A winter wheat light efficiency comprehensive index estimation method, system, device and medium

Abstract

The application provides a winter wheat light efficiency comprehensive index estimation method, system, device and medium, and belongs to the field of wheat breeding. The method comprises the following steps: determining the principal component score of each photosynthetic phenotype parameter on different principal components through principal component analysis; calculating the objective weight of each principal component through an objective weighting method, and obtaining a light efficiency comprehensive index representing the photosynthetic efficiency of winter wheat by weighting; obtaining a remote sensing image, calculating a vegetation index and extracting texture features from the remote sensing image; and training a machine learning regression model according to the vegetation index and the texture features to obtain a winter wheat light efficiency comprehensive index estimation model. The method reduces the dependence of photosynthetic efficiency evaluation on measured data, improves the acquisition flux of the winter wheat light efficiency comprehensive index, and provides reliable technical support for high light efficiency screening of large-scale breeding materials.

Description

Technical Field

[0001] This invention belongs to the field of wheat breeding, specifically relating to a method, system, equipment, and medium for estimating the comprehensive light efficiency index of winter wheat. Background Technology

[0002] Winter wheat is one of the world's major food crops. Breeding superior varieties is crucial not only for ensuring food security and stabilizing food supply, but also for supporting sustainable agricultural development. Crop phenotype is a comprehensive expression of a crop's appearance, physiological functions, and physicochemical properties under the combined influence of genotype and environment. Precise and efficient crop phenotyping is a key bottleneck in modern precision breeding. Photosynthetic phenotype, as a physiological functional phenotype, directly characterizes a plant's carbon assimilation capacity and environmental adaptability, and is a key trait determining dry matter accumulation, yield formation, and stress resistance. Therefore, screening superior materials and breeding high-photometric-efficiency varieties from the perspective of photosynthetic phenotype is considered an important direction for improving wheat yield potential and varietal traits.

[0003] To address these needs, and considering that high-photosynthetic-efficiency wheat typically exhibits stronger carbon assimilation capacity, superior stomatal regulation, and more coordinated photochemical energy conversion processes, current technologies often use photosynthetic phenotypic indicators such as net photosynthetic rate, stomatal conductance, and chlorophyll fluorescence as the basis for phenotypic identification and breeding practices in winter wheat. However, these photosynthetic phenotypic indicators can only reflect a specific local characteristic of the photosynthetic physiological process, resulting in an inaccurate characterization of the overall photosynthetic efficiency of wheat and making it difficult to guide wheat breeding practices. Summary of the Invention

[0004] To address the existing problems, this invention provides a method, system, computer equipment, and medium for estimating the comprehensive light efficiency index of winter wheat.

[0005] To achieve the above objectives, the present invention provides the following technical solution: A method for estimating the comprehensive light efficacy index of winter wheat includes: Acquire various photosynthetic phenotypic parameters of winter wheat samples, as well as remote sensing images of winter wheat planting areas; Principal components were extracted from various photosynthetic phenotypic parameters of winter wheat samples using principal component analysis, and the principal component scores of the winter wheat samples on each principal component were obtained. The objective weights of each principal component were calculated using the CRITIC objective weighting method. The comprehensive light efficiency index, which characterizes the photosynthetic efficiency of winter wheat samples, was obtained by weighting the objective weights and principal component scores. Vegetation indices are calculated based on remote sensing images, and texture features are extracted from the images. The vegetation index and texture features are used as input features, and the comprehensive light efficiency index is used as a supervised label to train a machine learning regression model, resulting in a comprehensive light efficiency index estimation model for winter wheat with the estimation results of the comprehensive light efficiency index of winter wheat as the output.

[0006] Optionally, in the method for estimating the comprehensive photosynthetic efficiency index of winter wheat provided by the present invention, the photosynthetic phenotypic parameters include net photosynthetic rate, transpiration rate, intercellular carbon dioxide concentration, stomatal conductance, photochemical quantum efficiency, excitation energy efficiency, and photochemical quenching coefficient.

[0007] Optionally, the method for estimating the comprehensive light efficacy index of winter wheat provided by the present invention further includes: Various photosynthetic phenotypic parameters were standardized to obtain a standardized photosynthetic phenotypic matrix. Principal component analysis was used to reduce the dimensionality of the standardized photosynthetic phenotype matrix to obtain the principal components. Determine the principal component scores of the winter wheat samples on each principal component.

[0008] Optionally, the method for estimating the comprehensive light efficacy index of winter wheat provided by the present invention further includes: Determine the standard deviation of the principal component scores on each principal component; Determine the correlation coefficients between each principal component, and use the correlation coefficients to determine the conflict between the principal components; The overall information content of each principal component is obtained from the standard deviation and conflict. The overall information content of each principal component is normalized to obtain the objective weight of each principal component.

[0009] Optionally, the method for estimating the comprehensive light efficacy index of winter wheat provided by the present invention further includes: Vegetation indices and texture features sensitive to photosynthetic phenotypes can be obtained by screening from vegetation indices and texture features using any one of the following methods: Pearson correlation coefficient, importance of projected variables, continuous projection algorithm, and elimination of uninformative variables.

[0010] Optionally, in the method for estimating the comprehensive light efficiency index of winter wheat provided by the present invention, the machine learning regression model can be any one of K-nearest neighbor algorithm, decision tree, support vector machine regression, random forest and extreme gradient boosting.

[0011] Optionally, the method for estimating the comprehensive light efficacy index of winter wheat provided by the present invention further includes: The hyperparameters of the machine learning regression model were tuned using a grid search method through 5-fold cross-validation.

[0012] This invention also provides a comprehensive light efficiency index estimation system for winter wheat, comprising: The light efficiency comprehensive index calculation module is used to obtain various photosynthetic phenotypic parameters of winter wheat samples, as well as remote sensing images of winter wheat planting areas; The comprehensive index construction module is used to extract principal components from various photosynthetic phenotypic parameters of winter wheat samples using principal component analysis, and obtain the principal component scores of winter wheat samples on each principal component; calculate the objective weights of each principal component using the CRITIC objective weighting method; and obtain the comprehensive light efficiency index characterizing the photosynthetic efficiency of winter wheat samples by weighting the objective weights and principal component scores. The model training module is used to calculate vegetation indices based on remote sensing images and extract texture features from the remote sensing images. The vegetation index and texture features are used as input features, and the comprehensive light efficiency index is used as a supervised label to train the machine learning regression model, so as to obtain the comprehensive light efficiency index estimation model of winter wheat with the estimation result of the comprehensive light efficiency index of winter wheat as the output.

[0013] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement any of the steps in a method for estimating the comprehensive light efficiency index of winter wheat.

[0014] The present invention also provides a computer-readable storage medium storing a computer program, which, when loaded by a processor, is capable of executing any step of a method for estimating the comprehensive light efficiency index of winter wheat.

[0015] The method for estimating the comprehensive light efficiency index of winter wheat provided by this invention has the following beneficial effects: This invention obtains various photosynthetic phenotypic parameters of winter wheat and then uses principal component analysis for dimensionality reduction and CRITIC objective weighting method for weighted fusion to construct a comprehensive light efficiency index that can fully characterize the photosynthetic efficiency level of winter wheat, providing a unified quantitative benchmark for the accurate evaluation of high light efficiency varieties of winter wheat. Considering the high physiological coupling and statistical collinearity among photosynthetic parameters such as net photosynthetic rate, stomatal conductance, and transpiration rate, this invention places principal component analysis (PCA) before CRITIC weighting. PCA first eliminates the collinearity redundancy among the original photosynthetic parameters, using uncorrelated principal components as the basis for CRITIC weighting. This allows CRITIC conflict calculations to be performed on truly independent dimensions, thereby obtaining objective weights that truly reflect the information carrying capacity of each principal component. Furthermore, the aforementioned photosynthetic parameters are subject to instantaneous fluctuations in environmental factors such as light, temperature, and wind speed in the field environment, resulting in significant random errors in single measurements. This invention uses CRITIC weighting to objectively evaluate the importance of principal components, ensuring that the constructed comprehensive light efficiency index eliminates information redundancy, avoids subjective weighting bias, and balances the information content and independence of each principal component.

[0016] Based on this, a machine learning regression model was trained using easily obtainable remote sensing features such as vegetation index and texture characteristics of winter wheat as input and a comprehensive photosynthetic efficiency index as a supervised signal. The result was an estimation model that could directly estimate the comprehensive photosynthetic efficiency index from remote sensing images. Once the model was trained, it eliminated the need for manual point-to-point measurement of various photosynthetic parameters using handheld devices. Only remote sensing images of winter wheat were required to quickly output estimation results reflecting the comprehensive photosynthetic efficiency level of wheat. This effectively reduced the dependence of photosynthetic efficiency evaluation on ground-based measured data, significantly increased the throughput of obtaining the comprehensive photosynthetic efficiency index of winter wheat, and provided reliable technical support for the early screening of high photosynthetic efficiency for large-scale breeding materials. Attached Figure Description

[0017] To more clearly illustrate the embodiments and design schemes of the present invention, the accompanying drawings required for this embodiment will be briefly described below. 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.

[0018] Figure 1 This is a schematic diagram of a method for estimating the comprehensive light efficiency index of winter wheat according to an embodiment of the present invention; Figure 2 Examples of the research area location and test area provided in this embodiment of the invention; Figure 3 This is an example of the field distribution of wheat varieties and sampling points provided in an embodiment of the present invention; Figure 4 This is an example of comparing CHPEI values ​​of different varieties provided in this embodiment of the invention; Figure 5 This is an example of comparing the correlation between CHPEI and different feature variables in an embodiment of the present invention; Figure 6 This is a comparative example of feature variable selection based on feature selection method provided in the embodiments of the present invention; Figure 7 This is a performance comparison example of CHPEI estimation models based on different feature selection methods provided in the embodiments of the present invention; Figure 8 This is an example of a CHPEI estimation scatter plot based on the XGBoost model provided in an embodiment of the present invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand and implement the technical solutions of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and should not be construed as limiting the scope of protection of the present invention.

[0020] Existing photosynthetic phenotypic indicators used to determine high photosynthetic efficiency in wheat not only rely on manual, handheld devices for fixed-point measurement, failing to meet the high-throughput phenotypic screening requirements for breeding materials, but also, a single indicator can only reflect a local characteristic of the photosynthetic physiological process, unable to comprehensively characterize the overall photosynthetic efficiency of wheat. To address these shortcomings, considering that Principal Component Analysis (PCA) can effectively eliminate collinear redundancy among multiple physiological parameters and extract core variation information through dimensionality reduction, and that the CRITIC objective weighting method can allocate weights based on the contrast strength and conflict between indicators, avoiding the subjectivity of traditional manual weighting, this invention organically combines these two methods to construct a Comprehensive High Photosynthetic Efficiency Index (CHPEI). Compared to a single indicator, this index can more comprehensively and robustly evaluate the overall photosynthetic productivity of winter wheat populations.

[0021] Furthermore, UAV remote sensing technology provides technical support for high-throughput crop phenotypic acquisition. Compared to UAV hyperspectral remote sensing, UAV visible light and multispectral remote sensing have significant advantages such as high cost-effectiveness, convenient data processing, and strong sensor stability. Moreover, the visible light vegetation index (RGBVI) and multispectral vegetation index (MSVI) constructed based on spectral bands can accurately respond to dynamic changes in crop canopy phenotypic characteristics. However, in dense canopies with high biomass, the asymptotic saturation effect of optical sensors may limit the ability to perceive changes in crop physiological states using only spectral information. To address this, this invention utilizes texture features (TF) to reflect the spatial physical properties of the canopy from a geometrical perspective. By combining texture features with vegetation indices to reflect crop changes, this effectively suppresses spectral saturation and improves the estimation accuracy of crop phenotypic parameters under complex cover conditions.

[0022] Furthermore, in the collaborative modeling process of multi-source remote sensing features, the original feature variables suffer from severe dimensional redundancy and collinearity, which greatly limits the model's accuracy and generalization ability. Therefore, this invention introduces an efficient feature selection method, which can not only effectively identify and extract feature subsets with strong driving power for the target variable, reducing model complexity, but also effectively suppress overfitting and enhance the robustness of model predictions. Regarding the modeling algorithm, considering the significant advantages of machine learning regression models in crop phenotypic estimation due to their excellent nonlinear mapping capabilities, especially with the application of ensemble learning, this invention integrates the advantages of multiple weak learners, providing stronger robustness while maintaining high estimation accuracy. By combining feature selection methods with the modeling strategy of machine learning regression models, the estimation accuracy of winter wheat CHEPI is improved, aiming to achieve accurate monitoring of high photosynthetic efficiency levels in winter wheat and effectively screen for varieties with high photosynthetic efficiency.

[0023] In summary, this invention combines PCA and CRITIC methods to construct a Comprehensive High Photosynthetic Efficiency Index (CHPEI), enabling quantitative characterization of winter wheat photosynthetic phenotypes. Visible vegetation index (RGBVI), multispectral vegetation index (MSVI), and texture features extracted from UAV spectral images, along with their combinations, are used to estimate the winter wheat CHPEI. This establishes a method for estimating wheat photosynthetic phenotypes based on UAV remote sensing, providing technical support for rapid and non-destructive screening of high-photosynthetic-efficiency winter wheat varieties.

[0024] Example 1 This invention provides a method for estimating the comprehensive light efficiency index of winter wheat, specifically as follows: Figure 1 As shown, it includes the following steps: Step 1: Obtain various photosynthetic phenotypic parameters of winter wheat samples, as well as remote sensing images of winter wheat planting areas.

[0025] Step 2: Principal components are extracted from various photosynthetic phenotypic parameters of winter wheat samples using principal component analysis (PCA) to obtain the principal component scores of the winter wheat samples on each principal component. The objective weights of each principal component are calculated using the CRITIC objective weighting method. A comprehensive photosynthetic efficiency index, characterizing the photosynthetic efficiency of the winter wheat samples, is obtained by weighting the objective weights and principal component scores. The photosynthetic phenotypic parameters include net photosynthetic rate, transpiration rate, intercellular carbon dioxide concentration, stomatal conductance, photochemical quantum efficiency, excitation energy efficiency, and photochemical quenching coefficient. Specifically, PCA involves standardizing various photosynthetic phenotypic parameters to obtain a standardized photosynthetic phenotypic matrix; reducing the dimensionality of the standardized photosynthetic phenotypic matrix using PCA to obtain the principal components; and determining the principal component scores of the winter wheat samples on each principal component. The CRITIC objective weighting method specifically includes: determining the standard deviation of the principal component scores on each principal component; determining the correlation coefficient between each principal component, and using the correlation coefficient to determine the conflict between each principal component; obtaining the comprehensive information content of each principal component from the standard deviation and conflict; and normalizing the comprehensive information content of each principal component to obtain the objective weight of each principal component.

[0026] Specifically, taking an ecological unmanned farm in a certain area as an example, this invention specifically introduces the method for estimating the comprehensive light efficiency index of winter wheat. Among other things, such as... Figure 2As shown, the area has an altitude of approximately 21 meters, an average annual sunshine duration of 2100-2500 hours, a frost-free period of 180-220 days, a temperate continental monsoon climate, an average annual precipitation of approximately 650 mm, and an average annual temperature of approximately 13.2℃. Summers are hot and rainy, with rain and heat occurring simultaneously; winters are cold and dry with little rain or snow. The soil in the ecological unmanned farm area is mainly clay loam, with a pH value of approximately 7.8, a soil bulk density of approximately 1.38 g / cm³, a field water holding capacity of approximately 23%, and a wilting coefficient of approximately 9.2%.

[0027] First, this embodiment uses 11 wheat materials, including 7 widely planted varieties and 4 domestically bred varieties that have not yet received official approval. These varieties were used in a randomized block design with 3 replicates, and the specific variety distribution is as follows: Figure 3 As shown in (1), there are a total of 33 communities, each community is 15m. 2 The plots were spaced 0.5m apart, and sowing began on October 13, 2023. Row spacing was 30cm. Base fertilizer consisted of 450kg / ha of compound fertilizer containing 15% nitrogen, phosphorus, and potassium. Top dressings of 46% pure urea were applied at the tillering stage and 300kg / ha at the jointing stage. Field management followed local high-yield programs, including regular irrigation, weeding, and integrated pest management. Harvesting took place on June 11, 2024, with a growth period of approximately 240 days. Field data were collected during four key growth stages: jointing, heading, flowering, and grain filling. Two sampling points were marked in each plot, as shown in the diagram. Figure 3 shown in (2).

[0028] During the wheat planting season, on sunny days from 10:00 to 12:00, specifically on April 6th (jointing stage), April 24th (heading stage), May 2nd (flowering stage), and May 11th (grain-filling stage), seven photosynthetic phenotypic parameters closely related to wheat photosynthesis were measured using a LI-6800 portable photosynthesis system (LI-COR, Lincoln, NE, USA). These parameters included net photosynthetic rate (Pn), transpiration rate (Tr), intercellular carbon dioxide concentration (Ci), stomatal conductance (Gs), photochemical quantum efficiency (PhiPS2), excitation energy efficiency (Fv' / Fm'), and photochemical quenching coefficient (qP). At sampling points in each experimental plot, flag leaves from the main stems of three winter wheat plants with consistent growth vigor were selected as the test samples. A total of six sets of repeated observation data were collected for each plot, and the average value was taken as the measured value of the photosynthetic parameters for that plot to ensure the robustness and representativeness of the experimental data.

[0029] Considering that evaluating high photosynthetic efficiency varieties using a single parameter is easily affected by collinearity and cannot fully reflect the comprehensive photosynthetic potential of crops, this invention adopts a comprehensive evaluation method (PCA-CRITIC) that combines principal component analysis (PCA) with the CRITIC objective weighting method to construct a comprehensive high photosynthetic efficiency index (CHPEI), aiming to provide a scientific basis for the rapid screening and quantitative characterization of high photosynthetic efficiency varieties of winter wheat.

[0030] For example, the measured photosynthetic phenotypic parameters are first standardized to obtain photosynthetic phenotypic parameters that eliminate differences in dimensions and orders of magnitude between different parameters and ensure data comparability, as shown in formula (1): (1) in, These are the standardized values ​​of the photosynthetic phenotypic parameters. For the first The first sample The original values ​​of each photosynthetic phenotypic parameter and The first The first sample The mean and standard deviation of each photosynthetic phenotypic parameter were then calculated. Next, PCA was used to reduce the dimensionality of the standardized matrix, transforming these photosynthetic phenotypic parameters with complex correlations into uncorrelated principal components to eliminate collinearity redundancy. The principal component scores of each sample on different principal components were then calculated, as shown in formulas (2), (3), and (4).

[0031] (2) (3) (4) (5) in, Let n be the covariance matrix, and n be the sample size. For the k-th eigenvalue, For the corresponding feature vector, Let k be the score vector of the principal component. The maximum value of the score vector of the k-th principal component. Let be the minimum value of the score vector of the k-th principal component. This is the principal component score after positive transformation of the principal component score. Specifically, formula (2) outputs the covariance matrix C; substituting C into formula (3) to perform eigenvalue decomposition, we obtain the eigenvectors. ;Bundle Substitute into formula (4) to calculate the original principal component scores. ;Will Substituting into formula (5), we obtain the normalized and normalized standard principal component scores. .

[0032] After obtaining the principal component scores, the contribution of each principal component in the overall system is determined using the CRITIC objective weighting method. For example, by comprehensively considering the information carrying capacity of the principal component scores and the information conflict between the components, the objective weight of each component is calculated, thereby effectively avoiding the subjective interference of artificial weighting in traditional evaluation methods. The specific weight calculation is shown in formulas (6), (7), and (8):

[0033] (6) (7) (8) in, The information conflict between the k-th principal component and other principal components is indicated, and m represents the total number of principal components extracted and retained. This represents the total information content of the k-th principal component. Let be the absolute correlation coefficient between the k-th and l-th principal components. This represents the standard deviation of the k-th principal component. Let the CRITIC weight of the k-th principal component satisfy the following condition: Finally, CHPEI is constructed using a weighted summation method, as shown in formula (9):

[0034] (9) in, The high light efficiency comprehensive score for the i-th sample is the light efficiency comprehensive index score that characterizes the photosynthetic efficiency of winter wheat. The higher the score, the better the high light efficiency level.

[0035] Step 2: Acquire remote sensing images, calculate vegetation indices from the images, and extract texture features. Vegetation indices include EXR, EXG, EXGR, INT, NPCI, NGBDI, NGRDI, WI, RGBVI, VARI, MGRVI, CIVI, NDVI, GNDVI, NDRE, CIre, CIg, DVI, SAVI, OSAVI, RDVI, RVI, TVI, and NNIR. Texture features include mean, variance, homogeneity, contrast, dissimilarity, entropy, second moment, and correlation.

[0036] Specifically, during the winter wheat planting process, a DJI Mavic 3M unmanned aerial vehicle (UAV) was used as a remote sensing data acquisition platform. This UAV integrates a high-resolution CMOS visible light (RGB) camera and a four-channel multispectral sensor, covering 560nm green light, 650nm red light, 730nm red edge, and 860nm near-infrared bands. The UAV aerial photography was conducted simultaneously with the ground-based data acquisition process. The UAV flew at an altitude of 30m and a speed of 2m / s, with a forward overlap of 80% and a lateral overlap of 70% to ensure high-precision stitching of orthophotos. The flight path was automatically planned using DJI Pilot software, and the camera used an equal-interval trigger mode. To ensure geometric accuracy and consistency of radiometric characteristics, 1m×1m ground control points (GCPs) were deployed around the test area, and their centimeter-level geographic coordinates were acquired using real-time kinematic (RTK) base stations. Furthermore, a standard radiometric calibration board was photographed before takeoff to eliminate radiometric distortion caused by the sensor and ambient light.

[0037] After acquiring raw RGB and multispectral (MS) remote sensing images, they were imported into Pix4Dmapper software. Aerial triangulation adjustment (BBA) was performed by introducing ground control points (GCPs) and manually marking tie points to generate dense point clouds and texture meshes. After radiometric and orthorectification, high-resolution RGB and MS orthorectified images were generated. These generated orthorectified images were imported into ENVI 5.6 software, where thresholding was performed using the Normalized Difference Vegetation Index (NDVI) to distinguish between the canopy and soil background. Statistical tools were used for comparative analysis to extract binarized vector masks, which were then used to crop regions of interest (ROIs) to effectively remove soil background interference.

[0038] Subsequently, vegetation indices were calculated based on 12 RGBVIs and 12 MSVIs closely related to wheat growth. The specific vegetation index calculations are shown in Table 1. Table 1. Calculation of Vegetation Index Based on UAV Remote Sensing Data Where r, g, and b represent the red, green, and blue bands in a visible light camera, respectively, and G, R, RE, and NIR represent the green, red, red-edge, and near-infrared bands in a multispectral sensor, respectively. Texture, as one of the fundamental visual features of an image, often manifests as a combination of similar patterns with varying intensities. In remote sensing image analysis, ground texture is an important characteristic distinguishing it from spectral features. In this invention, texture feature extraction is based on the Gray Level Co-occurrence Matrix (GLCM) method. By statistically analyzing the joint gray-level distribution between pixel pairs at certain directions and distances in the image, the spatial structural features of the texture are quantitatively described. For example, the "Co-occurrence Measures" tool in ENVI 5.6 software is used to construct GLCMs for each of the four bands. Each matrix can systematically extract eight types of texture metrics, including mean (MEA), variance (VAR), homogeneity (HOM), contrast (CON), dissimilarity (DIS), entropy (ENT), second moment (SEC), and correlation (COR). These features describe the texture from aspects such as image brightness, local variation, uniformity, structural contrast, degree of dissimilarity, information complexity, distribution uniformity, and linear correlation. In the specific calculation process, to balance feature stability and spatial detail preservation, the sliding window size is set to 5×5 pixels, and the step size is 1 pixel to ensure the continuity and integrity of the feature map.

[0039] Step 3: Select vegetation indices and texture features that are sensitive to photosynthetic phenotypes from vegetation indices and texture features using any one of the following methods: Pearson correlation coefficient, importance of projected variables, continuous projection algorithm, and elimination of uninformative variables.

[0040] Specifically, after extracting high-dimensional features such as RGBVI, MSVI, and TF, this invention employs various feature selection methods to filter features and obtain a subset of feature variables sensitive to photosynthetic phenotypes in order to eliminate redundant information. For example, the Pearson correlation coefficient method (PCC) is based on the linear correlation theory in statistics. It identifies sensitive features that directly contribute to the target variable by calculating the correlation coefficient between the independent and dependent variables and combining it with a significance test of P < 0.01. The projected variable importance method (VIP) constructs a partial least squares regression (PLSR) model and scores the explanatory variables according to their influence weights on the response variable. Variables with a VIP value greater than 1.0 are typically selected as predictors. This method has significant advantages in processing datasets with multicollinearity. Addressing the severe multicollinearity problem among feature variables, the continuous projection algorithm (SPA) utilizes vector space projection analysis to continuously minimize linear redundancy among variables during the filtering process, thereby extracting feature combinations with clear physical meaning and minimal redundancy while ensuring information integrity. Uninformative variable elimination (UVE) improves model robustness by introducing artificial random noise into the original variables and performing stability analysis to remove feature variables that contribute little or no to the prediction results. It is important to emphasize that the above feature selection methods can all be used in the processing stage after extracting high-dimensional features such as RGBVI, MSVI, and TF in this invention. The specific method should be determined by those skilled in the art based on actual needs, and this invention does not impose any limitations.

[0041] Step 4: Input the vegetation index and texture features into the untrained machine learning regression model to output the light efficiency index estimation result; then, using the vegetation index and texture features as input features and the comprehensive light efficiency index as the supervised label, train the machine learning regression model to obtain the winter wheat comprehensive light efficiency index estimation model, which outputs the winter wheat comprehensive light efficiency index estimation result. The machine learning regression model can be any one of the following: K-nearest neighbor algorithm, decision tree, support vector machine regression, random forest, and extreme gradient boosting. The hyperparameters of the machine learning regression model can be fine-tuned using grid search through 5-fold cross-validation.

[0042] Specifically, this invention can use one or more machine learning regression models to process nonlinear relationships in remote sensing data and construct a quantitative estimation model for wheat CHPEI. Among them, the K-Nearest Neighbor (KNN) algorithm, as a non-parametric learning method, finds the k known samples with the closest Euclidean distance to the test sample in the feature space and obtains the prediction result through local weighted averaging. Decision Tree (DT) constructs a tree-like decision logic by recursively segmenting sample features, exhibiting strong interpretability and insensitivity to outliers. Support Vector Machine Regression (SVR) uses kernel function technology to map input features to a high-dimensional feature space, seeking the globally optimal hyperplane to handle complex nonlinear regression tasks. In the field of ensemble learning, Random Forest (RF) uses the Bagging idea to construct multiple decision trees in parallel, effectively suppressing the overfitting tendency of a single model by averaging the results of multiple weak learners. Extreme Gradient Boosting (XGBoost), as an advanced boosting framework, precisely optimizes the loss function through second-order Taylor expansion and combines regularization terms to control model complexity, demonstrating excellent performance in the quantitative estimation of crop phenotypic parameters. To ensure optimal performance for each model, this invention uniformly employs 5-fold cross-validation to perform grid search tuning on the core hyperparameters of each algorithm.

[0043] For example, this invention can randomly select 2 / 3 of the samples as the training dataset for model training using the KS algorithm, and the remaining 1 / 3 as the test dataset for validation. During the validation phase, this invention selects the coefficient of determination (R²). 2 The R² values ​​(root mean square error (RMSE) and mean absolute error (MAE) are used to evaluate the performance of CHPEI estimation models constructed by combining different subsets of feature variables with machine learning regression models. A higher R² value and lower RMSE and MAE values ​​indicate higher model accuracy and more stable performance.

[0044] Example 2 Based on Example 1, this example can further verify the role of the comprehensive light efficiency index and provide preferred examples.

[0045] First, the CHPIE values ​​of the 11 winter wheat varieties in Example 1 at different growth stages were quantitatively evaluated. The comparison results of CHPIE values ​​of different varieties are as follows: Figure 4As shown, "05292", "05604", "06085", and "06144" are simplified identifiers, corresponding to the full test codes SH05292, SH05604, SH06085, and SH06144, respectively. It can be seen that the photosynthetic efficiency (CHEI) level of winter wheat exhibits significant fluctuations throughout its growth period and differences between varieties. Specifically, regarding the changes during the growth period, most varieties have high CHEI values ​​at the jointing stage, followed by a general decline in CHEI values ​​during the heading stage. However, after entering the flowering stage, the CHEI value shows a significant upward trend, with most varieties reaching or approaching their second peak during this stage. Subsequently, during the grain-filling stage, significant differences emerge based on varietal characteristics. Furthermore, a comparison of CHEI values ​​among multiple varieties shows that SH06144 and Liangxing 19 have the highest CHEI levels throughout the entire growth period, with CHEI values ​​of 0.479 and 0.471, respectively. In particular, the cultivar SH06144, with a CHPIE value of 0.589 during the grain-filling stage, far exceeded that of other varieties, indicating that it possesses extremely strong photosynthetic maintenance capacity in the later stages of growth. In contrast, Xinmai 916 and Jinong 114 exhibited the lowest high photosynthetic efficiency throughout the entire growth period, with CHPIE values ​​of 0.366 and 0.387, respectively, and their CHPIE levels remained low at multiple key growth stages. This clear varietal gradient validates the sensitivity of CHPIE in evaluating high photosynthetic efficiency in wheat, effectively distinguishing differences in photosynthetic productivity among varieties.

[0046] Next, a correlation analysis was performed between the CHPEI values ​​and different types of characteristic variables, and the correlation heatmap is shown below. Figure 5 As shown. By Figure 5 It can be seen that, compared with single photosynthetic parameters such as Pn, Tr, and Ci, CHPEI generally exhibits higher correlations with characteristic variables. In RGBVI, ExR (r=0.429) and RGBVI (r=-0.413) show the highest correlation coefficients with CHPEI, and are statistically significant at the 0.01 level. In MSVI, TVI and RDVI both show highly significant negative correlations with CHPEI (p<0.01), with correlation coefficients r=-0.453 and -0.42, respectively. Furthermore, in TI, G_COR and NIR_COR show higher correlations with CHPEI than with single photosynthetic parameters, with correlation coefficients r=-0.407 and 0.420, respectively, and p<0.01. This strongly demonstrates that CHPEI, by integrating multidimensional photosynthetic parameter information, can more robustly represent the high photosynthetic efficiency level of winter wheat and characterize the growth status of winter wheat under complex field conditions.

[0047] Furthermore, this embodiment also compared the impact of different feature variables, such as the Multispectral Vegetation Index (MSVI), the RGB Vegetation Index (RGBVI), Texture Features (TF), and Comprehensive Multispectral-Visible-TextureFeatures (CMVTF), on the performance of the estimation model. The results are shown in Table 2. Table 2 CHPEI estimation table based on different variable characteristics Table 2 shows that, among single-variable types, MSVI's estimation performance is generally better than RGBVI and TF. Further analysis reveals that CMVTF, which integrates spectral information and texture features, demonstrates significant superiority across all algorithms, with its overall accuracy far exceeding that of the CHPEI estimation model built from single data. This proves that the canopy spatial structure information provided by texture features can effectively supplement the deficiencies of spectral data in physiological monitoring. Furthermore, compared to SVR, DT, and KNN algorithms, the CHPEI estimation model built using ensemble learning algorithms such as XGBoost and RF achieves higher accuracy on the test set R. 2 The scores reached 0.725 and 0.703 respectively, and the RMSE decreased to 0.045 and 0.038 respectively. This indicates that the ensemble learning algorithm demonstrates stronger fitting ability when dealing with multi-type feature data such as CMVTF, and the estimation reliability of CHPEI is improved.

[0048] A comparison of the effectiveness of different feature selection methods in reducing information redundancy in the original features, for example... Figure 6 As shown, the PCC method focuses on retaining variables with strong linear driving force, selecting 33 highly significant related features. The VIP method, through partial least squares projection, selects 23 core features from the perspective of explanatory power. The SPA algorithm, through vector space projection, greatly reduces collinearity interference between features, retaining 31 features with complementary information. The UVE method evaluates the stability of features by adding noisy variables, ultimately identifying 27 feature variables that make robust contributions to CHPEI. These different types of feature selection strategies effectively reduce the number of feature variables, laying the foundation for building a concise and efficient estimation model.

[0049] The impact of feature selection strategies on the performance of estimation models, such as Figure 7 As shown, by Figure 7It can be seen that, compared to CMVTF, the CHPEI estimation models after feature selection all show varying degrees of performance improvement. Among the four selection methods, the variable combinations selected based on CMVTF-UVE exhibit the strongest robustness compared to CMVTF-PCC, CMVTF-VIP, and CMVTF-SPA. The XGBoost and RF algorithms show the highest accuracy on both the training and test sets. On the training set, R... 2 The R values ​​were 0.874 and 0.882 respectively, 1 / RMSE were 0.032 and 0.031 respectively, and MAE were 0.027 and 0.023 respectively. On the test set, R... 2 The scores were 0.862 and 0.842 respectively, RMSE were 0.032 and 0.034 respectively, and 1 / MAE were 0.027 and 0.029 respectively. The SVR algorithm was the second best, while the DT and KNN algorithms were the worst. This shows that the feature selection method combined with the ensemble learning algorithm has a greater advantage in estimating the model, resulting in a more accurate CHPEI estimation model.

[0050] Furthermore, taking the XGBoost algorithm as an example, the CHPEI estimation scatter plots obtained by different feature selection methods are as follows: Figure 8 As shown. By Figure 8 It can be seen that the feature subset selected using the UVE method is the optimal feature combination for estimating the CHPEI of winter wheat, and the CHPEI estimation model built based on CMVTF-UVE combined with the XGBoost algorithm (CMVTF-UVE-XGBoost) exhibits the best performance. This model shows high consistency on both the training and test sets, with predicted and measured values ​​closely distributed on either side of the 1:1 line, showing no obvious overfitting. Furthermore, compared to CMVTF-PCC, the CMVTF-UVE-XGBoost model has a higher R-value on both the training and test sets. 2 The results showed improvements of 8.84% and 12.98% respectively, while RMSE and MAE decreased by 17.95% and 23.81%, and 18.18% and 28.95% respectively. This further demonstrates that the UVE algorithm can effectively identify and retain feature subsets that are highly consistent with crop photosynthetic potential, and significantly improve the generalization ability and prediction accuracy of the CHPEI estimation model by eliminating collinear redundancy.

[0051] In summary, CHPEI exhibits significant varietal differences throughout the entire wheat growth period. SH06144 and Liangxing 19 maintained consistently high photosynthetic efficiency throughout the entire growth period, while Xinmai 916 and Jinong 114 performed poorly. This indicates that CHPEI is a more robust measure of photosynthetic productivity differences among wheat varieties compared to single photosynthetic parameters. Furthermore, compared to feature selection methods such as PCC, VIP, and SPA, the UVE method performs better in eliminating collinearity redundancy and information noise. The selected feature subset significantly improves the model's robustness and generalization ability, demonstrating that multi-source feature combination (CMVTF) outperforms single-type feature variables, including RGBVI, MSVI, and TF, in estimation performance. Moreover, the XGBoost and RF algorithms based on ensemble strategies outperform SVR, KNN, and DT in CHPEI estimation accuracy. Among them, the CMVTF-UVE-XGBoost model performs best, achieving an R² of 0.862 on the test set, with RMSE and MAE of 0.032 and 0.027, respectively. Furthermore, the spatial distribution estimation results of this model at different growth stages are highly consistent with the measured values, verifying its reliability in wheat CHPIE estimation. Therefore, this invention preferably adopts a CHPIE estimation model combining multi-source feature fusion, UVE feature optimization, and the XGBoost algorithm, which can effectively and accurately assess wheat photosynthetic efficiency levels and provides technical support for screening high-photosynthetic-efficiency wheat breeding materials.

[0052] This invention overcomes the limitations of evaluating crop photosynthetic efficiency using a single photosynthetic phenotypic parameter by constructing the CHPEI using a coupled PCA and CRITIC objective weighting method. In wheat varieties, the CHPEI value is high at the jointing stage, generally decreases after entering the heading stage, and then rebounds at the flowering stage. This trend may be because the plant's growth focus gradually shifts from the vegetative growth stage to the reproductive growth stage during the heading stage, leading to phased fluctuations in photosynthetic rate. The increase in CHPEI value at the flowering stage indicates a deep coordination between the plant's photosynthetic structure and function. Regarding varietal differences, SH06144 and Liangxing 19 exhibit significant photosynthetic robustness. In particular, the CHPEI value of SH06144 remains high in the later stages of growth, indicating that this variety can maintain high photosynthetic efficiency in the later stages of growth, far exceeding other varieties. This significant difference in photosynthetic productivity among varieties also verifies the application potential of CHPEI in the evaluation of high-efficiency breeding materials.

[0053] Furthermore, this invention significantly improves the accuracy of CHPEI estimation by integrating MSVI, RGBVI, and TF. MSVI often exhibits asymptotic saturation during the vigorous growth period of winter wheat due to dense canopy, limiting model estimation performance. RGBVI, on the other hand, can sensitively capture subtle pigment differences in the visible light band, while texture features reflect the physical properties of the canopy from a geometrical perspective, mitigating the loss of model accuracy caused by spectral saturation. Simultaneously, the four feature selection strategies employed in this invention—PCC, VIP, SPA, and UVE—effectively screened subsets of feature variables. Among them, the UVE method significantly improves the accuracy of the CHPEI estimation model by eliminating redundant variables with weak responses to photosynthetic physiological states and reducing multicollinearity among feature variables.

[0054] The machine learning regression model employed in this invention can reduce the multicollinearity problem among remote sensing data and capture the complex nonlinear relationship between remote sensing feature variables and CHPEI. Among them, ensemble learning models, represented by XGBoost and RF, outperform SVR, DT, and KNN in estimation accuracy. This is mainly attributed to their unique algorithmic architecture when processing high-dimensional remote sensing datasets. Specifically, XGBoost iteratively minimizes the residual loss function through a forward stepwise additive strategy, demonstrating a strong ability to capture complex nonlinear feature representations; meanwhile, RF, through its bagging strategy and random feature selection, significantly enhances robustness to remote sensing data noise while reducing model variance. In particular, the CMVTF-UVE-XGBoost model based on UVE optimization achieves the highest fitting accuracy on the test set, indicating that for the comprehensive photosynthetic phenotype of CHPEI, ensemble learning algorithms have more prominent nonlinear fitting capabilities, stronger anti-overfitting performance, and can achieve high-precision estimation of crop phenotypic parameters. Compared to expensive hyperspectral equipment, this study achieved satisfactory CHPEI estimation results using visible light and multispectral data, greatly reducing the cost of large-scale field applications.

[0055] Example 3 This invention also provides a comprehensive light efficiency index estimation system for winter wheat, comprising: The light efficiency comprehensive index calculation module is used to obtain various photosynthetic phenotypic parameters of winter wheat samples, as well as remote sensing images of winter wheat planting areas.

[0056] The comprehensive index construction module is used to extract principal components from various photosynthetic phenotypic parameters of winter wheat samples using principal component analysis, and obtain the principal component scores of winter wheat samples on each principal component; calculate the objective weights of each principal component using the CRITIC objective weighting method; and obtain the comprehensive light efficiency index characterizing the photosynthetic efficiency of winter wheat samples by weighting the objective weights and principal component scores.

[0057] The model training module is used to calculate vegetation indices based on remote sensing images and extract texture features from the remote sensing images. The vegetation index and texture features are used as input features, and the comprehensive light efficiency index of winter wheat is used as a supervised label to train the machine learning regression model, so as to obtain the comprehensive light efficiency index estimation model of winter wheat with the estimation result of the comprehensive light efficiency index as the output.

[0058] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the steps in an embodiment of a method for estimating the comprehensive light efficiency index of winter wheat. Specific implementation methods can be found in the method embodiments, and will not be repeated here.

[0059] Furthermore, the present invention also provides a non-transitory computer-readable storage medium containing instructions, on which a computer program is stored. For example, a memory containing instructions that can be executed by a processor of a computer device to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc. When the computer program is executed by the processor, it can implement the steps in an embodiment of a method for estimating the comprehensive light efficiency index of winter wheat. Specific implementation methods can be found in the method embodiments, which will not be repeated here.

[0060] Those skilled in the art will understand that embodiments of the present invention can provide methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0061] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0062] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0063] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0064] It should be noted that the specific embodiments described above enable those skilled in the art to more fully understand the present invention, but do not limit the present invention in any way. Therefore, although the present invention has been described in detail in this specification and embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the present invention; and all technical solutions and improvements that do not depart from the spirit and scope of the present invention are covered within the protection scope of the present invention patent. No reference numerals in the claims should be construed as limiting the scope of the claims. Any simple variations or equivalent substitutions of technical solutions that can be readily obtained by those skilled in the art within the scope of the technology disclosed in the present invention are within the protection scope of the present invention.

Claims

1. A method for estimating the comprehensive light efficacy index of winter wheat, characterized in that, include: Acquire various photosynthetic phenotypic parameters of winter wheat samples, as well as remote sensing images of winter wheat planting areas; Principal components were extracted from various photosynthetic phenotypic parameters of winter wheat samples using principal component analysis to obtain the principal component scores of the winter wheat samples on each principal component; the objective weights of each principal component were calculated using the CRITIC objective weighting method; and a comprehensive light efficiency index characterizing the photosynthetic efficiency of winter wheat samples was obtained by weighting the objective weights and principal component scores. The vegetation index is calculated based on the remote sensing image, and texture features are extracted from the remote sensing image; Using the vegetation index and texture features as input features, and the comprehensive light efficiency index as a supervised label, a machine learning regression model is trained to obtain a comprehensive light efficiency index estimation model for winter wheat, with the estimation result of the comprehensive light efficiency index for winter wheat as the output.

2. The method for estimating the comprehensive light efficiency index of winter wheat according to claim 1, characterized in that, The photosynthetic phenotypic parameters include net photosynthetic rate, transpiration rate, intercellular carbon dioxide concentration, stomatal conductance, photochemical quantum efficiency, excitation energy efficiency, and photochemical quenching coefficient.

3. The method for estimating the comprehensive light efficiency index of winter wheat according to claim 1, characterized in that, Principal component analysis was used to extract principal components from various photosynthetic phenotypic parameters of winter wheat samples. The principal component scores of the winter wheat samples on each principal component were obtained, including: Various photosynthetic phenotypic parameters were standardized to obtain a standardized photosynthetic phenotypic matrix. Principal component analysis was used to reduce the dimensionality of the standardized photosynthetic phenotype matrix, and the principal components were obtained. Determine the principal component scores of the winter wheat sample on each principal component.

4. The method for estimating the comprehensive light efficiency index of winter wheat according to claim 1, characterized in that, The objective weights of each principal component are calculated using the CRITIC objective weighting method, including: Determine the standard deviation of the principal component scores on each principal component; Determine the correlation coefficients between each principal component, and use the correlation coefficients to determine the conflict between the principal components; The overall information content of each principal component is obtained from the standard deviation and conflict. The comprehensive information of each principal component is normalized to obtain the objective weight of each principal component.

5. The method for estimating the comprehensive light efficiency index of winter wheat according to claim 1, characterized in that, After extracting texture features from the remote sensing image, the process also includes: Vegetation indices and texture features sensitive to photosynthetic phenotypes are obtained by filtering from the vegetation indices and texture features using any one of the following methods: Pearson correlation coefficient, importance of projected variables, continuous projection algorithm, and elimination of uninformative variables.

6. The method for estimating the comprehensive light efficiency index of winter wheat according to claim 1, characterized in that, The machine learning regression model can be any one of the following: K-nearest neighbor algorithm, decision tree, support vector machine regression, random forest, and extreme gradient boosting.

7. The method for estimating the comprehensive light efficiency index of winter wheat according to claim 1, characterized in that, The inputs for training a machine learning regression model include: The hyperparameters of the machine learning regression model were tuned using grid search through 5-fold cross-validation.

8. A comprehensive light efficiency index estimation system for winter wheat, characterized in that, include: The light efficiency comprehensive index calculation module is used to obtain various photosynthetic phenotypic parameters of winter wheat samples, as well as remote sensing images of winter wheat planting areas; The comprehensive index construction module is used to extract principal components from various photosynthetic phenotypic parameters of winter wheat samples using principal component analysis, and obtain the principal component scores of winter wheat samples on each principal component. The objective weights of each principal component were calculated using the CRITIC objective weighting method; and the comprehensive light efficiency index, which characterizes the photosynthetic efficiency of winter wheat samples, was obtained by weighting the objective weights and principal component scores. The model training module is used to calculate the vegetation index based on the remote sensing image and extract texture features from the remote sensing image; Using the vegetation index and texture features as input features, and the comprehensive light efficiency index as a supervised label, a machine learning regression model is trained to obtain a comprehensive light efficiency index estimation model for winter wheat, with the estimation result of the comprehensive light efficiency index for winter wheat as the output.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method for estimating the comprehensive light efficiency index of winter wheat as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is loaded by the processor, it is able to execute the steps of the method for estimating the comprehensive light efficacy index of winter wheat as described in any one of claims 1 to 7.

Images