A nondestructive testing method for multiple quality indexes of rice based on near infrared spectroscopy

Abstract

The application discloses a kind of based on near infrared spectroscopy rice multi-quality index nondestructive testing method, belong to grain quality detection technical field, including spectrum and environmental parameter acquisition, spectrum pretreatment enhancement, environmental adaptive multi-index prediction and quality index output.The application improves the consistency and comparability of spectral data through the spectrum pretreatment enhancement method, provides stable and reliable data basis for subsequent quality index prediction;Creatively adopt environmental adaptive multi-index prediction method, establish the dynamic correlation between spectral characteristics and quality index under the condition of environmental change, realize the adaptive prediction of rice multi-quality index under different temperature and humidity conditions, avoid the problem of repeated modeling or frequent calibration caused by environmental change in traditional method, thereby improve the generalization ability and prediction robustness of model, provide implementable technical means for rapid, nondestructive, reliable rice quality detection in field, grain depot and laboratory environment.

Description

Technical Field

[0001] This invention belongs to the field of grain quality testing technology, specifically referring to a non-destructive testing method for multiple quality indicators of rice based on near-infrared spectroscopy. Background Technology

[0002] Rice is one of the world's major food crops, and its quality indicators such as moisture, protein, and fat directly affect its processing performance, storage stability, and edible value. Traditional rice quality testing methods mostly rely on chemical analysis or single near-infrared spectroscopy measurements. These methods suffer from drawbacks such as sample destruction, cumbersome operation, and long testing cycles, making them unsuitable for large-scale, on-site rapid testing. Furthermore, existing spectroscopic detection methods face several technical challenges in practical applications: Firstly, due to differences in rice grain size, surface scattering effects, changes in packing state, and fluctuations in instrument measurement conditions, the original spectral signals often exhibit high-frequency noise, scattering shift, baseline drift, and amplitude inconsistencies, leading to unstable prediction accuracy for multiple indicators. Secondly, changes in environmental temperature and humidity alter the near-infrared absorption response characteristics of moisture and related chemical components in rice samples, causing differences in spectral performance obtained under the same quality conditions. This results in a shift in the mapping relationship between spectral characteristics and quality indicators, reducing the model's prediction accuracy under different environmental conditions. Frequent calibration or remodeling is required, increasing testing costs and operational complexity.

[0003] Therefore, there is an urgent need for a technical solution that can achieve rapid, non-destructive, and stable detection of multiple indicators in rice under different environmental conditions, in order to ensure prediction accuracy and consistency, while meeting the application needs of multiple scenarios such as fields, grain depots, and laboratories. Summary of the Invention

[0004] To address the above issues and overcome the shortcomings of existing technologies, this invention provides a non-destructive testing method for multiple quality indicators of rice based on near-infrared spectroscopy. This method includes the following steps:

[0005] Step S1: Acquisition of spectral and environmental parameters;

[0006] Step S2: Spectral preprocessing enhancement;

[0007] Step S3: Environmental Adaptive Multi-Indicator Prediction;

[0008] Step S4: Output of quality indicators.

[0009] Furthermore, in step S1, the acquisition of spectral and environmental parameters specifically involves sampling the rice sample to be tested using a near-infrared spectral acquisition device to obtain raw spectral data, and simultaneously acquiring the temperature and relative humidity parameters of the sampling environment to form environmental parameter data.

[0010] Further, in step S2, the enhanced spectral preprocessing is used to eliminate spectral noise, scattering shift, baseline drift, and amplitude differences to generate a standard spectral matrix for rice, including the following steps:

[0011] Step S21, spectral smoothing, specifically involves acquiring the original spectral data, smoothing each spectrum using a filter, reducing high-frequency random noise by applying a sliding window and polynomial fitting to the spectrum, while retaining the main absorption peak features, thus obtaining a smoothed spectral matrix.

[0012] Step S22 Multivariate scattering correction specifically involves calculating the average spectrum of all rice samples, performing linear regression on each smoothed spectrum to solve for the offset coefficient and amplitude scaling coefficient, and then performing scattering correction processing to obtain the scattering correction spectral matrix.

[0013] Step S23, baseline drift correction, is used to eliminate baseline changes in the spectrum caused by instrument drift or background light. Specifically, it involves fitting a low-frequency trend polynomial and subtracting it from each scattering correction spectrum to perform baseline correction and obtain a baseline-corrected spectral matrix.

[0014] Step S24 amplitude standardization is used to unify the spectral scale. Specifically, it involves normalizing each baseline correction spectrum to obtain the standard spectral matrix of rice.

[0015] Further, in step S3, the environmental adaptive multi-index prediction is used to achieve high-precision, non-destructive prediction of various quality indicators under different temperature and humidity conditions. Specifically, based on the rice standard spectral matrix and environmental parameter data, joint modeling and prediction training are performed through feature mapping, spectral-environment joint feature fusion, and multi-index regression prediction to obtain the environmental adaptive multi-index prediction result, including the following steps:

[0016] Step S31, spectral and environmental parameter feature mapping, specifically involves mapping the standard spectral matrix of rice by constructing a spectral branch, extracting the main spectral information, generating a spectral feature mapping matrix, and mapping the environmental parameter data by constructing an environmental branch to obtain an environmental feature mapping matrix.

[0017] Step S32, spectral-environment joint feature fusion, is used to fuse the output results of the spectral branch and the environmental branch. Specifically, an intermediate fusion layer is constructed, an attention mechanism is introduced to dynamically calculate the attention weights of the spectral feature mapping matrix under different temperature and humidity conditions, and the spectral feature mapping matrix and the environmental feature mapping matrix are fused according to the attention weights to obtain the spectral-environment joint feature matrix.

[0018] Step S33: Construction of a multi-index prediction model to achieve prediction of multiple quality indicators of rice. Specifically, a multi-index prediction model consisting of a spectral branch, an environmental branch, an intermediate fusion layer, and a multi-index regression layer is constructed. The joint feature matrix is ​​input into the multi-index regression layer to obtain the regression prediction results.

[0019] Step S34 predicts the output processing, specifically by obtaining the environmental compensation weight matrix through model training based on the difference between the current environmental parameters and the average environmental parameters of the training set, and then weighting and correcting the regression prediction results to generate environmental adaptive multi-index prediction results, ensuring stable prediction accuracy under different temperature and humidity conditions.

[0020] Furthermore, in step S4, the quality index output specifically involves summarizing, statistically analyzing, and storing multiple quality indicators for each rice sample based on the environmental adaptive multi-indicator prediction results, and displaying the distribution of quality indicators and batch trend changes in the form of charts, providing visualized data for production management and quality control.

[0021] The beneficial effects achieved by the present invention using the above scheme include: by using spectral preprocessing enhancement methods, high-frequency noise, scattering shift, baseline drift, and amplitude differences in the original spectrum are eliminated and corrected, improving the consistency and comparability of spectral data and providing a stable and reliable data foundation for subsequent quality index prediction; creatively employing an environmentally adaptive multi-index prediction method, a dynamic correlation between spectral characteristics and quality indicators under environmental change conditions is established, realizing adaptive prediction of multiple quality indicators of rice under different temperature and humidity conditions, effectively reducing the impact of environmental changes on the detection results, avoiding the problem of frequent repeated modeling or calibration caused by environmental changes in traditional methods, thereby improving the generalization ability and prediction robustness of the model, and providing an implementable technical means for rapid, non-destructive, and reliable rice quality detection in field, grain depot, and laboratory environments. Attached Figure Description

[0022] Figure 1 A flowchart illustrating a non-destructive testing method for multiple quality indicators of rice based on near-infrared spectroscopy provided by this invention;

[0023] Figure 2 This is a flowchart illustrating step S2;

[0024] Figure 3 This is a flowchart illustrating step S3.

[0025] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0026] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0027] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0028] Example 1, see Figure 1 This invention provides a non-destructive testing method for multiple quality indicators of rice based on near-infrared spectroscopy, the method comprising the following steps:

[0029] Step S1: Acquisition of spectral and environmental parameters;

[0030] Step S2: Spectral preprocessing enhancement;

[0031] Step S3: Environmental Adaptive Multi-Indicator Prediction;

[0032] Step S4: Output of quality indicators.

[0033] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, the spectral and environmental parameter acquisition is used to obtain the near-infrared spectral signal of the rice sample to be tested and the temperature and humidity information of the sampling environment. Specifically, the near-infrared spectral acquisition device is used to sample the rice sample to be tested to obtain the raw spectral data, and the temperature and relative humidity parameters of the sampling environment are collected simultaneously to form environmental parameter data. The near-infrared spectral acquisition device includes, but is not limited to, a desktop near-infrared spectrometer, a portable near-infrared spectrometer, a handheld near-infrared spectrometer, or an online near-infrared detection device.

[0034] In this embodiment, the rice samples to be tested do not require crushing or chemical treatment; bulk or single grains of rice can be used directly. Specifically, a near-infrared spectrometer with a wavelength range of 900–1700 nm is used to sample the rice samples. When applied in a laboratory setting, the rice samples can be placed in a sampling dish or sampling chamber for measurement; when applied in a grain depot setting, rice samples can be extracted from the grain pile and placed in the detection area for measurement; when applied in a field setting, a portable or handheld near-infrared spectrometer can be used to directly measure the collected rice samples. During the spectral acquisition process, ambient temperature and relative humidity parameters are recorded simultaneously, and an environmental parameter record table corresponding to the spectral data is established, thereby forming the original spectral data and environmental parameter data.

[0035] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S2, the enhanced spectral preprocessing is used to eliminate spectral noise, scattering shift, baseline drift, and amplitude differences to generate a standard spectral matrix for rice. This includes the following steps:

[0036] Step S21, spectral smoothing, is used to reduce random noise, improve the spectral signal-to-noise ratio, and smooth high-frequency fluctuations to retain the main absorption peak information. Specifically, it involves acquiring the original spectral data, using a filter to smooth each spectrum, and applying a sliding window and polynomial fitting to the spectrum to reduce high-frequency random noise while retaining the main absorption peak features, thus obtaining a smoothed spectral matrix.

[0037] The filter can specifically be a Savitzky-Golay filter, a moving average filter, or a wavelet denoising filter, and the sliding window size can be automatically adjusted according to the local signal-to-noise ratio of the spectrum.

[0038] Step S22, multivariate scattering correction, is used to eliminate spectral shifts caused by differences in rice sample particle size, surface scattering, and filling state. Specifically, it involves calculating the average spectrum of all rice samples, performing linear regression on each smoothed spectrum to solve for the shift coefficient and amplitude scaling coefficient, and then performing scattering correction processing to obtain the scattering correction spectral matrix.

[0039] The formula for calculating the average spectrum is:

[0040] ;

[0041] In the formula, This is the average spectrum, N is the number of rice samples, and k is the rice sample index. This is the smoothed spectrum of the k-th rice sample.

[0042] The calculation formula for the scattering correction process is as follows:

[0043] ;

[0044] In the formula, x msc is the scattering correction spectrum, b is the amplitude scaling factor, and a is the offset factor;

[0045] Step S23, baseline drift correction, is used to eliminate baseline changes in the spectrum caused by instrument drift or background light. Specifically, it involves fitting a low-frequency trend polynomial and subtracting it from each scattering correction spectrum to perform baseline correction and obtain a baseline-corrected spectral matrix.

[0046] The fitted low-frequency trend polynomial is specifically a low-frequency trend polynomial fitted for each scattering correction spectrum, and the calculation formula is as follows:

[0047] ;

[0048] In the formula, It is a low-frequency trend polynomial. It is the spectral wavelength. is the order of the fitted polynomial, preferably order 3, where p is the order index. These are the fitting coefficients, which can be obtained through least squares fitting.

[0049] The formula for calculating the baseline correction is as follows:

[0050] ;

[0051] In the formula, x bc It is a baseline-corrected spectrum;

[0052] Step S24 amplitude standardization is used to unify the spectral scale and eliminate spectral amplitude deviations caused by rice sample loading or measurement conditions. Specifically, it involves normalizing each baseline correction spectrum to obtain a standard spectral matrix of rice.

[0053] By performing the above operations, this scheme uses spectral preprocessing enhancement methods to eliminate and correct high-frequency noise, scattering shift, baseline drift, and amplitude differences in the original spectrum, thereby improving the consistency and comparability of spectral data. This provides a stable and reliable data foundation for subsequent quality index prediction and solves the problem that the original spectral signal often exhibits high-frequency noise, scattering shift, baseline drift, and amplitude inconsistency due to differences in rice grain size, surface scattering effects, changes in filling status, and fluctuations in instrument measurement conditions, leading to unstable prediction accuracy for multiple indicators.

[0054] Example 4, see Figure 1 and Figure 3This embodiment is based on the above embodiment. In step S3, the environmental adaptive multi-index prediction is used to achieve high-precision non-destructive prediction of various quality indicators under different temperature and humidity conditions. Specifically, based on the rice standard spectral matrix and environmental parameter data, joint modeling and prediction training are performed through feature mapping, spectral-environment joint feature fusion, and multi-index regression prediction to obtain the environmental adaptive multi-index prediction result, including the following steps:

[0055] Step S31, spectral and environmental parameter feature mapping, specifically involves mapping the standard spectral matrix of rice by constructing a spectral branch, extracting the main spectral information, generating a spectral feature mapping matrix, and mapping the environmental parameter data by constructing an environmental branch to obtain an environmental feature mapping matrix, so that the spectral features and environmental features have a unified feature expression form.

[0056] The spectral branch can specifically employ a one-dimensional convolutional neural network, principal component analysis, or an autoencoder to achieve spectral information compression and feature extraction.

[0057] The environmental branch can specifically employ linear mapping, single-layer or multi-layer fully connected networks to map environmental parameters to the feature space;

[0058] Step S32, spectral-environment joint feature fusion, is used to fuse the output results of the spectral branch and the environmental branch. Specifically, an intermediate fusion layer is constructed, an attention mechanism is introduced to dynamically calculate the attention weights of the spectral feature mapping matrix under different temperature and humidity conditions, and the spectral feature mapping matrix and the environmental feature mapping matrix are fused according to the attention weights to obtain the spectral-environment joint feature matrix.

[0059] The attention mechanism can specifically employ self-attention or dot-multiplication attention to obtain attention weights.

[0060] The formula for calculating the joint characteristic matrix of the spectral environment is:

[0061] ;

[0062] In the formula, F se It is the joint characteristic matrix of the spectral environment, F s It is a spectral feature mapping matrix. It is the attention weight, F e It is an environmental feature mapping matrix;

[0063] Step S33: Construction of a multi-index prediction model to achieve prediction of multiple quality indicators of rice. Specifically, a multi-index prediction model consisting of a spectral branch, an environmental branch, an intermediate fusion layer, and a multi-index regression layer is constructed. The joint feature matrix is ​​input into the multi-index regression layer to obtain the regression prediction results.

[0064] The multi-index regression layer specifically employs a multi-layer fully connected network combined with a nonlinear activation function to output predicted values ​​for multiple quality indices, and trains the model parameters using a mean squared error loss function.

[0065] The multiple quality indicators include, but are not limited to, moisture content, protein content, fat content, and starch content;

[0066] Step S34 predicts the output processing, specifically by obtaining the environmental compensation weight matrix through model training based on the difference between the current environmental parameters and the average environmental parameters of the training set, weighting and correcting the regression prediction results, and generating environmental adaptive multi-index prediction results to ensure stable prediction accuracy under different temperature and humidity conditions.

[0067] The training set is specifically a labeled dataset composed of raw spectral data, environmental parameter data, and multiple quality indicators obtained through laboratory chemical analysis, collected in step S1 under different environmental conditions.

[0068] The formula for weighted correction of the regression prediction results is as follows:

[0069] ;

[0070] In the formula, It is an environmental adaptive multi-indicator prediction result. It is the regression prediction result, W c It is the environmental compensation weight matrix, E new These are the current environmental parameters. It is the average environmental parameter of the training set.

[0071] By performing the above operations, this solution creatively adopts an environmentally adaptive multi-index prediction method to establish a dynamic correlation between spectral characteristics and quality indicators under environmental change conditions. This enables adaptive prediction of multiple quality indicators of rice under different temperature and humidity conditions, effectively reducing the impact of environmental changes on the detection results. It avoids the problem of frequent repetitive modeling or calibration caused by environmental changes in traditional methods, thereby improving the model's generalization ability and prediction robustness. It provides an feasible technical means for rapid, non-destructive, and reliable rice quality detection in field, grain depot, and laboratory environments. It solves the problem that changes in environmental temperature and humidity can alter the near-infrared absorption response characteristics of moisture and related chemical components in rice samples, causing differences in spectral performance obtained under the same quality conditions. This leads to a shift in the mapping relationship between spectral characteristics and quality indicators, thereby reducing the prediction accuracy of the model under different environmental conditions, requiring frequent calibration or remodeling, and increasing detection costs and operational complexity.

[0072] Example 5, see Figure 1This embodiment is based on the above embodiment. In step S4, the quality index output specifically involves summarizing, statistically analyzing, and storing multiple quality indicators of each rice sample based on the environmental adaptive multi-indicator prediction results, and displaying the distribution of quality indicators and batch trend changes in the form of charts to provide visualized data for production management and quality control.

[0073] The summary specifically involves arranging the predicted values ​​of multiple quality indicators for each rice sample according to the sample number to form a sample data table, and generating histograms, scatter plots, or box plots to visually reflect the distribution of each indicator among the samples. At the same time, preset target thresholds can be marked on the charts to quickly identify rice samples with abnormal indicators.

[0074] The statistics specifically involve calculating the average, standard deviation, minimum, and maximum values ​​of various indicators for multiple batches of rice samples. These values ​​are used for overall batch quality assessment and can generate line charts or bar charts to display the batch average values ​​and fluctuation ranges, reflecting batch trend changes.

[0075] Specifically, the storage involves saving the prediction results and statistical information as CSV, Excel, or JSON files, which can be uploaded to the production management system or cloud database via an interface to achieve historical traceability and data management. At the same time, it supports dynamic updates of charts to reflect newly added sampling results.

[0076] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0077] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0078] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A non-destructive testing method for multiple quality indicators of rice based on near-infrared spectroscopy, characterized in that: The method includes the following steps: Step S1: Acquisition of spectral and environmental parameters to obtain raw spectral data and environmental parameter data; Step S2, spectral preprocessing enhancement, is used to eliminate spectral noise, scattering shift, baseline drift, and amplitude differences to generate a standard spectral matrix for rice. It includes the following steps: Step S21, spectral smoothing; Step S22, multivariate scattering correction; Step S23, baseline drift correction; and Step S24, amplitude normalization. Step S3, Environmental Adaptive Multi-Indicator Prediction, is used to achieve high-precision, non-destructive prediction of various quality indicators under different temperature and humidity conditions. Specifically, based on the rice standard spectral matrix and environmental parameter data, joint modeling and prediction training are performed through feature mapping, spectral-environment joint feature fusion, and multi-indicator regression prediction to obtain environmental adaptive multi-indicator prediction results. The steps include: Step S31 Spectral and Environmental Parameter Feature Mapping, Step S32 Spectral-Environment Joint Feature Fusion, Step S33 Multi-Indicator Prediction Model Construction, and Step S34 Prediction Output Processing. Step S31, spectral and environmental parameter feature mapping, specifically involves mapping the standard spectral matrix of rice by constructing a spectral branch, extracting the main spectral information, generating a spectral feature mapping matrix, and mapping the environmental parameter data by constructing an environmental branch to obtain an environmental feature mapping matrix. Step S32, spectral-environment joint feature fusion, is used to fuse the output results of the spectral branch and the environmental branch. Specifically, an intermediate fusion layer is constructed, an attention mechanism is introduced to dynamically calculate the attention weights of the spectral feature mapping matrix under different temperature and humidity conditions, and the spectral feature mapping matrix and the environmental feature mapping matrix are fused according to the attention weights to obtain the spectral-environment joint feature matrix. Step S33: Construction of a multi-index prediction model to achieve prediction of multiple quality indicators of rice. Specifically, a multi-index prediction model consisting of a spectral branch, an environmental branch, an intermediate fusion layer, and a multi-index regression layer is constructed. The joint feature matrix is ​​input into the multi-index regression layer to obtain the regression prediction results. Step S34 predicts the output processing, specifically by obtaining the environmental compensation weight matrix through model training based on the difference between the current environmental parameters and the average environmental parameters of the training set, weighting and correcting the regression prediction results, and generating environmental adaptive multi-index prediction results to ensure stable prediction accuracy under different temperature and humidity conditions. Step S4, quality index output, specifically involves summarizing, statistically analyzing, and storing multiple quality indicators for each rice sample based on the environmental adaptive multi-indicator prediction results, and displaying the distribution of quality indicators and batch trend changes in the form of charts, providing visualized data for production management and quality control.

2. The non-destructive testing method for multiple quality indicators of rice based on near-infrared spectroscopy according to claim 1, characterized in that: Step S21 involves spectral smoothing, which specifically involves acquiring the original spectral data, smoothing each spectrum using a filter, reducing high-frequency random noise by applying a sliding window and polynomial fitting to the spectrum, while retaining the main absorption peak features, and obtaining a smoothed spectral matrix.

3. The non-destructive testing method for multiple quality indicators of rice based on near-infrared spectroscopy according to claim 2, characterized in that: Step S22, multivariate scattering correction, specifically involves calculating the average spectrum of all rice samples, performing linear regression on each smoothed spectrum to solve for the offset coefficient and amplitude scaling coefficient, and then performing scattering correction processing to obtain the scattering correction spectral matrix.

4. The non-destructive testing method for multiple quality indicators of rice based on near-infrared spectroscopy according to claim 3, characterized in that: Step S23, baseline drift correction, is used to eliminate baseline changes in the spectrum caused by instrument drift or background light. Specifically, it involves fitting a low-frequency trend polynomial and subtracting it from each scattering correction spectrum to perform baseline correction and obtain a baseline-corrected spectral matrix. Step S24 amplitude standardization is used to unify the spectral scale. Specifically, it involves normalizing each baseline correction spectrum to obtain the standard spectral matrix of rice.

5. The non-destructive testing method for multiple quality indicators of rice based on near-infrared spectroscopy according to claim 4, characterized in that: In step S1, the acquisition of spectral and environmental parameters specifically involves sampling the rice sample under test using a near-infrared spectral acquisition device to obtain raw spectral data, and simultaneously acquiring the temperature and relative humidity parameters of the sampling environment to form environmental parameter data.

Images