Noodle moisture online detection system and method

Abstract

The invention discloses a noodle moisture online detection system and method, and relates to the technical field of food detection, and the system comprises an image collection unit which is used for collecting RGB images of noodles in the production process of a noodle maker and converting the collected RGB images into grayscale images; the feature analysis unit is used for enhancing the grayscale image, calculating feature parameters of the enhanced grayscale image, constructing a feature matrix by using the feature parameters, and outputting a normalized feature matrix; and the detection unit is used for acquiring the water content of the noodle sample as calibration data, constructing a water content detection model, establishing a mapping relation between the image features and the water content of the noodles and outputting a detection value of the water content of the noodles. Real-time online detection of the water content of the noodles can be realized, a timely and accurate basis is provided for process regulation and control in the production process, and the production efficiency of noodle products is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of food detection, and in particular to an on-line noodle moisture detection system and method. Background Art

[0002] With the continuous expansion of the industrial production scale of noodles, the demand for accurate on-line detection of noodle moisture content has become increasingly prominent. In recent years, on-line moisture detection technology has gradually developed towards non-destructive and real-time directions, mainly covering means such as infrared spectroscopy, near-infrared (NIR) absorption method, microwave measurement, and resistance sensing. However, the high cost of sensors for spectral and microwave methods and their high sensitivity to material morphology and environmental conditions (such as temperature, humidity, production speed) have restricted their wide application on high-speed and batch production lines. At the same time, these technologies usually require complex signal de-spectroscopy and model calibration processes, and it is difficult to balance detection accuracy and system response speed.

[0003] In view of the deficiencies of the existing technologies, in recent years, scholars and engineers have tried to combine machine vision and image processing technologies to invert the moisture content by collecting noodle surface images and extracting features such as texture, grayscale, and morphology. Such methods have advantages such as non-contact, convenient installation, and relatively low cost, but still face multiple challenges in practical applications. Therefore, there is an urgent need for an on-line noodle moisture detection system that can not only efficiently process images and extract features, but also establish a robust mapping model to balance detection accuracy and on-line real-time performance. Summary of the Invention

[0004] In view of the problems existing in an existing on-line noodle moisture detection system, the present invention is proposed. Therefore, the problem to be solved by the present invention is how to provide an on-line noodle moisture detection system and method.

[0005] To solve the above technical problems, the present invention provides the following technical solutions:

[0006] In a first aspect, the present invention provides an on-line noodle moisture detection system, which includes an image acquisition unit, including an image acquisition module and an image preprocessing module. The image acquisition module is used to acquire RGB images of noodles during the production process of a noodle machine, and the image preprocessing module is used to convert the acquired RGB images into grayscale images;

[0007] A feature analysis unit, including an image enhancement module, a feature extraction module, and a parameter calculation module. The image enhancement module is used to enhance the grayscale image, the feature extraction module is used to calculate feature parameters for the enhanced grayscale image, and the parameter calculation module is used to construct a feature matrix using the feature parameters and output a normalized feature matrix;

[0008] The detection unit includes a data calibration module and a model construction module. The data calibration module is used to obtain the water content of the noodle sample as calibration data, and the model construction module is used to construct a water content detection model, establish a mapping relationship between the image features and the noodle water content, and output the detection value of the noodle water content.

[0009] As a preferred solution of the on-line noodle water detection system described in the present invention, the image enhancement module includes the following:

[0010] Obtain the input grayscale image as f(x,y) with a size of M×N; perform L-layer discrete wavelet decomposition on the grayscale image to obtain the lowest-frequency approximation sub-band A L (x,y) and the horizontal, vertical, and diagonal high-frequency sub-bands H i (x,y), V i (x,y), D i (x,y), where i = 1, 2,..., L; (x,y) represents the pixel coordinates and i represents the decomposition level;

[0011] Estimate the noise standard deviation of the high-frequency sub-bands. On the first-layer horizontal high-frequency sub-band H1, use the median absolute deviation MAD to estimate the noise standard deviation The expression is:

[0012]

[0013] In the formula, τ is the correction factor for converting MAD to the standard deviation of the Gaussian distribution;

[0014] Calculate the total number of coefficients N of each layer of high-frequency sub-bands i , and the expression is:

[0015]

[0016] Set the threshold scaling factor to α i , then the soft threshold T i is:

[0017]

[0018] For each high-frequency sub-band C i (x,y) ∈ {H i , V i , D i} perform the following correction, which is expressed as:

[0019]

[0020] In the formula, C i '(x,y) is the corrected high-frequency sub-band; keep the low-frequency sub-band A L unchanged;

[0021] Inverse wavelet transform is performed on all subbands {A L , H i ', V i ', D i '} to obtain the enhanced grayscale image f enh (x, y).

[0022] As a preferred solution of the online noodle moisture detection system described in the present invention, the feature extraction module includes the following:

[0023] The pixel values of the enhanced grayscale image are linearly mapped to G gray levels, and the expression is:

[0024]

[0025] In the formula, q(x, y) is the grayscale image after linear mapping;

[0026] The grayscale relationship of all pixel pairs in the image is statistically analyzed, and the gray-level co-occurrence matrix g(i, j) is generated according to the selected direction and distance; the probability of the occurrence of the normalized statistical gray-level pair (i, j) is expressed as:

[0027]

[0028] In the formula, P k (i, j) is the occurrence probability of the combination of gray levels i and j; g(i, j) is the count of the gray-level pair (i, j) in the gray-level co-occurrence matrix; L1 is the number of gray levels of the image;

[0029] The energy, contrast, homogeneity, and entropy values are calculated for the gray-level co-occurrence matrix respectively, and the parameters in each direction are averaged to obtain the final four-dimensional feature.

[0030] As a preferred solution of the online noodle moisture detection system described in the present invention, the parameter calculation module includes constructing a feature matrix for the feature parameters, extracting the four-dimensional features from multiple frames of images respectively, forming a feature matrix, and performing normalization processing, outputting the normalized feature matrix, and inputting it to the model construction module.

[0031] As a preferred solution of the online noodle moisture detection system described in the present invention, the data calibration module includes the following:

[0032] Take out multiple noodle samples from the production line and measure the fresh mass of the noodle samples;

[0033] Place the samples in an oven and dry them at 105 °C until they reach a constant weight state, and record the mass after drying;

[0034] The water content of the noodles is calculated based on the fresh quality, and the moisture content of the noodle sample is calculated using the formula:

[0035]

[0036] where w v is the moisture content of the v-th noodle sample;

[0037] The sample number, fresh quality, mass after drying, moisture content, and the image feature vectors collected synchronously are formed into a calibration data set.

[0038] As a preferred solution of the on-line noodle moisture detection system described in the present invention, wherein: the model construction module includes the following:

[0039] The calibration data set is randomly divided into a training set and a validation set, and the support vector regression algorithm with a kernel function is used. The regression function f(x) is expressed as:

[0040]

[0041] where {α j -α j *} are the Lagrange multipliers obtained by training, b is the bias term, K(·,·) is the kernel function, and M is the number of support vectors in the training set;

[0042] Optimal hyperparameters are searched for on the validation set through grid search combined with cross-validation. Using all the training set data, the elements in the calibration data set are input, and the support vector regression algorithm is trained with the optimal hyperparameters to obtain the final regression model f * (·);

[0043] The obtained final regression model is evaluated, and evaluation metrics are calculated on the validation set. The evaluation metrics include the root mean square error and the coefficient of determination; if the evaluation metrics do not meet the set standards, hyperparameter optimization is returned, and the hyperparameters are re-optimized for re-evaluation;

[0044] A moisture content detection model is obtained. The feature of the new frame image is input into the model, and the detected moisture content of the noodles during the production process is output Judge the moisture state of the noodles during the production process.

[0045] As a preferred solution of the on-line noodle moisture detection system described in the present invention, wherein: judging the moisture state of the noodles during the production process includes: if the detected moisture content < the lower limit of the moisture content, the moisture state of the noodle sample is dry; if the detected moisture content > the upper limit of the moisture content, the moisture state of the noodle sample is wet; otherwise, the moisture state of the noodle sample is normal.

[0046] In a second aspect, the present invention provides a method for on-line detection of noodle moisture content, which includes: collecting RGB images of noodles during the production process of a noodle machine and converting the RGB images into grayscale images;

[0047] Enhancing the grayscale image, calculating characteristic parameters for the enhanced grayscale image, and constructing a characteristic matrix using the characteristic parameters;

[0048] Obtaining the water content of noodle samples as calibration data;

[0049] Constructing a moisture content detection model, establishing a mapping relationship between image features and noodle moisture content, and outputting a detection value of noodle moisture content.

[0050] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the processor executes the computer program, the steps of a method for on-line detection of noodle moisture content are implemented.

[0051] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by a processor, the steps of a method for on-line detection of noodle moisture content are implemented.

[0052] The beneficial effects of the present invention are as follows: The present invention can enhance the reliability of detection results, improve the detection accuracy of the model and the consistency of practical applications. It realizes real-time on-line detection of noodle moisture content, provides timely and accurate basis for process control during production, and improves the consistency and overall production efficiency of noodle products. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for description in the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0054] Figure 1 It is a structural diagram of a system for on-line detection of noodle moisture content. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] In order to make the above objects, features and advantages of the present invention more understandable, the following will describe the specific embodiments of the present invention in detail with reference to the drawings of the specification. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0057] Secondly, as used herein, "one embodiment" or "an embodiment" refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or mutually exclusive embodiments selectively.

[0058] Referring to Figure 1 , which is the first embodiment of the present invention. This embodiment provides an on-line noodle moisture detection system, including:

[0059] An image acquisition unit, including an image acquisition module and an image preprocessing module. The image acquisition module is used to acquire the RGB images of the noodles during the production process of the noodle machine, and the image preprocessing module is used to convert the acquired RGB images into 8-bit grayscale images. This unit is used to obtain continuous image data on the noodle production line;

[0060] Specifically, the image acquisition module is responsible for obtaining the color images of the noodles in real time from the production line and directly transmitting the original frames to the preprocessing module.

[0061] The triggering methods include production line synchronous triggering and timing triggering; Production line synchronous triggering: Through a photoelectric switch or encoder pulse, a shooting signal is sent when the noodles reach the camera's field of view. Timing triggering: When the production rhythm is stable, automatic shooting is performed at a fixed frame rate (such as 30fps) to ensure the coherence of the image stream.

[0062] Frame capture: After the camera receives the trigger signal, it immediately acquires a complete RGB image (including R, G, and B channel data) and transmits the acquired RGB image to the image preprocessing module.

[0063] The image preprocessing module crops the color frame into the region of interest and converts it into a grayscale image, and outputs an image stream that can be directly used for moisture detection.

[0064] Obtain each frame of RGB image pushed by the image acquisition module, quickly identify the noodle area using a fixed threshold or a simple template, crop out the smallest rectangular area (ROI) containing only the noodles according to the recognition result, and remove the background.

[0065] Perform grayscale conversion on the image in the smallest rectangular area, calculate the weighted average for each pixel within the ROI, and output a single-channel grayscale image with a numerical range of 0–255.

[0066] Push each frame of grayscale image to the image enhancement module in the feature analysis unit through a message queue or callback, and at the same time save the image locally in parallel for subsequent offline backtracking and algorithm optimization.

[0067] The feature analysis unit includes an image enhancement module, a feature extraction module, and a parameter calculation module;

[0068] Specifically, the image enhancement module is used to enhance the grayscale image, and the wavelet transform combined with the adaptive threshold suppression method is used to enhance the grayscale image;

[0069] Select the wavelet basis and the decomposition level. Denote the used wavelet basis as ψ and the decomposition level as L. The input grayscale image is f(x, y) with a size of M×N.

[0070] Perform L-level discrete wavelet decomposition on f(x, y) to obtain the lowest-frequency approximation subband A L (x, y) and the horizontal, vertical, and diagonal high-frequency subbands H i (x, y), V i (x, y), D i (x, y), i = 1, 2, …, L.

[0071] In the formula, (x, y) represents the pixel coordinates, and i represents the decomposition level;

[0072] Estimate the noise standard deviation of the high-frequency subbands. On the first-level horizontal high-frequency subband H1, estimate the noise standard deviation using the median absolute deviation (MAD) The expression is:

[0073]

[0074] In the formula, τ is the correction factor for converting MAD to the standard deviation of the Gaussian distribution;

[0075] Perform adaptive threshold calculation, and calculate the total number N of coefficients of each high-frequency subband i , and the expression is:

[0076]

[0077] Set the threshold scaling factor as α i , then the soft threshold T i is:

[0078]

[0079] Perform soft threshold suppression, and perform the following correction on each high-frequency subband C i (x, y) ∈ {H i , V i , D i}, which is expressed as:

[0080]

[0081] Wherein, C i '(x, y) is the corrected high-frequency sub-band; the low-frequency sub-band A L remains unchanged;

[0082] For all sub-bands {A L , H i ', V i ', D i '}, perform inverse wavelet transform to obtain the enhanced grayscale image f enh (x, y).

[0083] The feature extraction module is used to calculate the feature parameters of the enhanced grayscale image, and obtain the feature parameters of the enhanced grayscale image by calculating the energy, contrast, homogeneity, entropy value and other feature parameters of the gray-level co-occurrence matrix;

[0084] Specifically, linearly map the pixel values of the enhanced grayscale image from [0, 255] to G gray levels:

[0085]

[0086] Wherein, q(x, y) is the gray-level image after linear mapping;

[0087] Construct a gray-level co-occurrence matrix, count the gray-level relationships of all pixel pairs in the image, and generate a matrix g(i, j) according to the selected directions (such as 0°, 45°, 90°, 135°) and distances (usually 1); g(i, j) is the count of the gray-level pair (i, j) in the gray-level co-occurrence matrix;

[0088] Normalize the probability of the occurrence of the gray-level pair (i, j), expressed as:

[0089]

[0090] Wherein, P(i, j) is the occurrence probability of the combination of gray levels i and j; g(i, j) is the count of the gray-level pair (i, j) in the gray-level co-occurrence matrix; L1 is the number of gray levels of the image (usually 256, that is, an 8-bit grayscale image);

[0091] Calculate the following parameters for each direction of the gray-level co-occurrence matrix respectively, and the calculation formula is:

[0092]

[0093] Among them, E is the energy of the gray-level co-occurrence matrix, Co is the contrast of the gray-level co-occurrence matrix, Ho is the homogeneity of the gray-level co-occurrence matrix, S is the entropy value of the gray-level co-occurrence matrix, and ∈ is a constant used to avoid the logarithm being zero;

[0094] Perform direction summarization, take the average of the parameters in each direction to obtain the final four-dimensional feature, and the expression is:

[0095]

[0096] where K is the number of offset directions.

[0097] The parameter calculation module is used to construct a feature matrix using feature parameters and output the normalized feature matrix;

[0098] Specifically, construct a feature matrix from the feature parameters, extract the four-dimensional features (E, Co, Ho, S) from the Q1 frame image respectively, and form the feature matrix X:

[0099]

[0100] Perform normalization processing, calculate the minimum value min of the j-th column j and the maximum value max j , and then map it to [0, 1], which is expressed as:

[0101]

[0102] Finally, output the normalized feature matrix X' for subsequent use in the moisture detection model.

[0103] The detection unit includes a data calibration module and a model construction module;

[0104] Specifically, the data calibration module is used to obtain the water content of the noodle sample as calibration data, and use the 105°C drying method to measure the water content of the noodle sample for providing calibration data for model training.

[0105] According to the production rhythm, regularly (or randomly) take multiple noodle samples from the production line, and measure the fresh mass m of the v-th noodle sample f，v (unit: g);

[0106] Place the sample in an oven and dry it at 105°C for a sufficient time (usually 4h) until the mass difference between two weighings is ≤ 0.1g, indicating that the constant weight state is reached, and record the mass m of the v-th noodle sample after drying d，v (unit: g);

[0107] Calculate the water content of the noodles based on the fresh mass, and calculate the moisture content of the v-th noodle sample. The formula is:

[0108]

[0109] In the formula, w v is the moisture content of the v-th noodle sample;

[0110] Construct a calibration data set, where each record contains: sample number v, fresh mass m f，v , mass m after drying d，v , moisture content w v and the image feature vector x collected synchronously v '=(E v ,Co v ,Ho v ,S v ); Finally, a calibration data set is formed

[0111] The model construction module is used to construct a moisture content detection model. The support vector regression (SVR) algorithm is used to construct the moisture content detection model, establish the mapping relationship between image features and moisture content, and output the detection value of the noodle moisture content

[0112] Randomly divide the calibration data set into a training set and a validation set. Use the support vector regression (SVR) algorithm with a kernel function. The regression function f(x) is expressed as:

[0113]

[0114] where {α j -α j *} are the Lagrange multipliers obtained by training, b is the bias term, K(·,·) is the kernel function (usually the radial basis function RBF), and M is the number of support vectors in the training set

[0115] The main hyperparameters include: penalty factor C: controlling the trade-off between fitting error and model complexity; kernel function parameter γ: controlling the bandwidth of the RBF kernel; ε: the size of the regression ε-insensitive loss interval. Use grid search combined with cross-validation to find the optimal (C,γ,ε) on the validation set

[0116] Use all the training set data, input {x v ',w v}, train the SVR with the optimal hyperparameters to obtain the final regression model f * (·)

[0117] Evaluate the obtained final regression model, calculate the evaluation metrics on the validation set. The evaluation metrics include root mean square error RMSE and coefficient of determination R 2 , and the formula is:

[0118]

[0119] where is the average moisture content of the noodle samples in the validation set, N valThe number of noodle samples in the validation set.

[0120] If the evaluation index does not meet the set standard, return to hyperparameter optimization, re-optimize the hyperparameters and conduct another evaluation.

[0121] Serialize and store the trained moisture content detection model, and load it into the online system. Input the new frame image features online, and the detected moisture content of the noodles during the production process can be output in real time. At the same time, judge the moisture state of the noodles during the production process.

[0122] Monitor the detected moisture content in real time, compare it with the preset target range, judge the moisture state of the noodle samples, and set the lower moisture content limit w min and the upper moisture content limit w max ,

[0123] If the detected moisture content is lower than the lower moisture content limit w min , the moisture state of the noodle samples is dry; if the detected moisture content is higher than the upper moisture content limit w max , the moisture state of the noodle samples is wet; otherwise, the moisture state of the noodle samples is normal.

[0124] Furthermore, this embodiment also provides a method for on-line detection of noodle moisture, including:

[0125] Collect the RGB images of the noodles during the production process by the noodle machine, and convert the RGB images into grayscale images;

[0126] Enhance the grayscale images, calculate the feature parameters for the enhanced grayscale images, and use the feature parameters to construct a feature matrix;

[0127] Obtain the moisture content of the noodle samples as calibration data;

[0128] Construct a moisture content detection model, establish a mapping relationship between the image features and the noodle moisture content, and output the detected value of the noodle moisture content.

[0129] This embodiment also provides a computer device applicable to the situation of a method for on-line detection of noodle moisture, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement all or part of the steps of the method described in the above-mentioned embodiments of the present invention.

[0130] This embodiment also provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, it executes the method in any optional implementation manner of the above embodiment. Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM for short), electrically erasable programmable read-only memory (EEPROM for short), erasable programmable read-only memory (EPROM for short), programmable read-only memory (PROM for short), read-only memory (ROM for short), magnetic memory, flash memory, magnetic disk or optical disc.

[0131] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. For the technical details not described in detail in this embodiment, reference can be made to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0132] In summary, the present invention can enhance the reliability of the detection result, improve the detection accuracy of the model and the consistency of practical applications. It realizes the real-time online detection of the water content of noodles, provides timely and accurate basis for process control in the production process, and improves the consistency and overall production efficiency of noodle products.

[0133] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not restrictive. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. An on-line noodle moisture detection system, characterized in that: including an image acquisition unit, including an image acquisition module and an image preprocessing module, where the image acquisition module is used to acquire the RGB image of the noodles during the production process of the noodle machine, and the image preprocessing module is used to convert the acquired RGB image into a grayscale image; a feature analysis unit, including an image enhancement module, a feature extraction module and a parameter calculation module, where the image enhancement module is used to enhance the grayscale image, the feature extraction module is used to calculate feature parameters for the enhanced grayscale image, and the parameter calculation module is used to construct a feature matrix using the feature parameters and output the normalized feature matrix; a detection unit, including a data calibration module and a model construction module, where the data calibration module is used to obtain the water content of the noodle sample as calibration data, and the model construction module is used to construct a water content detection model, establish the mapping relationship between the image features and the water content of the noodles, and output the detection value of the water content of the noodles.

2. The online noodle moisture detection system according to claim 1, characterized in that: The image enhancement module includes the following: The input grayscale image obtained is f(x, y) with a size of M×N; perform L-level discrete wavelet decomposition on the grayscale image to obtain the lowest-frequency approximation subband A L (x, y) and the horizontal, vertical, and diagonal high-frequency subbands H i (x, y), V i (x, y), D i (x, y), where i = 1, 2, …, L; (x, y) represents the pixel coordinates and i represents the decomposition level; Estimate the noise standard deviation for the high-frequency subbands. On the first-level horizontal high-frequency subband H1, use the median absolute deviation (MAD) to estimate the noise standard deviation The expression is as follows: where τ is the correction factor for converting MAD to the standard deviation of the Gaussian distribution; Calculate the total number N of coefficients in each high-frequency subband i , and the expression is: Set the threshold scaling factor to α i , then the soft threshold T i is: For each high-frequency subband C i (x,y) ∈ {H i , V i , D i} perform the following correction, denoted as: where C i '(x, y) is the corrected high-frequency subband; the low-frequency subband A L remains unchanged; Inverse wavelet transform is performed on all subbands {A L , H i , V i , D i} to obtain the enhanced grayscale image f enh (x, y).

3. The on-line noodle moisture detection system according to claim 2, characterized in that: The feature extraction module includes the following: Linearly map the pixel values of the enhanced grayscale image to G gray levels, and the expression is: where q(x,y) is the grayscale image after linear mapping; Statistically analyze the gray relationship of all pixel pairs in the image, and generate a gray-level co-occurrence matrix g(i,j) according to the selected direction and distance; normalize the probability of the occurrence of the gray pair (i,j), which is expressed as: where P k (i, j) is the occurrence probability of the combination of gray levels j and j; g(i, j) is the number of times the gray level pair (i, j) appears in the gray level co-occurrence matrix; L1 is the number of gray levels of the image; Calculate the energy, contrast, homogeneity and entropy values for the gray-level co-occurrence matrix respectively, and average the parameters in each direction to obtain the final four-dimensional feature.

4. The on-line moisture detection system for noodles according to claim 3, characterized in that: The parameter calculation module includes constructing a feature matrix for the feature parameters, extracting four-dimensional features for multiple frames of images respectively, forming a feature matrix, and performing normalization processing, outputting the normalized feature matrix, and inputting it into the model construction module.

5. The on-line noodle moisture detection system according to claim 4, characterized in that: The data calibration module includes the following: Take out multiple noodle samples from the production line and measure the fresh mass of the noodle samples; Place the samples in an oven and dry them at 105 °C until they reach a constant weight state, and record the mass after drying; Calculate the water content of the noodles based on the fresh mass, and calculate the moisture content of the noodle samples. The formula is: where w v is the moisture content of the v-th noodle sample; Form a calibration data set with the sample number, fresh mass, mass after drying, moisture content, and the image feature vector collected synchronously.

6. The on-line moisture detection system for noodles according to claim 5, characterized in that: The model construction module includes the following: Randomly divide the calibration data set into a training set and a validation set, and adopt a support vector regression algorithm with a kernel function. The regression function f(x) is expressed as: Among them, {α j -α j *} are Lagrange multipliers solved through training, b is the bias term, K(·,·) is the kernel function, and M is the number of support vectors in the training set; Find the optimal hyperparameters on the validation set through grid search combined with cross-validation. Use all the training set data, input the elements in the calibration dataset, and train the support vector regression algorithm with the optimal hyperparameters to obtain the final regression model f * (·); Evaluate the obtained final regression model, calculate the evaluation indexes on the validation set, and the evaluation indexes include the root mean square error and the coefficient of determination; if the evaluation indexes do not meet the set standards, return to hyperparameter optimization, re-optimize the hyperparameters and perform re-evaluation; Obtain a moisture content detection model, input the features of the new frame image into the model, and output the detected moisture content of the noodles during the production process. Judge the moisture state of the noodles during the production process.

7. The online noodle moisture detection system according to claim 6, characterized in that: Judging the water content state of the noodles during the production process includes: if the detected moisture content < the lower limit of the moisture content, the water content state of the noodle sample is dry; if the detected moisture content > the upper limit of the moisture content, the water content state of the noodle sample is wet; otherwise, the water content state of the noodle sample is normal.

8. A method for on-line detection of noodle moisture, based on the on-line noodle moisture detection system according to any one of claims 1 to 7, characterized in that: including Acquire the RGB image of the noodles during the production process of the noodle machine, and convert the RGB image into a grayscale image; Enhance the grayscale image, calculate the characteristic parameters for the enhanced grayscale image, and construct a feature matrix using the characteristic parameters; Obtain the water content of the noodle sample as calibration data; Construct a water content detection model, establish a mapping relationship between the image features and the water content of the noodles, and output the detection value of the water content of the noodles.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the online noodle moisture detection method described in claim 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the online noodle moisture detection method described in claim 8.

Images