Flexible resin plate AI intelligent detection equipment

Abstract

The application discloses a flexible resin plate AI intelligent detection equipment, and relates to the technical field of printing plate detection, which comprises an equipment rack, a feeding mechanism, a detection table and a discharging and sorting mechanism are sequentially arranged on the equipment rack along the material conveying direction, a visual detection unit is arranged above the detection table, a backlight source module is arranged below the detection table, an electrical control cabinet, an AI operation host and a man-machine interaction touch screen are further arranged on the rack; an electrical control unit is arranged in the electrical control cabinet, the man-machine interaction touch screen and the AI operation host are bidirectionally communicated and connected; the electrical control unit is electrically connected with the feeding mechanism, the discharging and sorting mechanism, the visual detection unit, the backlight source module, the AI operation host and the man-machine interaction touch screen and is communicated and linked. The flexible resin plate AI intelligent detection equipment is adopted, the performance is continuously optimized, high-integration electrical control is adopted, stable operation of the equipment is ensured, bidirectional communication and man-machine interaction are adopted, and the production cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of printing plate inspection technology, and in particular to an AI-powered intelligent inspection device for flexible resin plates. Background Technology

[0002] Flexible resin plates are the core consumables for flexographic printing, and are widely used in food packaging, pharmaceutical packaging, label printing, corrugated paper printing and other fields. The quality of the plate directly determines the precision, yield and batch consistency of the printed products.

[0003] Currently, the quality inspection of flexible resin plates in the industry mainly relies on manual visual inspection, which has the following core defects: extremely low inspection efficiency, with skilled quality inspectors taking more than 5 minutes to inspect a single plate, which cannot match the continuous production rhythm of plate-making lines and cannot achieve 24-hour uninterrupted operation; insufficient inspection accuracy, with minor scratches and other defects easily missed, leading to batch printing waste and serious economic losses; inconsistent judgment standards, with different quality inspectors having subjective differences in the classification of defects, making it impossible to achieve standardized quality control and easily causing customer quality disputes; high labor costs, with long training cycles for quality inspectors, and long-term visual inspection easily causing visual fatigue and high occupational disease risks. Summary of the Invention

[0004] The purpose of this invention is to provide a flexible resin plate AI intelligent inspection device that significantly improves inspection efficiency and accuracy, features autonomous iteration for continuous performance optimization, highly integrated electrical control to ensure stable operation, and two-way communication for human-machine interaction, thereby reducing production costs.

[0005] This invention provides an AI intelligent inspection device for flexible resin plates, including a frame. Along the material conveying direction, a feeding mechanism, an inspection table, and an output sorting mechanism are sequentially arranged on the frame. A vision inspection unit is located above the inspection table, and a backlight module is located below it. The frame also includes an electrical control cabinet, an AI computing host, and a human-machine interface touchscreen. The electrical control cabinet contains an electrical control unit, and the human-machine interface touchscreen and AI computing host are bidirectionally connected. The electrical control unit is electrically connected and communicates with the feeding mechanism, output sorting mechanism, vision inspection unit, backlight module, AI computing host, and human-machine interface touchscreen.

[0006] Preferably, the feeding mechanism includes a flattening roller assembly, a photoelectric correction unit, and a vacuum adsorption conveyor belt. The flattening roller assembly consists of two arc-shaped flattening rollers arranged symmetrically at the top and bottom, with the curvature of the arc-shaped flattening rollers perpendicular to the material conveying direction. The surface of the vacuum adsorption conveyor belt is uniformly distributed with micron-sized adsorption pores, and the photoelectric correction unit is located at the discharge end of the flattening roller assembly.

[0007] Preferably, the testing platform adopts a high-transmittance tempered glass platform, and the surface of the testing platform is arrayed with vacuum adsorption holes, which are synchronously controlled with the adsorption force of the vacuum adsorption conveyor belt; the backlight module is a high-brightness LED surface light source.

[0008] Preferably, the visual inspection unit includes a linear CCD camera group, a dual-sided telecentric lens group, a combined light source module, and a line laser contour sensor. The linear CCD camera group and the line laser contour sensor are triggered synchronously. The linear CCD camera group is composed of multiple 8K linear cameras horizontally stitched together. The scanning frequency of the linear CCD camera group and the conveying speed of the vacuum adsorption conveyor belt are synchronously controlled by an electrical control unit. The dual-sided telecentric lens group is set up one-to-one with the linear CCD camera group.

[0009] Preferably, the AI ​​computing host has a built-in AI intelligent detection system, which includes an image preprocessing module, a standard template matching module, a deep learning defect detection module, and a defect grading and judgment module; the AI ​​intelligent detection system also includes a model autonomous iterative optimization module.

[0010] Preferably, the standard template matching module supports importing vector design drafts in AI, CDR, and PDF formats.

[0011] Preferably, the deep learning defect detection module adopts an improved YOLOv8+Transformer dual-branch network architecture.

[0012] Preferably, the material sorting mechanism includes a qualified product conveyor belt, a non-qualified product storage unit, and a servo sorting push rod, which is communicatively connected to the AI ​​computing host.

[0013] Therefore, the present invention adopts the above-mentioned flexible resin plate AI intelligent inspection equipment, which significantly improves inspection efficiency and accuracy, independently iterates and continuously optimizes performance, has highly integrated electrical control to ensure stable operation of the equipment, and features two-way communication human-machine interaction to reduce production costs.

[0014] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall structure of the flexible resin plate AI intelligent detection device of the present invention; Figure 2 This is a schematic diagram of the linkage architecture between the visual inspection unit and the electrical control unit of the flexible resin plate AI intelligent inspection device of the present invention. Figure 3 This is a block diagram of the bidirectional communication architecture between the AI ​​intelligent detection system and the human-computer interaction touch screen of the flexible resin plate AI intelligent detection equipment of the present invention.

[0016] Figure Labels 1. Equipment frame; 2. Feeding mechanism; 21. Flattening roller assembly; 22. Photoelectric correction unit; 23. Vacuum adsorption conveyor belt; 3. Inspection table; 4. Discharge and sorting mechanism; 5. Vision inspection unit; 51. Linear CCD camera assembly; 52. Dual-sided telecentric lens assembly; 53. Combined light source module; 54. Line laser contour sensor; 6. Backlight module; 7. Electrical control cabinet; 71. Electrical control unit; 8. AI computing host; 9. Human-machine interaction touch screen. Detailed Implementation

[0017] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0018] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0019] The terms "first," "second," and similar words used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0020] Example 1 like Figures 1-3 As shown, the flexible resin AI intelligent detection device of the present invention includes a device frame 1. The device frame 1 is made of marble, which has excellent shock absorption performance and ensures imaging stability.

[0021] Along the material conveying direction, the equipment frame 1 is equipped with a feeding mechanism 2, a detection table 3, and a discharge sorting mechanism 4. A vision inspection unit 5 is installed above the detection table 3, and a backlight module 6 is installed below the detection table 3. The frame 1 is also equipped with an electrical control cabinet 7, an AI computing host 8, and a human-machine interaction touch screen 9. The electrical control cabinet 7 contains an electrical control unit, and the human-machine interaction touch screen 9 and the AI ​​computing host 8 are bidirectionally connected.

[0022] The electrical control unit is electrically connected and communicates with the feeding mechanism 2, the discharging sorting mechanism 4, the vision inspection unit 5, the backlight module 6, the AI ​​computing host 8, and the human-machine interaction touch screen 9 to realize the synchronous control of each unit of the equipment.

[0023] The feeding mechanism 2 includes a flattening roller assembly 21, a photoelectric correction unit 22, and a vacuum adsorption conveyor belt 23. The feeding mechanism 2 is used to achieve wrinkle-free, stretch-free, flat conveying and precise positioning of the flexible resin plate. The flattening roller assembly 21 consists of two symmetrically arranged arc-shaped flattening rollers. The roller surface of the arc-shaped flattening rollers is made of anti-static silicone material, and the curvature of the arc-shaped flattening rollers is perpendicular to the material conveying direction. The surface of the vacuum adsorption conveyor belt 23 is uniformly distributed with micron-level adsorption holes, and a vacuum adsorption force is provided by a negative pressure fan. The photoelectric correction unit 22 includes a through-beam photoelectric correction sensor and a servo correction actuator, which detects the edge position of the plate in real time. When a deviation occurs, the servo correction actuator automatically adjusts the angle of the conveyor belt within 0.1s to ensure that the straightness deviation of the plate conveying is ≤0.02mm. The photoelectric correction unit 22 is located at the discharge end of the flattening roller assembly 21. The surface of the vacuum adsorption conveyor belt 23 is uniformly distributed with 30μm micron-level adsorption pores. An adjustable vacuum adsorption force is provided by a silent negative pressure fan to completely and smoothly adhere the flexible resin plate to the surface of the conveyor belt, avoiding warping and stretching deformation of the plate, and adapting to flexible plates of different thicknesses.

[0024] The testing platform 3 uses a high-transmittance tempered glass platform. Vacuum adsorption holes are arrayed on the surface of the testing platform 3, and the adsorption force of the vacuum adsorption holes and the vacuum adsorption conveyor belt 23 is synchronously controlled. The backlight module 6 is a high-brightness LED surface light source, and the brightness can be automatically adjusted by the AI ​​computing host 8.

[0025] The visual inspection unit 5 includes a linear CCD camera group 51, a dual-sided telecentric lens group 52, a combined light source module 53, and a line laser contour sensor 54. The linear CCD camera group 51 and the line laser contour sensor 54 are synchronously triggered to simultaneously acquire full-width two-dimensional images and three-dimensional relief point cloud data of the flexible resin plate under test. The linear CCD camera group 51 is composed of multiple 8K linear cameras horizontally stitched together, with the stitching width adapted to the maximum detection width of the flexible resin plate under test. The scanning frequency of the linear CCD camera group 51 and the conveying speed of the vacuum adsorption conveyor belt 23 are synchronously controlled by the electrical control unit. The dual-sided telecentric lens group 52 is set one-to-one with the linear CCD camera group 51. The combined light source module 53 includes a coaxial incident light source and a ring low-angle light source, which can automatically switch the light source combination and brightness according to the type of plate under test.

[0026] The AI ​​computing host 8 has a built-in AI intelligent detection system, which includes an image preprocessing module, a standard template matching module, a deep learning defect detection module, and a defect grading and judgment module. This system fuses and processes acquired 2D images and 3D point cloud data to identify, classify, and determine the compliance of defects. The AI ​​intelligent detection system also includes a model autonomous iteration and optimization module, which has a manual review interface. Defect data reviewed manually is automatically added to the model training set, and the device can automatically fine-tune and iteratively update the model during idle periods.

[0027] The standard template matching module supports importing vector design drafts in AI, CDR, and PDF formats, automatically generating standard inspection templates, achieving sub-pixel-level registration between the plate to be tested and the standard template, and can automatically identify text and image areas, halftone areas, and blank areas, and set inspection thresholds for each area.

[0028] The deep learning defect detection module adopts an improved YOLOv8+Transformer dual-branch network architecture. The YOLO branch is used for fast defect location, and the Transformer branch is used for extracting subtle defect features. The types of defects that can be identified include pinholes, scratches, dirt spots, image and text defects, dot deformation, overprinting deviation, excessive relief depth, relief edge collapse, and page warping.

[0029] The material sorting mechanism includes a qualified product conveyor belt, a non-qualified product storage unit, and a servo sorting push rod. The servo sorting push rod is connected to the AI ​​computing host and automatically sorts qualified and non-qualified products based on the test results. The human-machine interaction unit is an industrial-grade touch screen all-in-one machine with a built-in data traceability and statistical analysis module. It can automatically generate test reports and supports integration with MES / ERP systems.

[0030] The complete workflow is as follows: Step S1, Pre-test parameter configuration: The operator imports the AI / CDR / PDF format vector design draft of the flexible resin plate to be tested through the human-machine interaction touch screen 9, and transmits the design draft data to the AI ​​computing host 8 through two-way communication. The AI ​​computing host 8 automatically parses the graphic information, generates a standard test template, and automatically identifies graphic areas, dot areas, and blank areas. The operator selects the corresponding plate model on the human-machine interaction touch screen 9, and the system automatically matches the preset test parameters, light source combination, and defect judgment threshold. Custom settings can also be made according to requirements. After configuration, the settings are saved with one click and sent to the electrical control unit 71 and the AI ​​computing host 8. The same type of plate can be directly called in the future.

[0031] S2. Feeding and leveling conveying: The operator issues a start command through the human-machine interaction touch screen 9, and the electrical control unit 71 controls the feeding mechanism 2 to start. The flexible resin plate to be tested is placed on the feeding platform. The plate is conveyed forward with the feeding mechanism 2. First, it is leveled in both directions by the flattening roller group 21 to eliminate wrinkles on the plate surface. Then, it is corrected in real time by the photoelectric correction unit 22 to ensure the straightness of the conveying. Finally, it is flattened and fixed by the vacuum adsorption conveyor belt 23 and conveyed to the testing table 3 without stretching or warping.

[0032] Step S3: Synchronous acquisition of full-dimensional data. After the plate enters the detection area, the electrical control unit 71 synchronously triggers the linear CCD camera group 51 and the line laser contour sensor 54. The linear CCD camera group 51, together with the combined light source module 53 and the backlight module 6, acquires high-definition two-dimensional images of the entire plate. The line laser contour sensor 54 scans synchronously to obtain three-dimensional point cloud data of the entire plate. The acquired images and point cloud data are transmitted to the AI ​​computing host 8 in real time via gigabit Ethernet.

[0033] Step S4, AI Intelligent Detection and Judgment: The AI ​​computing host 8 first uses the image preprocessing module to denoise, correct distortion, normalize grayscale, and stitch images on the original 2D image. It then denoises, filters, and reconstructs the 3D point cloud data to generate a page elevation map, aligning the 2D image and the 3D elevation map at the pixel level. Next, the standard template matching module performs sub-pixel registration between the test page data and the standard template, eliminating detection errors caused by overall page stretching and deformation. Subsequently, the deep learning defect detection module automatically identifies the type, location, size, and quantity of defects. Finally, the defect grading and judgment module classifies defects as fatal / serious / minor based on preset industry standards and user-defined thresholds, automatically determining whether the page is qualified or unqualified.

[0034] Step S5: Result Output and Automatic Sorting. The AI ​​computing host 8 transmits the detection results and judgment instructions back to the human-machine interface touch screen 9 and the electrical control unit 71 in real time. The defect location, size, and image are simultaneously displayed on the human-machine interface touch screen 9, allowing operators to quickly view defect details. Based on the AI ​​judgment results, the electrical control unit 71 controls the discharge sorting mechanism 4 to automatically sort qualified and unqualified products. Qualified products flow into the next process, while unqualified products are automatically collected without manual intervention.

[0035] Step S6, Data Storage and Model Iteration: The system automatically generates a complete inspection report for each plate, stores it in the AI ​​computing host 8, and synchronizes it to the human-machine interaction touch screen 9, achieving full-process traceability. Operators can manually review the AI ​​inspection results through the human-machine interaction touch screen 9. After review and confirmation, the defect data is transmitted back to the AI ​​computing host 8 through two-way communication and automatically added to the model training set. At night or during idle periods, the electrical control unit 71 controls the equipment to enter a low-power mode, and the AI ​​computing host 8 automatically completes the fine-tuning and iterative updates of the model without the need for professional algorithm personnel, continuously improving detection accuracy and reducing the missed detection rate and false detection rate.

[0036] Therefore, the present invention adopts the above-mentioned flexible resin plate AI intelligent inspection equipment, which significantly improves inspection efficiency and accuracy, independently iterates and continuously optimizes performance, has highly integrated electrical control to ensure stable operation of the equipment, and features two-way communication human-machine interaction to reduce production costs.

[0037] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A flexible resin plate AI intelligent inspection device, characterized in that, The equipment includes a frame on which a feeding mechanism, a detection platform, and a discharging and sorting mechanism are sequentially arranged along the material conveying direction. A vision inspection unit is located above the detection platform, and a backlight module is located below it. The frame also houses an electrical control cabinet, an AI computing host, and a human-machine interface touchscreen. The electrical control cabinet contains an electrical control unit, and the human-machine interface touchscreen and AI computing host are bidirectionally connected. The electrical control unit is electrically connected to and communicates with the feeding mechanism, discharging and sorting mechanism, vision inspection unit, backlight module, AI computing host, and human-machine interface touchscreen.

2. The flexible resin plate AI intelligent inspection equipment according to claim 1, characterized in that, The feeding mechanism includes a flattening roller assembly, a photoelectric correction unit, and a vacuum adsorption conveyor belt. The flattening roller assembly consists of two arc-shaped flattening rollers arranged symmetrically at the top and bottom, with the arc of the arc-shaped flattening rollers perpendicular to the material conveying direction. The surface of the vacuum adsorption conveyor belt is uniformly distributed with micron-sized adsorption pores, and the photoelectric correction unit is located at the discharge end of the flattening roller assembly.

3. The flexible resin plate AI intelligent inspection equipment according to claim 1, characterized in that, The testing platform uses a high-transmittance tempered glass platform, and the surface of the testing platform is arrayed with vacuum adsorption holes. The adsorption force of the vacuum adsorption holes and the vacuum adsorption conveyor belt are synchronously controlled; the backlight module is a high-brightness LED surface light source.

4. The flexible resin plate AI intelligent inspection equipment according to claim 1, characterized in that, The visual inspection unit includes a linear CCD camera group, a dual telecentric lens group, a combined light source module, and a line laser contour sensor. The linear CCD camera group and the line laser contour sensor are triggered synchronously. The linear CCD camera group is composed of multiple 8K linear cameras horizontally stitched together. The scanning frequency of the linear CCD camera group and the conveying speed of the vacuum adsorption conveyor belt are synchronously controlled by the electrical control unit. The dual telecentric lens group is set up one-to-one with the linear CCD camera group.

5. The flexible resin plate AI intelligent inspection equipment according to claim 1, characterized in that, The AI ​​computing host has a built-in AI intelligent detection system, which includes an image preprocessing module, a standard template matching module, a deep learning defect detection module, and a defect grading and judgment module; the AI ​​intelligent detection system also includes a model autonomous iterative optimization module.

6. The flexible resin plate AI intelligent inspection equipment according to claim 5, characterized in that, The standard template matching module supports importing vector design drafts in AI, CDR, and PDF formats.

7. The flexible resin plate AI intelligent inspection equipment according to claim 5, characterized in that, The deep learning defect detection module adopts an improved YOLOv8+Transformer dual-branch network architecture.

8. The flexible resin plate AI intelligent inspection equipment according to claim 1, characterized in that, The material sorting mechanism includes a qualified product conveyor belt, a non-qualified product storage unit, and a servo sorting push rod, which is connected to the AI ​​computing host.

Images