A PET finished product packaging quality detection method and system

Abstract

The present application discloses a PET finished product packaging quality detection method and system, the method comprising: obtaining a known process PET finished product data set and an unknown process PET finished product data set; performing image enhancement based on a generative adversarial network; obtaining a PET finished product recognition network model using a one-dimensional convolutional residual network; obtaining first image information through an image acquisition device; obtaining a first recognition quality result; obtaining a first material detection result based on a predetermined detection standard and a detection index set; obtaining a first predetermined weight distribution result through a principal component analysis method, and then obtaining a first PET finished product packaging quality detection result through weighted calculation. The method solves the problem that the prior art only detects conventional defects such as the appearance quality of PET finished products, but cannot simultaneously perform intelligent detection of material properties in PET finished product packaging, resulting in incomplete detection when detecting the quality of PET finished product packaging.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to a method and system for detecting the quality of PET finished product packaging. Background Art

[0002] PET finished product packaging is polyethylene terephthalate (PET) packaging. In actual production, it is often used in the packaging of food, beverages, medicines and other products in bottle shape. Therefore, the quality of PET finished product packaging directly affects the edible safety of the content of food, beverages, medicines and other products. In the prior art, the quality of PET finished product packaging is often tested by machine vision and other technologies, including conventional defect detection in the shape, size, positional relationship of the bottle body and bottle cap of PET finished product packaging, such as no cap, high cap, crooked cap, and complete breakage of the anti-theft ring. However, it is impossible to intelligently test the performance of the raw materials used in PET finished product packaging. Research on the use of computer technology to conduct integrated quality inspection of the appearance quality and internal materials of PET finished product packaging quality and improve the degree of automation of PET finished product packaging quality inspection is of great significance to ensure product quality and safety, improve the production efficiency of manufacturers, and reduce the production cost of manufacturers.

[0003] In the process of implementing the technical solution in the embodiments of the present application, the inventors of the present application found that the above technology has at least the following technical problems:

[0004] In the prior art, when the packaging quality of PET finished product packaging is tested by computer technology, only conventional defects such as the appearance quality of the PET finished product are tested, but the material properties in the PET finished product packaging cannot be intelligently tested at the same time, resulting in a technical problem of incomplete testing when testing the quality of PET finished product packaging. Summary of the invention

[0005] The purpose of the present application is to provide a PET finished product packaging quality detection method and system, so as to solve the problem that in the prior art, when the packaging quality of the PET finished product packaging is detected by computer technology, only conventional defects such as the appearance quality of the PET finished product are detected, but the material properties in the PET finished product packaging cannot be intelligently detected at the same time, resulting in incomplete detection when the PET finished product packaging quality is detected.

[0006] In view of the above problems, the embodiments of the present application provide a method and system for detecting the quality of PET finished product packaging.

[0007] In a first aspect, the present application provides a PET finished product packaging quality detection method, which is implemented by a PET finished product packaging quality detection system, wherein the method includes: obtaining a known process PET finished product data set and an unknown process PET finished product data set through a big data platform; performing image enhancement on the known process PET finished product data set and the unknown process PET finished product data set based on a generative adversarial network; modeling the PET finished product data through a one-dimensional convolutional residual network according to the known process PET finished product data set and the unknown process PET finished product data set after image enhancement to obtain a PET finished product recognition network model; obtaining first image information through the image acquisition device, the first image information including image information of a first PET finished product; inputting the first image information into the PET finished product recognition network model to obtain a first recognition quality result; detecting the material properties of the first PET finished product according to a predetermined detection standard and a detection index set to obtain a first material detection result; obtaining a first predetermined weight distribution result through a principal component analysis method, and performing weighted calculation on the first recognition quality result and the first material detection result according to the first predetermined weight distribution result to obtain a first PET finished product packaging quality detection result.

[0008] On the other hand, the present application also provides a PET finished product packaging quality detection system, which is used to execute a PET finished product packaging quality detection method as described in the first aspect, wherein the system includes: a first acquisition unit: the first acquisition unit is used to obtain a known process PET finished product data set and an unknown process PET finished product data set through a big data platform; a first execution unit: the first execution unit is used to perform image enhancement on the known process PET finished product data set and the unknown process PET finished product data set based on a generative adversarial network; a second acquisition unit: the second acquisition unit is used to model the PET finished product data through a one-dimensional convolutional residual network according to the known process PET finished product data set and the unknown process PET finished product data set after image enhancement, so as to obtain a PET finished product recognition network model; Three acquisition units: the third acquisition unit is used to obtain first image information through an image acquisition device, and the first image information includes image information of a first PET finished product; fourth acquisition unit: the fourth acquisition unit is used to input the first image information into the PET finished product recognition network model to obtain a first recognition quality result; fifth acquisition unit: the fifth acquisition unit is used to detect the material properties of the first PET finished product according to a predetermined detection standard and a set of detection indicators to obtain a first material detection result; sixth acquisition unit: the sixth acquisition unit is used to obtain a first predetermined weight distribution result through a principal component analysis method, and perform weighted calculation on the first recognition quality result and the first material detection result according to the first predetermined weight distribution result to obtain a first PET finished product packaging quality detection result.

[0009] In a third aspect, an embodiment of the present application further provides a PET finished product packaging quality inspection system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the method described in the first aspect are implemented.

[0010] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0011] 1. Obtain a known process PET finished product data set and an unknown process PET finished product data set through a big data platform; perform image enhancement on the known process PET finished product data set and the unknown process PET finished product data set based on a generative adversarial network; model the PET finished product data through a one-dimensional convolutional residual network according to the known process PET finished product data set and the unknown process PET finished product data set after image enhancement to obtain a PET finished product recognition network model; obtain first image information through the image acquisition device, the first image information including image information of a first PET finished product; input the first image information into the PET finished product recognition network model to obtain a first recognition quality result; detect the material properties of the first PET finished product according to a predetermined detection standard and a detection index set to obtain a first material detection result; obtain a first predetermined weight distribution result through a principal component analysis method, and perform weighted calculation on the first recognition quality result and the first material detection result according to the first predetermined weight distribution result to obtain a first PET finished product packaging quality detection result. By using the generative adversarial network to enhance the images of PET finished packaging with known and unknown processes, and then constructing a PET finished product recognition network model based on the image-enhanced PET finished packaging, the recognition accuracy of the model is improved, and the appearance quality of PET finished packaging is intelligently tested, thereby improving the detection efficiency of PET finished packaging. In addition, the material properties of PET finished packaging are intelligently tested, and the test results of PET finished packaging quality are further obtained by weighted calculation. The technical effect of comprehensively testing the material and appearance quality of PET finished packaging and improving the accuracy and efficiency of PET finished packaging quality detection is achieved.

[0012] 2. Image enhancement is performed by generating adversarial networks, which can be used for recognition in low signal-to-noise ratio environments and has good robustness, achieving the technical effect of improving model recognition accuracy and reducing interference from the surrounding environment.

[0013] 3. Through the one-dimensional convolutional residual network to automatically extract the input unknown process PET finished product data for identification, it can explore the deeper features of the unknown process PET finished product data and discover the connection and difference between different data signals. At the same time, by adding unknown process PET finished product data to the training data and adding the central error loss function to the fully connected layer of the network, the network can perform open set identification of PET finished products.

[0014] 4. By using principal component analysis to reduce the dimension of the detection quality attribute information, the system calculation complexity is reduced, thereby improving the speed of PET finished product packaging quality detection. In addition, by assigning weights to each attribute information through factor analysis, the accuracy and reliability of the PET finished product packaging quality detection results are improved.

[0015] 5. Through multiple rounds of data processing and calculation, the redundant processing of the detection quality attribute information is completed, making the retained data more universally applicable and reducing the complexity of subsequent data analysis.

[0016] 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. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0018] Figure 1 This is a schematic diagram of a process for testing the quality of PET finished product packaging according to an embodiment of the present application;

[0019] Figure 2 A schematic diagram of a process of performing image enhancement on the known process PET finished product data set and the unknown process PET finished product data set based on the parameter updating step in a PET finished product packaging quality detection method according to an embodiment of the present application;

[0020] Figure 3 This is a schematic diagram of a process for obtaining the PET finished product identification network model in a PET finished product packaging quality detection method according to an embodiment of the present application;

[0021] Figure 4 This is a schematic diagram of a process for obtaining a first predetermined weight distribution result in a PET finished product packaging quality detection method according to an embodiment of the present application;

[0022] Figure 5 This is a structural schematic diagram of a PET finished product packaging quality inspection system according to an embodiment of the present application;

[0023] Figure 6 It is a schematic diagram of the structure of an exemplary electronic device according to an embodiment of the present application.

[0024] Description of reference numerals:

[0025] A first obtaining unit 11 , a first executing unit 12 , a second obtaining unit 13 , a third obtaining unit 14 , a fourth obtaining unit 15 , a fifth obtaining unit 16 , a sixth obtaining unit 17 , a bus 300 , a receiver 301 , a processor 302 , a transmitter 303 , a memory 304 , and a bus interface 305 . DETAILED DESCRIPTION

[0026] The embodiment of the present application provides a PET finished product packaging quality detection method and system, which solves the problem that in the prior art, when the packaging quality of the PET finished product packaging is detected by computer technology, only conventional defects such as the appearance quality of the PET finished product are detected, but the material properties in the PET finished product packaging cannot be intelligently detected at the same time, resulting in the technical problem of incomplete detection when the PET finished product packaging quality is detected. By using a generative adversarial network to enhance the image of the PET finished product packaging of known and unknown processes, and then constructing a PET finished product recognition network model based on the PET finished product packaging after image enhancement, the recognition accuracy of the model is improved, and the appearance quality of the PET finished product packaging is intelligently detected, thereby improving the detection efficiency of the PET finished product packaging. In addition, the material properties of the PET finished product packaging are intelligently detected, and the detection results of the PET finished product packaging quality are further obtained by weighted calculation. The technical effect of comprehensively detecting the material and appearance quality of the PET finished product packaging and improving the accuracy and efficiency of the PET finished product packaging quality detection is achieved.

[0027] Below, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0028] Application Overview

[0029] PET finished product packaging is polyethylene terephthalate (PET) packaging. In actual production, it is often used in the packaging of food, beverages, medicines and other products in bottle shape. Therefore, the quality of PET finished product packaging directly affects the edible safety of the content of food, beverages, medicines and other products. In the prior art, the quality of PET finished product packaging is often tested by machine vision and other technologies, including conventional defect detection in the shape, size, positional relationship of the bottle body and bottle cap of PET finished product packaging, such as no cap, high cap, crooked cap, and complete breakage of the anti-theft ring. However, it is impossible to intelligently test the performance of the raw materials used in PET finished product packaging. Research on the use of computer technology to conduct integrated quality inspection of the appearance quality and internal materials of PET finished product packaging quality and improve the degree of automation of PET finished product packaging quality inspection is of great significance to ensure product quality and safety, improve the production efficiency of manufacturers, and reduce the production cost of manufacturers.

[0030] In the prior art, when the packaging quality of PET finished product packaging is tested by computer technology, only conventional defects such as the appearance quality of the PET finished product are tested, but the material properties in the PET finished product packaging cannot be intelligently tested at the same time, resulting in a technical problem of incomplete testing when testing the quality of PET finished product packaging.

[0031] In response to the above technical problems, the overall idea of ​​the technical solution provided by this application is as follows:

[0032] The present application provides a PET finished product packaging quality detection method, which is applied to a PET finished product packaging quality detection system, wherein the method includes: obtaining a known process PET finished product data set and an unknown process PET finished product data set through a big data platform; performing image enhancement on the known process PET finished product data set and the unknown process PET finished product data set based on a generative adversarial network; modeling the PET finished product data through a one-dimensional convolutional residual network according to the known process PET finished product data set and the unknown process PET finished product data set after image enhancement, and obtaining a PET finished product recognition network model; obtaining first image information through the image acquisition device, the first image information including image information of a first PET finished product; inputting the first image information into the PET finished product recognition network model to obtain a first recognition quality result; detecting the material properties of the first PET finished product according to a predetermined detection standard and a detection index set to obtain a first material detection result; obtaining a first predetermined weight distribution result through a principal component analysis method, and performing weighted calculation on the first recognition quality result and the first material detection result according to the first predetermined weight distribution result to obtain a first PET finished product packaging quality detection result.

[0033] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification.

[0034] Embodiment 1

[0035] Please refer to the attached Figure 1 The present application embodiment provides a PET finished product packaging quality detection method, wherein the method is applied to a PET finished product packaging quality detection system, and the method specifically includes the following steps:

[0036] Step S100: obtaining a data set of PET finished products with known processes and a data set of PET finished products with unknown processes through a big data platform;

[0037] Specifically, the PET finished product packaging quality detection method is applied to the PET finished product packaging quality detection system, and can enhance the image of the PET finished product packaging of known and unknown processes by using a generative adversarial network, and then construct a PET finished product recognition network model based on the PET finished product packaging after image enhancement, thereby improving the recognition accuracy of the model, and improving the detection efficiency of the PET finished product packaging by intelligently detecting the appearance quality of the PET finished product packaging. In addition, the material properties of the PET finished product packaging are intelligently detected, and the detection results of the PET finished product packaging quality are further obtained by weighted calculation.

[0038] The PET, or polyethylene terephthalate, is a milky white or slightly yellow polymer. It is non-toxic, odorless, and has good hygiene and safety, so it is often used to make packaging for food, beverages, medicines, etc. For example, mineral water bottles. Since PET finished product packaging is often packaged in direct contact with food, the quality of PET finished product packaging directly affects the safety of the corresponding products. It is very important to use a variety of methods to inspect the quality of PET finished product packaging during production. Using the big data platform, the packaging data of the complete production and processing technology of the PET finished product packaging and the PET finished product packaging with uncertain production and processing technology are collected, thereby obtaining the known process PET finished product data set and the unknown process PET finished product data set. The two collected data sets provide basic data for subsequent image enhancement of the PET finished product packaging.

[0039] Step S200: performing image enhancement on the known process PET finished product data set and the unknown process PET finished product data set based on a generative adversarial network;

[0040] Specifically, the Generative Adversarial Network (GAN) is simply to generate data that does not exist in the real world. The Generative Adversarial Network mainly includes a discriminator for judging whether the data is real data and a generator for fabricating data out of thin air. The known process PET finished product data set and the unknown process PET finished product data set are data collected by the system using big data, including a large number of PET finished product packaging process flow, finished product images and other related detailed data information. Through the discriminator and generator in the Generative Adversarial Network, the known process PET finished product data set and the unknown process PET finished product data set are image enhanced, that is, the generator in the Generative Adversarial Network is used to construct the basic data information set to be used when constructing the intelligent network model for subsequent PET finished product packaging quality identification. By using the Generative Adversarial Network to enhance the image, the technical effect of improving the accuracy of model recognition and reducing the interference of the surrounding environment is achieved.

[0041] Step S300: Modeling the PET finished product data through a one-dimensional convolutional residual network according to the known process PET finished product data set and the unknown process PET finished product data set after image enhancement, to obtain a PET finished product recognition network model;

[0042] Specifically, the known process PET finished product data set and the unknown process PET finished product data set obtained after generating the enhanced image through the adversarial network are used as the basic data for constructing the PET finished product recognition network model. The PET finished product recognition network model is used to intelligently perform targeted intelligent appearance quality detection on the quality of the PET finished product packaging to be detected, including quality detection of the packaging thickness, packaging shape, packaging size, color, etc. of the PET finished product packaging. In addition, since there are both PET finished product packaging images with known process flows and PET finished product packaging images with unknown process flows in the real environment, based on the known process PET finished product data set and the unknown process PET finished product data set after image enhancement, the PET finished product data is modeled by using a one-dimensional convolutional residual network to obtain the PET finished product recognition network model. The convolution kernel of the one-dimensional convolutional residual network is a one-dimensional convolutional neural network model, and the residual structure is used to improve the performance of the PET finished product recognition network model.

[0043] By using a one-dimensional convolutional residual network to automatically extract and input the image features of PET finished product packaging after image enhancement for recognition, the tedious steps of manually extracting image signal features in the traditional method of the prior art are avoided. The one-dimensional convolution is used instead of the two-dimensional convolution to accelerate the convergence speed of the model. At the same time, by adding PET finished product packaging data of unknown process to the training data, the technical effect of improving the generalization ability of the PET finished product recognition network model is achieved.

[0044] Step S400: obtaining first image information through the image acquisition device, wherein the first image information includes image information of a first PET finished product;

[0045] Step S500: inputting the first image information into the PET finished product recognition network model to obtain a first recognition quality result;

[0046] Specifically, the image acquisition device is built into the packaging quality inspection system, and can capture images of PET finished product packaging at multiple angles and distances. The image acquisition device is used to capture images of PET finished product packaging to be inspected for quality by the packaging quality inspection system in real time, thereby obtaining the first image information including image information of the first PET finished product. The first PET finished product is any PET finished product packaging to be inspected for quality.

[0047] Furthermore, the first image information collected by the image acquisition device is input into the PET finished product recognition network model, and the recognition result is the first recognition quality result. The first recognition quality result includes the recognition results of the packaging thickness, size, shape, color, etc. of the first PET finished product. Through the PET finished product recognition network model, the technical effect of intelligently detecting the appearance quality of the PET finished product packaging is achieved.

[0048] Step S600: testing the material properties of the first PET finished product according to a predetermined testing standard and a set of testing indicators to obtain a first material testing result;

[0049] Specifically, the detection index set refers to the indexes that need to be detected when the material properties of the first PET finished product are detected and evaluated, for example, performance indicators such as thermal stability, light resistance, and UV protection index. The predetermined detection standard refers to the detection information of each index standard that is pre-set based on the information of the indicators to be detected in the detection index set, after querying national standards, industry standards or self-designed standards, and comprehensively analyzing according to actual needs. Then, the material properties of the first PET finished product are detected based on the predetermined detection standard and the detection index set, and the detection result is the first material detection result of the first PET finished product. By detecting the material properties of the first PET finished product, the corresponding PET material performance results are obtained, thereby achieving the technical effect of comprehensively detecting the packaging quality of the first PET finished product and improving the accuracy and reliability of the detection results.

[0050] Step S700: Obtain a first predetermined weight distribution result by principal component analysis, and perform weighted calculation on the first recognition quality result and the first material detection result according to the first predetermined weight distribution result to obtain a first PET finished product packaging quality detection result.

[0051] Specifically, the principal component analysis method refers to analyzing the influence of appearance quality and material performance characteristics on the comprehensive evaluation of PET finished product packaging quality in PET finished product packaging, that is, the degree of correlation between the two variables of appearance quality and material performance characteristics on the comprehensive evaluation of PET finished product packaging quality, thereby comprehensively analyzing and determining the first predetermined weight allocation result. Further, based on the first predetermined weight allocation result, the weight coefficients corresponding to the first recognition quality result and the first material detection result are respectively assigned, and then the comprehensive detection result of the first PET finished product packaging quality, that is, the first PET finished product packaging quality detection result, is obtained by comprehensive calculation.

[0052] By using the generative adversarial network to enhance the images of PET finished packaging with known and unknown processes, and then constructing a PET finished product recognition network model based on the image-enhanced PET finished packaging, the recognition accuracy of the model is improved, and the appearance quality of the PET finished packaging is intelligently tested, thereby improving the detection efficiency of the PET finished packaging. In addition, the material properties of the PET finished packaging are intelligently tested, and the test results of the PET finished packaging quality are further obtained by weighted calculation. The technical effect of comprehensively testing the material properties and appearance quality of the PET finished packaging and improving the accuracy and efficiency of the PET finished packaging quality detection is achieved.

[0053] Further, as attached Figure 2 As shown, step S200 of the embodiment of the present application also includes:

[0054] Step S210: Initialize the parameters of the generator and the discriminator to obtain initial generator parameters and initial discriminator parameters;

[0055] Step S220: obtaining a predetermined threshold value of parameter update times of the generative adversarial network;

[0056] Step S230: updating the initial generator parameters and the initial discriminator parameters according to the parameter update times threshold, generating parameter update steps: the parameters of the generator remain unchanged, the discriminator is trained, and the parameters of the discriminator remain unchanged, the generator is trained;

[0057] Step S240: performing image enhancement on the known process PET finished product data set and the unknown process PET finished product data set based on the parameter updating step.

[0058] Specifically, the generative adversarial network mainly includes a discriminator for determining whether the data is real data and a generator for fabricating data out of thin air. The generative adversarial network converts the known process PET finished product data set and the unknown process PET finished product data set into denoised images with the same image effect by learning a nonlinear mapping function F. Initializing the parameters of the generator and the discriminator refers to the process of initializing the weights and biases of each node before network training. The initialization of parameters is related to whether the network can be trained to produce good results or how fast it converges, and whether good training results can be obtained.

[0059] The threshold of the number of parameter updates of the predetermined generative adversarial network is used to set the number of times the generator and the discriminator are updated. For example, when the number of parameter updates is set to 50, the parameters of the generator and the discriminator in the generative adversarial network are updated 50 times. The image of the known process PET finished product data set and the unknown process PET finished product data set is enhanced by updating the parameters, thereby achieving the effect of using GAN to enhance the image of the data set. Image enhancement by generative adversarial network can be recognized in a low signal-to-noise ratio environment and has good robustness.

[0060] Further, as attached Figure 3 As shown, step S300 of the embodiment of the present application also includes:

[0061] Step S310: constructing a PET finished product open set identification data set, wherein the PET finished product open set identification data set includes the known process PET finished product data set and the unknown process PET finished product data set after image enhancement;

[0062] Step S320: redistribute the data in the PET finished product open set identification data set, and divide the redistributed data into a training set and a validation set according to a predetermined ratio;

[0063] Step S330: inputting the training set into a 1D-ResNet network structure for training;

[0064] Step S340: Obtain a 1D-ResNet loss function by performing weighted summation of a multi-classification cross entropy loss function and a center error loss function;

[0065] Step S350: Calculate the loss value of the model on the training set and the validation set according to the 1D-ResNet loss function;

[0066] Step S360: When the loss value reaches a preset number of training times or the loss value of the network does not decrease for a preset number of consecutive times, the training process of the 1D-ResNet network is completed to obtain the PET finished product recognition network model.

[0067] Specifically, the PET finished product open set recognition data set includes both the known process PET finished product data set after image enhancement and the unknown process PET finished product data set after image enhancement. In the PET finished product open set recognition data set, each PET finished product data sample is randomly redistributed, and then divided into a training set and a verification set at a ratio of 8:2. Among them, 8:2 is the predetermined ratio, which is set in the system after prior analysis according to actual needs. Further, the input of 1D-ResNet is the training set for training. Among them, the 1D-ResNet network structure consists of two residual modules. The convolution layer of the residual module is composed of a one-dimensional convolution layer combined with a batch normalization layer operation and a Relu activation function, and a skip connection operation is added to the corresponding convolution layer, the convolution kernel size is 5, and the step size is 3. The output of each residual module is followed by a maximum pooling operation, and the pooling window size is 2. The parameters of the fully connected layer are set to 64 fully connected neurons in the first fully connected layer, 3 neurons in the second fully connected layer, and the output fully connected layer is set to 6 output neural units, and the activation function is a softmax function. A dropout operation is also added to the fully connected layer to prevent overfitting. Then the parameters of the 7 one-dimensional convolutional layers and 3 fully connected layers in 1D-ResNet are set, and the weighted sum of the multi-classification cross entropy loss function and the center error loss function is used as the loss function of 1D-ResNet. Furthermore, the optimization method of the model is selected as the adaptive optimization algorithm Adam. The number of network training is set, and the network parameters are trained by reverse error propagation by minimizing the loss function, as shown in the following formula:

[0068] J 1D-ResNet =J 1 +β·J 2

[0069] Among them, J 1 represents the classification cross entropy loss, J 2 represents the center error loss, β is a weight value between 0 and 1, and min(·) means training the network parameters by minimizing the loss function.

[0070] After each round of training, the loss value of the model on the training set and the validation set is calculated. When the preset number of training times is reached or the network loss value does not decrease for 5 consecutive times, the training process of the neural network is completed and the trained 1D-ResNet is obtained.

[0071] By using a one-dimensional convolutional residual network to automatically extract input unknown process PET finished product data for identification, it is possible to explore the deeper features of the unknown process PET finished product data and discover the connections and differences between different data signals. At the same time, by adding unknown process PET finished product data to the training data and adding a central error loss function to the fully connected layer of the network, the network can perform open set identification of PET finished products.

[0072] Furthermore, step S600 in the embodiment of the present application further includes:

[0073] Step S610: constructing a set of testing indicators, wherein the set of testing indicators includes barrier properties, impact strength, light transmittance index, and thermal stability performance;

[0074] Step S620: Score the material properties of the first PET finished product according to a predetermined testing standard to obtain a testing scoring matrix corresponding to each indicator in the testing indicator set;

[0075] Step S630: Obtain the correlation between the indicators in the detection indicator set, and construct an indicator evaluation matrix according to the evaluation coefficients of the indicators and the correlation;

[0076] Step S640: Obtain a first material detection result based on a product calculation result of the detection score matrix and the index evaluation matrix.

[0077] Specifically, according to the actual application scenario of PET finished product packaging, a detection index set is constructed. Among them, the detection index set includes barrier performance, impact strength, light transmittance index, and thermal stability. For example, when PET finished products are used for the packaging of medicines, in order to prevent the medicines from deteriorating after being exposed to light and affecting the effect of taking, the light transmittance of the PET finished product packaging should be tested. Further, the material performance of the first PET finished product is scored according to the predetermined detection standard to obtain the detection score matrix corresponding to each indicator in the detection index set. Among them, the predetermined detection standard is the minimum standard requirement for each performance feature of the PET finished product packaging set based on relevant national standards, industry standards and actual needs. By analyzing the degree of correlation between the indicators in the detection index set, an indicator evaluation matrix is ​​constructed according to the evaluation coefficients of each indicator and the correlation. Finally, based on the product calculation result of the detection score matrix and the indicator evaluation matrix, the first material detection result of the PET finished product packaging is obtained. The technical effect of intuitively evaluating the performance characteristics of each material in the PET finished product packaging and accurately evaluating the quality of the PET finished product packaging is achieved.

[0078] Further, as attached Figure 4 As shown, step S700 of the embodiment of the present application also includes:

[0079] Step S710: obtaining the inspection quality attribute information of the PET finished product packaging;

[0080] Step S720: performing principal component analysis on the detection quality attribute information to obtain first dimension reduction attribute information;

[0081] Step S730: Obtain a first predetermined weight distribution result based on factor analysis of the first dimension reduction attribute information.

[0082] Specifically, the inspection quality attribute information of the PET finished product packaging refers to a plurality of variables related to the quality of the PET finished product packaging proposed in order to comprehensively analyze the quality of the PET finished product packaging, wherein each variable reflects the quality of the PET finished product packaging to varying degrees. Further, the inspection quality attribute information of the PET finished product packaging is subjected to dimensionality reduction processing of principal component analysis, thereby obtaining attribute information that is strongly correlated with the quality of the PET finished product packaging, namely, the first dimensionality reduction attribute information. Finally, from the first dimensionality reduction attribute information, common features in each attribute information are extracted, thereby classifying the attribute information of the same nature into one attribute information. Among them, the attribute information with more common factors has a larger corresponding weight, and the attribute information with fewer common factors has a smaller corresponding weight, thereby obtaining the first predetermined weight distribution result.

[0083] By using principal component analysis to reduce the dimension of the detection quality attribute information, the system calculation complexity is reduced, thereby improving the speed of PET finished product packaging quality detection. In addition, by assigning weights to each attribute information through factor analysis, the accuracy and reliability of the PET finished product packaging quality detection results are improved.

[0084] Furthermore, step S720 of the embodiment of the present application further includes:

[0085] Step S721: After decentralizing the detection quality attribute information, a first covariance matrix is ​​obtained;

[0086] Step S722: Calculate the first covariance matrix to obtain a first eigenvalue and a first eigenvector of the first covariance matrix;

[0087] Step S723: Project the detection quality attribute information onto the first feature vector to obtain first dimension-reduced attribute information, wherein the first dimension-reduced attribute information is attribute information obtained after the detection quality attribute information is dimensionally reduced.

[0088] Specifically, the first covariance matrix is ​​obtained by decentralizing the detection quality attribute information. The centralization process refers to subtracting the mean of the data set from each item in the data set. For example, there are data sets 1, 2, 3, 6, 3, and their mean is 3. Then the data sets after centralization are 1-3, 2-3, 3-3, 6-3, 3-3, that is: -2, -1, 0, 3, 0. The data centralization process is to eliminate the influence of the dimension on the data structure, because the units of different variables are different, which will cause errors in various statistics. The covariance of the detection quality attribute information is further calculated to measure the overall error of the two variables and form the corresponding first covariance matrix. The eigenvalues ​​and eigenvectors corresponding to the first covariance matrix are further obtained. Finally, the corresponding eigenvalue data set is projected to the corresponding eigenvector to obtain the corresponding first dimension reduction data set. The first dimension reduction data set is the feature data set obtained after the first feature data set is reduced in dimension.

[0089] Through multiple rounds of data processing and calculation, the cumbersome processing of the detection quality attribute information is completed, making the retained data more universally applicable and reducing the complexity of subsequent data analysis.

[0090] Furthermore, the embodiment of the present application further includes step S800:

[0091] Step S810: obtaining the actual identification quality result of the first PET finished product;

[0092] Step S820: obtaining a first quality detection deviation based on a difference between the actual recognition quality result and the first recognition quality result;

[0093] Step S830: Correcting and training the PET finished product recognition network model according to the first quality detection deviation to obtain a PET finished product corrected recognition network model.

[0094] Specifically, the actual recognition quality result of the first PET finished product refers to the accurate quality result of the first PET finished product packaging obtained by testing the appearance quality, material performance indicators, etc. of the first PET finished product packaging based on professional institutions. Further comparing the actual recognition quality result with the first recognition quality result, the degree of intelligent detection deviation of the packaging quality detection system is obtained, which is the first quality detection deviation. Finally, the PET finished product recognition network model is corrected and trained according to the first quality detection deviation, and the model obtained after correction is the PET finished product corrected recognition network model. Through correction training, the quality recognition accuracy of the PET finished product recognition network model is improved, providing automated high-precision detection for the actual production and manufacturing of PET finished product packaging, reducing production costs, and achieving the technical effect of ensuring the safety of PET packaging products.

[0095] In summary, the PET finished product packaging quality inspection method provided in the embodiment of the present application has the following technical effects:

[0096] 1. Obtain a known process PET finished product data set and an unknown process PET finished product data set through a big data platform; perform image enhancement on the known process PET finished product data set and the unknown process PET finished product data set based on a generative adversarial network; model the PET finished product data through a one-dimensional convolutional residual network according to the known process PET finished product data set and the unknown process PET finished product data set after image enhancement to obtain a PET finished product recognition network model; obtain first image information through the image acquisition device, the first image information including image information of a first PET finished product; input the first image information into the PET finished product recognition network model to obtain a first recognition quality result; detect the material properties of the first PET finished product according to a predetermined detection standard and a detection index set to obtain a first material detection result; obtain a first predetermined weight distribution result through a principal component analysis method, and perform weighted calculation on the first recognition quality result and the first material detection result according to the first predetermined weight distribution result to obtain a first PET finished product packaging quality detection result. By using the generative adversarial network to enhance the images of PET finished packaging with known and unknown processes, and then constructing a PET finished product recognition network model based on the image-enhanced PET finished packaging, the recognition accuracy of the model is improved, and the appearance quality of PET finished packaging is intelligently tested, thereby improving the detection efficiency of PET finished packaging. In addition, the material properties of PET finished packaging are intelligently tested, and the test results of PET finished packaging quality are further obtained by weighted calculation. The technical effect of comprehensively testing the material and appearance quality of PET finished packaging and improving the accuracy and efficiency of PET finished packaging quality detection is achieved.

[0097] 2. Image enhancement is performed by generating adversarial networks, which can be used for recognition in low signal-to-noise ratio environments and has good robustness, achieving the technical effect of improving model recognition accuracy and reducing interference from the surrounding environment.

[0098] 3. Through the one-dimensional convolutional residual network to automatically extract the input unknown process PET finished product data for identification, it can explore the deeper features of the unknown process PET finished product data and discover the connection and difference between different data signals. At the same time, by adding unknown process PET finished product data to the training data and adding the central error loss function to the fully connected layer of the network, the network can perform open set identification of PET finished products.

[0099] 4. By using principal component analysis to reduce the dimension of the detection quality attribute information, the system calculation complexity is reduced, thereby improving the speed of PET finished product packaging quality detection. In addition, by assigning weights to each attribute information through factor analysis, the accuracy and reliability of the PET finished product packaging quality detection results are improved.

[0100] 5. Through multiple rounds of data processing and calculation, the redundant processing of the detection quality attribute information is completed, making the retained data more universally applicable and reducing the complexity of subsequent data analysis.

[0101] Embodiment 2

[0102] Based on the same inventive concept as the PET finished product packaging quality detection method in the aforementioned embodiment, the present invention also provides a PET finished product packaging quality detection system, see the attached Figure 5 , the system comprising:

[0103] A first obtaining unit 11, wherein the first obtaining unit 11 is used to obtain a data set of PET finished products with known processes and a data set of PET finished products with unknown processes through a big data platform;

[0104] A first execution unit 12, the first execution unit 12 is used to perform image enhancement on the known process PET finished product data set and the unknown process PET finished product data set based on a generative adversarial network;

[0105] A second obtaining unit 13, wherein the second obtaining unit 13 is used to model the PET finished product data through a one-dimensional convolutional residual network according to the known process PET finished product data set and the unknown process PET finished product data set after image enhancement, so as to obtain a PET finished product recognition network model;

[0106] A third obtaining unit 14, the third obtaining unit 14 is used to obtain first image information through an image acquisition device, the first image information includes image information of a first PET finished product;

[0107] A fourth obtaining unit 15, wherein the fourth obtaining unit 15 is used to input the first image information into the PET finished product recognition network model to obtain a first recognition quality result;

[0108] A fifth obtaining unit 16, the fifth obtaining unit 16 is used to detect the material properties of the first PET finished product according to a predetermined detection standard and a detection index set to obtain a first material detection result;

[0109] The sixth obtaining unit 17 is used to obtain a first predetermined weight distribution result through a principal component analysis method, and perform weighted calculation on the first recognition quality result and the first material detection result according to the first predetermined weight distribution result to obtain a first PET finished product packaging quality detection result.

[0110] Furthermore, the system also includes:

[0111] A seventh obtaining unit, wherein the seventh obtaining unit is used to initialize the parameters of the generator and the discriminator, and obtain initial generator parameters and initial discriminator parameters;

[0112] An eighth obtaining unit, wherein the eighth obtaining unit is used to obtain a threshold value of the number of parameter updates of a predetermined generative adversarial network;

[0113] A first generating unit, the first generating unit is used to update the initial generator parameters and the initial discriminator parameters according to the parameter update number threshold, and generate parameter updating steps: the parameters of the generator remain unchanged, the discriminator is trained, and the parameters of the discriminator remain unchanged, and the generator is trained;

[0114] A second execution unit is used to perform image enhancement on the known process PET finished product data set and the unknown process PET finished product data set based on the parameter updating step.

[0115] Furthermore, the system also includes:

[0116] A first construction unit, wherein the first construction unit is used to construct a PET finished product open set identification data set, wherein the PET finished product open set identification data set includes the known process PET finished product data set and the unknown process PET finished product data set after image enhancement;

[0117] A third execution unit, the third execution unit is used to redistribute the data in the PET finished product open set identification data set, and divide the redistributed data into a training set and a verification set according to a predetermined ratio;

[0118] A fourth execution unit, wherein the fourth execution unit is used to input the training set into a 1D-ResNet network structure for training;

[0119] A ninth obtaining unit, wherein the ninth obtaining unit is used to obtain a 1D-ResNet loss function by weighted summing a multi-classification cross entropy loss function and a center error loss function;

[0120] A first calculation unit, wherein the first calculation unit is used to calculate the loss value of the model on the training set and the validation set according to the 1D-ResNet loss function;

[0121] The tenth obtaining unit is used to complete the training process of the 1D-ResNet network to obtain the PET finished product recognition network model when the loss value reaches a preset number of training times or the loss value of the network has not decreased for a predetermined number of consecutive times.

[0122] Furthermore, the system also includes:

[0123] A second construction unit, the second construction unit is used to construct a set of detection indicators, the set of detection indicators includes barrier properties, impact strength, light transmittance index, and thermal stability performance;

[0124] an eleventh obtaining unit, the eleventh obtaining unit being used to score the material properties of the first PET finished product according to a predetermined testing standard, and obtain a testing scoring matrix corresponding to each indicator in the testing indicator set;

[0125] A twelfth obtaining unit, the twelfth obtaining unit is used to obtain the correlation between the indicators in the detection indicator set, and construct an indicator evaluation matrix according to the evaluation coefficients of the indicators and the correlation;

[0126] A thirteenth obtaining unit is used to obtain a first material detection result based on a product calculation result of the detection score matrix and the index evaluation matrix.

[0127] Furthermore, the system also includes:

[0128] A fourteenth obtaining unit, the fourteenth obtaining unit is used to obtain the detection quality attribute information of the PET finished product packaging;

[0129] a fifteenth obtaining unit, the fifteenth obtaining unit being used to perform principal component analysis on the detection quality attribute information to obtain first dimension reduction attribute information;

[0130] A sixteenth obtaining unit is used to obtain a first predetermined weight distribution result based on factor analysis of the first dimensionality reduction attribute information.

[0131] Furthermore, the system also includes:

[0132] A seventeenth obtaining unit, the seventeenth obtaining unit is used to obtain a first covariance matrix after performing decentralized processing on the detection quality attribute information;

[0133] an eighteenth obtaining unit, the eighteenth obtaining unit being used to calculate the first covariance matrix to obtain a first eigenvalue and a first eigenvector of the first covariance matrix;

[0134] A first projection unit, wherein the first projection unit is used to project the detection quality attribute information onto the first eigenvector to obtain first dimension reduction attribute information, wherein the first dimension reduction attribute information is attribute information obtained after the detection quality attribute information is dimensionally reduced.

[0135] Furthermore, the system also includes:

[0136] a nineteenth obtaining unit, the nineteenth obtaining unit being used to obtain an actual identification quality result of the first PET finished product;

[0137] a twentieth obtaining unit, configured to obtain a first quality detection deviation based on a difference between the actual recognition quality result and the first recognition quality result;

[0138] The twenty-first obtaining unit is used to perform correction training on the PET finished product recognition network model according to the first quality detection deviation to obtain a PET finished product corrected recognition network model.

[0139] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. Figure 1 The PET finished product packaging quality detection method and specific examples in the first embodiment are also applicable to the PET finished product packaging quality detection system of the present embodiment. Through the above detailed description of the PET finished product packaging quality detection method, those skilled in the art can clearly know the PET finished product packaging quality detection system of the present embodiment, so for the sake of brevity of the specification, it will not be described in detail here. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0140] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0141] Exemplary Electronic Devices

[0142] Reference below Figure 6 To describe the electronic device of the embodiment of the present application.

[0143] Figure 6 The figure shows a schematic structural diagram of an electronic device according to an embodiment of the present application.

[0144] Based on the inventive concept of a PET finished product packaging quality inspection method in the aforementioned embodiment, the present invention also provides a PET finished product packaging quality inspection system, on which a computer program is stored, and when the program is executed by a processor, the steps of any method of the PET finished product packaging quality inspection method described above are implemented.

[0145] Among them, Figure 6 In the embodiment of the present invention, a bus architecture (represented by bus 300) is shown, which may include any number of interconnected buses and bridges, and bus 300 links various circuits including one or more processors represented by processor 302 and memory represented by memory 304. Bus 300 may also link various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are not further described herein. Bus interface 305 provides an interface between bus 300 and receiver 301 and transmitter 303. Receiver 301 and transmitter 303 may be the same element, namely a transceiver, which provides a unit for communicating with various other devices over a transmission medium.

[0146] The processor 302 is responsible for managing the bus 300 and general processing, while the memory 304 may be used to store data used by the processor 302 when performing operations.

[0147] The present application provides a PET finished product packaging quality detection method, which is applied to a PET finished product packaging quality detection system, wherein the method includes: obtaining a known process PET finished product data set and an unknown process PET finished product data set through a big data platform; performing image enhancement on the known process PET finished product data set and the unknown process PET finished product data set based on a generative adversarial network; modeling the PET finished product data through a one-dimensional convolutional residual network according to the known process PET finished product data set and the unknown process PET finished product data set after image enhancement, and obtaining a PET finished product recognition network model; obtaining first image information through the image acquisition device, the first image information including image information of a first PET finished product; inputting the first image information into the PET finished product recognition network model to obtain a first recognition quality result; detecting the material properties of the first PET finished product according to a predetermined detection standard and a detection index set to obtain a first material detection result; obtaining a first predetermined weight distribution result through a principal component analysis method, and performing weighted calculation on the first recognition quality result and the first material detection result according to the first predetermined weight distribution result to obtain a first PET finished product packaging quality detection result. The invention solves the problem that in the prior art, when the packaging quality of PET finished product packaging is detected by computer technology, only conventional defects such as the appearance quality of the PET finished product are detected, but the material properties in the PET finished product packaging cannot be intelligently detected at the same time, resulting in the technical problem of incomplete detection when the quality of the PET finished product packaging is detected. By using a generative adversarial network to enhance the image of PET finished product packaging with known and unknown processes, and then constructing a PET finished product recognition network model based on the PET finished product packaging after image enhancement, the recognition accuracy of the model is improved, and the appearance quality of the PET finished product packaging is intelligently detected, thereby improving the detection efficiency of the PET finished product packaging. In addition, the material properties of the PET finished product packaging are intelligently detected, and the detection results of the PET finished product packaging quality are further obtained by weighted calculation. The technical effect of comprehensively detecting the material and appearance quality of the PET finished product packaging and improving the accuracy and efficiency of the PET finished product packaging quality detection is achieved.

[0148] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, devices, or computer program products. Therefore, the present application may adopt a complete software embodiment, a complete hardware embodiment, or a combination of software and hardware embodiments. In addition, the present application is in the form of a computer program product that can be implemented on one or more computer-usable storage media containing computer-usable program codes. The computer-usable storage medium includes, but is not limited to, various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), disk storage, compact disc read-only memories (CD-ROM), optical memories, etc.

[0149] The present invention is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products of the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A system that specifies the functions of a box or multiple boxes.

[0150] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction system, which is implemented in the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0151] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1The steps of the functions specified in a box or multiple boxes. Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic creative concepts. Therefore, the attached claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0152] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A PET finished product packaging quality detection method, characterized in that: The method is applied to a packaging quality detection system, the system includes an image acquisition device, and the method includes: Obtain a data set of PET finished products with known processes and a data set of PET finished products with unknown processes through a big data platform; Performing image enhancement on the known process PET finished product data set and the unknown process PET finished product data set based on a generative adversarial network; According to the known process PET finished product data set and the unknown process PET finished product data set after image enhancement, modeling the PET finished product data through a one-dimensional convolutional residual network to obtain a PET finished product recognition network model; Obtaining first image information through the image acquisition device, wherein the first image information includes image information of a first PET finished product; Inputting the first image information into the PET finished product recognition network model to obtain a first recognition quality result; Testing the material properties of the first PET finished product according to a predetermined testing standard and a set of testing indicators to obtain a first material testing result; A first predetermined weight distribution result is obtained by principal component analysis, and a weighted calculation is performed on the first recognition quality result and the first material detection result according to the first predetermined weight distribution result to obtain a first PET finished product packaging quality detection result.

2. The method according to claim 1, characterized in that The image enhancement of the known process PET finished product data set and the unknown process PET finished product data set based on the generative adversarial network includes: Initialize the parameters of the generator and discriminator to obtain the initial generator parameters and initial discriminator parameters; Obtaining a predetermined threshold value of parameter updates of a generative adversarial network; According to the parameter update times threshold, the initial generator parameters and the initial discriminator parameters are updated to generate parameter update steps: the parameters of the generator remain unchanged and the discriminator is trained, and the parameters of the discriminator remain unchanged and the generator is trained; Based on the parameter updating step, image enhancement is performed on the known process PET finished product data set and the unknown process PET finished product data set.

3. The method according to claim 1, characterized in that The method of obtaining the PET finished product identification network model comprises: Constructing a PET finished product open set identification data set, wherein the PET finished product open set identification data set includes the known process PET finished product data set and the unknown process PET finished product data set after image enhancement; Redistributing the data in the PET finished product open set identification data set, and dividing the redistributed data into a training set and a validation set according to a predetermined ratio; Input the training set into a 1D-ResNet network structure for training; The 1D-ResNet loss function is obtained by weighted summation of the multi-classification cross entropy loss function and the center error loss function; Calculate the loss value of the model on the training set and the validation set according to the 1D-ResNet loss function; When the loss value reaches a preset number of training times or the loss value of the network does not decrease for a predetermined number of consecutive times, the training process of the 1D-ResNet network is completed to obtain the PET finished product recognition network model.

4. The method according to claim 1, characterized in that The obtaining of the first material detection result comprises: Constructing a set of testing indicators, wherein the set of testing indicators includes barrier performance, impact strength, light transmittance index, and thermal stability performance; Score the material properties of the first PET finished product according to a predetermined testing standard to obtain a testing score matrix corresponding to each indicator in the testing indicator set; Obtaining the correlation between the indicators in the detection indicator set, and constructing an indicator evaluation matrix according to the evaluation coefficients of the indicators and the correlation; Based on the product calculation result of the detection score matrix and the index evaluation matrix, a first material detection result is obtained.

5. The method according to claim 1, characterized in that The step of obtaining the first PET finished product packaging quality test result also includes: Obtain information on the inspection quality attributes of finished PET packaging; Performing principal component analysis on the detection quality attribute information to obtain first dimension reduction attribute information; Based on factor analysis of the first dimension reduction attribute information, a first predetermined weight distribution result is obtained.

6. The method according to claim 5, characterized in that The performing principal component analysis on the detection quality attribute information to obtain first dimension reduction attribute information includes: After decentralizing the detection quality attribute information, a first covariance matrix is ​​obtained; Calculating the first covariance matrix to obtain a first eigenvalue and a first eigenvector of the first covariance matrix; The detection quality attribute information is projected onto the first eigenvector to obtain first dimension-reduced attribute information, wherein the first dimension-reduced attribute information is attribute information obtained after the detection quality attribute information is dimensionally reduced.

7. The method according to claim 1, characterized in that The method comprises: Obtaining an actual identification quality result of the first PET finished product; Obtaining a first quality detection deviation based on a difference between the actual recognition quality result and the first recognition quality result; The PET finished product recognition network model is corrected and trained according to the first quality detection deviation to obtain a PET finished product corrected recognition network model.

8. A PET finished product packaging quality inspection system, characterized in that: The system comprises: First acquisition unit: The first acquisition unit is used to obtain a data set of PET finished products with known processes and a data set of PET finished products with unknown processes through a big data platform; The first execution unit: the first execution unit is used to perform image enhancement on the known process PET finished product data set and the unknown process PET finished product data set based on a generative adversarial network; Second acquisition unit: The second acquisition unit is used to model the PET finished product data through a one-dimensional convolutional residual network according to the known process PET finished product data set and the unknown process PET finished product data set after image enhancement, so as to obtain a PET finished product recognition network model; A third obtaining unit: the third obtaining unit is used to obtain first image information through an image acquisition device, wherein the first image information includes image information of a first PET finished product; Fourth obtaining unit: the fourth obtaining unit is used to input the first image information into the PET finished product recognition network model to obtain a first recognition quality result; Fifth obtaining unit: the fifth obtaining unit is used to detect the material properties of the first PET finished product according to a predetermined testing standard and a set of testing indicators to obtain a first material testing result; Sixth acquisition unit: The sixth acquisition unit is used to obtain a first predetermined weight distribution result through a principal component analysis method, and perform weighted calculation on the first recognition quality result and the first material detection result according to the first predetermined weight distribution result to obtain a first PET finished product packaging quality detection result.

9. A PET finished product packaging quality inspection system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method according to any one of claims 1 to 7 are implemented.

Images