, yellow component value and black component value of each pixel point of the target sub-screen corresponding to the latest printed product.

Images