Intelligent breeding evaluation method and system for saline-alkali soil salt-tolerant crops

Abstract

The invention discloses a saline-alkali soil salt-tolerant crop intelligent breeding evaluation method and system. The method comprises the steps of crop data acquisition, crop data improvement, crop breeding evaluation model establishment and crop intelligent breeding evaluation. The invention belongs to the field of breeding evaluation, and particularly relates to an intelligent breeding evaluation method and system for saline-alkali soil salt-tolerant crops. According to the scheme, environmental suitability grade difference regulation is introduced, and it is avoided that only a single salt-tolerant grade sample is generated; target function optimization is carried out by introducing a stress level difference coefficient, and a function is adjusted through a phenotypic sample generator feature, so that a generated sample is highly matched with an actual salt stress rule of the saline-alkali soil; evaluating contributions of different hierarchical features of the crop breeding data according to a feature importance function through a hierarchical attention mechanism, and eliminating redundant features; by introducing environmental edge loss, the distinction degree of adjacent salt tolerance grades in the extreme saline-alkaline environment is improved; the distance between adjacent salt tolerance grade characteristic centers is increased, so that the crop breeding evaluation effect is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of breeding and evaluation, and in particular to an intelligent breeding and evaluation method and system for salt-tolerant crops in saline-alkali land. Background Art

[0002] The salt-tolerant crop breeding and assessment method for saline-alkali soils is a means of grading or judging the salt tolerance of crops by measuring their growth indicators and environmental parameters in saline-alkali environments. However, general salt-tolerant crop breeding and assessment methods for saline-alkali soils suffer from imbalanced crop breeding data, resulting in the omission of superior varieties; poor environmental adaptability, and a disconnect between crop breeding samples and the actual phenotypic patterns of crops under salt stress, leading to low reliability in breeding and assessment; general salt-tolerant crop breeding and assessment methods for saline-alkali soils suffer from redundant interference from crop characteristics, resulting in subtle phenotypic differences between adjacent salt tolerance levels, and extreme saline-alkali environments exacerbating trait overlap, leading to misjudgment of highly salt-tolerant varieties and poor crop breeding and assessment results. Summary of the Invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides an intelligent breeding and evaluation method and system for salt-tolerant crops in saline-alkali land. The general breeding and evaluation method for salt-tolerant crops in saline-alkali land has the problems of imbalance in crop breeding data, omission of excellent varieties, poor environmental adaptability, and disconnection between crop breeding samples and actual crop phenotypic laws under salt stress, which leads to low reliability of breeding evaluation. This solution avoids generating only a single salt-tolerant grade sample by introducing environmental adaptability differential regulation and integer feature range regulation; optimizes the objective function by introducing a stress differential coefficient, so that the generated integer data is more in line with actual physiological laws; and makes the generated samples highly adaptable to the actual salt stress laws of saline-alkali land by combining the phenotypic sample generator feature adjustment function with the salt tolerance grade attenuation function. , thereby improving the reliability of subsequent crop breeding evaluation; in view of the problem that the general salt-tolerant crop breeding evaluation method in saline-alkali land has redundant interference of crop characteristics, the phenotypic differences of adjacent salt-tolerant grades are subtle, and the extreme saline-alkali environment will aggravate the overlap of characteristics, which will lead to the misjudgment of high-salt-tolerant varieties and poor crop breeding evaluation results, this scheme uses a hierarchical attention mechanism to evaluate the contribution of different levels of features in crop breeding data according to the feature importance function, combines dynamic thresholds to screen high-value features, eliminate redundant features, and strengthen core features related to salt tolerance; by introducing environmental edge loss, the discrimination of adjacent salt-tolerant grades in extreme saline-alkali environments is improved; constructing salt-tolerant grade difference loss, widening the distance between the feature centers of adjacent salt-tolerant grades, and reducing the misjudgment of excellent varieties; thereby improving the effect of crop breeding evaluation.

[0004] The technical solution adopted by the present invention is as follows: The present invention provides a method for intelligent breeding and evaluation of salt-tolerant crops in saline-alkali land, the method comprising the following steps:

[0005] Step S1: crop data collection;

[0006] Step S2: crop data completion;

[0007] Step S3: establishing a crop breeding evaluation model;

[0008] Step S4: Crop intelligent breeding evaluation.

[0009] Furthermore, in step S1, the crop data collection is to collect crop breeding data, including crop salt tolerance related phenotypic data and saline-alkali land environment related data; and mark the crop salt tolerance grade label; standardize the acquired data to construct an initial crop breeding data set.

[0010] Furthermore, in step S2, the crop data improvement specifically includes the following contents:

[0011] Step S21: Floating-point data enhancement: embed the salt tolerance grade label into random noise z, input the standardized floating-point data, and generate new floating-point samples through the traditional phenotypic sample generation network, which is recorded as ;in, and They are the generated new floating point crop breeding data and the corresponding salt tolerance grade labels;

[0012] Step S22: Integer data enhancement: embed the salt tolerance level label into random noise z, input standardized integer data; introduce environmental adaptability differential control, and add integer feature range control, expressed as: ;in, It is the loss of differential regulation of environmental adaptability; is the environmental adaptability differential control coefficient; is the expectation on a specific distribution; is the interpolated sample distribution; It is an interpolation sample between the real sample and the generated sample, and is an intermediate sample used for differential regulation of environmental adaptability; is the distribution of the interpolated samples; is the gradient of the phenotype authenticity discriminator with respect to the interpolated sample; is the L2 norm;

[0013] Step S23: Objective function optimization; for integer data enhancement, construct a authenticity discrimination loss that includes environmental adaptability differential regulation , expressed as: ;in, is the distribution of real integer samples; is an integer sample; is the authenticity discriminant output of the phenotype authenticity discriminator; is the distribution of noise z; It is a phenotypic sample generator that labels the noise z and salt tolerance level. Generated integer samples; construct salt tolerance classification loss , expressed as: ; Generate new integer samples denoted as ;in, is the salt tolerance classification output of the phenotypic authenticity discriminator for the real integer samples; It is the salt tolerance classification output of the generated integer samples by the phenotypic authenticity discriminator; the stress difference coefficient is introduced, the stress difference coefficient Expressed as: ; Define the phenotypic sample generator feature adjustment function, expressed as: ;in, is the salt tolerance attenuation function, ;e is the soil electrical conductivity, the value before standardization; and are floating-point phenotypic traits before and after adjustment, respectively; is the crop salt tolerance level; the final integer data enhancement loss function Expressed as: ; and It is to enhance the weight; It is the sum of the differences in electrical conductivity of different soils; is the mean value of the floating-point phenotypic characteristic in the real sample when the soil conductivity is e;

[0014] Step S24: Fusion data: Based on the data splicing operation, the generated floating point samples are merged with the real floating point samples, which are recorded as ; Merge the generated integer samples with the real integer samples, denoted as ; Integrate crop breeding data Input DNN classifier, take salt tolerance level as classification target, optimize network parameters through back propagation, if , then the crop breeding fusion data is discarded; the enhancement effect is verified by classification accuracy; among them, and They are the floating-point samples and integer samples of the synthesized crop breeding dataset; and is the corresponding label; set the classification accuracy threshold. If the classification accuracy is higher than the classification accuracy threshold, the crop breeding fusion data obtained at this time will be used as the final crop breeding fusion data.

[0015] Furthermore, in step S3, the crop breeding evaluation model is established based on the deep learning model; the model updates parameters through the gradient descent algorithm and adjusts initial parameters through the particle swarm algorithm; specifically, the following contents are included:

[0016] Step S31: Feature extraction network construction; input data-enhanced crop breeding fusion data , perform multi-view processing: global average pooling view, extract the overall distribution of features, and generate a global view vector , including the overall trend of tiller number of high salt-tolerant crops; global maximum pooling perspective, extracting local significant features, generating local perspective vector , including the extremely salt-tolerant individuals that still maintain 80% leaf water content under salt stress; through the fully connected layer and Sigmoid activation, the channel weights of the two types of features are obtained, the salt-tolerant related features are highlighted, the weights are multiplied by the original features and then summed to output the fusion feature F(M i );

[0017] Step S32: Adaptive feature selection; based on the fusion feature F(M) output from step S31 through the hierarchical attention mechanism i ) to evaluate the importance and select high-contribution features to enter the fusion stage; construct the feature importance function I(·), the formula used is: ;in, It is F(M i ) low k features at layer l; is the feature activation strength within the level; ; is the set of k-th eigenvalues ​​of all samples in the l-th layer; is the kth eigenvalue of the i-th sample in the l-th layer; and dynamic threshold screening is performed, the dynamic threshold Expressed as: For the k-th feature of the i-th sample in the l-th layer, if the feature importance function value is higher than the dynamic threshold, it is determined to be a high-contribution feature and retained; otherwise, it is removed; is the number of training rounds;

[0018] Step S33: Salt tolerance feature layering and fine fusion; perform salt tolerance feature layering fusion on the high-contribution features of each level screened out in step S32. After unifying the dimensions through the fully connected layer, progressive fusion is used to allow the low-level basic features to provide detailed support for the high-level derived features, avoiding feature association breaks. The formula used for adjusting the dimensions of the fully connected layer is: ; The formula used for progressive fusion is: ; The formula used for the final feature output is: ;in, is the feature of the i-th sample after adjustment in the l-th layer; ReLU(·) is the ReLU function; FC(·) is the fully connected layer; is the high contribution feature of the i-th sample in the l-th layer after screening in step S32; is the fusion feature of the i-th sample at the highest level; is global average pooling; is the final feature vector of the sample;

[0019] Step S34: Network classification; use the data-enhanced samples to calculate the characteristic center of each salt tolerance grade, calculate the Euclidean distance between the sample feature to be predicted and each characteristic center, and the grade with the closest distance is the prediction result; the formula used to calculate the characteristic center of the salt tolerance grade is: ; The formula used to calculate the classification probability is: ;in, and are the characteristic centers of salt tolerance grade of the jth and uth salt tolerance grades, respectively; K is the total number of samples of the jth salt tolerance grade; is the final feature vector of the i-th sample in the j-th class output by step S33; x is the sample to be predicted; is the final feature vector of the sample to be predicted outputted in step S33; is the Euclidean distance; is the probability that the sample to be predicted belongs to the jth salt tolerance level;

[0020] Step S35: Loss function design; construct basic loss, using cross entropy loss Based on, it is expressed as: ; Where Q is the total number of samples participating in the training; b is the salt tolerance level index; and are the true label of the i-th sample and the probability of predicting it to be grade c; construct the salt tolerance grade difference loss , the formula used is: ; Introduce environmental edge loss, construct environmental difference adaptive edge, for each sample i and each , defining dynamic edges , expressed as: ; Environmental edge loss Expressed as: ;in, It is the characteristic center of salt tolerance grade of the kth salt tolerance grade; is the true level of the i-th sample The salt tolerance grade characteristic center; It is the basic edge; is the environmental scaling factor; 、 and is the environmental average of the sample, the normalized mean; yes The corresponding environmental average; the final loss L is expressed as: ;in, and These are the salt tolerance grade characteristic centers of the b-th and b+1-th salt tolerance grades, respectively; is the loss weight coefficient.

[0021] Furthermore, in step S4, the intelligent crop breeding assessment is based on the salt tolerance grade assessment model established in step S3, combined with the environmental characteristics of saline-alkali land and breeding goals, to complete the breeding assessment of high salt tolerance crops, specifically including the following contents:

[0022] Step S41: Evaluate the salt tolerance potential of candidate breeding materials; collect full-dimensional data of existing breeding materials, standardize them and input them into the evaluation model of step S3, and output the predicted salt tolerance level of each material as the basis for salt tolerance potential evaluation;

[0023] Step S42: Directed screening of environmental adaptability of saline-alkali land; combining the environmental data of the target saline-alkali area and based on the salt tolerance level prediction results of step S41, make material-environment recommendations.

[0024] The present invention provides an intelligent breeding system for salt-tolerant crops in saline-alkali land, comprising a crop data collection module, a crop data improvement module, a crop breeding evaluation model establishment module, and a crop intelligent breeding module;

[0025] The crop data collection module collects crop breeding data and constructs an initial crop breeding data set;

[0026] The crop data improvement module enhances the initial crop breeding data set through floating-point data enhancement, introduces integer data enhancement for environmental adaptability differential regulation, and introduces a stress differential coefficient to optimize the objective function to obtain the final crop breeding fusion data;

[0027] The crop breeding evaluation model establishment module is based on crop breeding fusion data and a deep learning model. It establishes a crop breeding evaluation model through multi-perspective extraction, adaptive selection, and hierarchical fusion of salt tolerance features, combined with a salt tolerance grade difference loss function.

[0028] The crop intelligent breeding module performs crop intelligent breeding based on a crop breeding evaluation model and in combination with the environmental characteristics of the target saline-alkali land.

[0029] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0030] (1) In view of the problems in the general salt-tolerant crop breeding evaluation method for saline-alkali land, such as imbalance in crop breeding data, omission of excellent varieties, poor environmental adaptability, and disconnection between crop breeding samples and the actual phenotypic laws of crops under salt stress, which leads to low reliability of breeding evaluation, this scheme introduces environmental adaptability differential control and integer feature range control to avoid generating only single salt-tolerant grade samples; optimizes the objective function by introducing stress differential coefficient, so that the generated integer data is more consistent with the actual physiological laws; and adjusts the phenotypic sample generator feature function in combination with the salt tolerance grade attenuation function to make the generated samples highly adapted to the actual salt stress laws of saline-alkali land, thereby improving the reliability of subsequent crop breeding evaluation.

[0031] (2) In view of the problems that the general salt-tolerant crop breeding and evaluation methods in saline-alkali lands have redundant interference of crop characteristics, subtle phenotypic differences between adjacent salt-tolerant grades, and extreme saline-alkali environments will aggravate feature overlap, which will lead to misjudgment of high-salt-tolerant varieties and poor crop breeding and evaluation results, this scheme uses a hierarchical attention mechanism to evaluate the contribution of features at different levels of crop breeding data according to the feature importance function, combines dynamic thresholds to screen high-value features, eliminate redundant features, and strengthen core features related to salt tolerance; by introducing environmental edge loss, the discrimination of adjacent salt-tolerant grades in extreme saline-alkali environments is improved; by constructing salt-tolerant grade difference loss, the distance between the feature centers of adjacent salt-tolerant grades is widened, and the misjudgment of excellent varieties is reduced; thereby improving the effect of crop breeding and evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 A schematic flow chart of a method for intelligent breeding and evaluation of salt-tolerant crops in saline-alkali land provided by the present invention;

[0033] Figure 2 This is a schematic diagram of an intelligent breeding system for salt-tolerant crops in saline-alkali land provided by the present invention.

[0034] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0036] In the description of the present invention, it should be understood that terms such as "up", "down", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0037] Example 1, see Figure 1 The present invention provides a method for intelligent breeding and evaluation of salt-tolerant crops in saline-alkali land, which comprises the following steps:

[0038] Step S1: Crop data collection: Collect crop breeding data and construct an initial crop breeding data set;

[0039] Step S2: Crop data improvement: The initial crop breeding dataset is enhanced by floating-point data, integer data enhancement is introduced for environmental adaptability differential regulation, and the stress differential coefficient is introduced to optimize the objective function to obtain the final crop breeding fusion data;

[0040] Step S3: Establishing a crop breeding evaluation model; Based on the crop breeding fusion data and the deep learning model, the crop breeding evaluation model is established through multi-perspective extraction, adaptive selection, and hierarchical fusion of salt tolerance features, combined with the salt tolerance grade difference loss function;

[0041] Step S4: Crop intelligent breeding evaluation: Based on the crop breeding evaluation model, crop intelligent breeding is carried out in combination with the environmental characteristics of the target saline-alkali land.

[0042] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, crop data collection is to collect crop breeding data, including crop salt tolerance related phenotypic data and saline-alkali land environment related data; and mark the crop salt tolerance grade label; the crop salt tolerance related phenotypic data includes integer phenotypic data: plant height, effective tiller number, fruit set rate, survival days, and floating-point phenotypic data: leaf relative water content, chlorophyll content, root length, and salt tolerance index; the saline-alkali land environment related data includes integer environmental data: soil salinity level, irrigation times, and floating-point environmental data: soil electrical conductivity, soil pH value; the crop salt tolerance grade label is 1-5, and the higher the grade, the better the salt tolerance; the acquired data is standardized to construct an initial crop breeding data set;

[0043] Data acquisition: Plant height, random sampling, measuring the vertical distance from the ground to the top of the main stem, and excluding deformed stems caused by salt damage during measurement;

[0044] Tiller number: randomly select 10 clusters of plants along an S-shaped route, retain 3 main stems in each cluster, and count the number of tillers that can produce ears and fruits in each plant;

[0045] The fruit setting rate was calculated by randomly sampling each plot, separating the full seeds from the empty seeds after threshing.

[0046] Chlorophyll content: on the 20th day after salt stress, three healthy leaves were selected from each plant, and three sites on each leaf were measured, avoiding the veins, and the average value was taken;

[0047] Root length: randomly sample each plot, dig out the whole root system, rinse it with running water, and spread it on a transparent glass plate to obtain the total root length;

[0048] Salt tolerance index: the soil EC value was set to 8 mS / cm, and the biomass ratio of the treatment group to the control group with a soil EC value of 2 mS / cm was calculated to obtain the salt tolerance index;

[0049] Soil conductivity: insert the probe into the soil to a depth of 20 cm, set a measuring point every 10 m, and record the temperature at the same time, and calibrate it to the standard value of 25°C;

[0050] Soil pH value: Use a strong alkali-resistant electrode to insert 5 cm into the soil, read the value after it stabilizes, repeat 3 times for each measurement point, and take the average value;

[0051] Soil salinity level, soil EC value corresponds to: Level 1, EC < 2mS / cm; Level 2, 2-4mS / cm; Level 3, 4-6mS / cm; Level 4, 6-8mS / cm; Level 5, EC > 8mS / cm; sampling is done using the plum blossom 5-point method, and the EC value is measured after mixing to represent the regional level.

[0052] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, the crop data is improved because most crops in saline-alkali land and high-salt environments have weak salt tolerance, while the number of varieties with strong salt tolerance is small and the artificial cultivation cycle is long, resulting in such samples accounting for a very low proportion in the data set. Therefore, by expanding the number of small samples, it is possible to avoid missing out on excellent varieties. Specifically, the following contents are included:

[0053] Step S21: Floating-point data enhancement: embed the salt tolerance grade label into random noise z, input the standardized floating-point data, and generate new floating-point samples through the traditional phenotypic sample generation network, which is recorded as ;in, and They are the generated new floating point crop breeding data and the corresponding salt tolerance grade labels;

[0054] Step S22: Integer data enhancement. To address the issues of strong discreteness of integer data in crop breeding data and easy distortion of traditional phenotypic sample generation networks, an improved phenotypic sample generation network with environmental adaptability differential regulation is adopted. The salt tolerance grade label is embedded in random noise z and the standardized integer data is input. Environmental adaptability differential regulation is introduced to avoid generating only single salt tolerance grade samples, and integer feature range regulation is added, which is expressed as: ;in, It is the loss of differential regulation of environmental adaptability; is the environmental adaptability differential control coefficient; is the expectation on a specific distribution; is the interpolated sample distribution; It is an interpolation sample between the real sample and the generated sample, and is an intermediate sample used for differential regulation of environmental adaptability; is the distribution of the interpolated samples; is the gradient of the phenotype authenticity discriminator with respect to the interpolated sample; is the L2 norm;

[0055] Step S23: Objective function optimization; for integer data enhancement, construct a authenticity discrimination loss that includes environmental adaptability differential regulation , expressed as: ;in, is the distribution of real integer samples; is an integer sample; is the authenticity discriminant output of the phenotype authenticity discriminator; is the distribution of noise z; It is a phenotypic sample generator that labels the noise z and salt tolerance level. Generated integer samples; construct salt tolerance classification loss , expressed as: ; Generate new integer samples denoted as ;in, is the salt tolerance classification output of the phenotypic authenticity discriminator for the real integer samples; It is the salt tolerance classification output of the phenotypic authenticity discriminator for the generated integer samples; the stress differential coefficient is introduced to adjust the characteristic distribution of the phenotypic sample generator according to the soil EC value, so that the generated samples are adapted to the phenotypic law of the specific salt differential. Expressed as: ; Define the phenotypic sample generator feature adjustment function, expressed as: ;in, is the salt tolerance attenuation function, ;e is the soil electrical conductivity, the value before standardization; and are floating-point phenotypic traits before and after adjustment, respectively; is the crop salt tolerance level; the final integer data enhancement loss function Expressed as: ; and is the enhanced weight; the phenotype authenticity discriminator is a single network dual output structure: the same phenotype authenticity discriminator outputs two results at the same time, and ; It is the sum of the differences in electrical conductivity of different soils; is the mean value of the floating-point phenotypic characteristic in the real sample when the soil conductivity is e;

[0056] By strengthening the matching between floating-point phenotypic characteristics and salinity differences in saline-alkali land, a physiologically reasonable basic regulation is provided for the generation of integer-type characteristics; and the generated samples are prevented from being out of line with the actual saline-alkali land stress law.

[0057] Step S24: Fusion data: Based on the data splicing operation, the pd.concat(·) function of the pandas library is used to merge the generated floating-point samples with the real floating-point samples, which is recorded as ; Merge the generated integer samples with the real integer samples, denoted as ; Integrate crop breeding data Input DNN classifier, take salt tolerance level as classification target, optimize network parameters through back propagation, if , then the crop breeding fusion data is discarded; the enhancement effect is verified by classification accuracy; among them, and They are the floating-point samples and integer samples of the synthesized crop breeding dataset; and is the corresponding label; set the classification accuracy threshold. If the classification accuracy is higher than the classification accuracy threshold, the crop breeding fusion data obtained at this time will be used as the final crop breeding fusion data.

[0058] By performing the above operations, we can address the problems of imbalance in crop breeding data and missed judgment of excellent varieties in general salt-tolerant crop breeding evaluation methods for saline-alkali land; poor environmental adaptability and disconnection between crop breeding samples and actual crop phenotypic laws under salt stress, which in turn leads to low reliability of breeding evaluation. This scheme introduces environmental adaptability differential control and integer feature range control to avoid generating only single salt-tolerant grade samples; optimizes the objective function by introducing the stress differential coefficient, so that the generated integer data is more in line with actual physiological laws; and adjusts the phenotypic sample generator feature function in combination with the salt tolerance grade attenuation function to make the generated samples highly adapted to the actual salt stress laws of saline-alkali land, thereby improving the reliability of subsequent crop breeding evaluation.

[0059] Example 4, see Figure 1 This embodiment is based on the above embodiment. In step S3, a crop breeding evaluation model is established based on a deep learning model. Salt tolerance-related features are captured through multi-perspective extraction of salt tolerance features, adaptive selection, and hierarchical fusion of salt tolerance features. Combined with the salt tolerance grade difference loss function, it includes two stages, training and testing, and finally outputs the crop salt tolerance evaluation grade to establish a crop breeding evaluation model. This solves the problems of complex phenotypic-environmental characteristics of salt-tolerant crops in saline-alkali land, few salt tolerance grade samples, and difficulty in classifying adjacent salt tolerance grades. The model updates parameters through a gradient descent algorithm and adjusts initial parameters through a particle swarm algorithm. Specifically, it includes the following contents:

[0060] Step S31: Feature extraction network construction; input data-enhanced crop breeding fusion data , perform multi-view processing: global average pooling view, extract the overall distribution of features, and generate a global view vector , including the overall trend of tiller number of high salt-tolerant crops; global maximum pooling perspective, extracting local significant features, generating local perspective vector , including the extremely salt-tolerant individuals that still maintain 80% leaf water content under salt stress; through the fully connected layer and Sigmoid activation, the channel weights of the two types of features are obtained, the salt-tolerant related features are highlighted, the weights are multiplied by the original features and then summed to output the fusion feature F(M i ); Through multi-perspective pooling, the overall distribution pattern and local key features are captured respectively, and the weights are dynamically strengthened to strengthen the core features of salt tolerance and filter out irrelevant features;

[0061] Global view vector ; Local view vector ; Where N is the number of features in the c-th channel and k is the feature index; is the kth eigenvalue of the cth channel; and are the first and Nth eigenvalues ​​of the cth channel respectively; the channel weight calculation formula is expressed as: ; ; Fusion features ;in, is the Sigmoid function; and is a learnable network parameter from a global perspective; and are the channel weights of the global perspective and the local perspective respectively; and are learnable network parameters of the local perspective;

[0062] Step S32: Adaptive feature selection; based on the fusion feature F(M) output from step S31 through the hierarchical attention mechanism i) to evaluate the importance and select high-contribution features to enter the fusion stage; construct the feature importance function I(·), the formula used is: ;in, It is F(M i ) low k features at layer l; is the feature activation strength within the level; ; is the set of k-th eigenvalues ​​of all samples in the l-th layer; is the kth eigenvalue of the i-th sample in the l-th layer; and dynamic threshold screening is performed, and the threshold adjusted with the training round is used to retain high contribution features; dynamic threshold Expressed as: For the k-th feature of the i-th sample in the l-th layer, if the feature importance function value is higher than the dynamic threshold, it is determined to be a high-contribution feature and retained; otherwise, it is removed; is the number of training rounds;

[0063] Different levels of characteristics have different values ​​for salt tolerance classification; the shallowest characteristic l=1 is the original phenotypic layer, corresponding to the basic phenotypic characteristics including plant height; l=2 is the environmental characteristic layer, which is a single type of original environmental characteristic, corresponding to the basic environmental characteristics including soil salinity; l=3 is the primary fusion layer, which is the preliminary fusion characteristic of phenotype-environment, corresponding to the preliminary phenotype-environment interaction characteristics including tiller number × pH value; the deepest characteristic l=4 is the advanced fusion layer, corresponding to the deep phenotypic-environment interaction characteristics including salinity × leaf water content;

[0064] Step S33: Salt tolerance feature stratification and fine fusion; The high-contribution features of each level screened out in step S32 are fused in a stratified manner for salt tolerance. Through the three-step process of dimension unification, hierarchical association, and feature concentration, the final feature vector that takes into account both the details of the basic indicators and the laws of the derived indicators is output; the classification of salt-tolerant crops needs to rely on both the basic phenotype of the individual plant and the derived laws of the group phenotype-environment, but the feature dimensions of different levels vary significantly; after unifying the dimensions through the fully connected layer, progressive fusion is used to allow the low-level basic features to provide detailed support for the high-level derived features to avoid feature association breaks; the formula used to adjust the dimensions of the fully connected layer is: ; The formula used for progressive fusion is: ; The formula used for the final feature output is: ;in, is the feature of the i-th sample after adjustment in the l-th layer; ReLU(·) is the ReLU function; FC(·) is the fully connected layer; is the high contribution feature of the i-th sample in the l-th layer after screening in step S32; is the fusion feature of the i-th sample at the highest level; is global average pooling; is the final feature vector of the sample;

[0065] Step S34: Network classification; use the data-enhanced samples to calculate the characteristic center of each salt tolerance grade, calculate the Euclidean distance between the sample feature to be predicted and each characteristic center, and the grade with the closest distance is the prediction result; the formula used to calculate the characteristic center of the salt tolerance grade is: ; The formula used to calculate the classification probability is: ;in, and are the characteristic centers of salt tolerance grade of the jth and uth salt tolerance grades, respectively; K is the total number of samples of the jth salt tolerance grade; is the final feature vector of the i-th sample in the j-th class output by step S33; x is the sample to be predicted; is the final feature vector of the sample to be predicted outputted in step S33; is the Euclidean distance; is the probability that the sample to be predicted belongs to the jth salt tolerance level. In the breeding of salt-tolerant crops, there are few samples of high-salt-tolerant varieties. The salt tolerance level information is concentrated by the salt tolerance level feature center to build a stable center. By projecting the samples into a feature space with controllable distance, the dependence on large-scale labeled data is reduced.

[0066] Step S35: Loss function design; construct basic loss, using cross entropy loss Based on, it is expressed as: ; Where Q is the total number of samples participating in the training; b is the salt tolerance level index; and are the true label of the i-th sample and the probability of predicting it to be level c; the true label is one-hot encoded, and when sample i is level 5, then , ; Construct salt tolerance level difference loss , used to strengthen spatial separation and increase the distance between adjacent salt tolerance level centers. The formula used is: ; Introduce environmental edge loss, construct environmental difference adaptive edge, for each sample i and each , defining dynamic edges , expressed as: ; Environmental edge loss Expressed as: ;in, It is the characteristic center of salt tolerance grade of the kth salt tolerance grade; is the true level of the i-th sample The salt tolerance grade characteristic center; It is the basic edge; is the environmental scaling factor; 、 and is the environmental average of the sample, the normalized mean; yes The corresponding environmental average value; adjacent salt tolerance levels are difficult to distinguish in extreme environments. Automatically enlarge the edge and significantly improve the resolution of high-salt areas; the final loss L is expressed as: ;in, and These are the salt tolerance grade characteristic centers of the b-th and b+1-th salt tolerance grades, respectively; is the loss weight coefficient.

[0067] By performing the above operations, we can address the problems of redundant interference of crop characteristics in general salt-tolerant crop breeding and evaluation methods in saline-alkali land, subtle phenotypic differences between adjacent salt-tolerant grades, and the fact that extreme saline-alkali environments will aggravate feature overlap, which in turn leads to misjudgment of high-salt-tolerant varieties and poor crop breeding and evaluation results. This scheme uses a hierarchical attention mechanism to evaluate the contribution of features at different levels of crop breeding data according to the feature importance function, combines dynamic thresholds to screen high-value features, eliminate redundant features, and strengthen core features related to salt tolerance; by introducing environmental edge loss, the discrimination of adjacent salt-tolerant grades in extreme saline-alkali environments is improved; by constructing salt-tolerant grade difference loss, the distance between feature centers of adjacent salt-tolerant grades is increased, and misjudgment of excellent varieties is reduced; thereby improving the effect of crop breeding and evaluation.

[0068] Example 5, see Figure 1 This embodiment is based on the above embodiment. In step S4, the intelligent crop breeding assessment is based on the salt tolerance assessment model established in step S3. The assessment is combined with the environmental characteristics of saline-alkali land and the breeding goals to complete the breeding assessment of high salt tolerance crops. Specifically, it includes the following contents:

[0069] Step S41: Evaluate the salt tolerance potential of candidate breeding materials; collect full-dimensional data of existing breeding materials, standardize them and input them into the evaluation model of step S3, and output the predicted salt tolerance level of each material as the basis for salt tolerance potential evaluation;

[0070] Step S42: Directed screening of adaptability to saline-alkali land environments; combining environmental data of the target saline-alkali area and based on the salt tolerance level prediction results of step S41, material-environment recommendations are made; when the target environment is a moderate to severe stress area with a salinity of 4-6 mS / cm, breeding materials of level 4 are selected as candidates for the main varieties; when the target environment is a severe stress area with a salinity >6 mS / cm, breeding materials of level 5 are selected as candidates for the main varieties and as salt tolerance gene donor parents for hybrid breeding.

[0071] Example 6, see Figure 2This embodiment is based on the above embodiment. The present invention provides an intelligent breeding system for salt-tolerant crops in saline-alkali land, including a crop data collection module, a crop data improvement module, a crop breeding evaluation model establishment module and a crop intelligent breeding module;

[0072] The crop data collection module collects crop breeding data and constructs an initial crop breeding data set;

[0073] The crop data improvement module enhances the initial crop breeding data set through floating-point data enhancement, introduces integer data enhancement for environmental adaptability differential regulation, and introduces a stress differential coefficient to optimize the objective function to obtain the final crop breeding fusion data;

[0074] The crop breeding evaluation model establishment module is based on crop breeding fusion data and a deep learning model. It establishes a crop breeding evaluation model through multi-perspective extraction, adaptive selection, and hierarchical fusion of salt tolerance features, combined with a salt tolerance grade difference loss function.

[0075] The crop intelligent breeding module performs crop intelligent breeding based on a crop breeding evaluation model and in combination with the environmental characteristics of the target saline-alkali land.

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

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

[0078] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. A method for intelligent breeding and evaluation of salt-tolerant crops in saline-alkali soil, characterized by: The method comprises the following steps: Step S1: Crop data collection: Collect crop breeding data and construct an initial crop breeding data set; Step S2: Crop data improvement: The initial crop breeding dataset is enhanced by floating-point data, integer data enhancement is introduced for environmental adaptability differential regulation, and the stress differential coefficient is introduced to optimize the objective function to obtain the final crop breeding fusion data; Step S3: Establishing a crop breeding evaluation model; Based on the crop breeding fusion data and the deep learning model, the crop breeding evaluation model is established through multi-perspective extraction, adaptive selection, and hierarchical fusion of salt tolerance features, combined with the salt tolerance grade difference loss function; Step S4: Crop intelligent breeding evaluation: Based on the crop breeding evaluation model, crop intelligent breeding is carried out in combination with the environmental characteristics of the target saline-alkali land.

2. The method for intelligent breeding and evaluation of salt-tolerant crops in saline-alkali soil according to claim 1, characterized in that: In step S2, the crop data improvement specifically includes the following: Step S21: Floating-point data enhancement: embed the salt tolerance grade label into random noise z, input the standardized floating-point data, and generate new floating-point samples through the traditional phenotypic sample generation network, which is recorded as ;in, and They are the generated new floating point crop breeding data and the corresponding salt tolerance grade labels; Step S22: Integer data enhancement: embed the salt tolerance level label into random noise z, input standardized integer data; introduce environmental adaptability differential control, and add integer feature range control, expressed as: ;in, It is the loss of differential regulation of environmental adaptability; is the environmental adaptability differential control coefficient; is the expectation on a specific distribution; is the interpolated sample distribution; It is an interpolation sample between the real sample and the generated sample, and is an intermediate sample used for differential regulation of environmental adaptability; is the distribution of the interpolated samples; is the gradient of the phenotype authenticity discriminator with respect to the interpolated sample; is the L2 norm; Step S23: Optimizing the objective function; Step S24: Fusion data: Based on the data splicing operation, the generated floating point samples are merged with the real floating point samples, which are recorded as ; Merge the generated integer samples with the real integer samples, denoted as ; Integrate crop breeding data Input DNN classifier, take salt tolerance level as classification target, optimize network parameters through back propagation, if , then the crop breeding fusion data is discarded; the enhancement effect is verified by classification accuracy; among them, and They are floating-point samples and integer samples of the synthesized crop breeding dataset; and is the corresponding label; set the classification accuracy threshold. If the classification accuracy is higher than the classification accuracy threshold, the crop breeding fusion data obtained at this time will be used as the final crop breeding fusion data.

3. The intelligent breeding and evaluation method for salt-tolerant crops in saline-alkali soil according to claim 2, characterized in that: In step S2, the objective function optimization is to enhance integer data and construct a authenticity discrimination loss including environmental adaptability differential regulation , expressed as: ;in, is the distribution of real integer samples; is an integer sample; is the authenticity discriminant output of the phenotype authenticity discriminator; is the distribution of noise z; It is a phenotypic sample generator that labels the noise z and salt tolerance level. Generated integer samples; construct salt tolerance classification loss , expressed as: ; Generate new integer samples denoted as ;in, is the salt tolerance classification output of the phenotypic authenticity discriminator for the real integer samples; It is the salt tolerance classification output of the generated integer samples by the phenotypic authenticity discriminator; the stress difference coefficient is introduced, the stress difference coefficient Expressed as: ; Define the phenotypic sample generator feature adjustment function, expressed as: ;in, is the salt tolerance attenuation function, ;e is the soil electrical conductivity, the value before standardization; and are floating-point phenotypic traits before and after adjustment, respectively; is the crop salt tolerance level; the final integer data enhancement loss function Expressed as: ; and It is to enhance the weight; It is the sum of the differences in electrical conductivity of different soils; is the mean value of the floating-point phenotypic characteristic in the real sample when the soil conductivity is e.

4. The method for intelligent breeding and evaluation of salt-tolerant crops in saline-alkali soil according to claim 3, characterized in that: In step S3, the crop breeding evaluation model is established based on the deep learning model. The model updates parameters using a gradient descent algorithm and adjusts initial parameters using a particle swarm algorithm. Specifically, the following contents are included: Step S31: Feature extraction network construction; input data-enhanced crop breeding fusion data , perform multi-view processing: global average pooling view, extract the overall distribution of features, and generate a global view vector , including the overall trend of tiller number of high salt-tolerant crops; global maximum pooling perspective, extracting local significant features, generating local perspective vector , including the extremely salt-tolerant individuals that still maintain 80% leaf water content under salt stress; through the fully connected layer and Sigmoid activation, the channel weights of the two types of features are obtained, the salt-tolerant related features are highlighted, the weights are multiplied by the original features and then summed to output the fusion feature F(M i ); Step S32: Adaptive feature selection; based on the fusion feature F(M) output from step S31 through the hierarchical attention mechanism i ) to evaluate the importance and select high-contribution features to enter the fusion stage; construct the feature importance function I(·), the formula used is: ;in, It is F(M i ) low k features at layer l; is the feature activation strength within the level; ; is the set of k-th eigenvalues ​​of all samples in the l-th layer; is the kth eigenvalue of the i-th sample in the l-th layer; and dynamic threshold screening is performed, the dynamic threshold Expressed as: For the k-th feature of the i-th sample in the l-th layer, if the feature importance function value is higher than the dynamic threshold, it is determined to be a high-contribution feature and retained; otherwise, it is removed; is the training round; Step S33: Salt tolerance feature layering and fine fusion; perform salt tolerance feature layering fusion on the high-contribution features of each level screened out in step S32. After unifying the dimensions through the fully connected layer, progressive fusion is used to allow the low-level basic features to provide detailed support for the high-level derived features, avoiding feature association breaks. The formula used for adjusting the dimensions of the fully connected layer is: ; The formula used for progressive fusion is: ; The formula used for the final feature output is: ;in, is the feature of the i-th sample after adjustment in the l-th layer; ReLU(·) is the ReLU function; FC(·) is the fully connected layer; is the high contribution feature of the i-th sample in the l-th layer after screening in step S32; is the fusion feature of the i-th sample at the highest level; is global average pooling; is the final feature vector of the sample; Step S34: network classification; Step S35: loss function design.

5. The method for intelligent breeding and evaluation of salt-tolerant crops in saline-alkali soil according to claim 4, characterized in that: In step S3, the network classification is to calculate the characteristic center of each salt tolerance grade using the data-enhanced samples, and calculate the Euclidean distance between the sample features to be predicted and the characteristic centers of each type. The grade with the closest distance is the prediction result. The formula used to calculate the characteristic center of the salt tolerance grade is: ; The formula used to calculate the classification probability is: ;in, and are the characteristic centers of salt tolerance grade of the jth and uth salt tolerance grades, respectively; K is the total number of samples of the jth salt tolerance grade; is the final feature vector of the i-th sample in the j-th class output by step S33; x is the sample to be predicted; is the final feature vector of the sample to be predicted outputted in step S33; is the Euclidean distance; is the probability that the sample to be predicted belongs to the jth salt tolerance level.

6. The method for intelligent breeding and evaluation of salt-tolerant crops in saline-alkali soil according to claim 5, characterized in that: In step S3, the loss function is designed to construct a basic loss with cross entropy loss Based on, it is expressed as: ; Where Q is the total number of samples participating in the training; b is the salt tolerance level index; and are the true label of the i-th sample and the probability of predicting it to be grade c; construct the salt tolerance grade difference loss , the formula used is: ; Introduce environmental edge loss, construct environmental difference adaptive edge, for each sample i and each , defining dynamic edges , expressed as: ; Environmental edge loss Expressed as: ;in, It is the characteristic center of salt tolerance grade of the kth salt tolerance grade; is the true level of the i-th sample The salt tolerance grade characteristic center; It is the basic edge; is the environmental scaling factor; 、 and is the environmental average of the sample, the normalized mean; yes The corresponding environmental average; the final loss L is expressed as: ;in, and These are the salt tolerance grade characteristic centers of the b-th and b+1-th salt tolerance grades, respectively; is the loss weight coefficient.

7. The method for intelligent breeding and evaluation of salt-tolerant crops in saline-alkali soil according to claim 6, characterized in that: In step S4, the intelligent crop breeding assessment is based on the salt tolerance grade assessment model established in step S3, combined with the environmental characteristics of saline-alkali land and breeding goals, to complete the breeding assessment of high salt tolerance crops, specifically including the following contents: Step S41: evaluating the salt tolerance potential of candidate breeding materials; Full-dimensional data collection is performed on existing breeding materials, and after standardization, it is input into the evaluation model of step S3 to output the salt tolerance grade prediction results of each material as the basis for salt tolerance potential evaluation; Step S42: Directed screening of environmental adaptability of saline-alkali land; combining the environmental data of the target saline-alkali area and based on the salt tolerance level prediction results of step S41, make material-environment recommendations.

8. The method for intelligent breeding and evaluation of salt-tolerant crops in saline-alkali soil according to claim 7, characterized in that: In step S1, the crop data collection is to collect crop breeding data, including crop salt tolerance related phenotypic data and saline-alkali land environment related data; The crops are labeled with salt tolerance levels; the acquired data are standardized to construct an initial crop breeding dataset.

9. A system for intelligent breeding and evaluation of salt-tolerant crops in saline-alkali soil, for implementing the method for intelligent breeding and evaluation of salt-tolerant crops in saline-alkali soil according to any one of claims 1 to 8, characterized in that: It includes crop data collection module, crop data improvement module, crop breeding evaluation model establishment module and crop intelligent breeding module; The crop data collection module collects crop breeding data and constructs an initial crop breeding data set; The crop data improvement module enhances the initial crop breeding data set through floating-point data enhancement, introduces integer data enhancement for environmental adaptability differential regulation, and introduces a stress differential coefficient to optimize the objective function to obtain the final crop breeding fusion data; The crop breeding evaluation model establishment module is based on crop breeding fusion data and a deep learning model. It establishes a crop breeding evaluation model through multi-perspective extraction, adaptive selection, and hierarchical fusion of salt tolerance features, combined with a salt tolerance grade difference loss function. The crop intelligent breeding module performs crop intelligent breeding based on a crop breeding evaluation model and in combination with the environmental characteristics of the target saline-alkali land.

Images