Soft package sealing performance detection method and system based on image recognition

Abstract

The invention discloses a soft package leakproofness detection method and system based on image recognition, and relates to the related field of soft package leakproofness detection.The method comprises the steps that a to-be-detected soft package is placed in a vacuumizing cavity, first-stage infrared irradiation is executed under the first pressure condition, and initial image data are collected; performing negative pressure treatment on the vacuumized cavity to a set pressure threshold value, executing second-stage infrared irradiation under a second pressure condition, and collecting deformation image data; performing registration alignment on the initial image data and the deformed image data, extracting image feature difference data, and calculating a deformation vector of the sealed area according to the image feature difference data; and constructing a soft package surface deformation distribution diagram according to the deformation vector to perform sealing judgment, and determining a package leakage detection result. The technical problem of low efficiency and accuracy of the existing soft package sealing performance detection is solved, and the technical effect of improving the sealing performance detection efficiency and accuracy is achieved.

Description

Technical Field

[0001] The present application relates to the field of soft package sealing detection, and in particular to a soft package sealing detection method and system based on image recognition. Background Art

[0002] Testing the sealability of soft packages is crucial for ensuring product quality and safety, especially in areas such as food, pharmaceuticals, and electronics. Poor sealability can lead to product deterioration, contamination, or damage, impacting consumer health and product lifespan. Currently, testing the sealability of soft packages primarily relies on traditional methods such as pressure testing, vacuum testing, and manual visual inspection. These methods are susceptible to human interference, have low detection efficiency, and struggle to accurately quantify subtle deformations in the seal area, resulting in inaccurate and unreliable test results.

[0003] In the current related technologies, soft package sealing detection has technical problems of low efficiency and accuracy. Summary of the Invention

[0004] The present application provides a soft package sealing detection method and system based on image recognition. By placing the soft package into a vacuum chamber, first collecting an initial infrared image under normal pressure, then vacuuming to a set negative pressure, collecting the deformed infrared image, comparing the two images, calculating the deformation of the sealing area, determining the sealing according to the deformation distribution diagram, and obtaining the leakage detection results, the application achieves the technical effect of improving the efficiency and accuracy of sealing detection.

[0005] The present application provides a soft package sealing detection method based on image recognition, comprising: placing the soft package to be tested in a vacuum chamber, performing a first stage of infrared irradiation under a first pressure condition, and collecting initial image data; subjecting the vacuum chamber to negative pressure treatment to a set pressure threshold, performing a second stage of infrared irradiation under a second pressure condition, and collecting deformed image data; aligning the initial image data with the deformed image data, extracting image feature difference data, and calculating a deformation vector of the sealing area based on the image feature difference data; constructing a soft package surface deformation distribution map based on the deformation vector to perform sealing judgment and determine a leak detection result.

[0006] In a possible implementation, the following processing is performed: the acquisition process of the first-stage infrared irradiation and the second-stage infrared irradiation includes: arranging multiple infrared sensor arrays in the vacuum chamber, setting the irradiation intensity of the infrared light source according to the characteristics of the soft package material, activating the multiple infrared sensor arrays to collect data according to the irradiation intensity, and obtaining a first reflection intensity distribution map; adding the first reflection intensity distribution map to the initial image data; after the negative pressure treatment is completed, dynamically adjusting the irradiation intensity of the second-stage infrared light source based on the first reflection intensity distribution map, activating the multiple infrared sensor arrays to collect data according to the irradiation intensity, and obtaining a second reflection intensity distribution map; adding the second reflection intensity distribution map to the deformed image data.

[0007] In a possible implementation, the initial image data and the deformed image data are registered and aligned, image feature difference data is extracted, and the following processing is performed: based on the first reflection intensity distribution map, key feature points in the initial image data are extracted to establish a first feature point set; based on the second reflection intensity distribution map, edge enhancement processing is performed on the deformed image data to extract a second feature point set; feature point matching is performed based on the first feature point set and the second feature point set to determine a two-stage infrared image pair, and the two-stage infrared image pair includes a spatial transformation matrix; affine transformation is performed on the initial image data and the deformed image data according to the spatial transformation matrix to determine the image feature difference data.

[0008] In a possible implementation, the deformation vector of the sealing area is calculated based on the image feature difference data, and the following processing is performed: historical leakage sample data is retrieved, the historical leakage sample data is traversed to perform image deformation annotation, and the deformation abnormality area label of the sealing area is obtained; convolution calculation is performed on the two-stage infrared image pair according to the deformation abnormality area label to construct a pixel-level deformation probability map; deformation calculation is performed on the sealing area according to the pixel-level deformation probability map to generate a deformation vector of the sealing area.

[0009] In a possible implementation, a soft package surface deformation distribution map is constructed based on the deformation vector to perform sealing judgment, determine the leakage detection result, and perform the following processing: construct a soft package surface coordinate system, map the deformation vector to the soft package surface coordinate system, and perform normalization processing in combination with the soft package material properties to construct the soft package surface deformation distribution map; traverse the soft package surface deformation distribution map to extract data, perform regional segmentation based on multiple deformation data sets, and generate a set of suspected leakage areas; perform morphological optimization on the set of suspected leakage areas, perform deformation analysis based on the optimized morphological data, and obtain a deformation vector amplitude; use the deformation vector amplitude as an index to search and match the historical leakage sample data to generate a leakage matching degree, and add the leakage matching degree to the leakage detection result.

[0010] In a possible implementation, the deformation vector amplitude is used as an index to retrieve and match the historical packet leakage sample data to generate a packet leakage matching degree, which is added to the packet leakage detection result, and the following processing is performed: the soft package surface deformation distribution map is grid-segmented to obtain a number of detection units; the deformation vector amplitude is combined with the detection units to calculate and generate a multidimensional feature vector, the multidimensional feature vector is analyzed and reduced in dimension to generate a deformation amplitude feature sequence; the deformation amplitude feature sequence is used as an index to traverse the historical packet leakage sample database to perform similarity retrieval and generate a packet leakage matching degree; the packet leakage matching degree is compared with a dynamic threshold, the leakage risk level is determined based on the comparison result, and the packet leakage matching degree and the leakage risk level are integrated into the packet leakage detection result.

[0011] In a possible implementation, the leaked package matching degree is compared with a dynamic threshold, and the following processing is performed: a query is performed based on the soft package material characteristics in combination with the detected ambient humidity to obtain a reference matching degree threshold table; a real-time pressure change rate is introduced to perform linear interpolation compensation on the reference matching degree threshold table to generate a dynamic threshold; when the leaked package matching degree exceeds the dynamic threshold, a high-confidence leakage alarm is triggered, and the comparison result is generated.

[0012] The present application also provides a soft package sealing detection system based on image recognition, including: an initial image data acquisition module, used to place the soft package to be tested into a vacuum cavity, perform a first stage of infrared irradiation under a first pressure condition, and collect initial image data; a deformation image data acquisition module, used to perform negative pressure treatment on the vacuum cavity to a set pressure threshold, perform a second stage of infrared irradiation under a second pressure condition, and collect deformation image data; an image feature difference data extraction module, used to align the initial image data with the deformed image data, extract image feature difference data, and calculate the deformation vector of the sealing area according to the image feature difference data; a sealing judgment module, used to construct a soft package surface deformation distribution map according to the deformation vector to perform sealing judgment and determine the leakage detection result.

[0013] The soft package sealing detection method and system based on image recognition proposed in this application first places the soft package to be tested in a vacuum chamber, performs a first stage of infrared irradiation under a first pressure condition, collects initial image data, then negatively pressurizes the vacuum chamber to a set pressure threshold, performs a second stage of infrared irradiation under a second pressure condition, collects deformed image data, then aligns the initial image data with the deformed image data, extracts image feature difference data, calculates the deformation vector of the sealing area based on the image feature difference data, and finally constructs a soft package surface deformation distribution map based on the deformation vector to determine the sealing and determine the leak detection result. This achieves the technical effect of improving the efficiency and accuracy of sealing detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention are briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in precise order. Instead, various steps may be processed in reverse order or simultaneously as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0015] Figure 1 A schematic flow chart of a method for detecting the sealing properties of a soft package based on image recognition provided in an embodiment of the present application.

[0016] Figure 2 This is a schematic diagram of the structure of a soft package sealing detection system based on image recognition provided in an embodiment of the present application.

[0017] Description of reference numerals: initial image data acquisition module 10 , deformed image data acquisition module 20 , image feature difference data extraction module 30 , sealing determination module 40 . DETAILED DESCRIPTION

[0018] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0019] In order to make the purpose, technical solutions and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limiting this application. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0020] In the following description, reference is made to “some embodiments”, which describes a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict, and the terms “first\second” involved are merely used to distinguish similar objects and do not represent a specific ordering of the objects. The terms “including” and “having” and any variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or that are inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs. The terms used herein are for the purpose of describing the embodiments of this application only.

[0021] The present application provides a method for detecting the sealing performance of a soft package based on image recognition. Figure 1 As shown, the method includes: In step S100 , the soft package to be tested is placed in a vacuum chamber, and a first stage of infrared irradiation is performed under a first pressure condition to collect initial image data.

[0022] Specifically, the soft package to be tested is placed in a sealed, evacuated chamber. This chamber is a sealed container used to vacuum the contents and is equipped with a pressure sensor and vacuum pump to control and monitor the pressure within the chamber. Under a first pressure condition (e.g., normal pressure or a slight negative pressure), the soft package is irradiated with an infrared irradiation device. The infrared device can be an infrared lamp or an infrared emitter, with its wavelength and power selected based on the characteristics of the soft package material. A high-resolution industrial camera is used to capture initial image data of the soft package. Camera parameters (such as exposure time and resolution) are optimized based on the size and material of the soft package.

[0023] For example, consider a food packaging bag placed in a vacuum chamber with a diameter of 50 cm and a height of 30 cm. A pressure sensor within the chamber detects a pressure of 1 atmosphere (normal pressure). An infrared lamp with a wavelength of 850 nm and a power of 50 watts is then activated, illuminating the surface of the bag for 10 seconds. An initial image of the bag is then captured using an industrial camera with a resolution of 2048 × 1536 pixels and an exposure time of 1 / 100 second.

[0024] In step S200 , the vacuum chamber is subjected to negative pressure treatment to a set pressure threshold, and a second stage of infrared irradiation is performed under a second pressure condition to collect deformation image data.

[0025] Specifically, a vacuum pump is used to apply negative pressure to the evacuated chamber, reducing the pressure within the chamber to a set pressure threshold (e.g., -0.5 atmospheres). A pressure sensor monitors pressure changes in real time to ensure that the set value is reached. Under a second pressure condition (negative pressure), the soft package is again irradiated with infrared light. The same industrial camera as in step S100 is used to capture image data of the soft package's deformation under negative pressure.

[0026] For example, after step S100 is completed, the vacuum pump is activated to reduce the pressure in the chamber from 1 atmosphere to -0.5 atmospheres. When the pressure sensor detects that the pressure has reached the set value, the infrared lamp is activated again, irradiating the soft bag with the same parameters (wavelength 850 nanometers, power 50 watts) for 10 seconds. Subsequently, an industrial camera is used with the same parameters (resolution 2048 × 1536 pixels, exposure time 1 / 100 second) to capture an image of the soft bag's deformation under negative pressure.

[0027] In one possible implementation, the acquisition process of the first-stage infrared irradiation and the second-stage infrared irradiation includes: arranging multiple infrared sensor arrays in the vacuum chamber, setting the irradiation intensity of the infrared light source according to the characteristics of the soft package material, activating the multiple infrared sensor arrays to collect data according to the irradiation intensity, and obtaining a first reflection intensity distribution map; adding the first reflection intensity distribution map to the initial image data; after the negative pressure treatment is completed, dynamically adjusting the irradiation intensity of the second-stage infrared light source based on the first reflection intensity distribution map, activating the multiple infrared sensor arrays to collect data according to the irradiation intensity, and obtaining a second reflection intensity distribution map; and adding the second reflection intensity distribution map to the deformed image data.

[0028] Specifically, multiple infrared sensor arrays are evenly arranged within the vacuum chamber. These sensor arrays are capable of detecting the reflected intensity of infrared light and converting it into electrical signals. The sensor arrays can be arranged in a grid or ring pattern to ensure comprehensive coverage of the soft package surface. For example, within a vacuum chamber with a diameter of 50 cm and a height of 30 cm, the height of the chamber is divided into five equal sections, each 6 cm high. A ring array is arranged at each equal section, with 10 sensors evenly spaced within each ring array. The angular spacing between the sensors is 36 degrees (360 degrees / 10 sensors). These sensors are capable of detecting the reflected intensity of infrared light with a wavelength range of 800 to 1000 nanometers.

[0029] The infrared light source's initial wavelength, intensity, and angle (the angle between the infrared light source and the surface of the soft package) are determined based on the soft package material's characteristics, such as its infrared absorptivity and reflectivity. These parameters can be predetermined through experimentation to ensure optimal image acquisition. For example, for a common plastic soft package material, the infrared light source's initial wavelength is set at 850 nanometers, the intensity at 50 watts, and the angle at 45 degrees. These parameters ensure that the infrared light effectively penetrates the soft package material and produces a clear reflection signal on the surface.

[0030] The infrared sensor array is activated and collects data according to the set illumination intensity and parameters. The sensor array converts the detected infrared reflection intensity into a digital signal and generates a first reflection intensity distribution map. This first reflection intensity distribution map shows the reflection intensity of different areas on the soft package surface. This data is added to the initial image data for subsequent image analysis.

[0031] After negative pressure treatment, the soft package deforms, causing changes in its surface optical properties. To ensure consistent spectral conditions during deformation image acquisition, the infrared light source's intensity, wavelength, and angle parameters must be dynamically adjusted. This adjustment improves the accuracy of image registration and feature extraction, leading to more precise calculation of deformation vectors.

[0032] For example, during the first phase of capturing initial image data, the soft package is at normal pressure. The infrared light source is set to 50 watts, a wavelength of 850 nanometers, and an illumination angle of 45 degrees. After the negative pressure treatment, the soft package deforms, and its surface optical properties (e.g., reflectivity, absorptivity, etc.) change. To ensure consistent spectral conditions during the second phase of capturing deformed images, the infrared light source parameters are dynamically adjusted based on a comparison of the first and second reflection intensity distributions. Suppose the first reflection intensity distribution shows that certain areas of the soft package have a specific distribution of reflection intensity at normal pressure. After the negative pressure treatment, the second reflection intensity distribution shows that the reflection intensity in these areas has changed significantly. This is likely due to changes in optical properties caused by surface deformation. Therefore, in the second phase, the infrared light source's illumination intensity is adjusted to 60 watts, the wavelength to 880 nanometers, and the illumination angle to 50 degrees. These adjustments compensate for the changes in optical properties caused by the soft package's deformation and ensure consistent spectral conditions during image capture under different pressure conditions. The infrared sensor array is then activated again to capture a second reflection intensity distribution map. This generated second reflection intensity distribution map shows the changes in reflection intensity of the soft package under negative pressure. This data is added to the deformation image data for subsequent image analysis. This implementation method dynamically adjusts the parameters of the infrared light source to compensate for the changes in optical properties caused by the soft package deformation, ensuring that the spectral conditions of image acquisition remain consistent under different pressure conditions. This improves the accuracy of image registration and feature extraction, allowing for more accurate calculation of deformation vectors and ultimately achieving precise detection of the soft package's sealing properties.

[0033] Step S300 : performing registration and alignment on the initial image data and the deformed image data, extracting image feature difference data, and calculating a deformation vector of a sealing area according to the image feature difference data.

[0034] Specifically, a computer vision algorithm (such as a feature point-based registration algorithm) is used to register and align the initial image and the deformed image. For example, the SIFT (Scale Invariant Feature Transform) algorithm can be used to extract feature points in the image, and the two images are aligned using these feature points. Feature difference data is extracted from the image, such as grayscale changes, edge intensity changes, etc. For example, an edge detection algorithm (such as the Canny algorithm) can be used to extract image edges, and then the difference between the two images is calculated. Combined with a preset deep learning model (such as a convolutional neural network, CNN), the extracted feature difference data is analyzed to calculate the deformation vector of the sealing area (a vector representing the degree of deformation of the object surface, used to describe the deformation of the soft bag under different pressure conditions). The deep learning model learns the deformation pattern of the soft bag under different pressure conditions through training data.

[0035] For example, use the SIFT algorithm to extract 100 feature points from each of the original and deformed images, and align the two images using these feature points. Then, use the Canny algorithm to extract image edges and calculate the difference in edge strength between the two images. This difference data is fed into a pretrained CNN model, which outputs a deformation vector for the sealed area. Assume that the deformation vector is represented as a two-dimensional array, where each element corresponds to the degree of deformation of a pixel on the soft bag surface.

[0036] In one possible implementation, the initial image data and the deformed image data are registered and aligned to extract image feature difference data. Step S300 further includes step S310 of extracting key feature points from the initial image data based on the first reflection intensity distribution map to establish a first feature point set. Specifically, key feature points are extracted from the first reflection intensity distribution map using a feature point detection algorithm in computer vision, such as Scale Invariant Feature Transform (SIFT), Speeded Up Robust Features (SURF), or Oriented Speeded Up Robust Features (ORB). These algorithms can detect significant feature points in an image, such as corners and edges. A descriptor is generated for each detected feature point. A descriptor is a vector that describes the image texture information surrounding the feature point. For example, the SIFT algorithm is used to extract key feature points from the first reflection intensity distribution map. The SIFT algorithm detects extreme points in a multi-scale space and finds stable feature points. A 128-dimensional descriptor is generated for each feature point. Assuming that 100 key feature points are extracted from the first reflection intensity distribution map, these feature points and their descriptors constitute the first feature point set.

[0037] Step S320, edge enhancement processing is performed on the deformed image data based on the second reflection intensity distribution map, and a second feature point set is extracted. Specifically, an edge enhancement algorithm, such as the Canny edge detection algorithm or the Sobel operator, is used to perform edge enhancement processing on the second reflection intensity distribution map. These algorithms can highlight the edge information in the image and make the feature points more obvious. On the edge-enhanced image, a feature point detection algorithm (such as SIFT, SURF or ORB) is used again to extract key feature points and generate descriptors. For example, the Canny edge detection algorithm is used to perform edge enhancement processing on the second reflection intensity distribution map. The Canny algorithm detects edges in the image through gradient calculation and non-maximum suppression. Then, the SIFT algorithm is used to extract key feature points from the edge-enhanced image. Assume that 120 key feature points are extracted from the edge-enhanced image, and these feature points and their descriptors constitute the second feature point set.

[0038] Step S330 determines a two-stage infrared image pair based on feature point matching performed on the first feature point set and the second feature point set. The two-stage infrared image pair includes a spatial transformation matrix. Specifically, a feature point matching algorithm, such as nearest neighbor matching (NN) or bidirectional nearest neighbor matching (NN+NNR), is used to match feature points in the first feature point set with feature points in the second feature point set. The matching algorithm compares the feature point descriptors to find the feature point pairs with the highest similarity. Using the matched feature point pairs, the spatial transformation matrix is calculated using the least squares method or other optimization algorithm. The spatial transformation matrix describes the geometric relationship between the two images. For example, the bidirectional nearest neighbor matching algorithm is used to match 100 feature points in the first feature point set with 120 feature points in the second feature point set. The matching algorithm finds matching feature point pairs by calculating the Euclidean distance between the descriptors. Assuming 80 feature point pairs are matched, the spatial transformation matrix is calculated using these matching point pairs using the least squares method. The spatial transformation matrix is a 3×3 matrix that describes the geometric transformation relationship between the two images.

[0039] Step S340: Perform an affine transformation on the initial image data and the deformed image data according to the spatial transformation matrix to determine the image feature difference data. Specifically, an affine transformation is performed on the initial image data and the deformed image data using the spatial transformation matrix. An affine transformation is a geometric transformation that can eliminate image offset errors caused by pressure changes and spatially align the two images. A pixel-by-pixel difference calculation is performed on the aligned images to obtain image feature difference data. The difference data can be expressed as the grayscale value difference or eigenvector difference between corresponding pixels in the two images. For example, an affine transformation is performed on the initial image data using the calculated spatial transformation matrix to spatially align it with the deformed image data. Then, a pixel-by-pixel difference calculation is performed on the two aligned images. Assuming that the resolution of the initial image and the deformed image is both 2048×1536 pixels, the grayscale value difference between the two images is calculated pixel by pixel to obtain a 2048×1536 difference image. This difference image reflects the deformation characteristics of the soft bag under different pressure conditions. This implementation method achieves precise alignment of the initial image and the deformed image and accurate extraction of feature difference data through feature point extraction, edge enhancement, feature point matching and affine transformation, thereby improving the reliability and accuracy of soft package sealing detection.

[0040] In one possible implementation, the deformation vector of the sealed area is calculated based on the image feature difference data. Step S300 further includes step S350, which retrieves historical packet leakage sample data, traverses the historical packet leakage sample data, and annotates image deformations to obtain labels for abnormal deformation regions in the sealed area. Specifically, historical packet leakage sample data is retrieved from a database, including images of known packets and their corresponding deformation information. Image annotation tools (such as LabelImg or VGG Image Annotator) are used to annotate the historical packet leakage sample data, including the location and extent of abnormal deformation regions. Machine learning or deep learning algorithms (such as convolutional neural networks (CNNs)) are then used to automatically annotate the historical packet leakage sample data to generate labels for abnormal deformation regions. For example, 1,000 images of known packets and their corresponding deformation information are retrieved from a database. These images are manually annotated using the LabelImg tool to identify the abnormal deformation regions in each packet leakage image. Simultaneously, these images are automatically annotated using a pretrained CNN model to generate labels for abnormal deformation regions. Assume that the annotation results show that there are obvious deformation anomalies in the edge areas and central areas of some images, and these areas are marked as deformation anomaly area labels.

[0041] Step S360 performs a convolution calculation on the two-stage infrared image pair based on the abnormal deformation region labels to construct a pixel-level deformation probability map. Specifically, a pre-trained CNN model is used to perform the convolution calculation on the two-stage infrared image pair. The CNN model extracts features from the image and calculates the deformation probability of each pixel through convolutional layers, pooling layers, and fully connected layers. A pixel-level deformation probability map is generated based on the output of the CNN model. The deformation probability value of each pixel indicates the probability of deformation at that pixel. For example, a pre-trained CNN model (such as U-Net) is used to perform the convolution calculation on the two-stage infrared image pair. Assuming the input resolution of the two-stage infrared image pair is 2048×1536 pixels, the CNN model outputs a pixel-level deformation probability map of the same resolution. In the deformation probability map, each pixel value ranges from 0 to 1, indicating the probability of deformation at that pixel. For example, a deformation probability value of 0.9 for some pixels indicates a high probability of deformation at these pixels.

[0042] Step S370, performs deformation calculation on the sealing area according to the pixel-level deformation probability map, and generates a deformation vector of the sealing area. Specifically, based on the pixel-level deformation probability map, a deformation calculation algorithm (such as optical flow method, block matching method, etc.) is used to perform deformation calculation on the sealing area. These algorithms calculate the displacement vector of each pixel by analyzing the pixel-level deformation probability map. The calculated pixel displacement vectors are summarized to generate a deformation vector of the sealing area. The deformation vector represents the overall deformation of the soft bag under different pressure conditions. For example, the optical flow method is used to perform deformation calculation on the pixel-level deformation probability map. Assuming that the deformation probability values of certain areas in the deformation probability map are higher, the optical flow method calculates the pixel displacement vectors of these areas. These displacement vectors are summarized to generate a deformation vector of the sealing area. Assume that the deformation vector shows that the maximum deformation of the soft bag under negative pressure conditions occurs in the central area, and the degree of deformation is 10%. This implementation method achieves precise analysis of soft package deformation and generation of deformation vectors by retrieving historical leaked package sample data, performing image annotation, convolution calculations, and deformation calculations. This can improve the accuracy and reliability of detection and help more accurately determine whether the soft package has problems such as leaks.

[0043] Step S400: constructing a soft package surface deformation distribution map according to the deformation vector to perform sealing determination and determine a leak detection result.

[0044] Specifically, based on the calculated deformation vector, a deformation distribution map of the soft package surface is constructed. A heat map can be used to represent areas with a larger degree of deformation in red and areas with a smaller degree of deformation in blue. Sealing judgment is made based on the preset deformation threshold rule. The deformation threshold rule is a pre-set rule for judging the sealing of the soft package. When the degree of deformation exceeds the threshold, the soft package is judged to have an air leak. For example, if the degree of deformation in a certain area exceeds the preset threshold (such as 10%), it is judged that there may be an air leak in the area. For example, the heat map constructed based on the deformation vector shows that the degree of deformation in most areas of the soft package surface is less than 5% (blue area), but in one corner of the soft package, the degree of deformation reaches 15% (red area). Based on the preset deformation threshold (10%), it is judged that there is an air leak in the corner area, and the soft package is ultimately determined to be a leaking package. The embodiment of the present application adopts technical means such as placing the soft package into a vacuum chamber, first collecting an initial infrared image under normal pressure, then vacuuming to a set negative pressure, collecting an infrared image after deformation, comparing the two images, calculating the deformation of the sealing area, determining the sealing according to the deformation distribution diagram, and obtaining the leakage detection result, thereby achieving the technical effect of improving the efficiency and accuracy of sealing detection.

[0045] In one possible implementation, a soft package surface deformation distribution map is constructed based on the deformation vectors to determine sealing and leak detection results. Step S400 further includes step S410: constructing a soft package surface coordinate system, mapping the deformation vectors to the soft package surface coordinate system, and normalizing them based on the soft package material properties to construct the soft package surface deformation distribution map. Specifically, a two-dimensional or three-dimensional coordinate system is established on the soft package surface to accurately locate the deformation vectors. The calculated deformation vectors are mapped to the soft package surface coordinate system, with each deformation vector corresponding to a point on the soft package surface. The deformation vectors are normalized based on the soft package material properties (such as elastic modulus and Poisson's ratio) to make the deformation data comparable. The normalized deformation vector data is visualized to construct a soft package surface deformation distribution map, representing the degree of deformation as a heat map.

[0046] For example, assume the soft package is a rectangular package 20 cm long and 10 cm wide. Establish a two-dimensional coordinate system on the soft package surface, with the origin at the lower left corner. Map the calculated deformation vectors to this coordinate system, with the origin of each deformation vector corresponding to a point on the soft package surface. Normalize the deformation vectors based on the elastic modulus (assumed to be 1.5 GPa) and Poisson's ratio (assumed to be 0.3) of the soft package material. Visualize the normalized deformation vector data to generate a heat map of the soft package surface deformation distribution, where red indicates a higher degree of deformation and blue indicates a lower degree of deformation.

[0047] Step S420, traverse the deformation distribution map of the soft package surface to extract data, perform regional segmentation based on multiple deformation data sets, and generate a set of suspected leakage areas. Specifically, traverse the deformation distribution map of the soft package surface to extract the deformation data of each pixel. Use an image segmentation algorithm (such as threshold segmentation, region growing, dam algorithm, etc.) to perform regional segmentation on the deformation distribution map to identify areas with a higher degree of deformation. Mark the segmented high deformation areas as suspected leakage areas to generate a set of suspected leakage areas. For example, use a threshold segmentation algorithm to perform regional segmentation on the deformation distribution map of the soft package surface. Assuming that the set threshold is 0.5 (normalized deformation degree), all areas with a deformation degree greater than 0.5 are identified as suspected leakage areas. By traversing the deformation distribution map, three areas with a higher degree of deformation are extracted, and these areas are marked as suspected leakage areas to generate a set of suspected leakage areas.

[0048] Step S430, morphological optimization is performed on the set of suspected leakage areas, deformation analysis is performed based on the optimized morphological data, and deformation vector amplitude is obtained. Specifically, morphological image processing techniques (such as dilation, erosion, opening operation, closing operation, etc.) are used to optimize the suspected leakage areas to remove noise and small interference areas. Deformation analysis is performed on the optimized suspected leakage areas, and the deformation vector amplitude of each area is calculated. For example, morphological optimization is performed on the set of suspected leakage areas, closing operation is used to remove small holes in the area, and opening operation is used to smooth the boundaries of the area. After optimization, the morphology of each suspected leakage area is clearer. Deformation analysis is performed on the optimized areas, and the deformation vector amplitude of each area is calculated. Assume that the deformation vector amplitude of the first suspected leakage area after optimization is 0.8, that of the second area is 0.6, and that of the third area is 0.7.

[0049] In step S440, the deformation vector amplitude is used as an index to search and match the historical leaked packet sample data, generate a leaked packet matching degree, and add the leaked packet matching degree to the leaked packet detection result. Specifically, a feature-based search and matching algorithm (such as nearest neighbor matching or cosine similarity matching) is used to match the deformation vector amplitude with the historical leaked packet sample data. Based on the matching results, a leaked packet matching degree is generated, indicating the degree of similarity between the current soft package and the historical leaked packet sample. The leaked packet matching degree is added to the leaked packet detection result to provide a reference for the final sealing determination. For example, the optimized deformation vector amplitudes (0.8, 0.6, and 0.7) of the suspected leak area are used as indexes to search and match the historical leaked packet sample data. Assume that in the historical leaked packet sample data, the sample with a deformation vector amplitude of 0.8 has a matching degree of 0.9, the sample with a deformation vector amplitude of 0.6 has a matching degree of 0.7, and the sample with a deformation vector amplitude of 0.7 has a matching degree of 0.8. These matching degrees are integrated, and the highest matching degree of 0.9 is taken as the final result. This implementation method achieves accurate detection of soft package sealing by constructing a deformation distribution map, identifying suspected leakage areas, optimizing the leakage area morphology, quantifying the deformation degree, and retrieving matching historical sample data.

[0050] In one possible implementation, the deformation vector amplitude is used as an index to search and match the historical missed packet sample data, generate a missed packet matching degree, and add the missed packet matching degree to the missed packet detection result. Step S440 further includes step S441, gridding and segmenting the soft package surface deformation distribution map to obtain a plurality of detection units. Specifically, image segmentation technology is used to divide the soft package surface deformation distribution map into a plurality of grid units of equal size. Each grid unit serves as an independent detection unit for subsequent deformation analysis. According to a preset grid size, the deformation distribution map is segmented into multiple detection units, each detection unit containing a certain number of pixels. For example, assuming that the resolution of the soft package surface deformation distribution map is 2048×1536 pixels, the image is segmented into grid units of 128×128 pixels, and each grid unit serves as a detection unit. In this way, the entire deformation distribution map is segmented into 16×12 detection units.

[0051] In step S442, the deformation vector amplitudes are combined with the calculations for the plurality of detection units to generate a multidimensional feature vector. The multidimensional feature vector is then analyzed and reduced in dimensionality to generate a deformation amplitude feature sequence. Specifically, for each detection unit, the average amplitude, maximum amplitude, and amplitude standard deviation of the deformation vector are calculated. These statistics can reflect the deformation characteristics within the detection unit. The calculated statistics are combined into a multidimensional feature vector. A dimensionality reduction algorithm (such as principal component analysis (PCA) or t-SNE) is used to analyze and reduce the multidimensional feature vector to generate a deformation amplitude feature sequence. The reduced feature sequence can be more effectively used for subsequent similarity retrieval. For example, for each detection unit, the average amplitude of the deformation vector is calculated to be 0.5, the maximum amplitude is 0.8, and the amplitude standard deviation is 0.2. These values are combined into a three-dimensional feature vector (0.5, 0.8, 0.2). PCA dimensionality reduction is performed on the multidimensional feature vectors of all detection units to generate a two-dimensional deformation amplitude feature sequence. Assume that the feature sequence after dimensionality reduction is (0.6, 0.7).

[0052] Step S443 uses the deformation amplitude feature sequence as an index to traverse the historical packet leakage sample database for similarity retrieval and generate a packet leakage matching degree. Specifically, a feature-based similarity retrieval algorithm (such as nearest neighbor matching, cosine similarity, etc.) is used to match the deformation amplitude feature sequence of the current detection unit with the feature sequence in the historical packet leakage sample database. Based on the matching results, the packet leakage matching degree of each detection unit is calculated. The higher the matching degree, the more similar the current detection unit is to the historical packet leakage sample, and the higher the possibility of packet leakage. For example, assume that there are 1000 samples in the historical packet leakage sample database, each with a deformation amplitude feature sequence. Using the cosine similarity algorithm, the deformation amplitude feature sequence (0.6, 0.7) of the current detection unit is matched with the historical samples. Assume that the matching degree of the sample with the highest matching degree is 0.85, and this matching degree is used as the packet leakage matching degree of the current detection unit.

[0053] In step S444, the leaked packet matching degree is compared with the dynamic threshold, the leakage risk level is determined based on the comparison result, and the leaked packet matching degree and the leakage risk level are integrated into the leaked packet detection result. Specifically, a dynamic threshold is set based on historical data and detection requirements. The dynamic threshold can be adjusted according to the detection environment and the characteristics of the soft package material. The leaked packet matching degree is compared with the dynamic threshold, and the leakage risk level is determined based on the comparison result. For example, a detection unit with a matching degree exceeding the threshold is judged as high risk, a detection unit with a matching degree below the threshold but close to the threshold is judged as medium risk, and a detection unit with a matching degree far below the threshold is judged as low risk. The leaked packet matching degree and the leakage risk level are integrated into the leaked packet detection result, and a detailed detection report is output. For example, assume that the dynamic threshold is set to 0.8. A detection unit with a matching degree of 0.85 is judged as high risk; a detection unit with a matching degree of 0.75 is judged as medium risk; and a detection unit with a matching degree of 0.6 is judged as low risk. These matching degrees and risk levels are integrated into the leak detection results, and a detailed inspection report is output, including the leak location, matching degree, and risk level. This implementation method achieves refined detection and risk assessment of soft package surface deformation distribution maps through grid segmentation, feature extraction and dimensionality reduction, similarity retrieval, and risk assessment.

[0054] In one possible implementation, the missing package matching degree is compared with a dynamic threshold. Step S444 further includes step S4441, where a query is performed based on the soft package material properties and the test environment humidity to obtain a benchmark matching degree threshold table. Specifically, a database containing different soft package material properties (such as elastic modulus, Poisson's ratio, material thickness, etc.) is established. A humidity sensor is installed in the test environment to monitor the ambient humidity during testing in real time. Based on the soft package material properties and the test environment humidity, a corresponding benchmark matching degree threshold table is queried from the database. This table stores matching degree thresholds for different materials and humidity conditions. For example, assuming the soft package material is polyethylene, with an elastic modulus of 1.5 GPa, a Poisson's ratio of 0.35, and a material thickness of 0.1 mm, and the test environment humidity is 60% RH. By querying the material property database, a benchmark matching degree threshold table for polyethylene material under 60% RH is found, which shows a benchmark matching degree threshold of 0.75.

[0055] Step S4442, introduce the real-time pressure change rate to perform linear interpolation compensation on the benchmark matching threshold table to generate a dynamic threshold. Specifically, a pressure sensor is installed in the vacuum chamber to monitor pressure changes in real time. Calculate the real-time pressure change rate, that is, the amount of pressure change per unit time. According to the pressure change rate, perform linear interpolation compensation on the benchmark matching threshold to generate a dynamic threshold. The compensation formula can be expressed as: dynamic threshold = benchmark matching threshold + k × pressure change rate, where k is the compensation coefficient, which is determined according to experimental data. For example, assuming that the benchmark matching threshold is 0.75, the real-time pressure change rate is 0.1kPa / s, and the compensation coefficient k is 0.05. According to the compensation formula: dynamic threshold = 0.75 + 0.05 × 0.1 = 0.755, therefore, the generated dynamic threshold is 0.755.

[0056] Step S4443, when the leaked package matching degree exceeds the dynamic threshold, a high-confidence leakage alarm is triggered and the comparison result is generated. Specifically, the leaked package matching degree is compared with the dynamic threshold to determine whether it exceeds the threshold. If the leaked package matching degree exceeds the dynamic threshold, a high-confidence leakage alarm is triggered. On the basis of the high-confidence leakage alarm, through further comprehensive judgment, a detailed report is generated to confirm the specific location, matching degree and risk level of the leaked package. The comparison result generation steps are as follows: According to historical data, the average deformation vector amplitude of the normal sealed soft package under negative pressure conditions is statistically analyzed and set as the first judgment threshold. According to the deformation distribution characteristics of the leakage sample, the maximum curvature parameter of the abnormal deformation area is extracted and set as the second judgment threshold. If there is a continuous area in the deformation distribution diagram that exceeds the first judgment threshold and the second judgment threshold at the same time, it is determined to be a leaked package.

[0057] For example, assume the leaked package matching degree is 0.8 and the dynamic threshold is 0.755. Since 0.8 > 0.755, a high-confidence leak alarm is triggered. Further analysis of the deformation distribution map reveals that the average deformation vector amplitude of a normally sealed soft package under negative pressure conditions is 0.5 (the first judgment threshold). The maximum curvature parameter of the abnormally deformed area is 0.9 (the second judgment threshold). In the deformation distribution map, a continuous area is found with a deformation vector amplitude of 0.6 (exceeding the first judgment threshold of 0.5), and the maximum curvature parameter of this area is 1.0 (exceeding the second judgment threshold of 0.9). The comprehensive judgment result is a leaked package, and a detailed comparison result report is generated, including information such as the leak location (the area with a deformation vector amplitude of 0.6 in the deformation distribution map), the matching degree (0.8), the risk level (high), and the judgment result (missed package). This implementation method achieves accurate detection of the sealing of soft bags by dynamically adjusting thresholds, triggering high-confidence leak alarms and comprehensive judgments. It can improve the adaptability, reliability and accuracy of detection, help more accurately determine whether the soft bag has air leakage problems, and provide detailed test reports.

[0058] In the above, refer to Figure 1 The soft package sealing detection method based on image recognition according to an embodiment of the present invention is described in detail. Figure 2 A soft package sealing detection system based on image recognition according to an embodiment of the present invention is described.

[0059] The soft package sealing inspection system based on image recognition according to an embodiment of the present invention is designed to address the technical issues of low efficiency and accuracy in existing soft package sealing inspections, thereby improving the efficiency and accuracy of sealing inspections. The soft package sealing inspection system based on image recognition includes: an initial image data acquisition module 10, a deformed image data acquisition module 20, an image feature difference data extraction module 30, and a sealing determination module 40.

[0060] The initial image data acquisition module 10 is used to place the soft package to be tested into the vacuum chamber, perform the first stage of infrared irradiation under the first pressure condition, and collect initial image data; the deformation image data acquisition module 20 is used to perform negative pressure treatment on the vacuum chamber to a set pressure threshold, perform the second stage of infrared irradiation under the second pressure condition, and collect deformation image data; the image feature difference data extraction module 30 is used to align the initial image data with the deformed image data, extract image feature difference data, and calculate the deformation vector of the sealing area based on the image feature difference data; the sealing judgment module 40 is used to construct a soft package surface deformation distribution map based on the deformation vector to perform sealing judgment and determine the leakage detection result.

[0061] In which, the system also includes an image acquisition unit for use when performing the first stage infrared irradiation and the second stage infrared irradiation. The image acquisition unit may further include: an infrared light source irradiation intensity setting subunit for arranging multiple infrared sensor arrays in the vacuum chamber, setting the irradiation intensity of the infrared light source according to the characteristics of the soft package material, activating the multiple infrared sensor arrays to perform data acquisition according to the irradiation intensity, and obtaining a first reflection intensity distribution map; a data adding subunit for adding the first reflection intensity distribution map to the initial image data; a second stage infrared light source irradiation intensity adjustment subunit for dynamically adjusting the irradiation intensity of the second stage infrared light source based on the first reflection intensity distribution map after the negative pressure treatment is completed, activating the multiple infrared sensor arrays to perform data acquisition according to the irradiation intensity, and obtaining a second reflection intensity distribution map; a data adding subunit for adding the second reflection intensity distribution map to the deformed image data.

[0062] The specific configuration of the image feature difference data extraction module 30 will be described in detail below. As described above, the initial image data and the deformed image data are registered and aligned to extract image feature difference data. The image feature difference data extraction module 30 may further include: a first feature point set establishment unit for extracting key feature points from the initial image data based on the first reflection intensity distribution map to establish a first feature point set; a second feature point set extraction unit for performing edge enhancement processing on the deformed image data based on the second reflection intensity distribution map to extract a second feature point set; a feature point matching unit for performing feature point matching based on the first feature point set and the second feature point set to determine a dual-stage infrared image pair, wherein the dual-stage infrared image pair includes a spatial transformation matrix; and an affine transformation unit for performing affine transformation on the initial image data and the deformed image data according to the spatial transformation matrix to determine the image feature difference data.

[0063] Among them, the deformation vector of the sealing area is calculated according to the image feature difference data, and the image feature difference data extraction module 30 can further include: a historical leakage sample data retrieval unit is used to retrieve the historical leakage sample data, traverse the historical leakage sample data to perform image deformation annotation, and obtain the deformation abnormality area label of the sealing area; a pixel-level deformation probability map construction unit is used to perform convolution calculation on the two-stage infrared image pair according to the deformation abnormality area label, and construct a pixel-level deformation probability map; a deformation calculation unit is used to perform deformation calculation on the sealing area according to the pixel-level deformation probability map, and generate a deformation vector of the sealing area.

[0064] The specific configuration of the sealing determination module 40 will be described in detail below. As described above, the sealing determination is performed by constructing a soft bag surface deformation distribution map based on the deformation vector to determine the sealing and leak detection results. The sealing determination module 40 may further include: a soft bag surface deformation distribution map construction unit for constructing a soft bag surface coordinate system, mapping the deformation vector to the soft bag surface coordinate system, and performing normalization processing based on the soft bag material properties to construct the soft bag surface deformation distribution map; a region segmentation unit for traversing the soft bag surface deformation distribution map to extract data, performing region segmentation based on multiple deformation data sets, and generating a set of suspected leakage regions; a morphological optimization unit for performing morphological optimization on the set of suspected leakage regions, performing deformation analysis based on the optimized morphological data, and obtaining deformation vector amplitudes; and a search and matching unit for using the deformation vector amplitudes as indexes to search and match the historical leak sample data to generate a leak matching degree, which is then added to the leak detection results.

[0065] Among them, the deformation vector amplitude is used as an index to retrieve and match the historical packet leakage sample data to generate a packet leakage matching degree, and the packet leakage matching degree is added to the packet leakage detection result. The retrieval and matching unit may further include: a grid segmentation subunit for gridding the soft package surface deformation distribution map to obtain a number of detection units; an analysis and dimensionality reduction subunit for calculating the several detection units in combination with the deformation vector amplitude to generate a multidimensional feature vector, analyzing and reducing the multidimensional feature vector to generate a deformation amplitude feature sequence; a similarity retrieval subunit for traversing the historical packet leakage sample database with the deformation amplitude feature sequence as an index to perform similarity retrieval and generate a packet leakage matching degree; a leakage risk level determination subunit for comparing the packet leakage matching degree with a dynamic threshold, determining the leakage risk level according to the comparison result, and integrating the packet leakage matching degree and the leakage risk level into the packet leakage detection result.

[0066] Among them, the leaked package matching degree is compared with the dynamic threshold, and the leakage risk level determination subunit may further include: a benchmark matching degree threshold table acquisition component is used to query according to the soft package material characteristics in combination with the detected environmental humidity to obtain a benchmark matching degree threshold table; a linear interpolation compensation component is used to introduce the real-time pressure change rate to perform linear interpolation compensation on the benchmark matching degree threshold table to generate a dynamic threshold; a comparison result generation component is used to trigger a high-confidence leakage alarm when the leaked package matching degree exceeds the dynamic threshold, and generate the comparison result.

[0067] The soft package sealing detection system based on image recognition provided by the embodiment of the present invention can execute the soft package sealing detection method based on image recognition provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0068] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other and are not used to limit the scope of protection of the present invention.

[0069] The above specific embodiments do not constitute a limitation to the scope of protection of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of this application should be included in the scope of protection of this application. In some cases, the actions or steps recorded in this application can be performed in an order different from that in the embodiments and can still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A soft package sealing detection method based on image recognition, characterized in that: The method comprises: The soft package to be tested is placed in a vacuum chamber, and the first stage of infrared irradiation is performed under the first pressure condition to collect initial image data; applying negative pressure to the vacuum chamber to a set pressure threshold, performing second-stage infrared irradiation under a second pressure condition, and collecting deformation image data; performing registration and alignment on the initial image data and the deformed image data, extracting image feature difference data, and calculating a deformation vector of the sealing area according to the image feature difference data; A soft package surface deformation distribution map is constructed based on the deformation vector to perform sealing judgment and determine the leakage detection result.

2. The soft package sealing detection method based on image recognition according to claim 1, characterized in that: The first stage infrared irradiation and the second stage infrared irradiation acquisition process include: Arranging a plurality of infrared sensor arrays in the vacuum chamber, setting the illumination intensity of the infrared light source according to the characteristics of the soft package material, activating the plurality of infrared sensor arrays to collect data according to the illumination intensity, and obtaining a first reflection intensity distribution map; adding the first reflection intensity distribution map to the initial image data; After the negative pressure treatment is completed, dynamically adjusting the illumination intensity of the infrared light source in the second stage based on the first reflection intensity distribution map, activating the multiple infrared sensor arrays to collect data according to the illumination intensity, and obtaining a second reflection intensity distribution map; The second reflection intensity distribution map is added to the deformed image data.

3. The soft package sealing detection method based on image recognition according to claim 2, characterized in that: The initial image data and the deformed image data are registered and aligned to extract image feature difference data, the method comprising: extracting key feature points in the initial image data based on the first reflection intensity distribution map to establish a first feature point set; performing edge enhancement processing on the deformed image data based on the second reflection intensity distribution map to extract a second feature point set; performing feature point matching based on the first feature point set and the second feature point set to determine a two-stage infrared image pair, wherein the two-stage infrared image pair includes a spatial transformation matrix; Affine transformation is performed on the initial image data and the deformed image data according to the spatial transformation matrix to determine the image feature difference data.

4. The soft package sealing detection method based on image recognition according to claim 3, characterized in that: Calculating a deformation vector of the sealing area according to the image feature difference data, the method comprising: Retrieving historical packet leakage sample data, traversing the historical packet leakage sample data to perform image deformation annotation, and obtaining deformation abnormality area labels of the sealed area; Performing convolution calculation on the dual-stage infrared image pair according to the deformation abnormality region label to construct a pixel-level deformation probability map; The deformation of the sealing area is calculated according to the pixel-level deformation probability map to generate a deformation vector of the sealing area.

5. The soft package sealing detection method based on image recognition according to claim 4, characterized in that: Constructing a soft package surface deformation distribution map based on the deformation vector to perform sealing determination and determine a leak detection result, the method comprising: Constructing a soft bag surface coordinate system, mapping the deformation vector to the soft bag surface coordinate system, performing normalization processing based on the soft bag material properties, and constructing a soft bag surface deformation distribution map; Traversing the deformation distribution map of the soft package surface to extract data, performing region segmentation based on multiple deformation data sets, and generating a set of suspected leakage regions; Performing morphological optimization on the set of suspected leakage areas, performing deformation analysis based on the optimized morphological data, and obtaining a deformation vector amplitude; The deformation vector amplitude is used as an index to search and match the historical packet leakage sample data to generate a packet leakage matching degree, and the packet leakage matching degree is added to the packet leakage detection result.

6. The soft package sealing detection method based on image recognition according to claim 5, characterized in that: The deformation vector amplitude is used as an index to search and match the historical packet leakage sample data, generate a packet leakage matching degree, and add the packet leakage matching degree to the packet leakage detection result. The method includes: Performing grid segmentation on the deformation distribution map of the soft package surface to obtain a plurality of detection units; Calculating the plurality of detection units in combination with the deformation vector amplitude to generate a multidimensional feature vector, analyzing and reducing the multidimensional feature vector to generate a deformation amplitude feature sequence; Using the deformation amplitude feature sequence as an index, traverse the historical packet leakage sample database to perform similarity retrieval and generate a packet leakage matching degree; The packet leakage matching degree is compared with a dynamic threshold, a leakage risk level is determined according to the comparison result, and the packet leakage matching degree and the leakage risk level are integrated into the packet leakage detection result.

7. The soft package sealing detection method based on image recognition according to claim 6, characterized in that: Comparing the missing packet matching degree with a dynamic threshold, the method includes: A reference matching threshold table is obtained by querying the soft package material characteristics in combination with the detection environment humidity; Introducing the real-time pressure change rate to perform linear interpolation compensation on the reference matching threshold table to generate a dynamic threshold; When the leaked packet matching degree exceeds the dynamic threshold, a high-confidence leak alarm is triggered and the comparison result is generated.

8. The soft package sealing detection system based on image recognition is characterized by: The system is used to implement the soft package sealing detection method based on image recognition according to any one of claims 1 to 7, and the system comprises: An initial image data acquisition module is used to place the soft package to be tested into the vacuum chamber, perform the first stage of infrared irradiation under the first pressure condition, and acquire initial image data; a deformation image data acquisition module, configured to subject the vacuum chamber to negative pressure treatment to a set pressure threshold, perform a second stage of infrared irradiation under a second pressure condition, and acquire deformation image data; an image feature difference data extraction module, configured to perform registration and alignment between the initial image data and the deformed image data, extract image feature difference data, and calculate a deformation vector of a sealing area based on the image feature difference data; The sealing determination module is used to construct a soft package surface deformation distribution map according to the deformation vector to perform sealing determination and determine the leakage detection result.

Images