Detection system and method for packaging box production and processing

Abstract

The invention discloses a detection system and method for packaging box production and processing, and belongs to the technical field of packaging production quality detection. The invention discloses a detection system and method for packaging box production and processing. The system comprises a dual-polarization-angle imaging module, a structured light three-dimensional reconstruction module, a temperature compensation module, a defect judgment module and a sorting execution module. According to the method, dual-polarization images and three-dimensional point clouds are synchronously collected, qualified standard values are dynamically calibrated in combination with temperature, accurate recognition and grading judgment of surface and internal defects of the transparent plastic packaging boxes are achieved, and finally unqualified products are removed through a high-speed sorting mechanism. The method solves the problem of misjudgment caused by reflective interference and temperature drift of the transparent material, and is suitable for online detection of a high-speed production line.

Description

Technical Field

[0001] The present invention belongs to the technical field of packaging production quality inspection, and in particular relates to an inspection system and method for packaging box production and processing. Background Art

[0002] In the packaging industry, the quality of transparent plastic packaging directly affects the product's appearance, transportation safety, and user experience. With the prevalence of automated production lines, efficient and accurate defect detection has become a key link in ensuring product qualification rates. Unqualified packaging may have defects such as surface scratches, internal bubbles, and structural wrinkles. If these defects enter the market, they will damage brand reputation, lead to customer complaints, and cause financial losses. Current traditional packaging inspection methods mostly rely on manual visual inspection or single-vision inspection technology, which has the following limitations: low inspection efficiency, manual inspection cannot adapt to high-speed production lines, and is easily affected by subjective factors; insufficient defect recognition capabilities, single two-dimensional vision cannot distinguish between surface mirror reflection interference and true defects, and cannot effectively detect three-dimensional defects such as internal bubbles and three-dimensional structural deformation; poor environmental adaptability, not considering the impact of temperature changes on material size, and significant defect measurement errors under different ambient temperatures; low intelligence level, lack of dynamic calibration mechanism, and detection accuracy easily degrades after long-term use, making it difficult to meet the inspection needs of packaging boxes made of diverse materials.

[0003] Therefore, the research and development of detection systems with high precision, real-time performance and environmental adaptability has important industrial application value. Summary of the Invention

[0004] In response to the above-mentioned pain points, the present invention provides a packaging box production and processing detection system and method that integrates dual-polarization angle imaging, structured light three-dimensional reconstruction, temperature compensation, multi-feature fusion judgment and high-speed sorting functions. It can accurately identify the surface and internal defects of transparent plastic packaging boxes in real time, and dynamically calibrate the detection standards according to the temperature to automatically reject unqualified products.

[0005] The scheme of the present invention is as follows:

[0006] A packaging box production and processing inspection system, characterized by including a dual-polarization angle imaging module, a structured light 3D reconstruction module, a temperature compensation module, a defect determination module, and a sorting execution module:

[0007] The dual-polarization angle imaging module is configured with an industrial camera for synchronously capturing the specular reflection image and the diffuse reflection image of the surface of the transparent plastic packaging box;

[0008] The structured light 3D reconstruction module integrates a fringe projector and a 3D point cloud solution unit to generate a 3D height map of the packaging box surface;

[0009] The temperature compensation module has a built-in temperature sensor and a dynamic calibration model; the temperature sensor collects the ambient temperature T in real time; the dynamic calibration model is based on the material preset reference temperature T0, the reference qualified standard value S0 and the material temperature coefficient α, through the formula:

[0010] S=S0×[1+α×(T-T0)]

[0011] Calculate the real-time qualified standard value S to eliminate the interference of temperature changes on defect size measurement; T-T0 is the temperature difference between the real-time temperature and the preset reference temperature. The real-time qualified standard value S is used for qualification comparison of the defect judgment module.

[0012] The defect determination module is in communication with the dual-polarization angle imaging module, the structured light 3D reconstruction module, and the temperature compensation module, and is used to extract the specular reflection features in the dual-polarization image and the defect geometric parameters in the 3D height map, compare them with the real-time qualified standard value S, and determine whether the packaging box is qualified;

[0013] The sorting execution module is in communication with the defect determination module and is configured to reject unqualified packaging boxes based on the determination results.

[0014] Preferably, in the dual-polarization angle imaging module, there are at least two industrial cameras, each of which is provided with a 0° and a 45° polarization filter to capture images of corresponding polarization angles.

[0015] Preferably, the polarization filter of the dual-polarization angle imaging module adopts an electrically adjustable phase retarder, which dynamically adjusts the polarization direction by controlling the voltage, and the adjustment range is 0° to 90° to adapt to the differences in reflective properties of transparent packaging boxes made of different materials.

[0016] Preferably, the dual-polarization angle imaging module is connected to the structured light 3D reconstruction module through a synchronous trigger circuit, and the synchronous trigger circuit uses a hardware timer to achieve microsecond synchronization, ensuring that when the conveyor belt linear speed is ≤5m / s, the single frame processing time is ≤50ms and the spatiotemporal error is ≤0.5mm.

[0017] Preferably, the temperature sensor is a PT100 platinum resistance sensor with a measurement accuracy of ±0.1°C and a response time of ≤20ms;

[0018] The dynamic calibration model establishes a material temperature coefficient α database based on historical test data, with an adjustment step of 0.0001 / °C, and supports input of the material preset reference temperature T0 and the reference qualified standard value S0 through the human-computer interaction interface.

[0019] Preferably, the defect determination module adopts a multi-feature fusion decision-making mechanism, including:

[0020] The feature weight allocation unit adjusts the weight coefficients of the dual-polarization image features and the three-dimensional height map features for different types of defects. When a surface defect is determined, the weight coefficient of the dual-polarization image features ranges from 0.7 to 1.0, and the weight coefficient of the three-dimensional height map features ranges from 0.0 to 0.3. When an internal defect is determined, the weight coefficient of the three-dimensional height map features ranges from 0.6 to 1.0, and the weight coefficient of the dual-polarization image features ranges from 0.0 to 0.4. For composite surface and internal defects, the weight is allocated according to the proportion of the two-dimensional projected area of ​​the defect on the packaging box surface. The surface defect weight = surface defect projected area / (surface defect projected area + projected area of ​​the corresponding area of ​​the internal defect), and the default value is not less than 0.3. The sum of the two weights is 1.0.

[0021] The hierarchical threshold judgment unit sets three-level judgment thresholds: warning threshold, qualified threshold and unqualified threshold. Among them: when the defect characteristic value is between the warning threshold and the qualified threshold, a manual review instruction is generated and sent to the manual review terminal; when the defect characteristic value exceeds the unqualified threshold, a rejection instruction is generated and sent to the sorting execution module.

[0022] Preferably, the defect judgment module includes a polarization feature processing unit and a three-dimensional feature extraction unit: the polarization feature processing unit calculates the phase difference of the dual polarization image and identifies the mirror reflection area; the three-dimensional feature extraction unit extracts bubbles with a depth greater than 0.2 mm or wrinkles with a height change greater than 0.1 mm from the three-dimensional height map; the processing results are input into the threshold comparison module as defect geometric parameters.

[0023] Preferably, it also includes an intelligent calibration module with a built-in standard defect sample, which is communicated with the dual-polarization angle imaging module and the structured light three-dimensional reconstruction module, and automatically generates a polarization compensation coefficient and a three-dimensional reconstruction error correction table every day, wherein the polarization compensation coefficient is input into the dual-polarization angle imaging module, and the three-dimensional reconstruction error correction table is input into the structured light three-dimensional reconstruction module, and the detection accuracy is periodically calibrated.

[0024] Preferably, the response time of the high-speed solenoid valve array of the sorting execution module is ≤30ms, which matches the single-frame processing time of the detection module, ensuring that unqualified products are removed in real time within the linear speed range of the conveyor belt.

[0025] A method for detecting packaging box production and processing, characterized by comprising the following steps:

[0026] S1. Synchronous polarization image acquisition: An industrial camera equipped with 0° and 45° polarization filters is used to synchronously capture dual-polarization angle images of the surface of the transparent plastic packaging box.

[0027] S2. 3D point cloud reconstruction: Project Gray code stripes and collect deformed stripe images. Calculate the 3D coordinates of the packaging box surface based on the phase shift method to generate a 3D height map with an accuracy of ≤ 0.1 mm.

[0028] S3. Defect feature extraction: Calculate the phase difference of the dual-polarization image, extract the mirror reflection area, and combine it with the 3D height map to identify bubble defects with a depth greater than 0.2 mm or wrinkle defects with a height change greater than 0.1 mm to obtain the defect geometric parameters;

[0029] S4. Dynamic standard calibration: Obtain the real-time ambient temperature T through the temperature sensor; input the preset reference temperature T0 and the reference qualified standard value S0 of the material through the human-computer interaction interface, retrieve the material temperature coefficient α from the database, and calculate the real-time qualified standard value S using the formula S = S0 × [1 + α × (T-T0)]; compare the defect geometric parameters with S to complete the qualification judgment;

[0030] S5. Sorting execution: Control the sorting execution module to remove unqualified packaging boxes based on the judgment results.

[0031] Compared with the prior art, the advantages of the present invention are:

[0032] (1) In the present invention, a dual-polarization angle imaging module is used to synchronously capture specular reflection and diffuse reflection images, and combined with structured light 3D reconstruction to generate a 0.1mm precision 3D height map. This can accurately identify different types of defects such as surface scratches and internal bubbles, solving the problem that traditional 2D vision cannot distinguish 3D defects.

[0033] (2) In the present invention, a temperature compensation module is used to collect the ambient temperature in real time through a PT100 sensor. Combined with the material temperature coefficient database, a dynamic calibration model is used to calibrate the qualified standard value, eliminating the interference of temperature changes on defect size measurement and ensuring the consistency of detection results under different environments;

[0034] (3) In the present invention, the defect judgment module adopts a multi-feature fusion decision-making mechanism to dynamically adjust the weight coefficients of the dual-polarization image and the three-dimensional height map according to the defect type, and sets three-level thresholds of warning, qualified, and unqualified to achieve accurate classification and processing of defects and reduce manual review costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a structural diagram of a detection system for packaging box production and processing.

[0036] Figure 2 The figure is a flow chart of a testing method for packaging box production and processing. DETAILED DESCRIPTION

[0037] The technical solutions of the embodiments of the present invention will be explained and described below, but the following embodiments are only preferred embodiments of the present invention and are not exhaustive. Based on the embodiments in the implementation manner, other embodiments obtained by those skilled in the art without creative work are all within the scope of protection of the present invention.

[0038] Example 1

[0039] 1. Application Scenario: This embodiment is applied to a production line for 1.2mm thick transparent PET plastic packaging boxes. The conveyor belt speed of the production line is stable at 4.2m / s. The detection system is installed in a metal frame 1.2m above the conveyor belt. The PLC is used to synchronize the control of each module. The overall detection process takes ≤45ms, meeting the inspection requirement of 3,600 pieces per hour. The system modules achieve microsecond-level coordinated operation through synchronized trigger circuits.

[0040] 2. System-related module configuration implementation

[0041] 1. Dual polarization angle imaging module:

[0042] 1-1. Two industrial cameras are used with a resolution of 2592×1944, a frame rate of 200fps, and a pixel depth of 12 bits. Each camera is equipped with an electrically adjustable phase-delay polarization filter with an adjustment range of 0° to 90°, a voltage control signal of 0-5V, and an adjustment step of 0.5°.

[0043] 0° polarization camera: collects specular reflection images when 0V voltage is applied, with an exposure time of 2ms;

[0044] 45° polarization camera: acquires diffuse reflection images when a 2.5V voltage is applied, and the polarization direction is dynamically adjusted through a PWM signal with a frequency of 100-500Hz;

[0045] 1-2. Equipped with a 12mm focal length lens, the field of view covers an area of ​​200mm×150mm. When the conveyor belt linear speed is 4.2m / s, the physical displacement corresponding to a single frame image is: displacement = 4.2m / s × 2ms = 8.4mm.

[0046] 2. Structured light 3D reconstruction module:

[0047] 2-1. 3D modeling hardware: A fringe projector projects Gray code fringes with a wavelength of 850nm, which is used in conjunction with the aforementioned industrial camera to collect deformed fringes. The 3D point cloud solution unit uses a processor with a main frequency of 2.3GHz and 16 cores. It uses a five-step phase shift method for phase solution, with a solution accuracy of ±0.05mm. At a working distance of 100mm, the point cloud spacing is 0.07mm, and the Z-axis accuracy is 0.08mm.

[0048] 2-2. Defect recognition capability: It can identify bubble defects with a diameter of ≥0.8mm and a depth of >0.2mm, or wrinkle defects with a height change of >0.1mm.

[0049] 3. Temperature compensation module:

[0050] 3-1. Temperature sensor: Four platinum resistance sensors are used, with a measurement accuracy of ±0.05°C, a response time of 12ms, a 4-wire connection method, connected to an analog module, and a 16-bit resolution;

[0051] 3-2. Dynamic calibration model verification:

[0052] 3-2-1. α value calibration experiment:

[0053] Scenario 1: Benchmark parameters: T0 = 25°C, S0 = 0.25 mm, PET sample;

[0054] The measured defect depth at 30°C is 0.252mm. According to the formula:

[0055] S=S0×[1+α×(T-T0)]

[0056] It is concluded that α = 0.0008 / °C;

[0057] Scenario 2: For the same PET sample, the measured defect depth at 10°C is 0.247mm. Substituting this into the formula yields α = 0.0008 / °C.

[0058] Scenario 3: For the same PET sample, the measured defect depth at 40°C is 0.254mm. Substituting this into the formula yields α = 0.0008 / °C.

[0059] 3-2-2. Model full temperature verification

[0060] Temperature scenario T(℃) S calculated value (mm) Measured defect depth (mm) Error rate high temperature 40 0.258 0.259 0.4％ Benchmark 25 0.250 0.250 0％ Low temperature 10 0.247 0.246 0.4％ Extremely low temperatures 0 0.246 0.245 0.4％

[0061] 3-2-4. Dynamic response experiment:

[0062] The heating plate raises the sample temperature from 25°C to 30°C within 10ms, the temperature sensor outputs a stable temperature value within 15ms, and the dynamic calibration model completes the calculation of S within 20ms. The false detection rate of 100 samples is 0 when continuously tested.

[0063] 4. Defect determination module

[0064] 4-1. Multi-feature fusion decision-making mechanism:

[0065] Surface defects: Dual polarization image feature weight 0.75, 3D height map feature weight 0.25, judgment conditions are dual polarization image phase difference ≥π / 3 and 3D height change >0.1mm;

[0066] Internal defects: 3D height map feature weight is 0.8, dual polarization image feature weight is 0.2, and the judgment condition is that the defect diameter in the 3D height map is greater than 1.2mm and the depth is greater than 0.2mm;

[0067] Composite defect weight calculation: If the surface defect projection area is 0.8mm 2 The projected area of ​​the internal defect is 1.2mm 2 , then the surface defect weight is:

[0068]

[0069] 4-2. Classification threshold setting (taking S = 0.251 mm as an example):

[0070] Warning threshold: 1.1×S=0.276mm, triggering manual review instructions;

[0071] Unqualified threshold: 1.3×S=0.326mm, generating a rejection instruction.

[0072] Sorting execution module:

[0073] 5-1. Using a solenoid valve array, the response time is 22ms; the driving cylinder has a stroke of 100mm and a thrust of 50N. The distance from the camera to the rejection port is 0.5m. The rejection delay time is calculated as follows:

[0074]

[0075] 6. Intelligent calibration module:

[0076] 6-1. Standard defect sample parameters:

[0077] Built-in standard defect templates include:

[0078] Wrinkle defect: depth 0.1mm, length 5mm, width 0.3mm;

[0079] Bubble defect: diameter 0.8mm, depth 0.25mm;

[0080] Scratch defect: length 1.5mm, depth 0.12mm;

[0081] 6-2. Automated calibration process: Polarization compensation: Automatically capture images of the standard sample at 3:00 a.m. every day. When the reflectivity deviation of the 0° polarization filter is greater than 5%, a compensation coefficient K1 = 1.03 is generated and written to the camera register via the SPI bus;

[0082] 3D reconstruction error correction: Compare the standard point cloud with the measured point cloud data. If the Z-axis error is greater than 0.06mm, generate the following correction table:

[0083]

[0084] Real-time calibration mechanism: When the detection error of 100 samples is continuously detected to be greater than 5%, the calibration process is automatically started.

[0085] 3. Implementation of the testing method process

[0086] S1. Synchronous polarization image acquisition: The 0° and 45° polarization cameras synchronously acquire images at a frame rate of 150 fps, with an exposure time of 2 ms and a conveyor speed of 4.2 m / s. The physical length corresponding to a single frame image is:

[0087]

[0088] S2. 3D point cloud reconstruction: The four-step phase shift method is used to solve the 3D coordinates of the packaging box surface. The generated 3D height map has an accuracy of 0.08mm and can identify height changes of 0.18mm.

[0089] S3. Defect feature extraction: Calculate the phase difference of the dual-polarization image and identify a bubble defect with a diameter of 1.8 mm and a depth of 0.22 mm. The depth value of the 3D height map is 0.22 mm, which is greater than the 0.2 mm threshold. The phase difference of the corresponding area in the dual-polarization image is π / 2, which is greater than the π / 4 threshold.

[0090] S4. Dynamic standard calibration: real-time ambient temperature T = 28°C, preset reference temperature T0 = 25°C, material temperature coefficient α = 0.0008 / °C, reference qualified standard value S0 = 0.3mm, calculate the real-time qualified standard value:

[0091] S=0.3×[1+0.0008×(28-25)]=0.30072mm

[0092] Since the defect depth is 0.22mm≤S, it is judged to be qualified;

[0093] S5. Sorting Execution: For unqualified packaging boxes, the sorting execution module triggers the solenoid valve within 22ms, and the cylinder push rod removes them to the waste area within 100ms. The system can process 3,600 packaging boxes per hour. When the conveyor speed is 4.2m / s, the rejection delay matches the action time.

[0094] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A detection system for packaging box production and processing, characterized in that: Including dual polarization angle imaging module, structured light 3D reconstruction module, temperature compensation module, defect judgment module, and sorting execution module: The dual-polarization angle imaging module is configured with an industrial camera for synchronously capturing the specular reflection image and the diffuse reflection image of the surface of the transparent plastic packaging box; The structured light 3D reconstruction module integrates a fringe projector and a 3D point cloud solution unit to generate a 3D height map of the packaging box surface; The temperature compensation module has a built-in temperature sensor and a dynamic calibration model; the temperature sensor collects the ambient temperature T in real time; the dynamic calibration model is based on the material preset reference temperature T0, the reference qualified standard value S0 and the material temperature coefficient α, through the formula: S=S0×[1+α×(T-T0)] Calculate the real-time qualified standard value S to eliminate the interference of temperature changes on defect size measurement; T-T0 is the temperature difference between the real-time temperature and the preset reference temperature. The real-time qualified standard value S is used for qualification comparison of the defect judgment module. The defect determination module is in communication with the dual-polarization angle imaging module, the structured light 3D reconstruction module, and the temperature compensation module, and is used to extract the specular reflection features in the dual-polarization image and the defect geometric parameters in the 3D height map, compare them with the real-time qualified standard value S, and determine whether the packaging box is qualified; The sorting execution module is in communication with the defect determination module and is configured to reject unqualified packaging boxes based on the determination results.

2. A packaging box production and processing detection system according to claim 1, characterized in that: In the dual-polarization angle imaging module, there are at least two industrial cameras, which are respectively provided with 0° and 45° polarization filters to capture images with corresponding polarization angles.

3. A packaging box production and processing detection system according to claim 2, characterized in that: The polarization filter of the dual-polarization angle imaging module adopts an electrically adjustable phase retarder, which dynamically adjusts the polarization direction by controlling the voltage. The adjustment range is 0° to 90° to adapt to the differences in reflective properties of transparent packaging boxes made of different materials.

4. A packaging box production and processing detection system according to claim 1, characterized in that: The dual-polarization angle imaging module and the structured light 3D reconstruction module are connected via a synchronization trigger circuit. The synchronization trigger circuit uses a hardware timer to achieve microsecond-level synchronization, ensuring that when the conveyor belt linear speed is ≤5m / s, the single-frame processing time is ≤50ms and the spatiotemporal error is ≤0.5mm.

5. A packaging box production and processing detection system according to claim 4, characterized in that: The temperature sensor is a PT100 platinum resistance sensor with a measurement accuracy of ±0.1°C and a response time of ≤20ms; The dynamic calibration model establishes a material temperature coefficient α database based on historical test data, with an adjustment step of 0.0001 / °C, and supports input of the material preset reference temperature T0 and the reference qualified standard value S0 through the human-computer interaction interface.

6. A packaging box production and processing detection system according to claim 1, characterized in that: The defect determination module adopts a multi-feature fusion decision-making mechanism, including: The feature weight allocation unit adjusts the weight coefficients of the dual-polarization image features and the three-dimensional height map features for different types of defects. When a surface defect is determined, the weight coefficient of the dual-polarization image features ranges from 0.7 to 1.0, and the weight coefficient of the three-dimensional height map features ranges from 0.0 to 0.

3. When an internal defect is determined, the weight coefficient of the three-dimensional height map features ranges from 0.6 to 1.0, and the weight coefficient of the dual-polarization image features ranges from 0.0 to 0.

4. For composite surface and internal defects, the weight is allocated according to the proportion of the two-dimensional projected area of ​​the defect on the packaging box surface. The surface defect weight = surface defect projected area / (surface defect projected area + projected area of ​​the corresponding area of ​​the internal defect), and the default value is not less than 0.

3. The sum of the two weights is 1.

0. The hierarchical threshold judgment unit sets three-level judgment thresholds: warning threshold, qualified threshold and unqualified threshold. Among them: when the defect characteristic value is between the warning threshold and the qualified threshold, a manual review instruction is generated and sent to the manual review terminal; when the defect characteristic value exceeds the unqualified threshold, a rejection instruction is generated and sent to the sorting execution module.

7. A packaging box production and processing detection system according to claim 1, characterized in that: The defect determination module includes a polarization feature processing unit and a three-dimensional feature extraction unit: the polarization feature processing unit calculates the phase difference of the dual-polarization image and identifies the mirror reflection area; the three-dimensional feature extraction unit extracts bubbles with a depth greater than 0.2mm or wrinkles with a height change greater than 0.1mm from the three-dimensional height map; the processing results are input into the threshold comparison module as defect geometric parameters.

8. The detection system for packaging box production and processing according to claim 1, characterized in that: It also includes an intelligent calibration module with a built-in standard defect sample, which communicates with the dual-polarization angle imaging module and the structured light 3D reconstruction module. It automatically generates polarization compensation coefficients and 3D reconstruction error correction tables every day. The polarization compensation coefficients are input into the dual-polarization angle imaging module, and the 3D reconstruction error correction table is input into the structured light 3D reconstruction module to periodically calibrate the detection accuracy.

9. The detection system for packaging box production and processing according to claim 1, characterized in that: The high-speed solenoid valve array response time of the sorting execution module is ≤30ms, which matches the single-frame processing time of the detection module, ensuring that unqualified products are removed in real time within the linear speed range of the conveyor belt.

10. A detection method for packaging box production and processing, characterized in that: The following steps are involved: S1. Synchronous polarization image acquisition: An industrial camera equipped with 0° and 45° polarization filters is used to synchronously capture dual-polarization angle images of the surface of the transparent plastic packaging box. S2. 3D point cloud reconstruction: Project Gray code stripes and collect deformed stripe images. Calculate the 3D coordinates of the packaging box surface based on the phase shift method to generate a 3D height map with an accuracy of ≤ 0.1 mm. S3. Defect feature extraction: Calculate the phase difference of the dual-polarization image, extract the mirror reflection area, and combine it with the 3D height map to identify bubble defects with a depth greater than 0.2 mm or wrinkle defects with a height change greater than 0.1 mm to obtain the defect geometric parameters; S4. Dynamic standard calibration: obtain the real-time ambient temperature T through the temperature sensor; input the preset reference temperature T0 and the reference qualified standard value S0 of the material through the human-computer interaction interface, retrieve the material temperature coefficient α from the database, and calculate the real-time qualified standard value S through the formula S = S0 × [1 + α × (T-T0)]; Compare the defect geometric parameters with S to complete the qualification judgment; S5. Sorting execution: Control the sorting execution module to remove unqualified packaging boxes based on the judgment results.

Images