Image recognition-based cigarette box printing waste screening and collecting system

Abstract

The application discloses a cigarette case printing waste screening and collecting system based on image recognition, and belongs to the technical field of image recognition and printing waste detection. The system comprises a conveying module, a multispectral polarization imaging module, a main control analysis module, a predictive pneumatic screening module and a waste collecting module. The multispectral polarization imaging module collects polarization and ultraviolet images, the main control analysis module determines waste through polarization reflection suppression and multi-source image fusion, the predictive pneumatic screening module dynamically calculates electromagnetic valve trigger delay and jet pulse width in combination with real-time vehicle speed, air pressure and temperature, and accurately removes waste, and the waste collecting module uses negative pressure air duct to suppress air vortex generated by jet airflow. The application solves the problems of low image recognition accuracy, inaccurate pneumatic screening delay and vortex interference on adjacent qualified products under high-speed and high-reflective surface, and realizes efficient and accurate screening and collecting of cigarette case waste.

Description

Technical Field

[0001] This invention belongs to the field of image recognition and printing waste detection technology, specifically relating to an image recognition-based screening and collection system for cigarette box printing waste. Background Technology

[0002] Cigarette box printing, as a typical example of high-end packaging printing, is characterized by its extremely fast production speed and complex processes. To enhance brand image and anti-counterfeiting features, modern cigarette box printing widely employs various special techniques such as metallic cardboard, holographic hot stamping, embossing, and UV varnishing. Real-time quality inspection and waste collection of printed cigarette boxes on high-speed production lines are crucial for ensuring high pass rates and reducing labor costs.

[0003] However, existing cigarette box printing waste detection systems face serious technical bottlenecks in practical applications. First, due to the extensive use of metal foil and holographic materials on the surface of cigarette boxes, strong specular reflection and light drift are easily generated under high-speed movement, leading to serious false detections and missed detections in traditional image recognition systems. Second, the spacing between cigarette boxes on high-speed conveyor lines (such as those reaching speeds of 600 sheets / minute or more) is extremely small. Traditional screening and collection devices often use fixed delay time control of pneumatic nozzles. However, fluctuations in air pressure in the air circuit system, changes in ambient temperature, and slight vibrations in the conveyor line speed can all cause millisecond-level deviations in the timing of airflow injection, resulting in the phenomenon of "missed blowing" of waste products or "mistaken blowing" of adjacent qualified products. In addition, the violent air vortex generated at the collection port when high-pressure airflow is injected can easily carry adjacent qualified cigarette boxes into the waste bin, causing serious paper waste.

[0004] Therefore, how to solve the problem of accurate image recognition under high-speed, high-reflectivity surfaces, eliminate the response lag of pneumatic actuators and environmental interference, and suppress the influence of airflow eddies on adjacent qualified products, so as to achieve efficient and accurate screening and collection of cigarette box printing waste, is a necessary and urgent technical problem that the cigarette box packaging printing industry needs to solve. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, the present invention provides a cigarette box printing waste screening and collection system based on image recognition, which aims to solve the technical problems of low image recognition accuracy, inaccurate pneumatic screening delay, and airflow vortex interference with adjacent qualified products under high-speed and high-reflectivity surfaces.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A cigarette box printing waste screening and collection system based on image recognition, comprising: The conveyor module is used for high-speed transport of printed cigarette boxes to be inspected; A multispectral polarization imaging module is positioned above the conveying module to simultaneously acquire multi-channel polarization images and ultraviolet excitation images as the cigarette box print passes through. The main control analysis module is electrically connected to the multispectral polarization imaging module. It is used to perform polarization reflection suppression and multi-source image fusion on the acquired images, extract the surface defect features of the cigarette box printing, and determine whether it is a defective product based on the preset defect evaluation logic. The predictive pneumatic screening module, located downstream of the multispectral polarization imaging module, includes a high-frequency solenoid valve, a nozzle, a pressure sensor, and a temperature sensor. It is used to dynamically calculate the pneumatic injection delay time and injection pulse width based on the judgment result of the main control analysis module, and to precisely blow the waste away from the conveying module with the injection airflow. The waste collection module is positioned in the direction of the predictive pneumatic screening module to collect the blown-away waste and to suppress the air vortex generated by the jet airflow using a negative pressure duct.

[0007] In a preferred embodiment of this application, the multispectral polarization imaging module includes a parallel polarization light source, a vertical polarization light source, an ultraviolet light source, and a high-speed industrial camera; the high-speed industrial camera synchronously acquires parallel polarization images, vertical polarization images, and ultraviolet excitation images under alternating pulse triggering from the parallel polarization light source, the vertical polarization light source, and the ultraviolet light source.

[0008] In a preferred embodiment of this application, the main control analysis module performs polarization reflection suppression and multi-source image fusion on the acquired images, and the specific processing steps are as follows: Step 1: Align the parallel polarization image and the vertical polarization image at the pixel level, calculate the degree of polarization of each pixel, in order to suppress specular reflection of the metal foil or holographic paper on the surface of the cigarette box and generate a reflection suppression image; Step 2: Weighted fusion of the reflection suppression image and the ultraviolet excitation image to generate a multispectral fused image; Step 3: Segment the various printing areas of the cigarette box print in the multispectral fusion image, including the hot stamping area, the embossing area, and the regular printing area.

[0009] In a preferred embodiment of this application, the main control analysis module extracts surface defect features of the cigarette box printing and determines whether it is a defective product based on a preset defect assessment logic. The specific processing steps are as follows: Step 1: For a regular printing area, calculate its color deviation from the standard template image in the CIELAB space; Step 2: For the area with uneven indentation, extract its edge contour and calculate the geometric deformation deviation of the contour; Step 3: Extract the texture features of the hot stamping area and calculate the texture anomaly degree; Step 4: Assign different dynamic weighting coefficients to color deviation, geometric deformation deviation, and texture anomaly degree according to different printing areas, and sum them up to obtain the comprehensive defect evaluation score. Step 5: Compare the comprehensive defect assessment score with the preset defect threshold. If it is greater than the defect threshold, the cigarette box print is determined to be a defective product, and the initial moment when it passes through the imaging module is recorded.

[0010] In a preferred embodiment of this application, the predictive pneumatic screening module dynamically calculates the pneumatic injection delay time and injection pulse width, and its specific processing steps are as follows: Step 1: The conveying speed of the conveying module is collected in real time by a high-precision encoder installed on the conveying module; Step 2: Calculate the physical travel time of the waste from the imaging position to the nozzle position based on the physical distance between the multispectral polarization imaging module and the nozzle, and the conveying speed. Step 3: Real-time collection of air supply pressure and ambient temperature in the air circuit system using air pressure and temperature sensors, and calculation of airflow propagation delay time in the pipeline; Step 4: Calculate the actual triggering delay time of the solenoid valve by combining its inherent mechanical response delay time, physical travel time, and airflow propagation delay time. Step 5: Calculate the injection pulse width of the solenoid valve dynamically based on the physical dimensions of the waste, the conveying speed, and the opening transition time of the solenoid valve.

[0011] In a preferred embodiment of this application, the waste collection module includes a collection box, a negative pressure fan, and a porous suction plate. The porous suction plate is located at the entrance of the collection box, and the negative pressure fan is connected to the inside of the collection box to form a negative pressure inside the collection box. This negative pressure is used to adsorb and eliminate the rebound airflow and air vortex generated by the nozzle jet airflow at the entrance of the collection box, preventing adjacent qualified cigarette box printing products from being disturbed by airflow and deviating from the conveying trajectory.

[0012] In a preferred embodiment of this application, the predictive pneumatic screening module is further configured to adaptively adjust the injection pressure of the high-frequency solenoid valve based on the comprehensive defect assessment score of the waste product; if the comprehensive defect assessment score exceeds the defect threshold by a greater than a preset ratio, it is determined that the waste product has severe warping or folding, triggering a high-pressure injection mode; if the exceedance is less than or equal to the preset ratio, the conventional air pressure injection mode is maintained.

[0013] As a preferred embodiment of this application, it also includes an adaptive light source calibration module, which is electrically connected to the main control analysis module. The adaptive light source calibration module is used to acquire an image of the background reference plate during the gap in the passage of the smokeless box of the conveying module, calculate the light source attenuation coefficient, and dynamically adjust the driving current of the parallel polarized light source, the vertical polarized light source, and the ultraviolet light source.

[0014] The beneficial effects of this invention are: (1) High-precision reflection suppression and multi-source fusion: This invention uses a multispectral polarization imaging module to calculate the degree of polarization using parallel and vertical polarization images, which perfectly eliminates the specular reflection interference of metal foil and holographic materials on the surface of cigarette boxes under high-speed movement; at the same time, it fuses ultraviolet excitation images to achieve synchronous high-precision detection of defects in invisible anti-counterfeiting ink and surface physical defects, which significantly reduces the false detection rate and the missed detection rate.

[0015] (2) Predictive Adaptive Pneumatic Control: This invention abandons the traditional fixed-delay injection method. It tracks the conveying speed in real time through an encoder and introduces air pressure and temperature sensors to dynamically compensate for the propagation delay of airflow in the pipeline and the mechanical lag of the solenoid valve. This ensures that the timing of airflow injection reaches microsecond precision under high-frequency and extremely small-pitch conveying conditions, completely eliminating the phenomenon of "missed blowing" and "mistaken blowing" of adjacent qualified products.

[0016] (3) Vortex suppression and energy saving and noise reduction: The waste collection module adopts a negative pressure air duct and a porous suction plate design to form a stable negative pressure barrier at the collection port, which instantly adsorbs and eliminates the rebound airflow and air vortex generated by the high-pressure jet airflow. This not only prevents adjacent qualified products from being carried into the waste bin by the vortex, but also reduces the operating noise of the system, reduces the ineffective consumption of compressed air, and realizes green and environmentally friendly production. Detailed Implementation

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] A waste screening and collection system for cigarette box printing based on image recognition includes a conveying module, a multispectral polarization imaging module, a main control analysis module, a predictive pneumatic screening module, and a waste collection module.

[0019] The conveying module uses a high-speed vacuum adsorption belt to smoothly and quickly transport the printed cigarette boxes to be inspected, ensuring that the cigarette boxes do not shake or shift during high-speed movement.

[0020] The multispectral polarization imaging module, located above the conveying module, includes a parallel polarization line light source, a vertical polarization line light source, an ultraviolet light source, and a high-speed polarization industrial camera. The light source controller controls the three light sources to flash in a high-frequency alternating pulse pattern. The high-speed industrial camera is triggered synchronously with the light source pulses, acquiring parallel polarization, vertical polarization, and ultraviolet excitation images of the same cigarette box print in a very short time.

[0021] The main control and analysis module adopts a GPU+FPGA heterogeneous computing architecture and is electrically connected to the multispectral polarization imaging module. The main control and analysis module performs real-time alignment and noise reduction on the acquired multi-channel images, and executes polarization reflection suppression and multi-source image fusion algorithms to extract the color, contour, and texture features of the cigarette box surface to determine whether it is a defective product.

[0022] The predictive pneumatic screening module, located downstream of the imaging module, includes a high-frequency solenoid valve, nozzles, a pressure sensor, a temperature sensor, and a high-precision rotary encoder. The encoder is mounted on the drive shaft of the conveyor belt to monitor the belt's physical displacement and speed fluctuations in real time. The pressure and temperature sensors are installed in the solenoid valve's inlet line to monitor the physical state of the compressed air in real time.

[0023] The waste collection module is located on the opposite side of the nozzle and includes a collection box, a negative pressure fan, a perforated suction plate, and a waste guide plate. The perforated suction plate is installed at the inlet of the collection box, and the negative pressure fan generates continuous negative pressure inside the box, forming an airflow barrier.

[0024] In this embodiment, the main control analysis module performs polarization reflection suppression and multi-source image fusion on the acquired images. The specific implementation process is as follows: Parallel polarization images are acquired synchronously by a high-speed industrial camera. Vertical polarization image and ultraviolet excitation images Because the metal foil and holographic paper on the surface of the cigarette box have extremely strong specular reflection properties, they can cause local image saturation (white spots). Furthermore, the depolarization effect of polarized light differs between specular and diffuse reflection surfaces. The main control analysis module calculates the polarization degree of each pixel to suppress specular reflection and fuses ultraviolet channel information to generate a fused image. The specific calculation formula is as follows: ; In the formula: For the merged image in pixel coordinates The grayscale value at that location; For parallel polarized images in pixel coordinates The grayscale value at that location; For vertically polarized images in pixel coordinates The grayscale value at that location; For ultraviolet-excited images in pixel coordinates The grayscale value at that point is used to highlight the characteristics of UV anti-counterfeiting ink and fluorescent impurities; This is the weighting coefficient for the polarization reflection suppression term, which is set to 0.65 in this embodiment; The weighting coefficient for the ultraviolet excitation characteristic term is 0.35 in this embodiment; It is a very small positive constant used to prevent the denominator from being zero, and its value is 10. -5 .

[0025] Using the above formula, the first term of the formula... It can extract polarization images that characterize surface roughness and physical defects very well, completely eliminating the specular reflection on the surface of metal cardboard; the second item retains anti-counterfeiting printing information. The fusion of the two provides high-quality, non-reflective image data for subsequent defect detection.

[0026] The main control analysis module acquires the fused image. Then, the conventional printing area, the embossing area, and the hot stamping area are segmented using a preset template matching algorithm, and features are extracted from each area to calculate the comprehensive defect assessment score. The specific calculation formula is as follows: ; In the formula: The overall defect assessment score for printed cigarette boxes; The color deviation value for the regular printing area is obtained by calculating the Euclidean distance between the image to be inspected and the standard template image in CIELAB space; The geometric deformation deviation value of the indentation area is obtained by calculating the Hausdorff distance between the edge contour of the image to be inspected and the contour of the standard template. The texture anomaly degree of the hot stamping area is obtained by calculating the chi-square distance between the image to be inspected and the standard template image on the local binary pattern (LBP) feature spectrum; , , These are the dynamic weighting coefficients for color deviation, geometric deformation deviation, and texture anomaly, respectively, and they satisfy... In this embodiment, for cigarette boxes with large-area hot stamping and embossing processes, the preset weighting coefficients are set to... , , .

[0027] The main control analysis module will calculate the results. Compared with the preset defect threshold Compare. If If the signal is received, the cigarette box is determined to be defective. The main control analysis module immediately sends a defective product trigger signal to the predictive pneumatic screening module and records the initial moment when the defective product passes through the imaging module. .

[0028] Upon receiving a defective product trigger signal, the predictive pneumatic sorting module must dynamically calculate the actual trigger delay time of the solenoid valve to ensure accurate rejection of defective products without interfering with adjacent qualified products during high-frequency conveying. Since the airflow velocity in the pipeline is significantly affected by air pressure and temperature, this invention introduces airflow condition compensation. The actual trigger delay time of the solenoid valve... The calculation formula is as follows: ; Among them, the physical travel time of the waste moving from the imaging position to the nozzle position. Determined by integrating the following formula: ; Propagation delay time of airflow in the pipeline The calculation formula is as follows: ; In the above formulas: The actual trigger delay time of the solenoid valve (in seconds); The physical travel time (in seconds) for the defective part to move from the imaging position to the nozzle position; The initial moment when a scrap product is determined to be a defective product; For the delivery module in The real-time transmission speed (unit: meters per second) is measured in real time by a high-precision encoder; The distance between the physical centers of the multispectral polarization imaging module and the nozzle (unit: meters) is given in this embodiment. rice; The inherent mechanical response delay time (in seconds) of the solenoid valve is used in this embodiment. =0.002 seconds (2 milliseconds); The propagation delay of airflow in the pipeline (unit: seconds); The length of the air pipe between the pneumatic nozzle and the solenoid valve (unit: meters) is given in this embodiment. =0.15 meters; is the gas constant, with a value of 287 J / (kg⋅K); The real-time ambient temperature (unit: Kelvin, K) within the gas path system is measured by a temperature sensor. The real-time air supply pressure (unit: Pascals, Pa) within the air circuit system is measured by a pressure sensor. Standard atmospheric pressure, with a value of 1.013 × 10⁻⁶. 5 Pa; This is the airflow propagation correction factor, which is set to 1.15 in this embodiment; The pressure compensation coefficient is 0.08 in this embodiment.

[0029] Using the above formula, the system can adjust the opening timing of the solenoid valve in microseconds according to real-time belt speed fluctuations, air pressure changes, and ambient temperature, ensuring that the airflow is ejected the instant the air jet reaches the nozzle in front of the center of the waste.

[0030] Meanwhile, to avoid excessively long spraying time interfering with the following qualified products, or excessively short spraying time failing to completely remove defective products, the solenoid valve's spray pulse width... Dynamic adaptive calculation is used, and the formula is as follows: ; In the formula: The pulse width of the solenoid valve (in seconds). The physical length (in meters) of the printed cigarette box along the conveying direction; The safe margin distance for spraying (unit: meters) is taken as 0.015 meters in this embodiment; Predicted conveying speed (in meters per second) when waste reaches the nozzle position; The opening transition time of the solenoid valve (unit: seconds) is taken as 0.001 seconds in this embodiment.

[0031] The above formulas all employ a standardization method to remove dimensions, eliminating the impact of dimensional differences on the calculations. The formulas are based on massive amounts of operational data sampled from the main control chip, and are obtained through parameter fitting optimization using a genetic algorithm. The preset parameters can be adaptively adjusted by technicians according to the actual workshop environment and the weight of the cigarette packs.

[0032] The hardware implementation in this embodiment is based on a high-performance industrial control platform. The main control and analysis module uses an Intel Core i7 processor in conjunction with the NVIDIA Jetson Orin Industrial computing platform, utilizing its integrated TensorRT acceleration engine for inference of image recognition algorithms. The high-speed industrial camera is a CMOS sensor camera with a polarization filter, a resolution of 2048×1084, and a frame rate of 200fps. The high-frequency solenoid valve is a MAC high-frequency solenoid valve with a response frequency of up to 300Hz. The software is developed based on the Linux real-time system kernel (RT-Preempt) to ensure extremely high real-time performance of control signal transmission.

[0033] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A cigarette box printing waste screening and collection system based on image recognition, characterized in that, include: The conveyor module is used for high-speed transport of printed cigarette boxes to be inspected; A multispectral polarization imaging module is positioned above the conveying module to simultaneously acquire multi-channel polarization images and ultraviolet excitation images as the cigarette box print passes through. The main control analysis module is electrically connected to the multispectral polarization imaging module. It is used to perform polarization reflection suppression and multi-source image fusion on the acquired images, extract the surface defect features of the cigarette box printing, and determine whether it is a defective product based on the preset defect evaluation logic. The predictive pneumatic screening module, located downstream of the multispectral polarization imaging module, includes a high-frequency solenoid valve, a nozzle, a pressure sensor, and a temperature sensor. It is used to dynamically calculate the pneumatic injection delay time and injection pulse width based on the judgment result of the main control analysis module, and to precisely blow the waste away from the conveying module with the injection airflow. The waste collection module is positioned in the direction of the predictive pneumatic screening module to collect the blown-away waste and to suppress the air vortex generated by the jet airflow using a negative pressure duct.

2. The image recognition-based waste screening and collection system for cigarette box printing as described in claim 1, characterized in that, The multispectral polarization imaging module includes a parallel polarization light source, a vertical polarization light source, an ultraviolet light source, and a high-speed industrial camera; the high-speed industrial camera synchronously acquires parallel polarization images, vertical polarization images, and ultraviolet excitation images under the alternating pulse triggering of the parallel polarization light source, the vertical polarization light source, and the ultraviolet light source.

3. The image recognition-based waste screening and collection system for cigarette box printing as described in claim 1, characterized in that, The main control analysis module performs polarization reflection suppression and multi-source image fusion on the acquired images. The specific steps include: Step 1: Align the parallel polarization image and the vertical polarization image at the pixel level, calculate the degree of polarization of each pixel, in order to suppress specular reflection of the metal foil or holographic paper on the surface of the cigarette box and generate a reflection suppression image; Step 2: Weighted fusion of the reflection suppression image and the ultraviolet excitation image to generate a multispectral fused image; Step 3: Segment the various printing areas of the cigarette box print in the multispectral fusion image, including the hot stamping area, the embossing area, and the regular printing area.

4. The image recognition-based waste screening and collection system for cigarette box printing as described in claim 3, characterized in that, The main control analysis module extracts surface defect features of the printed cigarette boxes and determines whether they are defective based on preset defect assessment logic. The specific steps include: Step 1: For a regular printing area, calculate its color deviation from the standard template image in the CIELAB space; Step 2: For the area with uneven indentation, extract its edge contour and calculate the geometric deformation deviation of the contour; Step 3: Extract the texture features of the hot stamping area and calculate the texture anomaly degree; Step 4: Assign different dynamic weighting coefficients to color deviation, geometric deformation deviation, and texture anomaly degree according to different printing areas, and sum them up to obtain the comprehensive defect evaluation score. Step 5: Compare the comprehensive defect assessment score with the preset defect threshold. If it is greater than the defect threshold, the cigarette box print is determined to be a defective product, and the initial moment when it passes through the imaging module is recorded.

5. The image recognition-based waste screening and collection system for cigarette box printing as described in claim 1, characterized in that, The predictive pneumatic screening module dynamically calculates the pneumatic injection delay time and injection pulse width, and its specific steps include: Step 1: The conveying speed of the conveying module is collected in real time by a high-precision encoder installed on the conveying module; Step 2: Calculate the physical travel time of the waste from the imaging position to the nozzle position based on the physical distance between the multispectral polarization imaging module and the nozzle, and the conveying speed. Step 3: Real-time collection of air supply pressure and ambient temperature in the air circuit system using air pressure and temperature sensors, and calculation of airflow propagation delay time in the pipeline; Step 4: Calculate the actual triggering delay time of the solenoid valve by combining its inherent mechanical response delay time, physical travel time, and airflow propagation delay time. Step 5: Calculate the injection pulse width of the solenoid valve dynamically based on the physical dimensions of the waste, the conveying speed, and the opening transition time of the solenoid valve.

6. The image recognition-based waste screening and collection system for cigarette box printing as described in claim 1, characterized in that, The waste collection module includes a collection box, a negative pressure fan, and a porous suction plate. The porous suction plate is located at the entrance of the collection box, and the negative pressure fan is connected to the inside of the collection box to form a negative pressure inside the collection box. This negative pressure is used to adsorb and eliminate the rebound airflow and air vortex generated by the nozzle jet airflow at the entrance of the collection box, preventing adjacent qualified cigarette box printing products from being disturbed by airflow and deviating from the conveying trajectory.

7. A cigarette box printing waste screening and collection system based on image recognition according to claim 5, characterized in that, The predictive pneumatic screening module is also used to adaptively adjust the injection pressure of the high-frequency solenoid valve based on the comprehensive defect assessment score of the waste product; if the comprehensive defect assessment score exceeds the defect threshold by a greater than a preset ratio, it is determined that the waste product has a serious warping or folding, triggering the high-pressure injection mode. If the deviation is less than or equal to the preset ratio, the normal air pressure injection mode will be maintained.

8. A cigarette box printing waste screening and collection system based on image recognition according to claim 1, characterized in that, The system also includes an adaptive light source calibration module, which is electrically connected to the main control analysis module. The adaptive light source calibration module is used to acquire images of the background reference board during the gap in the passage of the smokeless box of the conveying module, calculate the light source attenuation coefficient, and dynamically adjust the driving current of the parallel polarization light source, the vertical polarization light source, and the ultraviolet light source.