A Method for Detecting Wheat Grain Quality and Identifying Functional Genes Based on Hyperspectral Imaging Technology

Abstract

This invention discloses a method for wheat grain quality detection and functional gene analysis based on hyperspectral imaging technology, belonging to the field of agricultural testing technology. The method includes: extracting hyperspectral indices; measuring wheat grain quality indicators as manual measurements; assigning weight coefficients to each quality indicator to digitize the wheat quality phenotype; extracting characteristic wavelengths to establish predictive models for each indicator and wheat quality values; performing GWAS on hyperspectral indices and chemical indicators as phenotypic traits with genotypes to screen SNP loci, conduct co-location analysis, obtain candidate genes, and perform functional verification. This invention, through hyperspectral technology combined with chemometric modeling, is suitable for real-time monitoring during wheat quality breeding and processing; by combining spectral phenotype with GWAS gene mining, it overcomes the limitations of traditional quality detection that only serves phenotypic screening, forming a closed loop of "non-destructive detection - gene localization," providing targeted genes for wheat quality improvement.

Description

Technical Field

[0001] This invention belongs to the field of agricultural testing technology, specifically relating to a method for wheat grain quality detection and functional gene analysis based on hyperspectral imaging technology. Background Technology

[0002] Wheat is a major food crop in China and an important source of protein for humankind. Its protein and gluten content directly affects processing quality and nutritional value. High yield, high quality, and disease resistance have always been important breeding goals for wheat. With economic growth, consumers' demand for improved wheat quality is becoming increasingly urgent, and breeding goals have shifted from "enough to eat" to "eating well." The quality of wheat directly impacts the development of the agricultural economy and the food industry. Traditional manual methods for testing wheat quality suffer from low accuracy, time-consuming and labor-intensive processes, lack of objectivity, and destructive nature. In recent years, with the development of spectroscopic and computer technologies, non-destructive testing technologies for crop quality have emerged, improving the speed and accuracy of testing. Hyperspectral imaging technology can comprehensively acquire images and spectral information of the tested sample, enabling visualization of sample components and physicochemical properties, and has been widely used in the non-destructive testing of agricultural product quality-related indicators.

[0003] In the PLS model for moisture content in wheat flour established based on spectral preprocessing and wavelength optimization using a genetic algorithm, the R² and RMSEP of the model were 0.977 and 0.264, respectively. In the partial least squares (PLS) model for predicting the wet gluten content of wheat flour, the optimal model was obtained by subtracting a straight line from the preprocessing method, with a correlation coefficient R² of 0.934, PMSEC of 0.51, and RMSEP of 0.456.

[0004] In the prediction model of wheat flour rheological properties established based on information fusion using near-infrared and mid-infrared spectroscopy, the results showed that the prediction model was good, indicating that it is feasible to use near-infrared spectroscopy to detect the rheological properties of wheat flour. The original dataset was expanded using spectral data and oil content data generated by DCGAN, cleverly combining artificial intelligence GAN and NIR-HSI techniques for regression problems to predict the oil content of individual corn kernels, improving the model's predictive ability. Furthermore, using hyperspectral data of 515 japonica rice varieties, and using deep convolutional generative adversarial networks (DCGANs) to generate simulated data to improve model accuracy, a rice cereal protein content estimation model was constructed using partial least squares regression. The results showed that the simulated GPC values ​​generated after 8000 training iterations were closest to the measured values. The PSLR model established based on hyperspectral technology predicted micronutrients such as Ca, Mg, Mo, and Zn in wheat kernels and flour with an accuracy of 0.82, and the prediction effect was better for wheat kernels than for wheat flour.

[0005] Existing research and analysis on relevant quality parameters obtained through spectral imaging technology show that spectral imaging technology can reflect the content of physiological and biochemical indicators of agricultural products, enabling non-destructive and rapid detection of agricultural product quality. Summary of the Invention

[0006] Wheat quality traits are a combination of multiple traits, among which gluten quality, protein content, and starch are important indicators affecting wheat quality. Traditional methods for detecting wheat quality traits (such as the Dumas nitrogen determination method and hand-washing gluten) require sample destruction, are time-consuming, and require a large amount of flour, making it difficult to meet the needs of large-scale, multi-trait quality testing. Existing research mostly focuses on single traits (such as protein) and lacks optimization for joint modeling of multiple traits. Hyperspectral imaging technology, possessing both spectral and spatial information, is used for non-destructive testing of agricultural products. This invention proposes a method for wheat grain quality detection and functional gene analysis based on hyperspectral imaging technology. It employs a 400-2500nm hyperspectral imaging system, covering the visible, near-infrared, and short-wave infrared bands to enhance spectral information. Combining spectral technology, image processing technology, and artificial intelligence technology improves the stability and accuracy of the established model, enabling the simultaneous detection of 21 traits, including protein and gluten content. Simultaneously, a digital model of wheat quality Z is constructed to obtain comprehensive quality traits, achieving non-destructive detection of comprehensive wheat quality traits. Hyperspectral indices and true values ​​are used to jointly mine candidate genes for wheat quality, and hyperspectral features are used to rapidly analyze wheat quality functional genes.

[0007] To achieve the above objectives, the present invention provides the following solution:

[0008] A method for wheat grain quality detection and functional gene analysis based on hyperspectral imaging technology includes the following steps:

[0009] S1. A hyperspectral image of wheat grains is acquired using a hyperspectral imaging system, and the hyperspectral image is corrected, segmented, and masked, and the hyperspectral index of the processed hyperspectral image is extracted.

[0010] S2. The quality indicators of wheat grains are determined manually using chemical methods to obtain the artificially measured values;

[0011] S3. Preprocess the spectral reflectance data and extract the characteristic wavelengths;

[0012] S4. Perform correlation analysis on the quality indicators, and use principal component analysis to determine the weight coefficient of each quality indicator to obtain the wheat quality phenotype.

[0013] S5. Using the hyperspectral index of the characteristic wavelength as the independent variable, and the artificially measured value and the wheat quality phenotype as the dependent variables, a prediction model is established using partial least squares method and support vector machine; and based on the prediction model, the artificially measured value and the wheat quality phenotype of the unknown wheat grain sample are subjected to non-destructive testing.

[0014] S6. Based on the true values ​​of the hyperspectral index and the quality index as phenotypic traits, perform genome-wide association analysis and co-location analysis, and combine wheat genome annotation to mine candidate genes from candidate regions.

[0015] S7. By comparing transgenic plants with wild-type plants, the function of candidate genes and functional variation sites are verified.

[0016] More preferably, in S1, the band range of the hyperspectral imaging system is 400-2500nm, and the acquired hyperspectral images are stored as binary data streams;

[0017] Methods for correcting hyperspectral images include:

[0018] R= ;

[0019] In the formula, R represents the corrected image, R0 represents the original image, and R d The image represents the dark current, R w This represents the image displayed on the whiteboard.

[0020] More preferably, the method for segmenting a hyperspectral image includes:

[0021] The grayscale histogram of the corrected hyperspectral image is obtained, and the image of the band with obvious bimodal peaks in the grayscale histogram is obtained. The threshold is calculated using the OTSU method to perform image segmentation and obtain a binary image.

[0022] More preferably, the method for masking a hyperspectral image includes:

[0023] The obtained binary images are used to mask all hyperspectral images to obtain ROI hyperspectral images in the whole band; the total reflectance at each wavelength is calculated using the ROI hyperspectral images; primary hyperspectral indices and derived complex hyperspectral indices are extracted using LabVIEW; and the hyperspectral indices are preprocessed using SG convolution smoothing, multivariate scattering correction, variable standardization correction and normalization correction.

[0024] Primary hyperspectral indices include the original reflectance, pseudo-absorption coefficient, first derivative, and second derivative, while derived complex hyperspectral indices include the proportionality index and the normalization index.

[0025] More preferably, in S2, the quality indicators include: wet gluten content, dry gluten content, gluten index, protein content, grain length, grain width, thousand-grain weight, SDS sedimentation value, consistency, water absorption rate, formation time, stability time, and flour index.

[0026] More preferably, in S3, the preprocessing method includes: preprocessing the spectral reflectance data through multivariate scattering correction, SG convolution smoothing, standardization and normalization; and extracting feature wavelengths from the preprocessed spectral reflectance data.

[0027] More preferably, in S6, combined with wheat high-density SNP chip data, a hybrid linear model is used to correct the population structure, and correlation analysis is performed on the hyperspectral index and the true quality value to screen for significant correlation sites.

[0028] More preferably, in S7, gRNA targeting candidate genes is designed using the CRISPR / Cas9 system, the full-length CDS of the candidate genes is cloned, ligated into a strong promoter vector, and the protein content index of transgenic plants and wild-type grains is measured to quantify the effect of gene deletion / overexpression on quality.

[0029] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0030] This invention solves the problem of simultaneous detection of multiple quality traits; it predicts multiple quality indicators, accurately predicting phenotypes from phenotypes. This invention proposes a non-destructive detection method for wheat quality based on hyperspectral imaging, employing a 400-2500nm hyperspectral imaging system covering the visible-near-infrared-shortwave infrared bands to enhance spectral information. Combining spectral technology, image processing technology, and artificial intelligence technology, the stability and accuracy of the established model are improved, and a multi-trait joint prediction model is established to achieve simultaneous detection of multiple traits such as protein and gluten content. Furthermore, a precise prediction and inversion model for grain indicators is established. This method can achieve high-throughput non-destructive detection and distribution visualization of macromolecular substances in crop grains. Based on crop grain hyperspectral indices, this method can quickly identify candidate genes associated with quality traits, laying an important foundation for creating high-quality new varieties. Attached Figure Description

[0031] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a flowchart illustrating the overall technical process of the method proposed in this embodiment of the invention. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] Example 1:

[0036] like Figure 1 As shown, this embodiment provides a method for wheat grain quality detection and functional gene analysis based on hyperspectral imaging technology, including the following steps:

[0037] S1. A hyperspectral image of wheat grains is acquired using a hyperspectral imaging system, and the hyperspectral image is corrected, segmented, and masked. The hyperspectral index of the processed hyperspectral image is then extracted.

[0038] In this embodiment, the wavelength range of the hyperspectral imaging system is 400-2500nm, and the acquired hyperspectral images are stored as binary data streams.

[0039] Image processing was performed using LabVIEW 8.6 to reconstruct the corrected binary data stream into a hyperspectral image. The correction formula is as follows:

[0040] R= ;

[0041] In the formula, R represents the corrected image, R0 represents the original image, and R d The image represents the dark current, R w This represents the image under the whiteboard. Data reading and correction are performed by the LabVIEW sub-VIs Read from Binary File and Formula; subsequently, a dynamic link library written in C++ is called, combined with LabVIEW's IMAQ ArrayToImage to extract hyperspectral images at various wavelengths.

[0042] Gray-level histograms were obtained from the corrected hyperspectral images, and images of bands with obvious bimodal peaks in the gray-level histograms were acquired. Thresholds were calculated using the OTSU method for image segmentation to obtain binary images. Using the obtained binary images, LabVIEW's IMAQ Multiply function was used to mask all hyperspectral images, resulting in full-band ROI hyperspectral images. The total reflectance at each wavelength was calculated from the ROI hyperspectral images. Primary hyperspectral indices were extracted using LabVIEW's IMAQ Particle Analysis, and derived complex hyperspectral indices were extracted using the Formula subVI. Primary hyperspectral indices include original reflectance, pseudo-absorption coefficient, first derivative, and second derivative; derived complex hyperspectral indices include scaling indices and normalized indices. The hyperspectral indices were preprocessed using SG convolution smoothing, multivariate scattering correction, and variable standardization and normalization correction methods. The effectiveness of each method was compared, and the hyperspectral indices preprocessed using the optimal method were used for subsequent analysis.

[0043] S2. The quality indicators of wheat grains are determined artificially using chemical methods to obtain artificially measured values.

[0044] The quality indicators include nutritional quality, yield traits, and processing quality, specifically including 21 types such as wet gluten content, dry gluten content, gluten index, protein content, grain length, grain width, thousand-grain weight, SDS sedimentation value, consistency, water absorption rate, formation time, stability time, and flour index.

[0045] The quality values ​​of wheat were determined artificially using chemical methods, as follows: wet gluten content was determined according to GB / T5506.1-2008 standard; gluten index was determined according to SB / T 10248~10249-95; dry gluten content was determined according to GB / T5506.4-2008 / ISO 21415-4:2006; SDS-sedimentation value was determined according to the method described by Preston et al. (Preston et al. 1982); gluten swelling index was determined according to the method specified in GB / T 26627.1-2011; and GB / T14614-2006 was also referenced. The prescribed method involves using an electronic farinograph to determine farinograph parameters, including dough formation time, stability time, degree of weakening, and farinograph quality index; using an AM5100 rapid moisture analyzer to determine the moisture content of wheat samples; using the Dumas nitrogen determination method to determine protein content; and using a dual-wavelength colorimetric method to determine the content of amylose and amylopectin, in accordance with DB32 / T2265-2012.

[0046] S3. Preprocess the spectral reflectance data and extract the characteristic wavelengths.

[0047] Spectral reflectance data were preprocessed using multivariate scattering correction (MSC), SG convolution smoothing, standardization, and normalization. Feature wavelengths were extracted using recursive feature elimination (RFE, estimator=pls, n_features_to_select=100, step=10), principal component analysis (PCA), correlation analysis, and continuous projection algorithm (SPA), and the effects of various methods were compared.

[0048] S4. Perform correlation analysis on the quality indicators, and use principal component analysis to determine the weight coefficient of each quality indicator to obtain the wheat quality phenotype.

[0049] By analyzing the correlations among all wheat quality indicators and using Principal Component Analysis (PCA) for dimensionality reduction, a few principal components representing the main variation information of the original data were extracted. Based on the eigenvectors (loadings) of the PCA model, the weight coefficients of each original quality indicator in constituting the overall quality characteristics were determined. These weight coefficients were applied to the corresponding original indicator values, and a comprehensive digital wheat quality phenotypic score (Z = x) was obtained through weighted summation. Wet gluten content + y Dry gluten content + z Gluten index + ..., Z is the wheat quality phenotypic score, and x, y, z are coefficients), used to quantify and evaluate the overall quality level of wheat samples.

[0050] S5. Using the hyperspectral index of the characteristic wavelength as the independent variable, and the artificially measured value and the wheat quality phenotype as the dependent variables, a prediction model is established using partial least squares and support vector machine. The model parameters are optimized, the differences between models are compared, and independent verification is performed to evaluate the stability of the model. Based on the optimized model, the artificially measured value and the wheat quality phenotype of the unknown wheat grain sample are subjected to non-destructive testing.

[0051] Using the hyperspectral index as the independent variable and the manually measured values ​​of protein and gluten content as dependent variables, various machine learning algorithms, including partial least squares (PLSR, n_components: range(1, 100)) and support vector regression (SVR, C: [1, 10, 100], epsilon: [0.01, 0.1, 0.2], kernel: [rbf, linear]), were employed to establish prediction models based on the entire wavelength band and those based on characteristic wavelengths. Analysis of variance (ANOVA) and 5-fold cross-validation (cv = KFold(n_splits=5, shuffle=True, random_state=42)) were used to verify the established prediction models. The coefficient of determination (R²) was used for model validation.2 The accuracy of the model was evaluated using the root mean square error (RMSE); the model stability was tested by conducting 50 independent experiments based on random seeds (for _ in range(50):X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.25)).

[0052] S6. Based on the true values ​​of the hyperspectral index and the quality index as phenotypic traits, perform genome-wide association analysis (GWAS) and co-location analysis, and combine with wheat genome annotation to mine candidate genes from candidate regions.

[0053] By combining wheat high-density SNP chip data, a mixed linear model was used to correct the population structure, and correlation analysis was performed on hyperspectral index and true quality value to screen significant correlation sites.

[0054] S7. By comparing transgenic plants with wild-type plants, the function of candidate genes and functional variation sites are verified.

[0055] gRNAs targeting candidate genes were designed using the CRISPR / Cas9 system, the full-length CDS of the candidate genes were cloned, and they were ligated into a strong promoter vector. The protein content of transgenic plants and wild-type grains was measured to quantify the effects of gene deletion / overexpression on quality.

[0056] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for detecting wheat grain quality and analyzing functional genes based on hyperspectral imaging technology, characterized in that, Includes the following steps: S1. A hyperspectral image of wheat grains is acquired using a hyperspectral imaging system, and the hyperspectral image is corrected, segmented, and masked, and the hyperspectral index of the processed hyperspectral image is extracted. S2. The quality indicators of wheat grains are determined manually using chemical methods to obtain the artificially measured values; S3. Preprocess the spectral reflectance data and extract the characteristic wavelengths; S4. Perform correlation analysis on the quality indicators, and use principal component analysis to determine the weight coefficient of each quality indicator to obtain the wheat quality phenotype. S5. Using the hyperspectral index of the characteristic wavelength as the independent variable, and the artificially measured value and the wheat quality phenotype as the dependent variables, a prediction model is established using partial least squares method and support vector machine; and based on the prediction model, the artificially measured value and the wheat quality phenotype of the unknown wheat grain sample are subjected to non-destructive testing. S6. Based on the true values ​​of the hyperspectral index and the quality index as phenotypic traits, perform genome-wide association analysis and co-location analysis, and combine wheat genome annotation to mine candidate genes from candidate regions. S7. By comparing transgenic plants with wild-type plants, verify the function of candidate genes and functional variation sites; In S2, the quality indicators include: wet gluten content, dry gluten content, gluten index, protein content, particle length, particle width, thousand-particle weight, SDS sedimentation value, consistency, water absorption rate, formation time, stability time, and flour index. In S3, the preprocessing methods include: preprocessing the spectral reflectance data through multivariate scattering correction, SG convolution smoothing, standardization and normalization; and extracting feature wavelengths from the preprocessed spectral reflectance data. In S4, the correlation between all wheat quality indicators is analyzed, and principal component analysis is used for dimensionality reduction to extract a few principal components that can represent the main variation information of the original data. Based on the eigenvectors of the PCA model, the weight coefficients of each original quality indicator in constituting the comprehensive quality characteristics are determined. These weight coefficients are applied to the corresponding original indicator values, and a comprehensive digital wheat quality phenotypic score is obtained by weighted summation: Z = x wet gluten content + y dry gluten content + z gluten index + ..., where Z is the wheat quality phenotypic score, and x, y, and z are coefficients used to quantitatively evaluate the overall quality level of the wheat sample.

2. The method for wheat grain quality detection and functional gene analysis based on hyperspectral imaging technology according to claim 1, characterized in that, In S1, the band range of the hyperspectral imaging system is 400-2500nm, and the acquired hyperspectral images are stored as binary data streams. Methods for correcting hyperspectral images include: R= ； In the formula, R represents the corrected image, R0 represents the original image, and R d The image represents the dark current, R w This represents the image displayed on the whiteboard.

3. The method for wheat grain quality detection and functional gene analysis based on hyperspectral imaging technology according to claim 1, characterized in that, Methods for segmenting hyperspectral images include: The grayscale histogram of the corrected hyperspectral image is obtained, and the image of the band with obvious bimodal peaks in the grayscale histogram is obtained. The threshold is calculated using the OTSU method to perform image segmentation and obtain a binary image.

4. The method for wheat grain quality detection and functional gene analysis based on hyperspectral imaging technology according to claim 3, characterized in that, Methods for masking hyperspectral images include: The obtained binary images are used to mask all hyperspectral images to obtain ROI hyperspectral images in the whole band; the total reflectance at each wavelength is calculated using the ROI hyperspectral images; primary hyperspectral indices and derived complex hyperspectral indices are extracted using LabVIEW; and the hyperspectral indices are preprocessed using SG convolution smoothing, multivariate scattering correction, variable standardization correction and normalization correction. Primary hyperspectral indices include the original reflectance, pseudo-absorption coefficient, first derivative, and second derivative, while derived complex hyperspectral indices include the proportionality index and the normalization index.

5. The method for wheat grain quality detection and functional gene analysis based on hyperspectral imaging technology according to claim 1, characterized in that, In S6, combined with wheat high-density SNP chip data, a mixed linear model was used to correct the population structure, and correlation analysis was performed on hyperspectral index and true quality value to screen significant correlation sites.

6. The method for wheat grain quality detection and functional gene analysis based on hyperspectral imaging technology according to claim 1, characterized in that, In S7, gRNAs targeting candidate genes were designed using the CRISPR / Cas9 system, the full-length CDS of candidate genes were cloned, and they were ligated into a strong promoter vector. The protein content of transgenic plants and wild-type grains was measured to quantify the effects of gene deletion / overexpression on quality.

Images