Production quality dynamic early warning system and method based on deep learning

Abstract

The invention discloses a production quality dynamic early warning system and method based on deep learning, relates to the technical field of dairy product production quality early warning, and solves the technical problem that the prior art mainly depends on a single data source to carry out monitoring and early warning on the production quality of dairy products, and obviously, the monitoring and early warning precision is not high. The method comprises the following steps: acquiring a milk body image of a dairy product in a production process and production quality data of the dairy product; feature extraction is carried out on the milk body image of the milk product to obtain feature data corresponding to the milk product image; based on the historical image data, training an artificial intelligence model to obtain an image anomaly recognition model; identifying the image data based on an image anomaly identification model to obtain an identification result; calculating a production quality evaluation coefficient of the dairy product based on the production quality data of the dairy product; and judging whether the production quality of the dairy product needs to be pre-warned or not based on the identification result and the production quality evaluation coefficient. The problem is solved.

Description

Technical Field

[0001] The present invention belongs to the field of dairy product production quality early warning, and relates to deep learning technology, and specifically is a production quality dynamic early warning system and method based on deep learning. Background Art

[0002] Dairy products are products made from cow's milk or goat's milk and their processed products, by adding appropriate amounts of vitamins, minerals and other auxiliary materials, and using the production conditions required by the standards, such as liquid milk, milk powder, etc. The quality of dairy products directly affects the health of consumers and the reputation of the company, so quality management in the dairy production process is crucial.

[0003] The prior art (invention application with application number 2022101522450) discloses a method and device for early warning of dairy product production quality management; the method includes: obtaining dairy product quality inspection data of the business process from the business system; using the dairy product production process capability evaluation model based on the statistical process control rule to perform process capability analysis on the dairy product quality inspection data to obtain the corresponding process capability index; based on the statistical process control rule, the process capability index is monitored to determine the process capability index compliance rate corresponding to the dairy product production block within the preset time range; according to the process capability index compliance rate and the preset internal control standard, the dairy product process indicator control abnormality early warning information is output. This invention application provides support for the process capability management of dairy product quality business by monitoring and early warning of dairy product quality inspection data, quickly identifies management anomalies, reduces the quality risk of dairy product production, and improves the process capability, thereby improving the efficiency of dairy product production management. In the actual production process of dairy products, environmental factors and equipment operating conditions will affect the final production quality of dairy products. However, existing technologies mainly rely on a single data source to monitor and warn of the production quality of dairy products. Obviously, the accuracy of monitoring and warning is not high.

[0004] Therefore, the present invention proposes a production quality dynamic early warning system and method based on deep learning to solve the above problems. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes a dynamic early warning system and method for production quality based on deep learning, which is used to solve the problem that in the actual production process of dairy products, environmental factors and equipment operating conditions will affect the final production quality of dairy products, but the prior art mainly relies on a single data source to monitor and warn the production quality of dairy products, and it is obvious that the accuracy of monitoring and early warning is not high.

[0006] To achieve the above-mentioned object, a first aspect of the present invention provides a production quality dynamic early warning system based on deep learning, comprising: a data acquisition module, an image analysis module, a data analysis module and an early warning module;

[0007] Data acquisition module: used to collect milk images and production quality data of dairy products in real time during the production process; the production quality data includes: content data, equipment data of production equipment and environmental data of dairy products during the production process;

[0008] Image analysis module: extracting features from the milk body image of the dairy product to obtain feature data corresponding to the dairy product image; and,

[0009] Based on the historical image data, an artificial intelligence model is trained to obtain an image anomaly recognition model; based on the image anomaly recognition model, the image data is recognized to obtain a recognition result;

[0010] Data analysis module: Calculate the production quality assessment coefficient of dairy products based on the production quality data of dairy products;

[0011] Early warning module: Based on the recognition results and production quality assessment coefficient, it is determined whether an early warning is needed for the production quality of dairy products.

[0012] Preferably, the data acquisition module communicates with and / or is electrically connected to the image analysis module and the data analysis module respectively; the early warning module communicates with and / or is electrically connected to the image analysis module and the data analysis module respectively.

[0013] Preferably, the real-time collection of milk body images and production quality data of dairy products during the production process includes:

[0014] The image acquisition device is used to collect the milk body images of dairy products in the production process in real time;

[0015] The data acquisition equipment collects real-time content data of dairy products, equipment data of production equipment and environmental data of dairy products during the production process; the content data includes: pH value and density of dairy products during the production process; the equipment data includes: vibration frequency of the equipment and equipment temperature; the environmental data includes: ambient temperature, ambient humidity and dust concentration in the air.

[0016] Preferably, the feature extraction of the milk body image of the dairy product comprises:

[0017] The breast image is converted from the RGB color space to the HSV or HSB color space by using the color space conversion technology to extract the color feature data of the breast image; wherein the color feature data includes: hue, saturation and brightness;

[0018] The texture feature data of the breast image is extracted by image texture analysis technology; wherein the image texture analysis technology includes: Gabor filtering or gray level co-occurrence matrix; the texture feature data includes: roughness and uniformity.

[0019] It should be noted that hue is the basic attribute of color, which determines the type of color. For dairy products, the normal hue is usually white or light yellow, which depends on the fat and protein content of the dairy products. If the dairy products have abnormal hues such as red and blue, it may mean pollution or deterioration. Saturation refers to the depth of the color. For dairy products, the normal saturation is usually low, showing a soft color. If the saturation of dairy products is too high, it may mean the presence of additives or pigments. Abnormal conditions; brightness refers to the lightness and darkness of the color. Normal dairy products usually have a higher brightness, which is related to the gloss and reflectivity of their surface. If the brightness of dairy products decreases, it may mean oxidation, pollution or deterioration.

[0020] Roughness indicates the fineness of the surface texture of dairy products. Normal dairy products usually have a uniform and fine texture, while a rough surface may mean the presence of impurities, contamination or problems in the processing process. Uniformity indicates the consistency of the surface texture of dairy products. Normal dairy products usually have a uniform texture distribution, while abnormal texture distribution may mean quality problems.

[0021] Preferably, the identifying the image data based on the image anomaly identification model includes:

[0022] A1: extracting historical feature data corresponding to historical image data; wherein the historical image data refers to historical data of milk body images collected during the production process of dairy products; the historical feature data includes: color feature data and texture feature data corresponding to the historical image data;

[0023] A2: extracting reference feature data corresponding to a reference image; wherein the reference image refers to an image of a normal breast; the reference feature data includes: color feature data and texture feature data corresponding to the reference image;

[0024] A3: By comparing the historical feature data with the reference feature data, the historical image data is labeled and classified to obtain the labeling results; wherein the labeling results include: normal image and abnormal image;

[0025] A4: Integrate the historical feature data into standard input data, and integrate the annotation results corresponding to the historical feature data into standard output data; train an artificial intelligence model based on the standard input data and the standard output data to obtain an image anomaly recognition model; wherein the artificial intelligence model includes: a convolutional neural network or a deep belief network;

[0026] A5: The feature data corresponding to the real-time collected breast image is input into the image abnormality recognition model for recognition to obtain the recognition result.

[0027] Preferably, the production quality assessment coefficient of the dairy product is calculated based on the production quality data of the dairy product, comprising:

[0028] B1: The first evaluation coefficient is calculated based on the content data and marked as P1;

[0029] B2: The second evaluation coefficient is calculated based on the environmental data and marked as P2;

[0030] B3: The third evaluation coefficient is calculated based on the equipment data and marked as P3;

[0031] B4: The production quality assessment coefficient of dairy products is calculated by the formula: P = α × P1 + β × P2 + θ × P3; where P is the production quality assessment coefficient, α, β, θ are weight coefficients, and α + β + θ = 1.

[0032] It should be noted that the weight coefficient is set by experts based on experience.

[0033] Preferably, the first evaluation coefficient is calculated based on the content data, comprising:

[0034] Label pH value as Y and density as M;

[0035] By formula: P1 = C × e^[(Y-ZY)^2 / ZY]+D × e^[1 / ln(M+1)];

[0036] The first evaluation coefficient is calculated; wherein, C and D are proportional coefficients, and 0<C<1, 0<D<1, and ZY is the optimal production pH value of dairy products during the production process.

[0037] It should be noted that a high pH value will favor the growth and reproduction of certain microorganisms, causing the product to deteriorate faster and also making the taste of dairy products too sour and astringent; a low pH value will make the taste of dairy products too bland;

[0038] Low-density dairy products may mean a reduction in nutrients, especially the content of key ingredients such as protein and fat may be reduced, thus affecting the nutritional value of the product.

[0039] Preferably, the calculating the second evaluation coefficient based on the environmental data includes:

[0040] The ambient temperature is marked as W, the ambient humidity is marked as H, and the dust concentration in the air is marked as N;

[0041] By formula: P2 = X×e^[(W-ZW)^2 / ZW]+Y×e^[(H-ZH)^2 / ZH]+Z×e^[N / (N+1)];

[0042] The second evaluation coefficient is calculated; wherein X, Y, and Z are proportional coefficients, and 0<X<1, 0<Y<1, 0<Z<1, ZW is the optimum production temperature of dairy products during the production process, and ZH is the optimum production humidity of dairy products during the production process.

[0043] It should be noted that if the ambient temperature is too high, the reproduction rate of microorganisms such as bacteria and molds will be accelerated, which may cause the dairy products to deteriorate during processing or storage. At the same time, too high ambient temperature may accelerate the rate of chemical reactions, such as protein denaturation and fat oxidation, thus affecting the taste, texture and nutritional value of the product; low temperature environment may affect the stability of raw materials and finished products of dairy products, such as causing crystallization and coagulation of fat and protein, thus affecting the taste and texture of the product;

[0044] High humidity can easily lead to mold growth, which may contaminate dairy production equipment, raw materials and finished products, thus affecting the quality and safety of dairy products. Low humidity may cause raw materials and finished products to dry out too quickly during the dairy production process, affecting the taste and shelf life of the products.

[0045] Dust in the air may carry microorganisms such as bacteria and mold, which may contaminate dairy production equipment, raw materials and finished products, resulting in a decline in the quality of dairy products.

[0046] Preferably, the calculating the third evaluation coefficient based on the equipment data includes:

[0047] Mark the vibration frequency as L and the device temperature as T;

[0048] By formula: P3 = J × ln (L + 1) / tanh (L) + K × e^ [T / (T + 1)];

[0049] The third evaluation coefficient is calculated, wherein J and K are proportional coefficients, and 0<J<1, 0<K<1.

[0050] It should be noted that abnormalities in production equipment will directly affect the entire production process of dairy products. For example, if the equipment fails during the sterilization process, the dairy products will not be sterilized thoroughly and microbial residues will exist. This will not only reduce the quality of the dairy products but may also cause food safety issues. Equipment abnormalities will also affect the quality of dairy products. For example, if the production equipment fails during the processing, it may cause abnormalities in the color and taste of the dairy products. In addition, equipment failure may also affect the shelf life and stability of dairy products, making them more likely to deteriorate or lose their original nutritional value.

[0051] Preferably, judging whether it is necessary to issue an early warning for the production quality of dairy products based on the identification result and the production quality assessment coefficient includes:

[0052] If the recognition result is that the image is normal, no warning will be issued and continuous monitoring and judgment will be carried out;

[0053] If the recognition result is that the image is abnormal, a warning message is generated and sent to the client;

[0054] Determine whether the production quality assessment coefficient is greater than the preset threshold; if yes, generate an early warning message and send it to the client; if not, continue to monitor and judge.

[0055] It should be noted that the preset threshold is set by experts based on actual scenarios and experience.

[0056] A second aspect of the present invention provides a production quality dynamic early warning method based on deep learning, comprising:

[0057] Step 1: Real-time collection of milk images and production quality data of dairy products during the production process;

[0058] Step 2: extracting features from the milk body image of the dairy product to obtain feature data corresponding to the dairy product image; and training an artificial intelligence model based on historical image data to obtain an image anomaly recognition model; and recognizing the image data based on the image anomaly recognition model to obtain a recognition result;

[0059] Step 3: Calculate the production quality assessment coefficient of dairy products based on the production quality data of dairy products;

[0060] Step 4: Based on the identification results and production quality assessment coefficient, determine whether it is necessary to issue an early warning for the production quality of dairy products.

[0061] Compared with the prior art, the present invention has the following beneficial effects:

[0062] 1. In the actual production process of dairy products, environmental factors and equipment operating conditions will affect the final production quality of dairy products, but the existing technology mainly relies on a single data source to monitor and warn the production quality of dairy products, and it is obvious that the accuracy of monitoring and warning is not high. The present invention realizes multi-dimensional and high-precision monitoring of dairy product production quality by integrating a data acquisition module, an image analysis module, a data analysis module and an early warning module. The data acquisition module collects images and a variety of production quality data in the dairy product production process in real time, including content data, equipment data and environmental data; the image analysis module uses deep learning technology to extract features and identify anomalies in milk images, and the data analysis module calculates the production quality evaluation coefficient of dairy products based on the production quality data of dairy products; the early warning module combines the image recognition results and the production quality evaluation coefficient to monitor and warn the production quality in real time, overcoming the problem of insufficient monitoring accuracy of a traditional single data source, significantly improving the accuracy and reliability of the early warning, and ensuring the quality and safety of dairy product production.

[0063] 2. The present invention can fully capture the visual characteristics of the milk image by extracting the features of the milk image of the dairy product by color space conversion technology and image texture analysis technology, including converting the RGB color space into HSV or HSB color space to extract color feature data, and extracting texture feature data by using Gabor filtering or grayscale co-occurrence matrix and other technologies. This process can not only effectively identify the color changes of dairy products, but also sensitively detect the subtle differences in the surface texture of dairy products, which is crucial for judging whether the dairy products are contaminated, deteriorated or processed. Problems. Further, by constructing an image anomaly recognition model, that is, extracting historical feature data from historical image data and comparing it with reference feature data, annotating and classifying the historical image data, and then using the annotated data to train artificial intelligence models such as convolutional neural networks or deep belief networks, it is possible to achieve automatic recognition and anomaly detection of real-time collected milk images, which not only greatly improves the accuracy and efficiency of image recognition, but also can timely discover potential problems in the production process, and take measures in advance to avoid the occurrence of quality problems, thereby significantly improving the quality control level and safety of dairy production. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0065] Figure 1 A schematic diagram of a system module according to an embodiment of the present invention;

[0066] Figure 2 A schematic diagram of a specific process of an embodiment of the present invention;

[0067] Figure 3 Schematic diagram of the method steps of an embodiment of the present invention. DETAILED DESCRIPTION

[0068] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] See also Figure 1-Figure 2 , the first aspect of the present invention provides a production quality dynamic early warning system based on deep learning, including: a data acquisition module, an image analysis module, a data analysis module and an early warning module;

[0070] Data acquisition module: used to collect milk images and production quality data of dairy products in real time during the production process; the production quality data includes: content data, equipment data of production equipment and environmental data of dairy products during the production process;

[0071] Image analysis module: extracting features from the milk body image of the dairy product to obtain feature data corresponding to the dairy product image; and,

[0072] Based on the historical image data, an artificial intelligence model is trained to obtain an image anomaly recognition model; based on the image anomaly recognition model, the image data is recognized to obtain a recognition result;

[0073] Data analysis module: Calculate the production quality assessment coefficient of dairy products based on the production quality data of dairy products;

[0074] Early warning module: Based on the recognition results and production quality assessment coefficient, it is determined whether an early warning is needed for the production quality of dairy products.

[0075] Real-time collection of milk images and production quality data of dairy products during the production process, including:

[0076] The image acquisition device is used to collect the milk body images of dairy products in the production process in real time;

[0077] The data acquisition equipment collects real-time content data of dairy products, equipment data of production equipment and environmental data of dairy products during the production process; the content data includes: pH value and density of dairy products during the production process; the equipment data includes: vibration frequency of the equipment and equipment temperature; the environmental data includes: ambient temperature, ambient humidity and dust concentration in the air.

[0078] By extracting features from the milk body image of dairy products, including:

[0079] The breast image is converted from the RGB color space to the HSV or HSB color space by using the color space conversion technology to extract the color feature data of the breast image; wherein the color feature data includes: hue, saturation and brightness;

[0080] The texture feature data of the breast image is extracted by image texture analysis technology; wherein the image texture analysis technology includes: Gabor filtering or gray level co-occurrence matrix; the texture feature data includes: roughness and uniformity.

[0081] Identify image data based on image anomaly recognition models, including:

[0082] A1: extracting historical feature data corresponding to historical image data; wherein the historical image data refers to historical data of milk body images collected during the production process of dairy products; the historical feature data includes: color feature data and texture feature data corresponding to the historical image data;

[0083] A2: extracting reference feature data corresponding to a reference image; wherein the reference image refers to an image of a normal breast; the reference feature data includes: color feature data and texture feature data corresponding to the reference image;

[0084] A3: By comparing the historical feature data with the reference feature data, the historical image data is labeled and classified to obtain the labeling results; wherein the labeling results include: normal image and abnormal image;

[0085] A4: Integrate the historical feature data into standard input data, and integrate the annotation results corresponding to the historical feature data into standard output data; train an artificial intelligence model based on the standard input data and the standard output data to obtain an image anomaly recognition model; wherein the artificial intelligence model includes: a convolutional neural network or a deep belief network;

[0086] A5: The feature data corresponding to the real-time collected breast image is input into the image abnormality recognition model for recognition to obtain the recognition result.

[0087] Example 1: A dairy product processing plant hopes to improve its product quality monitoring capabilities, especially to monitor the status of dairy products in real time on the production line and promptly discover potential quality problems. The specific process is as follows:

[0088] 1. Data acquisition module;

[0089] Breast image acquisition:

[0090] High-definition cameras are installed at key locations on the production line to capture real-time images of dairy products during the production process. These cameras are configured to capture multiple frames per second to ensure that any subtle changes can be captured.

[0091] 2. Image analysis module;

[0092] Feature extraction:

[0093] The real-time collected breast images are converted from RGB color space to HSV color space, and color feature data such as hue, saturation and brightness are extracted.

[0094] The Gabor filtering technique is used to extract the texture feature data of breast images, including roughness and uniformity.

[0095] Image anomaly recognition model training:

[0096] A large amount of historical image data and its corresponding color feature data and texture feature data are extracted from the historical database.

[0097] A group of breast images known to be normal are selected as reference images, and their color feature data and texture feature data are extracted.

[0098] The historical feature data is compared with the reference feature data to distinguish normal and abnormal images, and the historical image data is labeled and classified (normal image or abnormal image).

[0099] The annotated historical feature data is used as standard input data, and the corresponding annotation results are used as standard output data to train an image anomaly recognition model based on convolutional neural network (CNN).

[0100] Real-time image recognition:

[0101] The feature data corresponding to the real-time collected breast image is input into the trained image abnormality recognition model for recognition.

[0102] The model outputs the recognition result, that is, whether the image is abnormal.

[0103] 3. Early warning module;

[0104] When the image analysis module identifies an abnormal image, the early warning module will receive the information and immediately trigger the early warning mechanism; by sending the early warning information, relevant personnel will be notified to take corresponding measures.

[0105] The production quality assessment coefficient of dairy products is calculated based on the production quality data of dairy products, including:

[0106] B1: The first evaluation coefficient is calculated based on the content data and marked as P1;

[0107] B2: The second evaluation coefficient is calculated based on the environmental data and marked as P2;

[0108] B3: The third evaluation coefficient is calculated based on the equipment data and marked as P3;

[0109] B4: The production quality assessment coefficient of dairy products is calculated by the formula: P = α × P1 + β × P2 + θ × P3; where P is the production quality assessment coefficient, α, β, θ are weight coefficients, and α + β + θ = 1.

[0110] The first evaluation coefficient is calculated based on the content data, including:

[0111] Label pH as Y and density as M;

[0112] By formula: P1 = C × e^[(Y-ZY)^2 / ZY]+D × e^[1 / ln(M+1)];

[0113] The first evaluation coefficient is calculated; wherein, C and D are proportional coefficients, and 0<C<1, 0<D<1, and ZY is the optimal production pH value of dairy products during the production process.

[0114] The second evaluation coefficient is calculated based on the environmental data, including:

[0115] The ambient temperature is marked as W, the ambient humidity is marked as H, and the dust concentration in the air is marked as N;

[0116] By formula: P2 = X×e^[(W-ZW)^2 / ZW]+Y×e^[(H-ZH)^2 / ZH]+Z×e^[N / (N+1)];

[0117] The second evaluation coefficient is calculated; wherein X, Y, and Z are proportional coefficients, and 0<X<1, 0<Y<1, 0<Z<1, ZW is the optimum production temperature of dairy products during the production process, and ZH is the optimum production humidity of dairy products during the production process.

[0118] The third evaluation coefficient is calculated based on the equipment data, including:

[0119] Mark the vibration frequency as L and the device temperature as T;

[0120] By formula: P3 = J × ln (L + 1) / tanh (L) + K × e^ [T / (T + 1)];

[0121] The third evaluation coefficient is calculated, wherein J and K are proportional coefficients, and 0<J<1, 0<K<1.

[0122] Based on the identification results and production quality assessment coefficients, it is determined whether an early warning is needed for the production quality of dairy products, including:

[0123] If the recognition result is that the image is normal, no warning will be issued and continuous monitoring and judgment will be carried out;

[0124] If the recognition result is that the image is abnormal, a warning message is generated and sent to the client;

[0125] Determine whether the production quality assessment coefficient is greater than the preset threshold; if yes, generate an early warning message and send it to the client; if not, continue to monitor and judge.

[0126] Example 2: Assume that a milk production line needs to be monitored to ensure that the quality of the milk reaches the optimal state. The specific process is as follows:

[0127] 1. Data collection;

[0128] Assume that at a certain moment, the system collects the following data:

[0129] Content data:

[0130] pH value (Y) = 6.6;

[0131] Density (M) = 1.035 g / cm 3 ;

[0132] Environmental data:

[0133] Temperature (W) = 22°C;

[0134] Humidity (H) = 55%;

[0135] Dust concentration (N) = 0.05 mg / m 3 ;

[0136] Device data:

[0137] Vibration frequency (L) = 12 Hz;

[0138] Equipment temperature (T) = 35°C;

[0139] 2. Data analysis;

[0140] S1: Calculate the first evaluation coefficient P1:

[0141] Assume that the optimum production pH value (ZY) = 6.8;

[0142] Assume that the proportionality coefficients C = 0.7, D ​​= 0.3;

[0143] The calculation formula is: P1 = C × e^[(Y-ZY)^2 / ZY]+D × e^[1 / ln(M+1)];

[0144] Calculation results:

[0145] P1=0.7×e^[(6.6-6.8)^2 / 6.8]+0.3×e^[1 / ln(1.035+1)]≈1.2737;

[0146] S2: Calculate the second evaluation coefficient P2;

[0147] Assume that the optimum production temperature (ZW) = 20°C and the optimum production humidity (ZH) = 60%;

[0148] Assume that the proportional coefficients X = 0.4, Y = 0.3, Z = 0.3;

[0149] The calculation formula is:

[0150] P2=X×e^[(W-ZW)^2 / ZW]+Y×e^[(H-ZH)^2 / ZH]+Z×e^[N / (N+1)];

[0151] Calculation results:

[0152] P2=0.4×e^[(26-20)^2 / 20]+0.3×e^[(55-60)^2 / 60]+0.3×e^[0.05 / (0.05+1)]≈3.1892;

[0153] S3: Calculate the third evaluation coefficient P3;

[0154] Assume that the proportionality coefficient J = 0.6, K = 0.4;

[0155] The calculation formula is: P3 = J × ln (L + 1) / tanh (L) + K × e^ [T / (T + 1)];

[0156] Calculation results:

[0157] P3=0.6×ln(12+1) / tanh(12)+0.4×e^[35 / (35+1)]≈2.5967;

[0158] S4: Calculate the production quality assessment coefficient P;

[0159] Assume that the weight coefficients α = 0.4, β = 0.3, θ = 0.3

[0160] The calculation formula is: P = α × P1 + β × P2 + θ × P3;

[0161] The calculation result is: P = 0.4 × 1.2737 + 0.3 × 3.1892 + 0.3 × 2.5967 ≈ 2.25;

[0162] 3. Early warning judgment;

[0163] Assume the preset threshold is 2.0;

[0164] Since the production quality assessment coefficient P (2.25) is greater than the preset threshold (2.0), an early warning is generated and sent to the client to remind the relevant person in charge to check the production process and ensure product quality;

[0165] When the client receives the early warning signal, the relevant person in charge can immediately take corresponding measures, such as suspending the production line, checking production equipment or raw materials, adjusting production process parameters, etc., to quickly respond to and correct potential quality problems, ensure the stability of the production process and the continuous compliance of product quality.

[0166] The present invention collects milk images and production quality data in real time during the dairy production process, and combines image analysis and data analysis technology to accurately judge the production quality status of dairy products and issue early warning signals in a timely manner when necessary, thereby effectively improving the quality control level and food safety assurance capabilities of the dairy production line, and bringing higher production efficiency and product reputation to the enterprise.

[0167] See also Figure 3 The second aspect of the present invention provides a production quality dynamic early warning method based on deep learning, comprising:

[0168] Step 1: Real-time collection of milk images and production quality data of dairy products during the production process;

[0169] Step 2: extracting features from the milk body image of the dairy product to obtain feature data corresponding to the dairy product image; and training an artificial intelligence model based on historical image data to obtain an image anomaly recognition model; and recognizing the image data based on the image anomaly recognition model to obtain a recognition result;

[0170] Step 3: Calculate the production quality assessment coefficient of dairy products based on the production quality data of dairy products;

[0171] Step 4: Based on the identification results and production quality assessment coefficient, determine whether it is necessary to issue an early warning for the production quality of dairy products.

[0172] Part of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is a formula closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.

[0173] Working principle of the present invention:

[0174] The present invention collects milk body images of dairy products during the production process and the production quality data of dairy products; extracts features from the milk body images of dairy products to obtain feature data corresponding to the dairy product images; and trains an artificial intelligence model based on historical image data to obtain an image anomaly recognition model; recognizes the image data based on the image anomaly recognition model to obtain a recognition result; calculates a production quality assessment coefficient of the dairy products based on the production quality data of the dairy products; and determines whether it is necessary to issue an early warning for the production quality of the dairy products based on the recognition result and the production quality assessment coefficient. The present invention solves the problem that in the actual production process of dairy products, environmental factors and equipment operating conditions will affect the final production quality of dairy products, but the prior art mainly relies on a single data source to monitor and warn the production quality of dairy products, and obviously the accuracy of monitoring and early warning is not high.

[0175] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A production quality dynamic early warning system based on deep learning, characterized in that: include: Data acquisition module, image analysis module, data analysis module and early warning module; Data acquisition module: used to collect milk images and production quality data of dairy products in real time during the production process; the production quality data includes: content data, equipment data of production equipment and environmental data of dairy products during the production process; Image analysis module: extracting features from the milk body image of the dairy product to obtain feature data corresponding to the dairy product image; and, Based on the historical image data, an artificial intelligence model is trained to obtain an image anomaly recognition model; based on the image anomaly recognition model, the image data is recognized to obtain a recognition result; Data analysis module: Calculate the production quality assessment coefficient of dairy products based on the production quality data of dairy products; Early warning module: Based on the recognition results and production quality assessment coefficient, it is determined whether an early warning is needed for the production quality of dairy products.

2. According to claim 1, a production quality dynamic early warning system based on deep learning is characterized in that: The real-time collection of milk body images and production quality data of dairy products during the production process includes: The image acquisition device is used to collect the milk body images of dairy products in the production process in real time; The data acquisition equipment collects real-time content data of dairy products, equipment data of production equipment and environmental data of dairy products during the production process; the content data includes: pH value and density of dairy products during the production process; the equipment data includes: vibration frequency of the equipment and equipment temperature; the environmental data includes: ambient temperature, ambient humidity and dust concentration in the air.

3. According to the deep learning-based production quality dynamic early warning system of claim 1, it is characterized in that: The method of extracting features from the milk body image of the dairy product comprises: The breast image is converted from the RGB color space to the HSV or HSB color space by using the color space conversion technology to extract the color feature data of the breast image; wherein the color feature data includes: hue, saturation and brightness; The texture feature data of the breast image is extracted by image texture analysis technology; wherein the image texture analysis technology includes: Gabor filtering or gray level co-occurrence matrix; the texture feature data includes: roughness and uniformity.

4. According to the deep learning-based production quality dynamic early warning system of claim 1, it is characterized in that: The identifying of the image data based on the image anomaly identification model includes: A1: extracting historical feature data corresponding to historical image data; wherein the historical image data refers to historical data of milk body images collected during the production process of dairy products; the historical feature data includes: color feature data and texture feature data corresponding to the historical image data; A2: extracting reference feature data corresponding to a reference image; wherein the reference image refers to an image of a normal breast; the reference feature data includes: color feature data and texture feature data corresponding to the reference image; A3: By comparing the historical feature data with the reference feature data, the historical image data is labeled and classified to obtain the labeling results; wherein the labeling results include: normal image and abnormal image; A4: Integrate the historical feature data into standard input data, and integrate the annotation results corresponding to the historical feature data into standard output data; train an artificial intelligence model based on the standard input data and the standard output data to obtain an image anomaly recognition model; wherein the artificial intelligence model includes: a convolutional neural network or a deep belief network; A5: The feature data corresponding to the real-time collected breast image is input into the image abnormality recognition model for recognition to obtain the recognition result.

5. According to the deep learning-based production quality dynamic early warning system of claim 1, it is characterized in that: The production quality assessment coefficient of the dairy product is calculated based on the production quality data of the dairy product, including: B1: The first evaluation coefficient is calculated based on the content data and marked as P1; B2: The second evaluation coefficient is calculated based on the environmental data and marked as P2; B3: The third evaluation coefficient is calculated based on the equipment data and marked as P3; B4: The production quality assessment coefficient of dairy products is calculated by the formula: P = α × P1 + β × P2 + θ × P3; where P is the production quality assessment coefficient, α, β, θ are weight coefficients, and α + β + θ = 1.

6. A production quality dynamic early warning system based on deep learning according to claim 5, characterized in that: The step of calculating the first evaluation coefficient based on the content data includes: Label pH value as Y and density as M; By formula: P1 = C × e^[(Y-ZY)^2 / ZY]+D × e^[1 / ln(M+1)]; The first evaluation coefficient is calculated; wherein, C and D are proportional coefficients, and 0<C<1, 0<D<1, and ZY is the optimal production pH value of dairy products during the production process.

7. The production quality dynamic early warning system based on deep learning according to claim 5 is characterized in that: The step of calculating the second evaluation coefficient based on the environmental data includes: The ambient temperature is marked as W, the ambient humidity is marked as H, and the dust concentration in the air is marked as N; By formula: P2 = X×e^[(W-ZW)^2 / ZW]+Y×e^[(H-ZH)^2 / ZH]+Z×e^[N / (N+1)]; The second evaluation coefficient is calculated; wherein X, Y, and Z are proportional coefficients, and 0<X<1, 0<Y<1, 0<Z<1, ZW is the optimum production temperature of dairy products during the production process, and ZH is the optimum production humidity of dairy products during the production process.

8. The deep learning-based production quality dynamic early warning system according to claim 5 is characterized in that: The calculating the third evaluation coefficient based on the equipment data includes: Mark the vibration frequency as L and the device temperature as T; By formula: P3 = J × ln (L + 1) / tanh (L) + K × e^ [T / (T + 1)]; The third evaluation coefficient is calculated, wherein J and K are proportional coefficients, and 0<J<1, 0<K<1.

9. The production quality dynamic early warning system based on deep learning according to claim 1 is characterized in that: The determining whether it is necessary to issue an early warning for the production quality of dairy products based on the identification result and the production quality assessment coefficient includes: If the recognition result is that the image is normal, no warning will be issued and continuous monitoring and judgment will be carried out; If the recognition result is that the image is abnormal, a warning message is generated and sent to the client; Determine whether the production quality assessment coefficient is greater than the preset threshold; if yes, generate an early warning message and send it to the client; if not, continue to monitor and judge.

10. A production quality dynamic early warning method based on deep learning, applied to a production quality dynamic early warning system based on deep learning according to any one of claims 1 to 9, characterized in that: include: Step 1: Real-time collection of milk images and production quality data of dairy products during the production process; Step 2: extracting features from the milk body image of the dairy product to obtain feature data corresponding to the dairy product image; and training an artificial intelligence model based on historical image data to obtain an image anomaly recognition model; Identify the image data based on the image anomaly recognition model to obtain a recognition result; Step 3: Calculate the production quality assessment coefficient of dairy products based on the production quality data of dairy products; Step 4: Based on the identification results and production quality assessment coefficient, determine whether it is necessary to issue an early warning for the production quality of dairy products.

Images