An intelligent food packaging defect detection system
Through multi-layer and multi-scene analysis of the intelligent food packaging defect detection system, the problems of inefficiency and inaccuracy of traditional manual detection methods are solved, and the accurate identification and positioning of food packaging defects are achieved, and the accuracy and efficiency of detection are improved. .
Patent Information
- Application Number
- CN202411217412.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-02
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-09-02
AI Technical Summary
Traditional manual testing methods are inefficient and susceptible to human factors, resulting in unstable and inaccurate results of food packaging defect detection, making it difficult to fully identify and detect packaging defects.
Design an intelligent food packaging defect detection system, including feature search module, layer division module, dynamic detection module, review and counter-check module and classification processing module, and achieve accurate food packaging defects through multi-layer detection and multi-scene analysis. Identify and locate.
It realizes multi-layer inspection of food packaging in multiple scenarios, improves the accuracy and efficiency of detection, can accurately identify and locate packaging defects, and reduces the influence of human factors.
Smart Images

Figure CN118735917B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of intelligent packaging detection, and in particular to an intelligent food packaging defect detection system. Background Art
[0002] With the increasing attention of consumers to food safety and quality, and the rapid development of the food industry, it is particularly important. During the production and transportation of food, food packaging is prone to various defects, such as cracks, stains, uneven colors, label errors, etc., and the traditional food packaging defect detection method mainly relies on manual inspection, but the manual detection method is not only inefficient, but also easily affected by human factors, resulting in inaccurate and unreliable detection results, and there is a technical problem that it is difficult to fully identify and detect packaging defects. Summary of the invention
[0003] The present application provides an intelligent food packaging defect detection system to solve the technical problems that traditional manual detection methods are inefficient and susceptible to human factors, resulting in unstable and inaccurate detection results, and it is difficult to fully identify and detect packaging defects. It realizes multi-layer detection of packaging in multiple scenarios, achieving the technical effect of accurately identifying and locating food packaging defects.
[0004] The present application provides an intelligent food packaging defect detection system, comprising: a feature search module, the feature search module is used to construct a feature solution space based on target food packaging data, traverse the feature solution space for searching, and extract a food packaging feature set; a layer division module, the layer division module is used to perform multi-layer division according to the food packaging feature set to generate N packaging layers, where N is an integer greater than 1; a dynamic detection module, the dynamic detection module is used to set multiple detection scenarios, perform dynamic detection on the N packaging layers according to the multiple detection scenarios combined with the food packaging feature set, and generate N packaging layer detection results; a review and reverse query module, the review and reverse query module is used to review and reverse query the N packaging layer detection results, update the N packaging layer detection results according to the review and reverse query results, and generate N packaging layer detection optimization results; a classification processing module, the classification processing module is used to send the N packaging layer detection optimization results to a remote control terminal to perform intelligent classification processing on the target food packaging.
[0005] In a possible implementation, the feature search module performs the following processing: randomly selecting multiple initial features to be detected according to packaging detection requirement information; constructing a packaging feature point array set through the multiple initial features to be detected; performing dimensionality reduction processing based on the packaging feature point array set to generate multiple reduced dimensionality feature arrays; and combining the multiple reduced dimensionality feature arrays to construct the feature solution space.
[0006] In a possible implementation, the feature search module performs the following processing: calculates the distribution probability of multiple packaging feature points and preset feature points in the packaging feature point array set to obtain a feature distribution array; constructs a probability distribution function, performs dimensionality reduction processing on the lattice distribution of the feature distribution array, and generates the divergence of the dimensionality reduction similarity of the feature distribution array; and constructs the multiple dimensionality reduction feature arrays based on the divergence of the dimensionality reduction similarity.
[0007] In a possible implementation, the feature search module performs the following processing: retrieve historical defect detection data, perform correlation analysis based on the target food packaging data and the historical defect detection data, and generate a detection correlation factor; based on the detection correlation factor, traverse the feature solution space to perform feature evaluation and generate a correlation feature score; based on the correlation feature score, perform search weighting on the feature solution space in combination with greedy search to determine multiple feature weight assignment values; sort the multiple feature weight assignment values in descending order, and extract the food packaging feature set according to the weight assignment value sequence.
[0008] In a possible implementation, the layer division module performs the following processing: performing cluster analysis based on the food packaging feature set to determine multiple packaging feature categories; setting a layer structure according to the target food packaging data, and creating multiple feature layers according to the layer structure and in combination with the multiple packaging feature categories; determining a layer overlay relationship based on the weight allocation value sequence; adjusting the multiple feature layers according to the layer overlay relationship to determine the N packaging layers.
[0009] In a possible implementation, the dynamic detection module performs the following processing: based on multiple food packaging usage scenarios combined with the packaging detection demand information, multiple detection scenarios are set; based on the multiple detection scenarios, a first detection scenario is randomly extracted, and defect detection is performed on the N packaging layers in combination with the food packaging feature set to determine whether the packaging features in the N packaging layers meet the preset packaging standard features; if the packaging features in the N packaging layers do not meet the preset packaging standard features, defects are marked for the N packaging layers, and according to the first defect identification result, the M packaging layers are added to the first detection scenario defect data record group, thereby iterating to obtain multiple detection scenario defect data record groups, where M is an integer less than or equal to N; according to the multiple detection scenario defect data record groups, detection identification is performed in combination with the N packaging layers to obtain the detection results of the N packaging layers.
[0010] In a possible implementation, the dynamic detection module performs the following processing: if the packaging features in the N packaging layers meet the preset packaging standard features, a switching instruction is generated; according to the switching instruction, the first detection scene is switched to the second detection scene, and defect detection is performed on the N packaging layers according to the second detection scene, thereby iterating until the multiple detection scenes are traversed and the scene detection results are obtained; based on the scene detection results, the i-th defect data and the i-th detection scene are extracted; based on the i-th detection scene, the N packaging layers are defect-identified using the i-th defect data to generate a second defect identification result; and the second defect identification result is added to the N packaging layer detection results.
[0011] In a possible implementation, the classification processing module performs the following processing: retrieve historical food packaging sample files for defect classification and determine the defect detection level; set defect classification standards through the remote control terminal in combination with the defect detection level, and determine whether the N packaging layer detection optimization results meet the preset classification threshold based on the defect classification standard and in combination with the self-inspection cycle; when the N packaging layer detection optimization results meet the preset classification threshold, generate a repairable label, synchronize the repairable label to the remote control terminal to repair the target food packaging; when the N packaging layer detection optimization results do not meet the preset classification threshold, generate an unrepairable label, backtrack based on the unrepairable label, determine the defect source information of the target food packaging, and formulate a correction plan based on the defect source information.
[0012] One or more technical solutions provided in this application have at least the following technical effects or advantages:
[0013] The intelligent food packaging defect detection system provided in the present application solves the technical problems that traditional manual detection methods are inefficient and susceptible to human factors, resulting in unstable and inaccurate detection results, and it is difficult to fully identify and detect packaging defects. It realizes multi-layer detection of packaging in multiple scenarios, achieving the technical effect of accurately identifying and locating food packaging defects. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the accompanying drawings of the embodiment of the present invention will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.
[0015] Figure 1A schematic diagram of the structure of an intelligent food packaging defect detection system provided in an embodiment of the present application;
[0016] Figure 2 A schematic diagram of a dynamic detection structure of an intelligent food packaging defect detection system provided in an embodiment of the present application.
[0017] Explanation of the accompanying drawings: feature search module 10, layer division module 20, dynamic detection module 30, review and reverse query module 40, classification processing module 50. 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 the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the 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 describe a subset of all possible embodiments, but it is 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 to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, 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 inherent to these processes, systems, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.
[0021] The present application embodiment provides an intelligent food packaging defect detection system, such as Figure 1 As shown, the system comprises:
[0022] A feature search module 10, the feature search module 10 is used to construct a feature solution space based on target food packaging data, traverse the feature solution space to search, and extract a food packaging feature set;
[0023] Furthermore, the feature search module 10 in the device provided in the embodiment of the application is also used for:
[0024] A plurality of initial features to be detected are randomly selected according to packaging detection requirement information; a packaging feature point array set is constructed by the plurality of initial features to be detected; a dimensionality reduction process is performed based on the packaging feature point array set to generate a plurality of reduced-dimensionality feature arrays; and the plurality of reduced-dimensionality feature arrays are combined to construct the feature solution space.
[0025] In this embodiment, according to the specific needs of packaging detection, the material, shape, size, color, texture, etc. of the food packaging may be included, and multiple initial features are randomly selected from a large number of possible features as the features to be detected. The multiple initial features to be detected may include image pixel values, color histograms, shape descriptors (such as perimeter, area, aspect ratio, etc.), texture features (such as grayscale co-occurrence matrix, Fourier transform coefficients, etc.) and other data. Furthermore, each initial feature to be detected is represented as one or more numerical values (or vectors), and these numerical values (or vectors) can constitute a feature point. Then, the feature points of all the initial features to be detected are combined to construct a packaging feature point array set, and the packaging feature point array set can contain information on all the initial features to be detected.
[0026] Since the original packaging feature point array set may contain a large number of feature points, directly using it for detection may lead to problems such as high computational complexity and poor real-time performance. Therefore, it is necessary to perform dimensionality reduction processing to reduce the number of features while retaining as much information as possible that is useful for detection. Through dimensionality reduction processing, multiple dimensionality reduction feature arrays can be obtained, and each dimensionality reduction feature array in the multiple dimensionality reduction feature arrays is a low-dimensional representation of the original feature point array set, and different dimensionality reduction methods or different dimensionality reduction parameters may obtain different dimensionality reduction feature arrays. The multiple dimensionality reduction feature arrays may describe the original data from different angles, so combining them can construct a more comprehensive feature solution space, and the feature solution space can be a set containing multiple dimensionality reduction feature arrays, and the feature solution space provides multiple low-dimensional representations of the original data. In packaging detection, different dimensionality reduction feature arrays can be selected or combined for detection according to specific detection tasks.
[0027] Furthermore, the feature search module 10 in the device provided in the embodiment of the application is also used for:
[0028] Calculate the distribution probability of multiple packaging feature points and preset feature points in the packaging feature point array set to obtain a feature distribution array; construct a probability distribution function, perform dimensionality reduction processing on the lattice distribution of the feature distribution array, and generate the divergence of the dimensionality reduction similarity of the feature distribution array; construct the multiple dimensionality reduction feature arrays according to the divergence of the dimensionality reduction similarity.
[0029] In this embodiment, by comparing the similarity or distance between two feature points, the distribution probability of each packaging feature point in the packaging feature point array set and the preset feature point is calculated in turn, thereby obtaining the distribution probability of multiple packaging feature points and the preset feature points, and constructing a feature distribution array on this basis. Each row or column of the feature distribution array can represent a packaging feature point, and the elements therein are the distribution probability values of these points and the preset feature points.
[0030] A probability distribution function is further constructed to describe the dot distribution in the feature distribution array, and the probability distribution function is as follows:
[0031] ;
[0032] in, To characterize the measure of the closeness between the dot lattice distribution of the characteristic distribution array and the reduced-dimensional dot lattice distribution of the characteristic distribution array, let The dimension reduction process is performed with the constraint condition of tending to 0. To characterize the divergence of the similarity between the dot lattice distribution of the feature distribution array and the reduced-dimensional dot lattice distribution of the feature distribution array, is a dimension reduction dot matrix distribution based on the feature distribution array. With dot matrix The similarity probability of is a dot matrix distribution based on the characteristic distribution array With dot matrix The similarity probability.
[0033] The feature distribution array is subjected to dimensionality reduction processing by using a probability distribution function. The goal of the dimensionality reduction processing is to convert the high-dimensional feature distribution array into a low-dimensional representation while retaining the key information in the original data as much as possible, and to calculate and obtain the divergence of the dimensionality reduction similarity of the feature distribution array. The divergence of the dimensionality reduction similarity of the feature distribution array is an indicator used to measure the degree of dispersion between data points, and finally multiple dimensionality reduction feature arrays are constructed according to the divergence of the dimensionality reduction similarity. Specifically, since the data points represented by the dimensionality reduction feature array with a smaller divergence are more compact and similar in the low-dimensional space, the dimensionality reduction feature array with a smaller divergence can be selected as the final feature representation, thereby selecting different numbers of dimensionality reduction feature arrays according to the divergence of the dimensionality reduction similarity to construct multiple dimensionality reduction feature arrays.
[0034] Furthermore, the feature search module 10 in the device provided in the embodiment of the application is also used for:
[0035] Retrieve historical defect detection data, perform correlation analysis based on the target food packaging data and the historical defect detection data, and generate a detection correlation factor; based on the detection correlation factor, traverse the feature solution space to perform feature evaluation and generate a correlation feature score; based on the correlation feature score, perform search weighting on the feature solution space in combination with greedy search to determine multiple feature weight assignment values; sort the multiple feature weight assignment values in descending order, and extract the food packaging feature set according to the weight assignment value sequence.
[0036] In this embodiment, first, historical defect detection data related to food packaging is retrieved from a database, and the historical defect detection data may include information such as defect type, defect description, conditions under which the defect occurs, and features related to the defect. Then, the target food packaging data (such as images, physical parameters, etc.) is associated with the historical defect detection data for analysis, which means that by comparing the features in the target food packaging data with those in the historical data, possible similarities or differences are identified. The degree of association between the target food packaging data and different defect types may be calculated by using statistical analysis or machine learning algorithms such as decision trees, random forests, and logistic regression, and based on the results of the association analysis, one or more detection association factors are generated for each defect type.
[0037] The detection correlation factor is further used to traverse each reduced-dimensional feature array in the feature solution space, and the matching degree or correlation between each feature array and the detection correlation factor is calculated. The feature evaluation can be performed based on factors such as the feature's discrimination, importance, and stability to generate an associated feature score. Then, a greedy search strategy (such as a greedy algorithm) is used to search for the optimal feature combination in the feature solution space. This means that in each iteration, the current optimal feature is selected according to the associated feature score, and a weight value is assigned to it, and this process is repeated until the preset number of features is reached. The weight value is used to reflect the importance or contribution of the feature in defect detection.
[0038] Furthermore, the weight assignment values of all features are sorted in descending order from large to small to generate a weight assignment value sequence. According to the weight assignment value sequence after descending order, the first p features are selected as the food packaging feature set, where p is an integer less than the weight assignment value sequence and greater than 1, so as to improve the accuracy and efficiency of food packaging defect detection.
[0039] A layer division module 20, the layer division module 20 is used to perform multi-layer division according to the food packaging feature set to generate N packaging layers, where N is an integer greater than 1;
[0040] Furthermore, the layer division module 20 in the device provided in the embodiment of the application is also used for:
[0041] A cluster analysis is performed based on the food packaging feature set to determine a plurality of packaging feature categories; a layer structure is set according to the target food packaging data, and a plurality of feature layers are created according to the layer structure and in combination with the plurality of packaging feature categories; a layer overlay relationship is determined based on the weight allocation value sequence; and the plurality of feature layers are adjusted according to the layer overlay relationship to determine the N packaging layers.
[0042] In this embodiment, a clustering algorithm (such as K-means, hierarchical clustering, DBSCAN, etc.) is first used to perform cluster analysis on the food packaging feature set, which means that the video packaging feature set is grouped based on the similarity between the video packaging feature sets, so as to obtain multiple packaging feature categories. Then, a target food packaging data is used as basic information to design a layer structure. The layer structure can be a hierarchical structure, wherein each level represents a feature category or feature group, which is used to organize and manage different features in the target food packaging data. Furthermore, based on the layer structure, multiple feature layers are created in combination with multiple packaging feature categories, which means that according to the multiple packaging feature categories obtained by cluster analysis, a feature layer is created for each category, and each feature layer contains all feature information belonging to the feature category, wherein the feature layer can be represented by an image, a chart, a text or other visual form, so as to obtain multiple feature layers.
[0043] Furthermore, determining the layer overlay relationship based on the weight allocation value sequence means firstly assigning a weight value to each feature layer based on the weight allocation value sequence, wherein the weight value is used to reflect the importance or contribution of the corresponding feature layer in defect detection, and determining the overlay relationship between the feature layers according to the weight value, that is, the layer with a higher weight value should be located at the top of the overlay order so that it can be given priority in defect detection, thereby determining the layer overlay relationship, and finally adjusting and arranging the multiple feature layers according to the determined layer overlay relationship to ensure that the layer with a higher weight value is located at the top so that it can be considered and identified first during defect detection, and outputting N packaging layers, wherein the N packaging layers can intuitively obtain the distribution and performance of different feature categories in the target food packaging data, which is helpful for subsequent defect detection and analysis.
[0044] A dynamic detection module 30, the dynamic detection module 30 is used to set a plurality of detection scenarios, perform dynamic detection on the N packaging layers according to the plurality of detection scenarios combined with the food packaging feature set, and generate N packaging layer detection results;
[0045] Further, such as Figure 2 As shown, the dynamic detection module 30 in the device provided in the embodiment of the application is also used for:
[0046] Based on multiple food packaging usage scenarios combined with the packaging detection demand information, multiple detection scenarios are set; based on the multiple detection scenarios, a first detection scenario is randomly extracted, and defect detection is performed on the N packaging layers in combination with the food packaging feature set to determine whether the packaging features in the N packaging layers meet the preset packaging standard features; if the packaging features in the N packaging layers do not meet the preset packaging standard features, defects are marked for the N packaging layers, and according to the first defect marking result, the M packaging layers are added to the first detection scenario defect data record group, thereby iterating to obtain multiple detection scenario defect data record groups, where M is an integer less than or equal to N; according to the multiple detection scenario defect data record groups, detection marking is performed in combination with the N packaging layers to obtain the detection results of the N packaging layers.
[0047] In this embodiment, firstly, according to a plurality of food packaging usage scenarios and in combination with packaging detection demand information, a plurality of different detection scenarios are set, wherein each of the plurality of detection scenarios covers different factors such as packaging types, usage environments, and detection focuses. Since random selection helps to comprehensively test the effectiveness of each detection scenario and feature set, one of the plurality of detection scenarios is randomly selected as the first detection scenario, and the above-constructed food packaging feature set is used to perform defect detection on N packaging layers. Then, according to the feature points in the food packaging feature set and the preset detection algorithm, it is determined whether the packaging features in each packaging layer meet the preset packaging standard features. When the packaging features in the N packaging layers do not meet the preset packaging standard features, it is determined that there are M packaging layers with defects in the first detection scenario. On this basis, the N packaging layers are defectively marked, which means that only the packaging layers with defects in the first detection scenario are defectively marked. The defect marking may include but is not limited to highlighting, marking positions, recording defect types, and other information, thereby generating a first defect marking result. According to the first defect marking result, Add M defective packaging layers (M is less than or equal to N) to the defect data record group of the first detection scene, that is, the first detection scene defect data record group, which contains detailed information of the defective packaging layer in the first detection scene, that is, the defect type, location, degree, etc., and repeat the above steps to iteratively detect other detection scenes in multiple detection scenes, and in each detection scene, randomly extract and use the food packaging feature set for defect detection, and record the defect data. After iterative detection, defect data record groups of multiple detection scenes will eventually be obtained, that is, multiple detection scene defect data record groups, and each record group in the multiple detection scene defect data record groups contains the defective layer found in the corresponding detection scene and its detailed information. According to all the detection scene defect data record groups, combined with N packaging layers, the defect frequency, severity, common defect type and other information of each layer are counted for overall detection identification, so as to obtain N packaging layer detection results, realize comprehensive evaluation of the quality and performance of food packaging in different usage scenarios, and provide strong support for subsequent improvement and optimization.
[0048] Furthermore, the dynamic detection module 30 in the device provided in the embodiment of the application is also used for:
[0049] If the packaging features in the N packaging layers meet the preset packaging standard features, a switching instruction is generated; according to the switching instruction, the first detection scene is switched to the second detection scene, and defect detection is performed on the N packaging layers according to the second detection scene, and this is iterated until the multiple detection scenes are traversed and the scene detection results are obtained; based on the scene detection results, the i-th defect data and the i-th detection scene are extracted; based on the i-th detection scene, defects are marked for the N packaging layers using the i-th defect data to generate a second defect marking result; and the second defect marking result is added to the N packaging layer detection results.
[0050] In this embodiment, when the packaging features in the N packaging layers meet the preset packaging standard features, a switching instruction is generated, and the switching instruction is used to indicate that the current detection scene (the first detection scene) has been completed, that is, no defects in any packaging layer are found in the first detection scene, and it is allowed to switch to the next detection scene, that is, according to the switching instruction, the current first detection scene is switched to the second detection scene, and further in the second detection scene, by using the same food packaging feature set and detection algorithm as the first detection scene, the N packaging layers are again defect-detected, and defect detection is iteratively performed in each detection scene until all preset detection scenes are traversed. After all detection scenes are traversed, the detection results of each scene are integrated to generate a scene detection result, and the scene detection result may include defect data, defect type, frequency, etc. found in each scene.
[0051] Then, based on the scene detection results, the i-th detection scene and its corresponding i-th defect data are extracted, wherein, since no defects are found in any packaging layer in the first detection scene, i is any positive integer other than one of the total number of multiple detection scenes, which can be 0, and i can be an index representing the position of the scene and defect data in the result list. Furthermore, based on the extracted i-th detection scene, the i-th defect data is used to perform defect identification operations such as marking defect locations and recording defect types on N packaging layers on the image, thereby generating a second defect identification result, which can include all layers identified as defects and their detailed information. Finally, the second defect identification result is added to the overall detection result of the N packaging layers. If there are detection results before, they are merged or updated to ensure that the latest defect information is included. When all detection scenes have completed defect identification and result updates, outputting the final N packaging layer detection results is helpful to understand the performance of the product in different usage environments and make improvements accordingly.
[0052] A review and reverse checking module 40, the review and reverse checking module 40 is used to review and reverse check the N packaging layer detection results, update the N packaging layer detection results according to the review and reverse checking results, and generate N packaging layer detection optimization results;
[0053] In this embodiment, the methods, standards and processes for the review and reverse checking are determined based on the historical inspection record data, and then the inspection results of N packaging layers are checked one by one. At the same time, according to the preset review and reverse checking standards, the accuracy, completeness and consistency of the inspection results are evaluated, and the problems, omissions or misjudged defects found in the review and reverse checking process are recorded in detail, that is, the opinions, suggestions and conclusions of the review and reverse checking are recorded, and then the review and reverse checking results are summarized and analyzed to determine the inspection results that need to be updated or revised, thereby obtaining the review and reverse checking results.
[0054] Further, updating and optimizing the N packaging layer detection results according to the review and reverse checking results means that the misjudged defects can be removed from the original detection results or corrected to the correct defect type, and the omitted defects can be added to the detection results, and the corresponding defect information and description can be supplemented, and additional annotations or marks can be added to the suspected defects that need further attention for subsequent processing. After completing the update and optimization of the N packaging layer detection results, it is necessary to compare with the judgment of experts or quality inspectors, or use other reliable detection methods to verify the updated N packaging layer detection results to ensure that the updated detection results are more accurate, reliable and complete. After verification, the updated detection results are organized into N packaging layer detection optimization results. The N packaging layer detection optimization results can not only contain the detailed detection results, defect type, location, degree and other information of each packaging layer, but also add some additional information, such as detection time, detection personnel, detection algorithm used, etc., for subsequent tracking and tracing, and ensure the accuracy and reliability of the detection results through verification and feedback.
[0055] The classification processing module 50 is used to send the N packaging layer detection optimization results to the remote control terminal to perform intelligent classification processing on the target food packaging.
[0056] Furthermore, the classification processing module 50 in the device provided in the embodiment of the application is also used for:
[0057] Retrieve historical food packaging sample files for defect grading and determine the defect detection level; set defect classification standards through the remote control terminal in combination with the defect detection level, and determine whether the N packaging layer detection optimization results meet the preset classification threshold based on the defect classification standard and in combination with the self-inspection cycle; when the N packaging layer detection optimization results meet the preset classification threshold, generate a repairable label, synchronize the repairable label to the remote control terminal to repair the target food packaging; when the N packaging layer detection optimization results do not meet the preset classification threshold, generate an unrepairable label, backtrack based on the unrepairable label, determine the defect source information of the target food packaging, and formulate a correction plan based on the defect source information.
[0058] In this embodiment, the historical food packaging sample archive is first analyzed to classify various possible defects, and a corresponding level is set for each type of defect, that is, a corresponding defect detection level is set for the current food packaging inspection. The defect detection level can be used to determine which defects should be paid attention to during the inspection process and the severity of these defects. In combination with the defect detection level, a defect classification standard is set through the remote control terminal. The remote control terminal can be a user interface or a software platform, which is used to remotely manage and control the inspection process of food packaging. In the remote control terminal, according to the defect detection level determined above, a specific defect classification standard is set so that the detected defects can be automatically classified during the inspection process, and according to the defect classification standard, combined with the self-inspection cycle, it is judged whether the optimization results of the N packaging layer inspections meet the preset classification threshold, which means that the N packaging layers are first inspected, and the inspection results are optimized according to the set defect classification standard, and then combined with the set self-inspection cycle (for example, self-inspection every hour, every day or every week), it is judged whether these optimized inspection results meet the preset classification threshold.
[0059] Furthermore, food packaging can be divided into two categories, repairable and unrepairable, according to the severity and nature of the defects. For repairable defects, such as small stains and slight deformation, corresponding repair measures can be taken; for unrepairable defects, such as large-area damage and severe deformation, they need to be scrapped directly, that is, when the detection optimization results of N packaging layers meet the preset classification threshold, it means that these defects can be repaired by simple repair means. At this time, repairable labels are generated and synchronized to the remote control terminal to repair the target food packaging. When the detection optimization results of N packaging layers do not meet the preset classification threshold, it means that these defects are serious and cannot be repaired by simple repair means. At this time, unrepairable labels are generated and the backtracking mechanism is triggered. Through the backtracking mechanism, the food packaging with unrepairable labels is inspected in various links of the production process, such as raw materials, production equipment, process parameters, etc., and the defect source information is determined. Finally, according to the defect source information obtained by backtracking, a corresponding correction plan is formulated. The correction plan includes measures such as replacing raw materials, adjusting production equipment or process parameters, and strengthening quality inspections to prevent similar defects from happening again. Through continuous testing, analysis, repair and correction, the quality of food packaging is ensured to meet standards, thereby improving the safety of food packaging.
[0060] The embodiments of the present application solve the technical problems that traditional manual inspection methods are inefficient and easily affected by human factors, resulting in unstable and inaccurate inspection results, and it is difficult to fully identify and detect packaging defects. It realizes multi-layer inspection of packaging in multiple scenarios, achieving the technical effect of accurately identifying and locating food packaging defects.
[0061] The above specific implementation manner does not constitute a limitation to the protection scope of the present 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 substitutions and improvements made within the spirit and principles of the present application should be included in the protection scope of the present application. In some cases, the actions or steps recorded in the present application can be performed in an order different from that in the embodiment and can still achieve the desired results. In addition, the process depicted in the accompanying drawings does 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. An intelligent food packaging defect detection system, characterized in that: The system comprises: A feature search module, the feature search module is used to construct a feature solution space based on target food packaging data, traverse the feature solution space to search, and extract a food packaging feature set; A layer division module, the layer division module is used to perform multi-layer division according to the food packaging feature set to generate N packaging layers, where N is an integer greater than 1; A dynamic detection module, the dynamic detection module is used to set a plurality of detection scenarios, perform dynamic detection on the N packaging layers according to the plurality of detection scenarios combined with the food packaging feature set, and generate N packaging layer detection results; A review and reverse check module, the review and reverse check module is used to review and reverse check the N packaging layer detection results, update the N packaging layer detection results according to the review and reverse check results, and generate N packaging layer detection optimization results; A classification processing module, the classification processing module is used to send the N packaging layer detection optimization results to the remote control terminal to perform intelligent classification processing on the target food packaging; The layer division module is also used for: Performing cluster analysis based on the food packaging feature set to determine multiple packaging feature categories; Setting a layer structure according to the target food packaging data, and creating a plurality of feature layers according to the layer structure and in combination with the plurality of packaging feature categories; Determine the layer overlay relationship based on the weight distribution value sequence; The multiple feature layers are adjusted according to the layer overlay relationship to determine the N packaging layers.
2. The intelligent food packaging defect detection system according to claim 1, characterized in that: The feature search module is also used for: Randomly select multiple initial features to be tested according to packaging testing requirement information; Constructing a packaging feature point array set by using the multiple initial features to be detected; Performing dimensionality reduction processing based on the packaging feature point array set to generate multiple dimensionality reduction feature arrays; The multiple dimension reduction feature arrays are combined to construct the feature solution space.
3. The intelligent food packaging defect detection system according to claim 2, characterized in that: The feature search module is also used for: Calculating the distribution probabilities of a plurality of packaging feature points and preset feature points in the packaging feature point array set to obtain a feature distribution array; Constructing a probability distribution function, performing dimensionality reduction processing on the dot distribution of the feature distribution array, and generating a divergence of the dimensionality reduction similarity of the feature distribution array; The plurality of dimensionality reduction feature arrays are constructed according to the divergence of the dimensionality reduction similarities.
4. The intelligent food packaging defect detection system according to claim 3, characterized in that: The feature search module is also used for: Retrieving historical defect detection data, performing correlation analysis based on the target food packaging data and the historical defect detection data, and generating a detection correlation factor; Based on the detection correlation factor, traverse the feature solution space to perform feature evaluation and generate a correlation feature score; According to the associated feature scores, the feature solution space is searched weighted in combination with greedy search to determine a plurality of feature weight assignment values; The plurality of feature weight allocation values are sorted in descending order, and the food packaging feature set is extracted according to the weight allocation value sequence.
5. The intelligent food packaging defect detection system according to claim 2, characterized in that: The dynamic detection module is also used for: Based on multiple food packaging usage scenarios and the packaging detection requirement information, multiple detection scenarios are set; Randomly extracting a first detection scene based on the multiple detection scenes, performing defect detection on the N packaging layers in combination with the food packaging feature set, and determining whether packaging features in the N packaging layers meet preset packaging standard features; If the packaging features in the N packaging layers do not meet the preset packaging standard features, the N packaging layers are defect-marked, and the M packaging layers are added to the first detection scene defect data record group according to the first defect identification result, thereby iterating to obtain multiple detection scene defect data record groups, where M is an integer less than or equal to N; According to the plurality of detection scene defect data record groups, detection identification is performed in combination with the N packaging layers to obtain the detection results of the N packaging layers.
6. The intelligent food packaging defect detection system according to claim 5, characterized in that: The dynamic detection module is also used for: If the packaging features in the N packaging layers meet the preset packaging standard features, a switching instruction is generated; Switching the first detection scene to a second detection scene according to the switching instruction, and performing defect detection on the N packaging layers according to the second detection scene, thereby iterating until the multiple detection scenes are traversed and a scene detection result is obtained; Extracting the i-th defect data and the i-th detection scene based on the scene detection result; Based on the i-th detection scenario, the i-th defect data is used to perform defect identification on the N packaging layers to generate a second defect identification result; The second defect identification result is added to the N packaging layer detection results.
7. The intelligent food packaging defect detection system according to claim 1, characterized in that: The classification processing module is also used for: Retrieve historical food packaging sample files for defect grading and determine defect detection levels; Defect classification standards are set through the remote control terminal in combination with the defect detection level, and according to the defect classification standards and in combination with the self-checking cycle, it is determined whether the N packaging layer detection optimization results meet the preset classification threshold; When the N packaging layer detection optimization results meet the preset classification threshold, a repairable label is generated, and the repairable label is synchronized to the remote control terminal to perform repair processing on the target food package; When the optimization results of the N packaging layer inspections do not meet the preset classification threshold, an irreparable label is generated, backtracking is performed based on the irreparable label, defect source information of the target food packaging is determined, and a correction plan is formulated based on the defect source information.
Citation Information
Patent Citations
Intelligent control method and system for automatic optical detection of apparent defects and medium
CN116188475A
Portable airplane surface corrosion defect detection and evaluation integrated system and method
CN117805133A
Casting quality tracking system of numerical control machine tool
CN118134062A
Flaw inspection device
JP1993101164A
