Multi-dimensional detection method and system for product defects of automatic stewed noodle production line

Abstract

The invention discloses a multi-dimensional detection method and system for product defects of an automatic stewed noodle production line, and relates to the related field of data processing, and the method comprises the steps: obtaining a plurality of dough products corresponding to a dough kneading control scheme according to an automatic dough kneading line of the automatic stewed noodle production line; according to the dough surface defect index set, performing surface variation detection on the plurality of dough products; introducing a health change detection and correction mechanism to carry out health change analysis on the plurality of dough products; performing gluten strength variation detection on the plurality of dough products according to the dough kneading control scheme; and performing global defect evaluation on the plurality of dough products according to the dough variation detection first matrix, the dough variation detection second matrix and the dough variation detection third matrix to obtain a dough defect detection report. The technical problems of inaccurate and incomplete detection results of product defect detection of an existing stewed noodle automatic production line are solved, and the technical effect of improving the accuracy and comprehensiveness of defect detection is achieved.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to a multi-dimensional detection method and system for product defects in an automated production line for braised noodles. Background Technology

[0002] As a popular traditional noodle dish, the quality of braised noodles directly impacts consumers' eating experience and market reputation. Defect detection in the dough during the production process is a crucial step in ensuring product quality; therefore, accurate detection of defects in braised noodle dough is essential. Currently, the main methods for detecting defects in braised noodle dough rely on simple observation and judgment based on human experience, or on some single-dimensional automated detection methods. However, these methods suffer from problems such as strong subjectivity, low efficiency, and susceptibility to fatigue errors in manual judgment. Furthermore, single-dimensional automated detection methods cannot comprehensively cover the multifaceted defect characteristics of the dough, leading to inaccurate and incomplete test results, and ultimately failing to effectively guarantee the quality of braised noodle dough.

[0003] Currently, the technology for detecting product defects in automated noodle production lines suffers from inaccurate and incomplete results. Summary of the Invention

[0004] This application provides a multi-dimensional detection method and system for product defects in an automated noodle production line. It employs a method based on an automated dough kneading line to acquire multiple dough products, and then performs surface anomaly detection (resulting in a first matrix), hygiene anomaly analysis (resulting in a second matrix), and gluten strength anomaly detection (resulting in a third matrix) on each dough. The three matrices are then combined to perform a global defect assessment, generating a dough defect detection report. This approach solves the technical problems of inaccurate and incomplete detection results in existing automated noodle production lines, achieving a significant improvement in the accuracy and comprehensiveness of defect detection.

[0005] This application provides a multi-dimensional detection method for product defects in an automated production line for braised noodles, comprising: obtaining multiple dough products corresponding to a dough kneading control scheme based on the automated dough kneading line of the automated production line for braised noodles; performing surface anomaly detection on the multiple dough products based on a dough surface defect index set to obtain a first dough anomaly detection matrix; introducing a hygiene anomaly detection correction mechanism to perform hygiene anomaly analysis on the multiple dough products to obtain a second dough anomaly detection matrix; performing gluten strength anomaly detection on the multiple dough products based on the dough kneading control scheme to obtain a third dough anomaly detection matrix; and performing a global defect assessment on the multiple dough products based on the first dough anomaly detection matrix, the second dough anomaly detection matrix, and the third dough anomaly detection matrix to obtain a dough defect detection report.

[0006] In a possible implementation, based on a set of dough surface defect indicators, surface anomaly detection is performed on the multiple dough products to obtain a first dough anomaly detection matrix. The following processing is then performed: pairwise similarity evaluation is performed on the set of dough surface defect indicators to obtain a defect indicator similarity evaluation set; decoupling optimization is performed on the set of dough surface defect indicators based on the defect indicator similarity evaluation set to obtain multiple independent defect indicators; convolutional feature learning is performed on the multiple independent defect indicators to obtain multiple defect convolutional capture models; and surface anomaly capture is performed on the multiple dough products based on the multiple defect convolutional capture models to generate the first dough anomaly detection matrix.

[0007] In a possible implementation, surface anomaly capture is performed on the multiple dough products using the multiple defect convolutional capture models, and the following processing is performed: A Kth dough product is extracted from the multiple dough products, and multi-angle image acquisition is performed on the Kth dough product using an industrial camera to obtain a Kth dough image, where K is a positive integer; Simultaneously, the Kth image acquisition scene data corresponding to the Kth dough image is retrieved, and anomaly identification is performed based on the Kth image acquisition scene data to obtain the Kth acquisition scene anomaly features; The Kth dough image is corrected based on the Kth acquisition scene anomaly features to obtain a Kth dough corrected image; The Kth dough corrected image is input into the multiple defect convolutional capture models to obtain multiple dough defect capture results, and the multiple dough defect capture results are fused to generate a Kth surface anomaly detection result.

[0008] In a possible implementation, the following processing is performed: the hygiene mutation detection and correction mechanism includes: performing colony mutation detection and correction on the Kth dough product according to the dough kneading control scheme to obtain a first hygiene mutation detection feature; performing pH value mutation detection and correction on the Kth dough product according to the dough kneading control scheme to obtain a second hygiene mutation detection feature; performing odor mutation detection and correction on the Kth dough product according to the dough kneading control scheme to generate a third hygiene mutation detection feature; generating a hygiene mutation detection result for the Kth dough product based on the first hygiene mutation detection feature, the second hygiene mutation detection feature, and the third hygiene mutation detection feature, and adding the hygiene mutation detection result for the Kth dough product to the second dough mutation detection matrix.

[0009] In a possible implementation, the Kth dough product is subjected to colony mutation detection correction according to the dough kneading control scheme to obtain a first feature for hygiene mutation detection, and the following processing is performed: Colony detection is performed on the Kth dough product to obtain Kth colony detection data; normal samples for dough colony detection are retrieved according to the dough kneading control scheme to construct a dough colony detection space; mutation capture is performed on the Kth colony detection data according to the dough colony detection space to obtain a Kth colony mutation capture result; detection interference analysis is performed based on the colony detection operation data and colony detection environment data corresponding to the Kth colony detection data to determine the Kth colony detection interference feature; the Kth colony detection interference feature is used to correct the Kth colony mutation capture result to generate the first feature for hygiene mutation detection.

[0010] In a possible implementation, the dough products are subjected to gluten strength variation detection according to the dough kneading control scheme to obtain a third matrix for dough variation detection. The following processes are then performed: gluten strength detection is performed on the multiple dough products to obtain gluten strength detection data for each dough, and gluten strength detection scenario data is recorded simultaneously; normal samples for dough gluten strength detection are retrieved according to the dough kneading control scheme to construct a gluten strength detection space; variation capture is performed on the gluten strength detection data of each dough according to the gluten strength detection space to obtain gluten strength variation features of each dough; interference compensation is performed on the gluten strength variation features of each dough according to the gluten strength detection scenario data to generate the third matrix for dough variation detection.

[0011] In a possible implementation, interference compensation is performed on the dough gluten strength variation features based on the data from each gluten strength detection scenario to generate the third dough variation detection matrix. The following processing is then performed: anomaly detection is performed based on the data from each gluten strength detection scenario to obtain anomaly features for each gluten strength detection scenario; integrated training is performed based on the historical set of gluten strength variation detection compensation to obtain a gluten strength variation detection compensation model; the anomaly features from each gluten strength detection scenario and the dough gluten strength variation features are input into the gluten strength variation detection compensation model to obtain compensation results for each gluten strength variation; and the third dough variation detection matrix is ​​constructed based on the compensation results for each gluten strength variation.

[0012] In a possible implementation, a global defect assessment is performed on the multiple dough products based on the first dough anomaly detection matrix, the second dough anomaly detection matrix, and the third dough anomaly detection matrix to obtain a dough defect detection report. The following processing is then performed: surface defect evaluation is conducted based on the first dough anomaly detection matrix to obtain the surface defect coefficient for each dough; hygiene defect evaluation is conducted based on the second dough anomaly detection matrix to obtain the hygiene defect coefficient for each dough; gluten strength defect evaluation is conducted based on the third dough anomaly detection matrix to obtain the gluten strength defect coefficient for each dough; and based on the multidimensional defect weights of the dough, the surface defect coefficients, hygiene defect coefficients, and gluten strength defect coefficients of each dough are weighted and fused to generate the dough defect detection report, which includes the global defect coefficients for each dough.

[0013] In a possible implementation, the following process is performed: determining whether the global defect coefficient of each dough is greater than or equal to the global defect threshold of the dough, obtaining the judgment result of each global defect, and generating a dough defect early warning signal based on the judgment result of each global defect.

[0014] This application also provides a multi-dimensional detection system for product defects in an automated production line for braised noodles, comprising: a dough product acquisition module, used to acquire multiple dough products corresponding to a dough kneading control scheme based on the automated dough kneading line of the automated braised noodles production line; a surface anomaly detection module, used to perform surface anomaly detection on the multiple dough products based on a dough surface defect index set, and obtain a first dough anomaly detection matrix; a hygiene anomaly analysis module, used to introduce a hygiene anomaly detection correction mechanism to perform hygiene anomaly analysis on the multiple dough products, and obtain a second dough anomaly detection matrix; a gluten strength anomaly detection module, used to perform gluten strength anomaly detection on the multiple dough products based on the dough kneading control scheme, and obtain a third dough anomaly detection matrix; and a global defect assessment module, used to perform a global defect assessment on the multiple dough products based on the first dough anomaly detection matrix, the second dough anomaly detection matrix, and the third dough anomaly detection matrix, and obtain a dough defect detection report.

[0015] This application proposes a multi-dimensional defect detection method and system for an automated noodle production line. First, based on the automated dough mixing line of the production line, multiple dough products corresponding to the dough mixing control scheme are obtained. Then, based on a set of dough surface defect indicators, surface anomaly detection is performed on these multiple dough products to obtain a first dough anomaly detection matrix. Next, a hygiene anomaly detection correction mechanism is introduced to analyze the hygiene anomalies of the multiple dough products, obtaining a second dough anomaly detection matrix. Then, based on the dough mixing control scheme, gluten strength anomaly detection is performed on the multiple dough products to obtain a third dough anomaly detection matrix. Finally, a global defect assessment is performed on the multiple dough products based on the first, second, and third dough anomaly detection matrices to obtain a dough defect detection report. This achieves the technical effect of improving the accuracy and comprehensiveness of defect detection. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a flowchart illustrating the multi-dimensional detection method for product defects in an automated noodle production line provided in this application embodiment.

[0018] Figure 2 A schematic diagram of the structure of the multi-dimensional detection system for product defects in the automatic production line of braised noodles provided in this application embodiment.

[0019] Figure labeling: 10 Dough product acquisition module, 20 Surface anomaly detection module, 30 Hygiene anomaly analysis module, 40 Gluten strength anomaly detection module, 50 Global defect assessment module. Detailed Implementation

[0020] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below.

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0023] This application provides a multi-dimensional detection method for product defects in an automated production line for braised noodles, such as... Figure 1 As shown, the method includes: Step S100: Based on the automatic dough mixing line of the braised noodles automatic production line, obtain multiple dough products corresponding to the dough mixing control scheme.

[0024] Specifically, a computer control system precisely controls the operating parameters of the dough mixer, such as mixing time, stirring speed, and water addition. These parameters are set according to a preset dough mixing control scheme to ensure the consistency of dough from different batches. The dough mixing control scheme is a pre-set combination of various parameters during the dough mixing process, including mixing time, stirring speed, and water addition, based on the requirements of the braised noodles production process, to ensure the quality and consistency of the dough. For example, the mixing speed of the dough mixer can be controlled by a variable frequency motor, the water addition is precisely controlled by a high-precision flow sensor and solenoid valve, and the mixing time is controlled by a timer. After mixing, an automated dough forming and dividing device cuts the large dough into multiple smaller dough balls. This device is equipped with high-precision cutting blades and weighing sensors to ensure that the weight of each small dough ball is within a specified range. For example, the dough dividing device can be set to divide the dough into small dough balls weighing 100 grams ± 2 grams each.

[0025] For example, the dough kneading control scheme requires a kneading time of 10 minutes, a stirring speed of 60 revolutions per minute, and a water addition of 40% of the flour weight. The automated dough kneading equipment operates according to these parameters. After kneading, the dough forming and dividing device divides the dough into multiple small dough balls, each weighing 100 grams, with a weight error controlled within ±2 grams.

[0026] Step S200: Based on the dough surface defect index set, perform surface anomaly detection on the multiple dough products to obtain the first matrix for dough anomaly detection.

[0027] Specifically, the dough surface defect index set is a set of quantitative indicators used to describe and measure surface defects in dough, such as crack length and width, area of ​​discolored spots, and color differences, to guide the implementation of surface anomaly detection. A high-resolution industrial camera is used to photograph the surface of each dough product, and then image processing algorithms are used to analyze the surface defects. For example, edge detection algorithms can be used to identify cracks on the dough surface, and color recognition algorithms can be used to detect discolored spots. Image processing software can convert the detected defect information into numerical data, forming the first matrix for dough anomaly detection. To improve the accuracy of image detection, the light source and background of the detection environment need to be controlled. For example, a uniform diffuse light source is used to avoid misjudgments caused by uneven lighting; a background with high contrast to the dough color is selected to more clearly identify surface defects.

[0028] For example, a 5-megapixel industrial camera equipped with a ring-shaped LED light source can be used to photograph the surface of the dough. Image processing algorithms detect a crack on the dough surface, 2 mm long and 0.5 mm wide; simultaneously, a discolored spot with an area of ​​1 square millimeter is also detected. This data is recorded to form the first matrix for dough anomaly detection; for example, one column in the matrix represents the crack length, and another column represents the area of ​​the discolored spot.

[0029] In one possible implementation, based on a set of dough surface defect indicators, surface anomaly detection is performed on the multiple dough products to obtain a first dough anomaly detection matrix. Step S200 further includes step S210, where pairwise similarity evaluation is performed on the set of dough surface defect indicators to obtain a defect indicator similarity evaluation set. Specifically, algorithms such as cosine similarity, Euclidean distance, or Pearson correlation coefficient are used to evaluate the similarity of each pair of indicators in the dough surface defect indicator set. For example, assuming the dough surface defect indicator set includes crack length, crack width, area of ​​discolored spots, and color difference of discolored spots, a similarity matrix is ​​obtained by calculating the similarity between these indicators. A large amount of sample data of dough surface defects is collected, and the data is standardized to facilitate similarity calculation. For example, the units of crack length and width are standardized to millimeters, the units of discolored spot area are standardized to square millimeters, and color differences are quantified using a color space (such as RGB or HSV).

[0030] For example, the defect index set includes: crack length (L), crack width (W), area of ​​discolored spots (A), and color difference of discolored spots (C). The similarity between each pair of indices is calculated using the cosine similarity algorithm, resulting in the similarity matrix shown in Table 1.

[0031] Table 1: Example of a similarity matrix Indicators L W A C L 1.00 0.85 0.20 0.10 W 0.85 1.00 0.15 0.05 A 0.20 0.15 1.00 0.90 C 0.10 0.05 0.90 1.00 Step S220: Decouple and optimize the dough surface defect index set based on the defect index similarity evaluation set to obtain multiple independent defect indices. Specifically, the PCA algorithm is used to reduce the dimensionality of the defect index similarity evaluation set and extract the principal components, thereby obtaining independent defect indices. PCA transforms the original data into a new coordinate system through linear transformation, making the new coordinate axes (principal components) independent of each other. Through decoupling optimization, redundancy and interference between indices are reduced.

[0032] Step S230: Convolutional feature learning is performed based on the multiple independent defect indicators to obtain multiple defect convolutional capture models. Specifically, a convolutional neural network model is constructed, using the independent defect indicators as input, and extracting features of surface defects in the dough through convolutional layers, pooling layers, and fully connected layers. For example, for crack features, a dedicated convolutional network is designed to capture the shape and distribution of cracks; for discolored spot features, another convolutional network is designed to capture the size and color of spots. The CNN model is trained using a large amount of labeled data (including dough images containing known defect types), and the network parameters are adjusted using the backpropagation algorithm to optimize model performance.

[0033] For example, for the crack feature, a CNN model with 3 convolutional layers and 2 pooling layers was constructed. The training dataset contained 1000 images of dough with cracks, each labeled with the location and length of the crack. After training, this model achieved 95% accuracy on the validation set. For the heterochromatic spot feature, another CNN model with 4 convolutional layers and 3 pooling layers was constructed. The training dataset contained 800 images of dough with heterochromatic spots, each labeled with the location, area, and color of the spot. After training, this model achieved 93% accuracy on the validation set.

[0034] Step S240: Surface anomaly capture is performed on the multiple dough products using the multiple defect convolutional capture models to generate the first dough anomaly detection matrix. Specifically, the trained defect convolutional capture models are applied to actual dough product detection. For each dough product, an surface image is captured using an industrial camera, and then the image is input into a CNN model. The model outputs the detected defect features and location information. The detected defect features and location information are integrated into the first dough anomaly detection matrix. For example, each row of the matrix represents a dough product, and each column represents a defect feature (such as crack length, area of ​​discolored spots, etc.).

[0035] For example, 100 dough products were inspected, and the two CNN models mentioned above were used to capture crack features and heterochromatic spot features, respectively. The detection results are shown in Table 2 (partial data).

[0036] Table 2: Example of the first matrix for dough anomaly detection Dough number Crack length (mm) <![CDATA[Area of heterochromatic spots (mm 2 )]]> 1 2.5 1.2 2 0.0 0.8 3 1.8 0.0 ... ... ... 100 0.0 1.5 This implementation separates highly correlated defect indicators through pairwise similarity evaluation and decoupling optimization, reducing redundancy and interference between indicators, making convolutional feature learning more accurate, and enabling the CNN model to focus more on learning its own features, thereby improving detection accuracy.

[0037] In one possible implementation, surface anomaly capture is performed on the multiple dough products according to the multiple defect convolution capture models. Step S240 further includes step S241, extracting the Kth dough product from the multiple dough products, and acquiring multi-angle images of the Kth dough product using an industrial camera to obtain the Kth dough image, where K is a positive integer. Specifically, the Kth dough product is extracted sequentially from the multiple dough products on the automatic noodle production line. For example, a robotic arm or a positioning device on a conveyor belt can be used to ensure that the extracted dough product is accurate each time. The Kth dough product is photographed from multiple angles (such as top, side, and bottom) using a high-resolution industrial camera. One or more images are taken from each angle to comprehensively capture the details of the dough surface.

[0038] For example, suppose there are 100 dough products on an automated noodle production line. The 1st, 2nd, ... 100th dough product is extracted sequentially. For each dough product, an industrial camera is used to take one image each from the top, side, and bottom angles, resulting in a total of 3 images.

[0039] Step S242: Synchronously retrieve the Kth image acquisition scene data corresponding to the Kth dough image, and perform anomaly identification based on the Kth image acquisition scene data to obtain the Kth acquisition scene anomaly features. Specifically, while capturing the image, record scene data related to image acquisition, such as light source intensity, background color, and shooting distance. This data is used for image processing and anomaly identification. Image processing algorithms (such as edge detection, color segmentation, and texture analysis) are used to analyze the acquired scene data and identify potential anomaly features. For example, uneven light source intensity may lead to inconsistent image brightness, and changes in background color may lead to target recognition errors.

[0040] For example, the image acquisition scene data for the first dough product is shown in Table 3. Through anomaly detection algorithm analysis, it was found that the light source intensity of the side image of the first dough product was low, resulting in uneven image brightness. The anomaly feature was identified as "insufficient light source intensity".

[0041] Table 3: Examples of Image Acquisition Scene Data Dough number Shooting angle Light source intensity (lux) Background color (RGB) Shooting distance (mm) Image resolution Image size (MB) 1 top 5000 (255,255,255) 300 1920×1080 2.5 1 side 4800 (255,255,255) 300 1920×1080 2.5 1 bottom 4900 (255,255,255) 300 1920×1080 2.5 Step S243: Correct the Kth dough image based on the abnormal features of the Kth acquisition scene to obtain the corrected Kth dough image. Specifically, based on the identified abnormal features, a corresponding image correction algorithm is used to process the image. For example, if the light source intensity is insufficient, histogram equalization or brightness adjustment algorithms can be used to correct the image brightness; if the background color changes, a background segmentation algorithm can be used to remove background interference.

[0042] For example, for the side image of the first dough product, due to insufficient light source intensity, a histogram equalization algorithm is used for brightness correction. The average brightness of the image before correction is 120, and the average brightness of the image after correction is 180.

[0043] Step S244: The corrected image of the Kth dough is input into the multiple defect convolutional capture models to obtain multiple dough defect capture results, and these multiple dough defect capture results are fused to generate the Kth surface anomaly detection result. Specifically, the corrected image is input into multiple defect convolutional capture models, each model detects different defect types (such as cracks, discolored spots, etc.) and outputs its own defect capture result. The defect capture results of multiple models are integrated to generate a detection result containing all defect information. This implementation method, by synchronously retrieving scene data and performing anomaly recognition during image acquisition, can effectively cope with environmental changes (such as fluctuations in light source intensity, changes in background color, etc.), enhancing the robustness of the detection system.

[0044] Step S300: Introduce a hygiene anomaly detection and correction mechanism to analyze the hygiene anomalies of the multiple dough products and obtain a second matrix for dough anomaly detection.

[0045] Specifically, microbial detection sensors, such as ATP fluorescence detectors, are installed at key locations on the automated noodle production line. These sensors can quickly detect the microbial content on the dough surface, judging the hygiene status of the dough surface by changes in fluorescence intensity. The detection results are output in numerical form and used to construct a second matrix for dough anomaly detection. Chemical sensors are used to detect potential chemical residues on the dough surface, such as pesticide residues and detergent residues. For example, gas chromatography-mass spectrometry (GC-MS) is used to analyze the chemical composition of sampled dough and convert the results into numerical data. The detection results are corrected by comparing the data of standard samples and actual test samples. For example, if the detected microbial content is higher than the threshold of the standard sample, it indicates a hygiene problem on the dough surface; if the detected chemical residue content exceeds the safety standard, it also indicates a hygiene anomaly in the dough.

[0046] For example, using an ATP fluorescence detector to analyze the surface of each dough ball, the results showed that the ATP fluorescence intensity on the dough surface was 500 relative optical units (RLU), while the threshold for the standard sample was 300 RLU, indicating that the microbial content on the dough surface exceeded the standard. Simultaneously, GC-MS analysis revealed the presence of a certain detergent residue on the dough surface at a concentration of 0.5 ppm, while the safety standard is 0.1 ppm. These data were recorded to form a second matrix for dough anomaly detection; for example, one column in the matrix represents the ATP fluorescence intensity, and another column represents the detergent residue concentration.

[0047] In one possible implementation, step S300 further includes step S310, performing colony anomaly detection correction on the Kth dough product according to the dough kneading control scheme to obtain a first characteristic for hygiene anomaly detection. Specifically, a colony counter or microbial detector is used to detect the total colony count of the Kth dough product. These devices can quickly detect the microbial content on the surface and inside of the dough. The detection results are corrected according to the standard colony count range set in the dough kneading control scheme. If the detected colony count exceeds the standard range, it is recorded as an anomaly.

[0048] For example, suppose the standard colony count range set in the dough control scheme is no more than 1000 colonies per gram of dough. Colony testing was performed on the dough products, and the results are shown in Table 4.

[0049] Table 4: Examples of Colony Mutation Detection Results Dough number Colony count (cells / gram) Is it abnormal? 1 850 no 2 1200 yes ... ... ... 100 950 no Step S320: Perform pH value anomaly detection correction on the Kth dough product according to the dough kneading control scheme to obtain the second characteristic for hygiene anomaly detection. Specifically, a pH meter is used to detect the acidity or alkalinity of the Kth dough product. The pH meter can accurately measure the pH value of the dough. The detection result is corrected according to the standard pH value range set in the dough kneading control scheme. If the detected pH value exceeds the standard range, it is recorded as an anomaly.

[0050] For example, suppose the standard pH range set in the dough mixing control scheme is 5.5-6.5. The pH value of the dough products was tested, and the results are shown in Table 5.

[0051] Table 5: Examples of pH value anomaly detection results Dough number pH value Is it abnormal? 1 6.0 no 2 5.0 yes ... ... ... 100 6.2 no Step S330: The Kth dough product is subjected to odor anomaly detection correction according to the dough kneading control scheme, generating a third hygiene anomaly detection feature. Specifically, the odor of the Kth dough product is detected using an electronic nose or gas chromatography-mass spectrometry (GC-MS). These devices can detect the types and concentrations of volatile organic compounds (VOCs) in the dough. The detection results are corrected according to the standard odor characteristics set in the dough kneading control scheme. If the detected odor characteristic does not match the standard characteristic, it is recorded as an anomaly.

[0052] For example, suppose the standard odor characteristics set in the dough control scheme are a specific combination of VOCs. Odor testing was performed on the dough products, and the results are shown in Table 6.

[0053] Table 6: Examples of Odor Anomaly Detection Results Dough number Odor characteristics (types of VOCs) Is it abnormal? 1 Standard features no 2 Abnormal characteristics yes ... ... ... 100 Standard features no Step S340: Generate the hygiene mutation detection result for the Kth dough based on the first, second, and third hygiene mutation detection features, and add the Kth dough hygiene mutation detection result to the second dough mutation detection matrix. Specifically, integrate the results of colony mutation detection, pH value mutation detection, and odor mutation detection to generate the hygiene mutation detection result for the Kth dough, and add the generated hygiene mutation detection result to the second dough mutation detection matrix. This implementation comprehensively assesses the hygiene status of the dough product through three detection methods: colony, pH value, and odor, ensuring the comprehensiveness and accuracy of the detection results.

[0054] In one possible implementation, the Kth dough product is corrected for bacterial colony variation detection according to the dough kneading control scheme to obtain a first characteristic for hygiene variation detection. Step S310 further includes step S311, performing bacterial colony detection on the Kth dough product to obtain Kth bacterial colony detection data. Specifically, a colony counter or microbial detector is used to detect the total bacterial count of the Kth dough product. These devices can quickly detect the microbial content on the surface and inside of the dough.

[0055] Step S312: Based on the dough kneading control scheme, normal dough samples are retrieved for bacterial colony detection, constructing a dough colony detection space. Specifically, normal dough samples conforming to the dough kneading control scheme are retrieved from historical data; the bacterial counts of these samples are within the standard range. The bacterial count distribution of these normal samples is used to construct a detection space for anomaly detection.

[0056] Step S313: Based on the dough colony detection space, perform anomaly capture on the Kth colony detection data to obtain the Kth colony anomaly capture result. Specifically, statistical analysis methods are used to identify outliers in the colony detection data.

[0057] Step S314: Based on the colony detection operation data and colony detection environment data corresponding to the Kth colony detection data, analyze the detection interference to determine the interference characteristics of the Kth colony detection. Specifically, analyze the operation data (such as detection time and detection equipment status) and environmental data (such as temperature and humidity) during the colony detection process to identify possible interference factors. For example, interference analysis was performed on the colony detection data of dough products, and the results are shown in Table 7.

[0058] Table 7: Examples of Interference Detection Analysis Dough number Colony count (cells / gram) Detection time Detecting equipment status Ambient temperature (°C) Ambient humidity (%) Interference characteristics 1 850 08:00 normal 25 60 No interference 2 1200 09:00 normal 26 65 High humidity interference ... ... ... ... ... ... ... 100 950 17:00 normal 24 55 No interference Step S315: Correct the Kth colony mutation capture result based on the Kth colony detection interference feature to generate the first feature for hygiene mutation detection. Specifically, if no interference features are found during the detection process, the colony mutation capture result is directly used as the final first feature for hygiene mutation detection. If high humidity interference is found during the detection process, the colony mutation capture result is corrected based on historical data and empirical formulas. For example, high humidity may lead to an inflated colony count, which can be corrected by subtracting an empirical correction value (e.g., 150 CFU / g). If low temperature interference is found during the detection process, the colony mutation capture result is corrected based on historical data and empirical formulas. For example, low temperature may lead to an inflated colony count, which can be corrected by adding an empirical correction value (e.g., 100 CFU / g). If the detection equipment is in an abnormal state (e.g., equipment aging, inaccurate calibration), the colony mutation capture result is corrected based on the deviation of the equipment state. For example, equipment aging may lead to inaccurate colony count detection, which can be adjusted using an equipment calibration coefficient (e.g., 1.1). If multiple interference characteristics are found during the detection process, the colony mutation capture results should be corrected by comprehensively considering various interference factors. For example, if both high humidity and low temperature interference exist simultaneously, a comprehensive correction value can be used for correction. Examples of correction results based on different interference characteristics are shown in Table 8.

[0059] Table 8: Examples of Colony Variation Correction Dough number Colony detection data (cells / gram) Mutation capture results Interference characteristics Correction method Corrected colony count (cells / gram) First characteristic of hygiene mutation detection 1 850 normal No interference Uncorrected 850 normal 2 1200 abnormal High humidity interference Subtract 150 per gram 1050 normal 3 700 normal Low temperature interference Add 100 per gram 800 normal 4 900 normal Equipment malfunction Multiply by the calibration factor of 1.1 990 normal 5 1100 abnormal High humidity + Subtract 50 per gram 1050 normal ... ... ... ... ... ... ... 100 950 normal No interference Uncorrected 950 normal This method is also applicable to pH and odor detection. Through similar calibration steps, accurate second and third features for detecting hygiene anomalies can be generated and ultimately integrated into the second matrix for dough anomaly detection, thereby improving the reliability of the detection results.

[0060] Step S400: Detect gluten strength variation in the multiple dough products according to the dough kneading control scheme to obtain a third matrix for dough variation detection.

[0061] Specifically, gluten strength variation detection refers to the detection of changes in the gluten strength characteristics of dough. This is mainly achieved through methods such as tensile testing to measure indicators like tensile strength and elongation, determining whether the dough's gluten strength meets production requirements. A dough stretching instrument is used to perform a tensile test on each dough product, measuring the tensile strength and elongation. By precisely controlling the stretching speed and measuring force changes, gluten strength characteristic data of the dough is obtained. For example, the stretching speed can be set to 50 mm per minute, measuring the maximum tensile force and breaking elongation of the dough during the stretching process. The data obtained from the tensile test is compared and analyzed with the gluten strength standards set in the dough mixing control scheme. If the dough's tensile strength is lower than the standard value, or the elongation exceeds the standard range, it indicates that the dough's gluten strength has changed. The gluten strength detection results are converted into a numerical matrix, namely the third matrix for dough variation detection.

[0062] For example, a tensile test was conducted on each dough product, with the stretching speed set at 50 mm per minute. The test results showed that the maximum tensile force of one dough was 10 Newtons, and the elongation at break was 150 mm. However, the gluten strength standard set in the dough control scheme was a maximum tensile force of 12 Newtons and an elongation at break of 120 mm ± 10 mm. Data analysis revealed that the tensile strength of this dough was lower than the standard value, and the elongation exceeded the standard range, indicating an anomaly in gluten strength. These data were recorded to form a third matrix for dough anomaly detection; for example, one column in the matrix represents tensile strength, and another column represents elongation at break.

[0063] In one possible implementation, according to the dough kneading control scheme, gluten strength variation detection is performed on the multiple dough products to obtain a third matrix for dough variation detection. Step S400 further includes step S410, performing gluten strength detection on the multiple dough products to obtain gluten strength detection data for each dough, and simultaneously recording scene data for each gluten strength detection. Specifically, a dough extensometer or texture analyzer is used to perform gluten strength detection on each dough product, measuring the tensile strength and elongation of the dough. Scene data during the detection process, such as detection time, detection equipment status, ambient temperature, and ambient humidity, are recorded simultaneously. For example, gluten strength detection is performed on 100 dough products, and the detection results are shown in Table 9.

[0064] Table 9: Examples of Tension Test Data Dough number Tensile strength (N) Elongation (mm) Detection time Detecting equipment status Ambient temperature (°C) Ambient humidity (%) 1 10 150 08:00 normal 25 60 2 8 120 08:10 normal 25 60 3 12 160 08:20 normal 25 60 ... ... ... ... ... ... ... 100 9 130 09:00 normal 25 60 Step S420: Based on the dough kneading control scheme, normal dough samples for gluten strength testing are retrieved to construct a gluten strength testing space. Specifically, normal dough samples conforming to the dough kneading control scheme are retrieved from historical data; the gluten strength data of these samples are within the standard range. The distribution of gluten strength data from these normal samples is used to construct a testing space for anomaly detection. For example, based on the retrieved normal sample gluten strength data distribution, the determined gluten strength testing space is: tensile strength 8-12N, elongation 120-160mm.

[0065] Step S430: Based on the gluten strength detection space, the anomaly capture of the gluten strength detection data of each dough is performed to obtain the gluten strength anomaly characteristics of each dough. Specifically, statistical analysis methods are used to identify outliers in the detection data.

[0066] Step S440: Based on the gluten strength detection scenario data, interference compensation is performed on the gluten strength variation characteristics of each dough, generating the third matrix for dough variation detection. Specifically, compensation is performed on the gluten strength variation characteristics based on the detection scenario data (such as detection time, equipment status, ambient temperature, humidity, etc.). For example, a low ambient temperature may lead to a lower gluten strength detection value, which can be compensated for using empirical formulas. For example, the detection scenario data for the fourth dough product is shown in Table 10.

[0067] Table 10: Example of testing scenario data for the 4th dough product Dough number Tensile strength (N) Elongation (mm) Mutation capture results Detection time Detecting equipment status Ambient temperature (°C) Ambient humidity (%) 4 6 100 abnormal 08:30 normal 20 55 Based on historical data, for every 1°C decrease in ambient temperature, tensile strength decreases by an average of 0.5N, and elongation decreases by an average of 10mm. Therefore, the compensated tendon strength data are: tensile strength compensation value: (25-20)×0.5=2.5N, elongation compensation value: (25-20)×10=50mm, compensated tensile strength: 6+2.5=8.5N, compensated elongation: 100+50=150mm. The compensated tendon strength data is within the standard range, therefore the final tendon strength variation characteristic is "normal". This implementation method improves the reliability of the test results by analyzing the test scene data, identifying possible interference factors, and compensating for them.

[0068] In one possible implementation, interference compensation is performed on the dough strength variation characteristics based on the data from each dough strength detection scenario to generate the third matrix for dough variation detection. Step S440 further includes step S441, which involves performing anomaly detection based on the data from each dough strength detection scenario to obtain the abnormal features of each dough strength detection scenario. Specifically, statistical analysis methods (such as Z-score and IQR) are used to perform anomaly detection on the dough strength detection scenario data to identify possible interference features.

[0069] Step S442 involves ensemble training based on the historical set of tendon anomaly detection and compensation data to obtain a tendon anomaly detection and compensation model. Specifically, historical tendon anomaly detection data and their corresponding compensation results are collected to form a historical set of tendon anomaly detection and compensation data. Machine learning algorithms (such as random forest and gradient boosting tree) are used to train the historical data to construct the tendon anomaly detection and compensation model. The historical tendon anomaly detection and compensation data is shown in Table 11. By using this historical data to train the tendon anomaly detection and compensation model, the model can learn the compensation rules under different anomaly characteristics.

[0070] Table 11: Examples of Historical Tension Anomaly Detection Compensation Data Historical Sample Number Tensile strength (N) Elongation (mm) Abnormal characteristics Compensated tensile strength (N) Elongation after compensation (mm) 1 6 100 Low temperature interference 8.5 150 2 7 110 High humidity interference 8 130 3 5 90 Equipment aging 7.5 120 ... ... ... ... ... ... Step S443: Input the abnormal features of each gluten strength detection scene and the gluten strength variation features of each dough into the gluten strength variation detection and compensation model to obtain the compensation results for each gluten strength variation. Specifically, the abnormal features of the gluten strength detection scene and the gluten strength variation features are input into the trained gluten strength variation detection and compensation model, and the model outputs the compensated gluten strength data.

[0071] Step S444: Based on the compensation results for each gluten strength variation, construct the third matrix for dough variation detection. Specifically, integrate the compensated gluten strength data into the third matrix for subsequent analysis and reporting. This implementation method, through integrated training, enables the constructed gluten strength variation detection and compensation model to learn the compensation rules under different abnormal characteristics, thereby improving the accuracy of compensation.

[0072] Step S500: Perform a global defect assessment on the multiple dough products based on the first dough variation detection matrix, the second dough variation detection matrix, and the third dough variation detection matrix to obtain a dough defect detection report.

[0073] Specifically, the first, second, and third matrices for dough anomaly detection are fused together. Algorithms such as weighted summation and neural networks can be used to comprehensively consider the impact of surface anomalies, hygiene anomalies, and gluten strength anomalies on product quality, resulting in a comprehensive defect assessment score for each dough product. Based on the comprehensive defect assessment score, the defect severity of the dough products is categorized into different levels, such as minor, moderate, and severe defects. Then, a detailed dough defect detection report is generated based on the defect level, including information such as defect type, defect location, and defect severity.

[0074] For example, a weighted summation algorithm is used to fuse the three matrices, assuming a surface anomaly weight of 0.4, a hygiene anomaly weight of 0.3, and a gluten strength anomaly weight of 0.3. For a certain dough product, its surface anomaly detection score is 60 points (out of 100), its hygiene anomaly detection score is 70 points, and its gluten strength anomaly detection score is 80 points. The comprehensive defect assessment score is: 0.4×60+0.3×70+0.3×80=69 points. According to the established defect grading standard, 60-70 points is a medium defect. Therefore, this dough product is rated as a medium defect, and the defect detection report records in detail the presence of cracks and discolored spots on its surface, excessive microbial content in terms of hygiene, and insufficient tensile strength in terms of gluten strength.

[0075] In one possible implementation, a global defect assessment is performed on the multiple dough products based on the first dough anomaly detection matrix, the second dough anomaly detection matrix, and the third dough anomaly detection matrix to obtain a dough defect detection report. Step S500 further includes step S510, where surface defect evaluation is performed based on the first dough anomaly detection matrix to obtain the surface defect coefficient for each dough. Specifically, a preset evaluation model is used to calculate the surface defect coefficient for each dough based on data (such as crack length, area of ​​discolored spots, etc.) in the first dough anomaly detection matrix. For example, the surface defect coefficient can be expressed as a weighted sum of crack length and discolored spot area.

[0076] Step S520: Hygiene defect evaluation is performed based on the second dough variation detection matrix to obtain the hygiene defect coefficient for each dough. Specifically, a preset evaluation model is used to calculate the hygiene defect coefficient for each dough based on data from the second dough variation detection matrix (such as colony variation, pH variation, odor variation, etc.). For example, the hygiene defect coefficient can be expressed as a weighted sum of colony variation, pH variation, and odor variation.

[0077] Step S530: Evaluate gluten strength defects based on the third matrix for dough variation detection to obtain the gluten strength defect coefficient for each dough. Specifically, using a preset evaluation model, calculate the gluten strength defect coefficient for each dough based on data (such as tensile strength, elongation, etc.) in the third matrix for dough variation detection. For example, the gluten strength defect coefficient can be expressed as a weighted sum of tensile strength and elongation.

[0078] Step S540: Based on the multidimensional defect weights of the dough, the surface defect coefficients, hygiene defect coefficients, and gluten strength defect coefficients of each dough are weighted and fused to generate a dough defect detection report. The dough defect detection report includes the global defect coefficients of each dough. Specifically, the weights of surface defects, hygiene defects, and gluten strength defects are set according to actual needs. For example, the weight of surface defects is 0.4, the weight of hygiene defects is 0.3, and the weight of gluten strength defects is 0.3. The surface defect coefficient, hygiene defect coefficient, and gluten strength defect coefficient of each dough are multiplied by their respective weights and then summed to obtain the global defect coefficient. This implementation method comprehensively considers surface defects, hygiene defects, and gluten strength defects to conduct a comprehensive defect assessment of each dough product, ensuring the comprehensiveness and accuracy of the assessment results.

[0079] In one possible implementation, the method further includes: determining whether the global defect coefficient of each dough is greater than or equal to the global defect threshold of the dough, obtaining the judgment result of each global defect, and generating a dough defect early warning signal based on the judgment result of each global defect.

[0080] Specifically, a global defect threshold for dough is set based on production standards and quality requirements. For example, a threshold of 0.5 means that when the global defect coefficient reaches or exceeds 0.5, the dough is considered to have a serious defect. The global defect coefficient of each dough product is judged. If the global defect coefficient is greater than or equal to the threshold, it is marked as "defect exceeds the standard"; otherwise, it is marked as "normal." Based on the judgment result, a corresponding early warning signal is generated. If the dough defect exceeds the standard, an early warning signal is issued to notify production personnel to handle the situation. The early warning signal can take the form of an audible and visual alarm, SMS notification, or system prompt, etc., to promptly notify production personnel to handle the situation. This implementation method, by setting a global defect threshold and making judgments, can promptly detect dough products with serious defects, preventing unqualified products from entering subsequent production stages. Generating early warning signals can quickly notify production personnel, enabling them to take timely measures, such as adjusting production processes and checking equipment status, thereby reducing the defect rate and improving product quality.

[0081] This application employs a method based on an automated dough production line that acquires multiple dough products and performs surface anomaly detection (obtaining a first matrix), hygiene anomaly analysis (obtaining a second matrix), and gluten strength anomaly detection (obtaining a third matrix) on the dough. By combining the three matrices, a global defect assessment is performed, resulting in a dough defect detection report. This method solves the technical problem of inaccurate and incomplete detection results in existing automated noodle production lines, thereby improving the accuracy and comprehensiveness of defect detection.

[0082] In the above text, refer to Figure 1 A multi-dimensional detection method for product defects in an automated noodle production line according to an embodiment of the present invention is described in detail. Next, reference will be made to... Figure 2 A multi-dimensional detection system for product defects in an automated noodle production line according to an embodiment of the present invention is described.

[0083] The multi-dimensional detection system for product defects in an automated noodle production line according to an embodiment of the present invention addresses the technical problems of inaccurate and incomplete detection results in existing automated noodle production line product defect detection systems, thereby improving the accuracy and comprehensiveness of defect detection. The multi-dimensional detection system for product defects in an automated noodle production line includes: a dough product acquisition module 10, a surface anomaly detection module 20, a hygiene anomaly analysis module 30, a gluten strength anomaly detection module 40, and a global defect assessment module 50.

[0084] The dough product acquisition module 10 is used to acquire multiple dough products corresponding to the dough kneading control scheme based on the automatic dough kneading line of the braised noodles automatic production line; the surface anomaly detection module 20 is used to perform surface anomaly detection on the multiple dough products based on the dough surface defect index set to obtain a dough anomaly detection first matrix; the hygiene anomaly analysis module 30 is used to introduce a hygiene anomaly detection correction mechanism to perform hygiene anomaly analysis on the multiple dough products to obtain a dough anomaly detection second matrix; the gluten strength anomaly detection module 40 is used to perform gluten strength anomaly detection on the multiple dough products based on the dough kneading control scheme to obtain a dough anomaly detection third matrix; and the global defect assessment module 50 is used to perform a global defect assessment on the multiple dough products based on the dough anomaly detection first matrix, the dough anomaly detection second matrix, and the dough anomaly detection third matrix to obtain a dough defect detection report.

[0085] The specific configuration of the surface anomaly detection module 20 will be described in detail below. As mentioned above, surface anomaly detection is performed on the multiple dough products based on the dough surface defect index set to obtain a dough anomaly detection first matrix. The surface anomaly detection module 20 may further include: a similarity evaluation unit for pairwise similarity evaluation of the dough surface defect index set to obtain a defect index similarity evaluation set; a decoupling optimization unit for decoupling and optimizing the dough surface defect index set based on the defect index similarity evaluation set to obtain multiple independent defect indices; a convolutional feature learning unit for performing convolutional feature learning based on the multiple independent defect indices to obtain multiple defect convolutional capture models; and a surface anomaly capture unit for capturing surface anomalies on the multiple dough products based on the multiple defect convolutional capture models to generate the dough anomaly detection first matrix.

[0086] The surface anomaly capture unit further includes: a multi-angle image acquisition subunit for extracting the Kth dough product from the multiple dough products and acquiring multi-angle images of the Kth dough product using an industrial camera to obtain the Kth dough image, where K is a positive integer; an anomaly recognition subunit for synchronously retrieving the Kth image acquisition scene data corresponding to the Kth dough image and performing anomaly recognition based on the Kth image acquisition scene data to obtain the Kth acquisition scene anomaly features; a correction processing subunit for correcting the Kth dough image based on the Kth acquisition scene anomaly features to obtain the Kth dough corrected image; and a defect convolution capture subunit for inputting the Kth dough corrected image into the multiple defect convolution capture models to obtain multiple dough defect capture results and fusing the multiple dough defect capture results to generate the Kth surface anomaly detection result.

[0087] The specific configuration of the hygiene variation analysis module 30 will be described in detail below. As mentioned above, the hygiene variation analysis module 30 may further include: a colony variation detection and correction unit for performing colony variation detection and correction on the Kth dough product according to the dough kneading control scheme to obtain a first hygiene variation detection feature; a pH value variation detection and correction unit for performing pH value variation detection and correction on the Kth dough product according to the dough kneading control scheme to obtain a second hygiene variation detection feature; an odor variation detection and correction unit for performing odor variation detection and correction on the Kth dough product according to the dough kneading control scheme to generate a third hygiene variation detection feature; and a dough variation detection second matrix acquisition unit for generating a Kth dough hygiene variation detection result based on the first hygiene variation detection feature, the second hygiene variation detection feature, and the third hygiene variation detection feature, and adding the Kth dough hygiene variation detection result to the dough variation detection second matrix.

[0088] The process involves performing colony mutation detection and correction on the Kth dough product according to the dough kneading control scheme to obtain a first feature for hygiene mutation detection. The colony mutation detection and correction unit may further include: a colony detection subunit for performing colony detection on the Kth dough product to obtain Kth colony detection data; a dough colony detection normal sample retrieval subunit for retrieving normal samples for dough colony detection according to the dough kneading control scheme to construct a dough colony detection space; a mutation capture subunit for capturing mutations in the Kth colony detection data based on the dough colony detection space to obtain a Kth colony mutation capture result; a detection interference analysis subunit for analyzing detection interference based on the colony detection operation data and colony detection environment data corresponding to the Kth colony detection data to determine the Kth colony detection interference feature; and a correction subunit for correcting the Kth colony mutation capture result based on the Kth colony detection interference feature to generate the first feature for hygiene mutation detection.

[0089] The specific configuration of the gluten strength variation detection module 40 will be described in detail below. As mentioned above, the gluten strength variation detection module 40 can further include: a gluten strength detection unit for performing gluten strength detection on the multiple dough products according to the dough kneading control scheme, obtaining gluten strength detection data for each dough, and synchronously recording gluten strength detection scenario data for each dough; a dough gluten strength detection normal sample retrieval unit for retrieving normal samples of dough gluten strength detection according to the dough kneading control scheme, and constructing a gluten strength detection space; a variation capture unit for capturing variations in the gluten strength detection data of each dough according to the gluten strength detection space, and obtaining gluten strength variation characteristics of each dough; and an interference compensation unit for compensating for interference in the gluten strength variation characteristics of each dough according to the gluten strength detection scenario data, and generating the dough variation detection third matrix.

[0090] Specifically, the interference compensation unit performs interference compensation on the dough gluten strength variation features based on the data from each gluten strength detection scenario to generate the third dough variation detection matrix. The interference compensation unit may further include: an anomaly detection subunit for performing anomaly detection based on the data from each gluten strength detection scenario to obtain anomaly features of each gluten strength detection scenario; an integration and fusion training subunit for performing integration and fusion training based on the historical set of gluten strength variation detection compensation to obtain a gluten strength variation detection compensation model; a gluten strength variation detection compensation subunit for inputting the anomaly features of each gluten strength detection scenario and the gluten strength variation features of each dough into the gluten strength variation detection compensation model to obtain compensation results for each gluten strength variation; and a third dough variation detection matrix construction subunit for constructing the third dough variation detection matrix based on the compensation results for each gluten strength variation.

[0091] The specific configuration of the global defect assessment module 50 will be described in detail below. As mentioned above, the global defect assessment of the multiple dough products is performed based on the first dough anomaly detection matrix, the second dough anomaly detection matrix, and the third dough anomaly detection matrix to obtain a dough defect detection report. The global defect assessment module 50 may further include: a surface defect evaluation unit for evaluating surface defects based on the first dough anomaly detection matrix to obtain surface defect coefficients for each dough; a hygiene defect evaluation unit for evaluating hygiene defects based on the second dough anomaly detection matrix to obtain hygiene defect coefficients for each dough; a gluten strength defect evaluation unit for evaluating gluten strength defects based on the third dough anomaly detection matrix to obtain gluten strength defect coefficients for each dough; and a weighted fusion unit for weighted fusion of the surface defect coefficients, hygiene defect coefficients, and gluten strength defect coefficients of each dough based on the multidimensional defect weights of the dough, to generate the dough defect detection report, which includes the global defect coefficients of each dough.

[0092] The system may further include: a dough defect early warning signal generation module for determining whether the global defect coefficient of each dough is greater than or equal to the global defect threshold of the dough, obtaining the judgment result of each global defect, and generating a dough defect early warning signal based on the judgment result of each global defect.

[0093] The multi-dimensional detection system for product defects in the automatic production line of braised noodles provided in this embodiment of the invention can execute the multi-dimensional detection method for product defects in the automatic production line of braised noodles provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0094] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0095] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand 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 this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A multi-dimensional detection method for product defects in an automated noodle production line, characterized in that, The method includes: Based on the automatic dough mixing line of the braised noodles production line, obtain multiple dough products corresponding to the dough mixing control scheme; Based on the dough surface defect index set, surface anomaly detection is performed on the multiple dough products to obtain the first matrix of dough anomaly detection; A hygiene anomaly detection and correction mechanism is introduced to analyze hygiene anomalies in the multiple dough products, and a second matrix for dough anomaly detection is obtained. Based on the dough kneading control scheme, the gluten strength variation detection is performed on the multiple dough products to obtain a third matrix for dough variation detection. A global defect assessment of the multiple dough products is performed based on the first dough mutation detection matrix, the second dough mutation detection matrix, and the third dough mutation detection matrix to obtain a dough defect detection report.

2. The multi-dimensional detection method for product defects in an automated noodle production line as described in claim 1, characterized in that, Based on the dough surface defect index set, surface anomaly detection is performed on the multiple dough products to obtain a first matrix for dough anomaly detection, including: The set of surface defect indicators of the dough is subjected to pairwise similarity evaluation to obtain a similarity evaluation set of defect indicators; The dough surface defect index set is decoupled and optimized based on the defect index similarity evaluation set to obtain multiple independent defect indices. Multiple defect convolutional feature learning methods are performed based on the multiple independent defect indicators to obtain multiple defect convolutional capture models. The surface anomaly detection matrix of the dough products is generated by capturing surface anomalies based on the multiple defect convolution capture models.

3. The multi-dimensional detection method for product defects in an automated noodle production line as described in claim 2, characterized in that, Surface variation capture of the multiple dough products is performed based on the multiple defect convolution capture models, including: Extract the Kth dough product from the multiple dough products, and acquire multi-angle images of the Kth dough product using an industrial camera to obtain the Kth dough image, where K is a positive integer; Synchronously retrieve the Kth image acquisition scene data corresponding to the Kth dough image, and perform anomaly identification based on the Kth image acquisition scene data to obtain the abnormal features of the Kth acquisition scene; The Kth dough image is corrected based on the abnormal features of the Kth acquisition scene to obtain the corrected Kth dough image. The Kth dough correction image is input into the multiple defect convolution capture models to obtain multiple dough defect capture results, and the multiple dough defect capture results are fused to generate the Kth surface anomaly detection result.

4. The multi-dimensional detection method for product defects in an automated noodle production line as described in claim 1, characterized in that, The hygiene anomaly detection and correction mechanism includes: Based on the dough mixing control scheme, the Kth dough product is subjected to colony mutation detection and correction to obtain the first feature of hygiene mutation detection; Based on the dough mixing control scheme, the pH value of the Kth dough product is detected and corrected for abnormality to obtain a second feature for hygiene abnormality detection. Based on the dough mixing control scheme, the Kth dough product is subjected to odor change detection and correction to generate a third feature for hygiene change detection. The Kth dough hygiene mutation detection result is generated based on the first hygiene mutation detection feature, the second hygiene mutation detection feature, and the third hygiene mutation detection feature, and the Kth dough hygiene mutation detection result is added to the second dough mutation detection matrix.

5. The multi-dimensional detection method for product defects in an automated noodle production line as described in claim 4, characterized in that, According to the dough mixing control scheme, the Kth dough product is subjected to colony mutation detection correction to obtain the first feature for hygiene mutation detection, including: Colony detection was performed on the Kth dough product to obtain Kth colony detection data; Based on the dough kneading control scheme, normal samples for dough colony detection are retrieved to construct a dough colony detection space; Based on the dough colony detection space, the Kth colony detection data is subjected to mutation capture to obtain the Kth colony mutation capture result; Based on the colony detection operation data and colony detection environment data corresponding to the Kth colony detection data, the detection interference is analyzed to determine the Kth colony detection interference characteristics; The first feature for detecting hygiene anomalies is generated by correcting the Kth colony mutation capture result based on the interference feature of the Kth colony detection.

6. The multi-dimensional detection method for product defects in an automated noodle production line as described in claim 1, characterized in that, According to the dough kneading control scheme, gluten strength variation detection is performed on the multiple dough products to obtain a third matrix for dough variation detection, including: The gluten strength of the multiple dough products is tested to obtain gluten strength test data for each dough, and the data for each gluten strength test scenario is recorded simultaneously. Based on the dough kneading control scheme, normal samples for dough gluten strength testing are retrieved to construct a gluten strength testing space. Based on the gluten strength detection space, the gluten strength detection data of each dough are captured to obtain the gluten strength variation characteristics of each dough. Based on the data from each gluten strength detection scenario, interference compensation is performed on the gluten strength variation characteristics of each dough to generate the third matrix for dough variation detection.

7. The multi-dimensional detection method for product defects in an automated noodle production line as described in claim 6, characterized in that, Based on the gluten strength detection scenario data, interference compensation is performed on the gluten strength variation characteristics of each dough to generate the third matrix for dough variation detection, including: Anomaly detection is performed based on the data from each tendon strength detection scenario to obtain the abnormal characteristics of each tendon strength detection scenario. The model for detecting and compensating tendon and force variations is obtained by integrating and training the historical set of tendon and force variation detection and compensation. The abnormal features of each gluten strength detection scenario and the abnormal features of each dough gluten strength variation are input into the gluten strength variation detection compensation model to obtain the compensation results of each gluten strength variation. Based on the compensation results of each gluten strength variation, a third matrix for dough variation detection is constructed.

8. The multi-dimensional detection method for product defects in an automated noodle production line as described in claim 1, characterized in that, A global defect assessment is performed on the multiple dough products based on the first dough anomaly detection matrix, the second dough anomaly detection matrix, and the third dough anomaly detection matrix to obtain a dough defect detection report, including: Surface defects are evaluated based on the first matrix for dough anomaly detection to obtain the surface defect coefficient for each dough. Hygiene defect evaluation is performed based on the second matrix for dough anomaly detection to obtain the hygiene defect coefficient for each dough; Based on the third matrix for dough anomaly detection, gluten strength defects are evaluated to obtain the gluten strength defect coefficient for each dough. Based on the multidimensional defect weights of the dough, the surface defect coefficients, hygiene defect coefficients, and gluten strength defect coefficients of each dough are weighted and fused to generate a dough defect detection report, which includes the global defect coefficients of each dough.

9. The multi-dimensional detection method for product defects in an automated noodle production line as described in claim 8, characterized in that, Determine whether the global defect coefficient of each dough is greater than or equal to the global defect threshold of the dough, obtain the judgment result of each global defect, and generate a dough defect early warning signal based on the judgment result of each global defect.

10. A multi-dimensional detection system for product defects in an automated noodle production line, characterized in that: The system is used to implement the multi-dimensional detection method for product defects in the automatic production line of braised noodles according to any one of claims 1-9, and the system includes: The dough product acquisition module is used to obtain multiple dough products corresponding to the dough kneading control scheme based on the automatic dough kneading line of the braised noodles automatic production line. The surface anomaly detection module is used to perform surface anomaly detection on the multiple dough products according to the dough surface defect index set, and obtain the first matrix of dough anomaly detection. The hygiene anomaly analysis module is used to introduce a hygiene anomaly detection and correction mechanism to perform hygiene anomaly analysis on the multiple dough products and obtain a second matrix for dough anomaly detection. The gluten strength variation detection module is used to detect gluten strength variation in the multiple dough products according to the dough kneading control scheme, and obtain a third matrix for dough variation detection. The global defect assessment module is used to perform a global defect assessment on the multiple dough products based on the first dough mutation detection matrix, the second dough mutation detection matrix, and the third dough mutation detection matrix, and to obtain a dough defect detection report.

Images