Pork quality grade nondestructive evaluation method based on hyperspectral imaging

Abstract

The invention discloses a pork quality grade lossless evaluation method based on hyperspectral imaging, and relates to the technical field of pork detection.The pork quality grade lossless evaluation method comprises the following steps that pork samples of different qualities are prepared, and standardization processing is conducted on the pork samples; performing chromatic aberration analysis on the pork samples with different qualities by using a colorimeter, and performing quality grade division on the pork samples with different qualities based on the analysis results of the muscle color, the fat color and the gloss of the pork. According to the pork quality grade lossless evaluation method based on hyperspectral imaging, reflection or transmission information of a pork sample in a plurality of continuous spectral wavebands is captured by using a hyperspectral imaging technology, accurate evaluation of the pork quality grade is realized by analyzing spectral data, higher-dimension data is provided, and the evaluation accuracy of the pork quality grade is improved. Meanwhile, rapid scanning can be achieved, the detection time is greatly shortened, the overall detection efficiency is improved, and pork quality monitoring is more efficient and timely.

Description

Technical Field

[0001] The invention relates to the technical field of pork detection, and in particular to a non-destructive evaluation method for pork quality grade based on hyperspectral imaging. Background Art

[0002] In order to ensure the sanitary quality of pork, it is necessary to test it in order to obtain objective and accurate pork quality information to meet the market demand for high-quality pork. In the pork quality testing process, the testing technology is required to be able to quickly and accurately complete the assessment of pork quality grades in order to reduce production cycles and costs. The non-destructive testing technology using hyperspectral imaging technology can perform testing without destroying the sample, obtain the spectral and spatial information of the pork, and achieve non-destructive evaluation of the pork quality grade, thus avoiding the damage to the sample by traditional testing methods.

[0003] In current technology, hyperspectral imaging is used to detect pork. Due to the different surface conditions of pork samples, errors in the detection effect will occur, which will in turn affect the accuracy of the test results. Therefore, how to comprehensively detect the spectral information of various characteristics of the samples to improve the evaluation accuracy of the pork quality grade is the problem we need to solve. To this end, a non-destructive evaluation method for pork quality grade based on hyperspectral imaging is proposed. Summary of the invention

[0004] To achieve the above objectives, the present invention is implemented by the following technical scheme: a non-destructive evaluation method for pork quality grade based on hyperspectral imaging, comprising the following steps:

[0005] Prepare pork samples of different qualities and standardize them. Cut the pork samples into uniform sizes and shapes to eliminate the effects of shape and size on the test results. At the same time, ensure that the sample surface is clean to reduce the interference of impurities on hyperspectral imaging.

[0006] Use a colorimeter to analyze the color difference of pork samples of different qualities, and classify the pork samples of different qualities into different quality grades based on the analysis results of pork muscle color, fat color and gloss;

[0007] Configure the hyperspectral imaging system to calibrate the system, preheat the hyperspectral imager, and adjust the environmental parameters of the detection environment;

[0008] Perform hyperspectral imaging on pork samples, collect hyperspectral image information of pork samples at different wavelengths, correct the collected hyperspectral images, and use image segmentation technology to segment the spectral images of pork samples from the background to obtain complete pork image information;

[0009] The color feature vector is extracted from the acquired pork image information through image analysis, and the correlation between pork quality and image color features is analyzed using a pork quality evaluation model.

[0010] Deploy a pork quality evaluation model, collect hyperspectral images of the pork to be inspected in real time, and determine the quality grade of the current pork to be inspected by combining the quality evaluation results and the classified pork quality grades.

[0011] Preferably, the standardized processing of the pork sample specifically includes:

[0012] According to the testing requirements, select pork samples of different qualities to ensure that the samples can cover different qualities, including fresh pork samples and pork samples under different deterioration conditions. Among them, pork samples include tenderloin, leg meat, ham, etc.;

[0013] Place the pork sample on the workbench, use a ruler and a marker to mark the area to be cut on the pork, and cut the pork sample along the mark to ensure that the thickness and weight of each sample are consistent to facilitate the penetration and reflection of light during hyperspectral imaging;

[0014] Mark the cut pork samples, record their quality, processing date, storage conditions, and store pork samples of different qualities in corresponding sealed containers to avoid cross-infection;

[0015] Inspect the surface of the pork sample and clean any blood, hair, and moisture impurities adhering to the surface of the pork.

[0016] Preferably, the color difference analysis and quality grade classification process specifically includes:

[0017] Calibrate the colorimeter and use it to perform colorimetric analysis on the muscle and fat parts of the pork samples, measure the L value (brightness), a value (red-green value) and b value (yellow-blue value) of the pork samples, and create a data record table for each sample to record its L, a, b values ​​and corresponding quality information, and perform statistical analysis to evaluate the color characteristics of the pork samples;

[0018] Based on the measured L value (brightness), a value (red-green value) and b value (yellow-blue value), the pork samples were scored and measured according to the marbling score standard and the determination method of intramuscular fat content. The marbling score and intramuscular fat content were combined with the color difference data to evaluate the muscle color, fat color and gloss of pork. The color difference data of pork samples of different qualities were compared, and the differences in L, a and b values ​​of pork samples of different qualities were analyzed. The differences in muscle color, fat color and gloss were further analyzed in combination with the marbling score and intramuscular fat content.

[0019] The quality grade classification standard of pork samples is formulated based on the color difference analysis results, muscle color, fat color, gloss analysis results and the degree of deviation of color difference data;

[0020] According to the established classification standards, pork samples of different qualities are divided into quality grades, namely low quality grade, medium quality grade and high quality grade.

[0021] Preferably, the calibration process of the hyperspectral imaging system specifically includes:

[0022] Assemble the hyperspectral imager, camera, lens and light source to form a hyperspectral imaging system, ensure stable connection, avoid looseness or poor contact, connect the control software and data processing software of the hyperspectral imaging system, and configure the spectral range, resolution and acquisition speed parameters of the hyperspectral imaging system;

[0023] Use standard light sources to calibrate the spectral response of the hyperspectral imaging system, measure the spectral response of each pixel and the response changes at different wavelengths, eliminate the spectral response differences between different pixels, use calibration objects to measure the focal length of the camera, lens distortion, and the relationship parameters between pixels, calibrate the geometric parameters of the hyperspectral imaging system, and ensure that the imaging results are true and reliable;

[0024] Use a color card to perform color calibration, measure the difference between the color of the color card in the image and its actual color, and adjust the color response function of the hyperspectral imaging system based on the measurement results so that the color in the imaging process can be accurately restored and corrected;

[0025] Measure the output electrical signals of the camera of the hyperspectral imaging system under different exposure conditions and the output electrical signals in the absence of light. Based on the measurement results, reduce the influence of noise and eliminate the influence of dark current by taking multiple shots of the same scene and taking the average value.

[0026] Perform a warm-up procedure on the hyperspectral imager. The warm-up time is between 30 minutes and 1 hour to ensure the stability of the equipment and the imaging quality. During the warm-up process, pay attention to monitor the status of the imager to ensure that the warm-up is complete and the system is stable.

[0027] According to the detection requirements, adjust the brightness, color temperature and angle of the light source to ensure stable and uniform lighting conditions to avoid interference with the imaging results. Use a thermometer and hygrometer to monitor the temperature and humidity of the detection environment, and adjust the temperature and humidity according to the detection requirements of the hyperspectral imaging system.

[0028] Preferably, the process of obtaining the pork image information specifically includes:

[0029] According to the characteristics of the pork sample and the detection requirements, the imaging parameters are set, wherein the imaging parameters include spectral range, resolution, acquisition speed and exposure time;

[0030] Place the pork sample in the detection area, start the hyperspectral imaging system, image the pork sample, collect the hyperspectral image information of the pork sample at different wavelengths, and form a hyperspectral data cube;

[0031] Import the collected hyperspectral image information into the data processing software and perform correction processing on it, including spectral response correction, geometric correction and color correction. Among them, use a standard light source to correct the spectral response of the hyperspectral imaging system to ensure that the response of each pixel at different wavelengths is consistent. By measuring the spectral characteristics of the standard light source, the spectral response curve of the system is adjusted to eliminate the spectral response differences between pixels. The hyperspectral image is geometrically corrected using known feature points in the image. By establishing a corresponding relationship between the original image space and the coordinate space used for mapping, the image is accurately corrected in geometry to ensure that the shape and position of the object in the image are consistent with the actual one. Use a color card for color correction to ensure that the hyperspectral imaging system can accurately restore and correct the color in the image. By measuring the difference between the color of the color card in the image and its true color, the color response function of the system is adjusted;

[0032] The corrected hyperspectral image is preprocessed, including denoising and contrast enhancement, to improve the accuracy of image segmentation. According to the spectral characteristics and spatial distribution characteristics of the pork sample and the background, the image segmentation method of edge detection segmentation is selected to segment the hyperspectral image, and the spectral image of the pork sample is separated from the background.

[0033] The segmented images are post-processed, including removing small area noise and filling holes, to obtain complete pork image information.

[0034] Preferably, the process of constructing the pork quality evaluation model specifically includes:

[0035] Integrate the collected pork image information to form a data set, and attach the corresponding quality evaluation data of each pork sample, analyze the saturation, freshness, tenderness, flavor and water retention of each pork sample, and clarify its quality grade;

[0036] The pork image is converted from RGB color space to Lab color space by image analysis technology, and relevant color feature data is extracted in the converted Lab color space, including statistical features of brightness and hue and color moment features, where brightness is the L component and hue is the a and b components;

[0037] A pork quality evaluation model was constructed using a neural network algorithm combined with the extracted color feature data. The extracted color feature data was used as the input of the model, and the quality grade of pork was used as the output of the model. The model was verified using the image information in the data set, and the model was optimized to improve its prediction performance.

[0038] Combining the output of the pork quality evaluation model and the extracted color feature data, the benchmark values ​​of the color features, saturation and freshness of the pork samples were preset, and the quality evaluation coefficient was calculated through comprehensive analysis to quantify the quality level of pork. The impact of color features on pork quality was then analyzed, and the correlation between pork quality and image color features was evaluated.

[0039] According to the output results of the quality evaluation coefficient and the divided quality levels, the quality evaluation thresholds corresponding to each quality level are determined.

[0040] Preferably, the calculation process of the quality evaluation coefficient specifically includes:

[0041] Traverse the color characteristics, saturation and freshness data of each pork sample, and determine the base values ​​of color characteristics, saturation and freshness respectively. For color characteristics, the base value of brightness is determined according to the meat standard. L , based on the standard color of fresh pork, the base values ​​of the a and b components of the hue are determined as Base a and Base b , according to the maximum allowable range of color change of fresh pork, determine the range of the reference value of the hue a and b components ab , based on the storage conditions and shelf life of pork, the baseline values ​​of freshness and saturation are determined as Base S and Base P ;

[0042] Calculate the maximum value max(L) and minimum value min(L) of brightness from the extracted color features, and calculate the ratio of the brightness range to the reference value Calculate the brightness change function T L , analyze the uniformity of the brightness range;

[0043] Calculate the absolute deviation of the mean of hue a and b components from the reference value |μ a -Base a | and |μ b -Base b |, and add the two deviations and divide them by the hue range reference value Range ab , calculate the hue mean influence function T ab , analyze the closeness between the hue mean and the reference value;

[0044] Use an exponential decay function Calculate the effect of saturation and use the limit function Calculate the impact of freshness, the ratio of comprehensive saturation and freshness, and calculate the quality impact function T SP , which measures the freshness and color saturation of pork;

[0045] The brightness change function, hue mean influence function and quality influence function were integrated to calculate the quality evaluation coefficient Q and quantitatively analyze the quality level of pork.

[0046] Preferably, the calculation expression of the quality evaluation coefficient is:

[0047]

[0048] Among them, Q is the quality evaluation coefficient, T L is the brightness change function, T ab is the hue mean influence function, T SP is the quality impact function, and the value of Q ranges from 0 to 1.

[0049] Preferably, the value of the quality evaluation coefficient and the evaluation threshold preset for the quality level are obtained to match the three quality levels, which are:

[0050] Low quality level: 0≤Q dz Pork samples of this grade usually show poor color characteristics, with obvious color differences and unevenness, and low saturation and freshness;

[0051] Medium quality grade: Q dz ≤Q gz ; The pork samples of this grade are relatively good in color and texture, with hue and brightness close to the benchmark values, and saturation and freshness at a medium level;

[0052] High quality grade: Q gz ≤Q<1; pork samples of this grade show excellent color characteristics, with uniform hue and brightness close to ideal, and high saturation and freshness;

[0053] Among them, Q is the quality evaluation coefficient, Q dz is the lower threshold corresponding to the medium quality level and the upper threshold corresponding to the low quality level, Q gz is the lower threshold corresponding to the high quality level and the upper threshold corresponding to the medium quality level, Q dz =0.4, Q gz =0.7.

[0054] Preferably, the process of determining the quality grade of the pork to be inspected specifically includes:

[0055] ​​The pork quality evaluation model is deployed into the quality inspection system. The pork to be inspected is placed in the field of view of the imager. The parameters of the imager are adjusted to ensure that the quality of the collected image meets the requirements. The hyperspectral image data of the pork to be inspected is collected in real time and transmitted to the quality inspection system.

[0056] Preprocess the collected hyperspectral images, extract features related to pork quality from the preprocessed images, and input the extracted features into the deployed pork quality evaluation model for quality evaluation;

[0057] According to the output of the pork quality evaluation model and the calculated value of the quality evaluation coefficient, combined with the preset quality grade classification standard, the quality grade of the pork to be inspected is determined;

[0058] Record the pork quality assessment results and mark the pork samples according to the assessment results. At the same time, as needed, issue warnings for pork with lower quality grades so that they can be culled.

[0059] The present invention provides a non-destructive evaluation method for pork quality grade based on hyperspectral imaging. It has the following beneficial effects:

[0060] 1. This non-destructive evaluation method for pork quality grade based on hyperspectral imaging uses hyperspectral imaging technology to capture the reflection or transmission information of pork samples in multiple continuous spectral bands, and through analyzing spectral data, it can achieve accurate evaluation of pork quality grade, provide higher-dimensional data, significantly improve the accuracy and reliability of detection, and at the same time, can achieve rapid scanning, greatly shorten the detection time, improve the overall detection efficiency, and make pork quality monitoring more efficient and timely.

[0061] Second, this non-destructive evaluation method for pork quality grade based on hyperspectral imaging establishes an accurate pork quality grade classification system through data analysis of a large number of samples, providing a scientific basis for pork grading and quality control, effectively promoting the standardization and normalization of pork quality, and by training machine learning models, it can achieve real-time prediction and classification of pork quality grades, thereby automatically eliminating unqualified products on the production line and improving production efficiency and product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a flow chart of a non-destructive evaluation method of pork quality grade based on hyperspectral imaging according to the present invention;

[0063] Figure 2 The flowchart of color difference analysis and quality grade classification of the present invention is as follows;

[0064] Figure 3 The figure is a flow chart for constructing the pork quality evaluation model of the present invention. DETAILED DESCRIPTION

[0065] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are provided for the purpose of illustration and description, and are not intended to be exhaustive or to limit the present invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention, and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

[0066] The first embodiment, as Figure 1 , Figure 2 As shown, the present invention provides a technical solution: a non-destructive evaluation method for pork quality grade based on hyperspectral imaging, comprising the following steps:

[0067] Prepare pork samples of different qualities and standardize them. Cut the pork samples into uniform sizes and shapes to eliminate the influence of shape and size on the test results. At the same time, ensure that the sample surface is clean and reduce the interference of impurities on hyperspectral imaging. Select pork samples of different qualities according to the test requirements to ensure that the samples can cover different qualities, including fresh pork samples and pork samples under different deterioration conditions. Pork samples include tenderloin, leg meat, ham, etc. Place the pork samples on the workbench, use a ruler and a marker to mark the area to be cut on the pork, and cut the pork samples along the marks to ensure that the thickness and weight of each sample are consistent to facilitate the penetration and reflection of light during hyperspectral imaging. Mark the cut pork samples, record their quality, processing date, and storage condition information, and place pork samples of different qualities in corresponding sealed containers for storage to avoid cross infection. Inspect the surface of the pork samples and clean the blood, hair, and moisture impurities attached to the surface of the pork.

[0068] Use a colorimeter to perform color difference analysis on pork samples of different qualities, classify pork samples of different qualities into quality grades based on the analysis results of pork muscle color, fat color and gloss, calibrate the colorimeter, use the colorimeter to perform color difference analysis on the muscle and fat parts of the pork samples, measure the L value (brightness), a value (red-green value) and b value (yellow-blue value) of the pork samples, create a data record table for each sample, record its L, a, b values ​​and corresponding quality information, perform statistical analysis to evaluate the color characteristics of the pork samples, and based on the measured L value (brightness), a value (red-green value) and b value (yellow-blue value), according to the marbling score standard and the intramuscular fat content. Determination method: score and determine the pork samples, combine the marbling score and intramuscular fat content with the color difference data, evaluate the muscle color, fat color and gloss of the pork, compare the color difference data of pork samples of different qualities, analyze the differences in L, a, and b values ​​of pork samples of different qualities, further analyze the differences in muscle color, fat color and gloss in combination with the marbling score and intramuscular fat content, formulate the quality grade classification standard for pork samples based on the comprehensive color difference analysis results, muscle color, fat color, gloss analysis results and the degree of deviation of color difference data, and classify the pork samples of different qualities into low quality grade, medium quality grade and high quality grade according to the formulated classification standard;

[0069] Configure the hyperspectral imaging system for system calibration, preheat the hyperspectral imager, and adjust the environmental parameters of the detection environment. Assemble the hyperspectral imager, camera, lens and light source to form a hyperspectral imaging system to ensure stable connection and avoid looseness or poor contact. The model of the hyperspectral imager is V10E of the Imspector series spectrometer. This model of spectrometer collects spectral images with a wavelength range of 400nm-1000nm with a spectral resolution of 2.8nm and a light flux of less than 9nm in radius. Connect the control software and data processing software of the hyperspectral imaging system, configure the spectral range, resolution, and acquisition speed parameters of the hyperspectral imaging system, use a standard light source, calibrate the spectral response of the hyperspectral imaging system, measure the spectral response of each pixel and the response changes at different wavelengths, eliminate the spectral response differences between different pixels, use calibration objects to measure the focal length of the camera, lens distortion, and the relationship parameters between pixels, calibrate the geometric parameters of the hyperspectral imaging system to ensure success. The image results are true and reliable. Use color cards for color calibration, measure the difference between the color of the color card in the image and its real color, and adjust the color response function of the hyperspectral imaging system according to the measurement results, so that the color in the imaging process can be accurately restored and corrected. Measure the output electrical signals of the camera of the hyperspectral imaging system under different exposure conditions, as well as the output electrical signals in the absence of light. According to the measurement results, reduce the influence of noise and eliminate the influence of dark current by taking multiple shots of the same scene and taking the average value. Perform a preheating procedure on the hyperspectral imager. The preheating time is between 30 minutes and 1 hour to ensure the stability of the equipment and the imaging quality. During the preheating process, pay attention to monitor the status of the imager to ensure that the preheating is completed and the system is stable. According to the detection requirements, adjust the brightness, color temperature and angle of the light source to ensure that the lighting conditions are stable and uniform to avoid interference with the imaging results. Use a thermometer and hygrometer to monitor the temperature and humidity of the detection environment, and adjust the temperature and humidity according to the detection requirements of the hyperspectral imaging system.

[0070] Perform hyperspectral imaging on pork samples, collect hyperspectral image information of pork samples at different wavelengths, and correct the collected hyperspectral images. Use image segmentation technology to segment the spectral image of pork samples from the background to obtain complete pork image information. Set imaging parameters according to the characteristics of pork samples and detection requirements, where the imaging parameters include spectral range, resolution, acquisition speed and exposure time. Place pork samples in the detection area, start the hyperspectral imaging system, image the pork samples, collect hyperspectral image information of pork samples at different wavelengths, and form a hyperspectral data cube. Import the collected hyperspectral image information into the data processing software and perform correction processing on it, including spectral response correction, geometric correction and color correction. Use a standard light source to correct the spectral response of the hyperspectral imaging system to ensure that the response of each pixel at different wavelengths is consistent. The spectral response of the system is adjusted by measuring the spectral characteristics of the standard light source. The hyperspectral image is geometrically corrected by using the known feature points in the image. The geometric precision correction of the image is achieved by establishing the corresponding relationship between the original image space and the coordinate space used for mapping, ensuring that the shape and position of the objects in the image are consistent with the actual ones. The color card is used for color correction to ensure that the hyperspectral imaging system can accurately restore and correct the color in the image. The color response function of the system is adjusted by measuring the difference between the color of the color card in the image and its true color. The corrected hyperspectral image is preprocessed, including denoising and contrast enhancement, to improve the accuracy of image segmentation. According to the spectral characteristics and spatial distribution characteristics of the pork sample and the background, the image segmentation method of edge detection segmentation is selected to segment the hyperspectral image, and the spectral image of the pork sample is separated from the background. The segmented image is post-processed, including removing small area noise and filling holes, to obtain complete pork image information.

[0071] The color feature vector is extracted from the acquired pork image information through image analysis, and the correlation between pork quality and image color features is analyzed using a pork quality evaluation model.

[0072] Deploy a pork quality evaluation model, collect hyperspectral images of the pork to be inspected in real time, and determine the quality grade of the current pork to be inspected by combining the quality evaluation results and the classified pork quality grades.

[0073] The second embodiment is based on the first embodiment. Figure 3 As shown in the figure, the construction process of the pork quality evaluation model specifically includes:

[0074] Integrate the collected pork image information to form a data set, and attach the corresponding quality evaluation data of each pork sample, analyze the saturation, freshness, tenderness, flavor and water retention of each pork sample, and clarify its quality grade. Convert the pork image from RGB color space to Lab color space through image analysis technology, and extract relevant color feature data in the converted Lab color space, including statistical features of brightness and hue and color moment features, where brightness is the L component and hue is the a and b components. Use a neural network algorithm combined with the extracted color feature data to build a pork quality evaluation model, use the extracted color feature data as the input of the model, and the quality grade of pork as the output of the model. Use the image information in the data set to verify the model, optimize the model to improve its prediction performance, combine the output of the pork quality evaluation model and the extracted color feature data, preset the benchmark values ​​of color features, saturation and freshness of pork samples, comprehensively analyze and calculate the quality evaluation coefficient, quantify the quality level of pork, and then analyze the influence of color features on pork quality, evaluate the correlation between pork quality and image color features, and determine the quality evaluation threshold corresponding to each quality grade according to the output result of the quality evaluation coefficient and the divided quality grades;

[0075] Furthermore, the calculation process of the quality evaluation coefficient specifically includes:

[0076] Traverse the color characteristics, saturation and freshness data of each pork sample, and determine the base values ​​of color characteristics, saturation and freshness respectively. For color characteristics, the base value of brightness is determined according to the meat standard. L , based on the standard color of fresh pork, the base values ​​of the a and b components of the hue are determined as Base a and Base b , according to the maximum allowable range of color change of fresh pork, determine the range of the reference value of the hue a and b components ab , based on the storage conditions and shelf life of pork, the baseline values ​​of freshness and saturation are determined as Base S and Base P , calculate the maximum value max(L) and minimum value min(L) of brightness from the extracted color features, and calculate the ratio of the brightness range to the reference value Calculate the brightness change function T L , analyze the uniformity of the brightness range, and calculate the absolute deviation of the mean of the hue a and b components from the reference value |μ a -Base a | and |μ b -Base b |, and add the two deviations and divide them by the hue range reference value Range ab , calculate the hue mean influence function T ab, analyze the closeness of the hue mean to the reference value, using an exponential decay function Calculate the effect of saturation and use the limit function Calculate the impact of freshness, the ratio of comprehensive saturation and freshness, and calculate the quality impact function T SP , measure the freshness and color saturation of pork, integrate the brightness change function, hue mean influence function and quality influence function, calculate the quality evaluation coefficient Q, and quantitatively analyze the quality level of pork;

[0077] Furthermore, the calculation expression of the quality evaluation coefficient is:

[0078]

[0079] Among them, Q is the quality evaluation coefficient, T L is the brightness change function, T ab is the hue mean influence function, T SP is the quality influence function, and the value range of Q is between 0 and 1. When the brightness range is uniform, the hue mean is close to the reference value, and the saturation and freshness indexes are high, the value of Q is close to 1, indicating that the pork quality is high. When the brightness range is uneven, the hue mean is far from the reference value, and the saturation and freshness indexes are low, the value of Q is close to 0, indicating that the pork quality is low.

[0080] Furthermore, the values ​​of the quality evaluation coefficients and the preset evaluation thresholds of the quality levels are obtained to match the three quality levels, which are:

[0081] Low quality level: 0≤Q dz Pork samples of this grade usually show poor color characteristics, with obvious color differences and unevenness, and low saturation and freshness;

[0082] Medium quality grade: Q dz ≤Q gz ; The pork samples of this grade are relatively good in color and texture, with hue and brightness close to the benchmark values, and saturation and freshness at a medium level;

[0083] High quality grade: Q gz ≤Q<1; pork samples of this grade show excellent color characteristics, with uniform hue and brightness close to ideal, and high saturation and freshness;

[0084] Among them, Q is the quality evaluation coefficient, Q dz is the lower threshold corresponding to the medium quality level and the upper threshold corresponding to the low quality level, Q gz is the lower threshold corresponding to the high quality level and the upper threshold corresponding to the medium quality level, Q dz =0.4, Q gz =0.7;​​

[0085] The process of determining the quality grade of pork to be inspected includes:

[0086] The pork quality evaluation model is deployed to the quality inspection system. The pork to be inspected is placed in the field of view of the imager. The parameters of the imager are adjusted to ensure that the quality of the collected image meets the requirements. The hyperspectral image data of the pork to be inspected is collected in real time and transmitted to the quality inspection system. The collected hyperspectral image is preprocessed, and features related to pork quality are extracted from the preprocessed image. The extracted features are input into the deployed pork quality evaluation model for quality evaluation. According to the output of the pork quality evaluation model and the calculated value of the quality evaluation coefficient, combined with the preset quality grade classification standard, the quality grade of the current pork to be inspected is determined, the pork quality assessment results are recorded, and the pork samples are marked according to the assessment results. At the same time, according to needs, pork with lower quality grades is warned so that it can be culled.

[0087] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field and related fields without creative work should fall within the scope of protection of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention are implemented according to the conventional means in the field unless otherwise specified and limited.

Claims

1. A non-destructive evaluation method for pork quality grade based on hyperspectral imaging, characterized in that: The following steps are involved: Prepare pork samples of different qualities and standardize the pork samples; Use a colorimeter to analyze the color difference of pork samples of different qualities, and classify the pork samples of different qualities into different quality grades based on the analysis results of pork muscle color, fat color and gloss; Configure the hyperspectral imaging system to calibrate the system and preheat the hyperspectral imager; Perform hyperspectral imaging on pork samples, collect hyperspectral image information of pork samples at different wavelengths, correct the collected hyperspectral images, and use image segmentation technology to segment the spectral images of pork samples from the background to obtain complete pork image information; The color feature vector is extracted from the acquired pork image information through image analysis, and the correlation between pork quality and image color features is analyzed using a pork quality evaluation model. Deploy a pork quality evaluation model, collect hyperspectral images of the pork to be inspected in real time, and determine the quality grade of the current pork to be inspected by combining the quality evaluation results and the classified pork quality grades.

2. The method for nondestructive evaluation of pork quality grade based on hyperspectral imaging according to claim 1, characterized in that: The standardized processing of the pork samples specifically includes: According to the testing requirements, pork samples of different qualities were selected, including fresh pork samples and pork samples under different deterioration conditions; Place the pork sample on a workbench, use a ruler and a marker to mark the area to be cut on the pork, and cut the pork sample along the mark; Mark the cut pork samples, record their quality, processing date, storage conditions, and place pork samples of different qualities in corresponding sealed containers for storage; Inspect the surface of the pork sample and clean any blood, hair, and moisture impurities adhering to the surface of the pork.

3. The method for nondestructive evaluation of pork quality grade based on hyperspectral imaging according to claim 2 is characterized in that: The color difference analysis and quality grade classification process specifically includes: Calibrate the colorimeter, use the colorimeter to perform colorimetric analysis on the muscle and fat parts of the pork samples, determine the L value, a value, and b value of the pork samples, and create a data record table for each sample to record its L, a, b values ​​and corresponding quality information, and perform statistical analysis to evaluate the color characteristics of the pork samples; Based on the measured L value, a value and b value, the pork samples were scored and measured according to the marbling score standard and the determination method of intramuscular fat content, and the marbling score and intramuscular fat content were combined with the color difference data to evaluate the muscle color, fat color and gloss of pork. The color difference data of pork samples of different qualities were compared, and the differences in L, a and b values ​​of pork samples of different qualities were analyzed. Combined with the marbling score and intramuscular fat content, the differences in muscle color, fat color and gloss were further analyzed. The quality grade classification standard of pork samples is formulated based on the color difference analysis results, muscle color, fat color, gloss analysis results and the degree of deviation of color difference data; According to the established classification standards, pork samples of different qualities are divided into quality grades, namely low quality grade, medium quality grade and high quality grade.

4. The method for nondestructive evaluation of pork quality grade based on hyperspectral imaging according to claim 3 is characterized in that: The calibration process of the hyperspectral imaging system specifically includes: Assemble the hyperspectral imager, camera, lens and light source to form a hyperspectral imaging system, connect the control software and data processing software of the hyperspectral imaging system, and configure the spectral range, resolution and acquisition speed parameters of the hyperspectral imaging system; Use standard light sources to calibrate the spectral response of the hyperspectral imaging system, measure the spectral response of each pixel and the response changes at different wavelengths, eliminate the spectral response differences between different pixels, use calibration objects to measure the focal length of the camera, lens distortion, and the relationship parameters between pixels, and calibrate the geometric parameters of the hyperspectral imaging system; Use a color card to perform color calibration, measure the difference between the color of the color card in the image and its actual color, and adjust the color response function of the hyperspectral imaging system based on the measurement results; Measure the output electrical signals of the camera of the hyperspectral imaging system under different exposure conditions and the output electrical signals in the absence of light. Based on the measurement results, reduce the influence of noise and eliminate the influence of dark current by taking multiple shots of the same scene and taking the average value. Perform a warm-up procedure on the hyperspectral imager, with the warm-up time ranging from 30 minutes to 1 hour; According to the detection requirements, adjust the brightness, color temperature and angle of the light source, use a thermometer and hygrometer to monitor the temperature and humidity of the detection environment, and adjust the temperature and humidity according to the detection requirements of the hyperspectral imaging system.

5. The method for nondestructive evaluation of pork quality grade based on hyperspectral imaging according to claim 4, characterized in that: The process of obtaining the pork image information specifically includes: According to the characteristics of the pork sample and the detection requirements, the imaging parameters are set, wherein the imaging parameters include spectral range, resolution, acquisition speed and exposure time; Place the pork sample in the detection area, start the hyperspectral imaging system, image the pork sample, collect the hyperspectral image information of the pork sample at different wavelengths, and form a hyperspectral data cube; Import the collected hyperspectral image information into the data processing software and perform correction processing on it, including spectral response correction, geometric correction and color correction; The corrected hyperspectral image is preprocessed, including denoising and contrast enhancement. According to the spectral characteristics and spatial distribution characteristics of the pork sample and the background, the image segmentation method of edge detection segmentation is selected to segment the hyperspectral image, and the spectral image of the pork sample is separated from the background. The segmented images are post-processed, including removing small area noise and filling holes, to obtain complete pork image information.

6. The method for nondestructive evaluation of pork quality grade based on hyperspectral imaging according to claim 5, characterized in that: The construction process of the pork quality evaluation model specifically includes: Integrate the collected pork image information to form a data set, and attach the corresponding quality evaluation data of each pork sample, analyze the saturation, freshness, tenderness, flavor and water retention of each pork sample, and clarify its quality grade; The pork image is converted from RGB color space to Lab color space by image analysis technology, and relevant color feature data is extracted in the converted Lab color space, including statistical features of brightness and hue and color moment features, where brightness is the L component and hue is the a and b components; A pork quality evaluation model was constructed using a neural network algorithm combined with the extracted color feature data. The extracted color feature data was used as the input of the model, and the quality grade of pork was used as the output of the model. The model was verified using the image information in the data set. Combining the output of the pork quality evaluation model and the extracted color feature data, the benchmark values ​​of the color features, saturation and freshness of the pork samples were preset, and the quality evaluation coefficient was calculated through comprehensive analysis to quantify the quality level of the pork. The impact of color features on pork quality was then analyzed, and the correlation between pork quality and image color features was evaluated. According to the output results of the quality evaluation coefficient and the divided quality levels, the quality evaluation thresholds corresponding to each quality level are determined.

7. The method for nondestructive evaluation of pork quality grade based on hyperspectral imaging according to claim 6, characterized in that: The process of presetting the reference values ​​of the pork sample color characteristics, saturation and freshness specifically includes: Traversing the color characteristics, saturation and freshness data of each pork sample, and determining the benchmark values ​​of the color characteristics, saturation and freshness respectively; For color characteristics, the base value of brightness is determined according to the meat standard. L ; Based on the standard color of fresh pork, the base values ​​of a and b components of hue are determined as Base a and Base b ; Determine the range of the reference values ​​of the hue a and b components according to the maximum allowable range of color change of fresh pork ab ; Based on the storage conditions and shelf life of pork, the baseline values ​​of freshness and saturation are determined as Base S and Base P .

8. The method for nondestructive evaluation of pork quality grade based on hyperspectral imaging according to claim 7, characterized in that: The calculation process of the quality evaluation coefficient specifically includes: Calculate the maximum value max(L) and minimum value min(L) of brightness from the extracted color features, and calculate the ratio of the brightness range to the reference value Calculate the brightness change function T L , analyze the uniformity of the brightness range; Calculate the absolute deviation of the mean of hue a and b components from the reference value |μ a -Base a | and |μ b -Base b |, and add the two deviations and divide them by the hue range reference value Range ab , calculate the hue mean influence function T ab , analyze the closeness between the hue mean and the reference value; Use an exponential decay function Calculate the effect of saturation and use the limit function Calculate the impact of freshness, the ratio of comprehensive saturation and freshness, and calculate the quality impact function T SP , a measure of pork’s freshness and color saturation; The brightness change function, hue mean influence function and quality influence function were integrated to calculate the quality evaluation coefficient Q and quantitatively analyze the quality level of pork.

9. The method for nondestructive evaluation of pork quality grade based on hyperspectral imaging according to claim 8, characterized in that: The value of the quality evaluation coefficient and the preset evaluation threshold of the quality level are obtained to match the three quality levels, which are: Low quality level: 0≤Q dz ;​ Medium quality grade: Q dz ≤Q gz ;​ High quality grade: Q gz ≤Q<1; Among them, Q is the quality evaluation coefficient, Q dz is the lower threshold corresponding to the medium quality level and the upper threshold corresponding to the low quality level, Q gz is the lower threshold corresponding to the high quality level and the upper threshold corresponding to the medium quality level, Q dz =0.4, Q gz =0.

7.

10. The method for nondestructive evaluation of pork quality grade based on hyperspectral imaging according to claim 9, characterized in that: The process of determining the quality grade of the pork to be inspected specifically includes: The pork quality evaluation model is deployed into the quality inspection system. The pork to be inspected is placed in the field of view of the imager, and the hyperspectral image data of the pork to be inspected is collected in real time and transmitted to the quality inspection system. Preprocess the collected hyperspectral images, extract features related to pork quality from the preprocessed images, and input the extracted features into the deployed pork quality evaluation model for quality evaluation; According to the output of the pork quality evaluation model and the calculated value of the quality evaluation coefficient, combined with the preset quality grade classification standard, the quality grade of the current pork to be inspected is determined; Record the pork quality assessment results, label the pork samples according to the assessment results, and issue warnings for pork of lower quality grades based on demand.

Images