A method for predicting salt tolerance of rice

Abstract

The application discloses a rice salt tolerance prediction method, and relates to the technical field of agricultural information processing, and comprises the following steps: for each salt stress intensity interval, a plurality of yield salt tolerance prediction sub-models are constructed by using corresponding seedling salt tolerance characteristics and rice salt tolerance indexes; environment similarity indexes of test environment information and historical test environment information are calculated; based on the environment similarity indexes, the applicability of the yield salt tolerance prediction sub-models is determined, scene control parameters are generated, the seedling salt tolerance characteristics are input into the corresponding yield salt tolerance prediction sub-models according to the scene control parameters, and yield salt tolerance index prediction values are output; the yield salt tolerance index prediction values are assigned corresponding prediction credibility marks through credibility evaluation, and output decision of the yield salt tolerance index prediction values is performed according to the prediction credibility marks, and a salt tolerance prediction report is output. Through the construction of the plurality of prediction sub-models and the output decision mechanism, the prediction stability and applicability under different salt stress scenes are improved.

Description

Technical Field

[0001] This invention relates to the field of agricultural information processing technology, and in particular to a method for predicting the salt tolerance of rice. Background Technology

[0002] With the continuous expansion of saline-alkali arable land globally and the increasing demands for stable food supply in agricultural production, the evaluation and screening of rice salt tolerance has become a key technical link in crop genetic improvement and saline-alkali land utilization. Existing research shows that the growth inhibition, yield reduction, and physiological metabolic disorders of rice under salt stress conditions have obvious stage-specific and environmentally dependent characteristics. Therefore, how to reliably predict the salt tolerance of rice in the early stages of growth is of great significance for breeding selection, cultivation management, and variety layout. To achieve this goal, related technologies have undergone a development process from field phenotypic observation to multi-index comprehensive evaluation, and from manual experience judgment to data-driven analysis.

[0003] However, existing technologies still have certain shortcomings. In the process of model building, existing prediction methods often incorporate data under different salt stress conditions into a single prediction model, failing to fully consider the impact of differences in salt stress intensity on the rice response mechanism. This results in insufficient stability of the model when applied across environments. When there is a deviation between the experimental environment and the training environment, the prediction results are prone to inaccuracy. Most existing technologies focus on outputting a single predicted value or grade result, lacking an evaluation mechanism for the reliability of the prediction results. They cannot effectively distinguish between stable prediction results and prediction results with high uncertainty, thereby increasing the risk of misjudgment in breeding decisions or production management. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for predicting rice salt tolerance to solve the problems of poor stability in cross-environment prediction and low reliability of application decisions.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a method for predicting rice salt tolerance, comprising: collecting seedling growth data of rice under salt stress test conditions and performing time-series analysis to form seedling agronomic trait data, while recording corresponding test environment information; extracting seedling salt tolerance characteristics from the seedling agronomic trait data, dividing the salt stress intensity intervals of the salt stress test conditions according to the test environment information, and calculating the rice salt tolerance index in each salt stress intensity interval; for each salt stress intensity interval, constructing multiple yield salt tolerance prediction sub-models using the corresponding seedling salt tolerance characteristics and rice salt tolerance index, and calculating the environmental similarity index between the test environment information and historical test environment information; determining the applicability of the yield salt tolerance prediction sub-models based on the environmental similarity index, generating scenario control parameters, inputting the seedling salt tolerance characteristics into the corresponding yield salt tolerance prediction sub-models according to the scenario control parameters, and outputting the predicted yield salt tolerance index value; assigning corresponding prediction credibility labels to the predicted yield salt tolerance index value through credibility assessment, and performing output adjudication on the predicted yield salt tolerance index value according to the prediction credibility labels, and outputting a salt tolerance prediction report.

[0008] As a preferred embodiment of the method for predicting rice salt tolerance according to the present invention, the specific steps for generating seedling agronomic trait data and simultaneously recording corresponding experimental environment information are as follows.

[0009] Under salt stress test conditions, seedling growth status data were organized and divided into stages according to the seedling growth time sequence to form seedling time sequence data;

[0010] The growth status of rice seedlings was structurally summarized based on the seedling time series data to form agronomic trait data of the seedling stage, while the corresponding experimental environment information was recorded.

[0011] As a preferred embodiment of the method for predicting rice salt tolerance according to the present invention, the specific steps for extracting the salt tolerance characteristics of seedling agronomic traits from seedling stage data are as follows:

[0012] The variation range, rate of change, and stage differences of agronomic traits during the seedling growth stage are statistically analyzed to generate response degree indicators.

[0013] Based on the response degree index, correlation screening and structured mapping were performed on the agronomic trait data of the seedling stage to generate salt tolerance characteristics of the seedling stage.

[0014] As a preferred embodiment of the method for predicting rice salt tolerance according to the present invention, the specific steps of dividing the salt stress intensity range of the salt stress test conditions according to the experimental environment information and calculating the rice salt tolerance index in each salt stress intensity range are as follows.

[0015] Extract salt stress concentration parameters, salt stress duration parameters, and salt stress application method parameters from the experimental environment information to form environmental parameters;

[0016] Based on environmental parameters, a partitioning mapping process is performed on the salt stress test conditions to divide the salt stress test conditions into multiple salt stress intensity intervals and assign corresponding salt stress intensity interval identifiers.

[0017] Based on the seedling agronomic traits and corresponding seedling salt tolerance characteristics within each salt stress intensity interval, the interval salt tolerance index is calculated. The interval salt tolerance indices corresponding to each salt stress intensity interval are then stored in a structured manner to form the rice salt tolerance index.

[0018] As a preferred embodiment of the method for predicting rice salt tolerance according to the present invention, the specific steps for constructing multiple yield and salt tolerance prediction sub-models are as follows:

[0019] We statistically analyzed the time-series derived indices corresponding to each salt tolerance feature dimension at each seedling stage, and performed window aggregation and vector concatenation to construct a seedling stage response path representation layer.

[0020] Extract interval statistics from the rice salt tolerance index corresponding to the salt stress intensity interval markers, and perform vector splicing and conditional modulation processing on the salt stress intensity interval markers and interval statistics to construct a salt intensity conditional gating layer.

[0021] Within each salt stress intensity range, by establishing and solidifying the mapping rules from time-derived indices to rice salt tolerance indices, multiple yield salt tolerance prediction sub-models corresponding one-to-one with each salt stress intensity range are constructed.

[0022] As a preferred embodiment of the method for predicting rice salt tolerance according to the present invention, the specific steps for calculating the environmental similarity index between experimental environment information and historical experimental environment information are as follows:

[0023] The experimental environment information and the historical experimental environment information corresponding to each salt stress intensity range are parameterized to form an environmental input feature vector and a historical environmental feature vector.

[0024] Based on the degree of difference between the environmental input feature vector and the historical environmental feature vector, the corresponding environmental similarity index is calculated.

[0025] In a preferred embodiment of the method for predicting rice salt tolerance according to the present invention, the specific steps for generating the scenario control parameters are as follows:

[0026] Based on the environmental similarity index, the matching degree of each yield salt tolerance prediction sub-model is ranked to form a set of candidate prediction sub-models;

[0027] Identify the similarity concentration interval and similarity decay interval in the candidate prediction sub-model set, perform interval mapping processing, and generate scene control parameters.

[0028] As a preferred embodiment of the method for predicting rice salt tolerance according to the present invention, the specific steps for outputting the predicted yield salt tolerance index are as follows:

[0029] The seedling response path representation layer performs temporal derivation processing on the salt tolerance characteristics of the seedling stage to generate the seedling response path representation vector.

[0030] The salt intensity condition gating layer performs interval mapping processing on the seedling response path representation vector based on the scenario control parameters, generating a controllable seedling response path vector.

[0031] Calculate the predicted yield salt tolerance index of the controlled seedling stage response path vector under the current salt stress test conditions.

[0032] In a preferred embodiment of the method for predicting rice salt tolerance according to the present invention, the specific steps for assigning corresponding prediction confidence labels to the predicted yield salt tolerance index values ​​through confidence assessment are as follows:

[0033] According to the seedling stage time window, the time-series derived indicators corresponding to each seedling stage salt tolerance characteristic dimension are subjected to stage consistency judgment processing to generate a prediction consistency metric.

[0034] Extract the baseline interval for the predicted stability of the yield salt tolerance index and perform interval determination processing with the predicted consistency metric to generate a prediction confidence label.

[0035] In a preferred embodiment of the method for predicting rice salt tolerance according to the present invention, the specific steps for outputting the salt tolerance prediction report are as follows:

[0036] Extract the predicted value field and the confidence status field of the predicted value of the predicted confidence marker and the predicted value of the salt tolerance index, respectively, and perform consistency verification to generate the prediction result record;

[0037] The predicted results are recorded and processed for output format adjudication and structured encapsulation to generate a salt tolerance prediction report.

[0038] The beneficial effects of this invention are as follows: by constructing multiple prediction sub-models according to the salt stress intensity range, explicit modeling of the differences in rice response mechanisms under different salt stress intensities is achieved; this enables the selection of the most suitable prediction sub-model based on environmental similarity during the prediction stage, improving the prediction stability and applicability under different salt stress scenarios; at the same time, the credibility assessment and output adjudication mechanism that combines prediction consistency and stability upgrades the prediction results from "numerical output" to "numerical + credible semantic" decision information, reducing the risk of prediction misjudgment and improving the usability of prediction results in breeding decisions and production management. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of a method for predicting salt tolerance in rice.

[0041] Figure 2 A flowchart for generating agronomic trait data during the seedling stage and recording experimental environmental information.

[0042] Figure 3 A flowchart for generating seedling salt tolerance characteristics and forming a rice salt tolerance index.

[0043] Figure 4 A flowchart for outputting the salt tolerance index prediction value and generating a salt tolerance prediction report. Detailed Implementation

[0044] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0045] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0046] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0047] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a method for predicting the salt tolerance of rice, comprising the following steps:

[0048] S1. Collect seedling growth data of rice under salt stress test conditions and perform time-series analysis to form seedling agronomic trait data, and record the corresponding test environment information.

[0049] S1.1 Under salt stress test conditions, seedling growth status data were organized and divided into stages according to the seedling growth time sequence to form seedling time sequence data.

[0050] It should be noted that seedling growth status data is used to characterize the growth and development performance of rice during the seedling stage, including agronomic traits reflecting growth vigor and physiological status such as plant height, biomass increase, number of leaves, leaf area index, aboveground dry weight, root length, root volume, root activity, and leaf color or chlorophyll content. Seedling growth status data is obtained by continuously observing and recording rice samples at the same growth stage at fixed time intervals (e.g., every 3 days) under salt stress test conditions. The collection methods include a combination of manual measurement and automated collection. The obtained data are uniformly archived according to timestamps and rice object identifiers to ensure comparability and continuity between different sampling times.

[0051] Salt stress test conditions refer to the controlled test environment artificially set to simulate the growth state of rice in a saline-alkali environment. Salt stress test conditions are usually formed by introducing a certain concentration (such as 100 mmol / L) of soluble salts into irrigation water or culture medium.

[0052] Under salt stress test conditions, the seedling growth status data of the same rice species at different sampling times during the seedling growth stage were arranged in chronological order of collection time. The arranged seedling growth status data were then divided into stages so that the seedling growth status data at each sampling time could be mapped to a specific seedling growth stage, thus forming seedling time-series data.

[0053] S1.2. Based on the seedling stage time series data, the growth status of rice seedlings is summarized in a structured manner to form seedling agronomic trait data, and the corresponding experimental environment information is recorded at the same time.

[0054] It should be noted that the seedling time-series data corresponding to different sampling times during the seedling growth stage of the same rice species are collected according to the categories of seedling growth status indicators such as plant height, biomass, growth organ status, and root system status. The time-series arrangement relationship of each indicator category with the change of sampling time is maintained to form seedling agronomic trait data. In the process of completing the structured summary of seedling agronomic trait data, the experimental environment information corresponding to the seedling time-series data is recorded simultaneously, and the experimental environment information is associated and stored with the seedling agronomic trait data.

[0055] S2. Extract the salt tolerance characteristics of seedling agronomic traits from the seedling stage data, divide the salt stress intensity range of the salt stress test conditions according to the experimental environment information, and calculate the rice salt tolerance index in each salt stress intensity range.

[0056] S2.1 Statistically analyze the variation range, rate of change, and stage differences of agronomic traits during the seedling growth stage, and generate response degree indicators.

[0057] It should be noted that, for each agronomic trait dimension of each rice species in the seedling stage agronomic trait data, the numerical sequence sorted by sampling time within the seedling growth stage is extracted as the trait time series. The difference between the maximum and minimum values ​​of the trait time series is used as the magnitude of change index; the mean absolute value of the ratio of the trait difference at adjacent sampling times to the corresponding time difference in the trait time series is used as the rate of change index; the trait time series is grouped according to the seedling growth stage, and the stage mean of each stage is calculated separately. The mean absolute value of the difference between the stage means of adjacent stages is used as the stage difference index; range normalization is performed on the magnitude of change index, rate of change index, and stage difference index within the same batch of rice species, and weighted summation is performed to generate the response degree index.

[0058] It should also be noted that the weights of the three categories of indicators are calculated based on the three normalized values ​​of the same batch of rice subjects. The weights are determined using the coefficient of variation method, which includes calculating the ratio of the standard deviation to the mean of the three normalized values ​​on the subject set to obtain the three coefficients of variation, and then normalizing the three coefficients of variation to obtain the amplitude weight, rate weight, and difference weight.

[0059] S2.2 Based on the response degree index, correlation screening and structured mapping processing are performed on the agronomic trait data of the seedling stage to generate salt tolerance characteristics of the seedling stage.

[0060] It should be noted that, within the same rice species, the response degree indices of each seedling stage agronomic trait dimension are mapped one-to-one with the corresponding seedling stage agronomic trait data, constructing a "trait dimension - response degree index" association data pair. The rank correlation coefficient between the response degree index sequence and the corresponding seedling stage agronomic trait change sequence in the association data pair is calculated as the correlation measure. The correlation measures corresponding to each seedling stage agronomic trait dimension are sorted, and the seedling stage agronomic trait dimensions with correlation measures in the top interval (e.g., the top 30%) are retained to form a candidate salt tolerance trait set. Structured mapping processing is performed on each seedling stage agronomic trait dimension in the candidate salt tolerance trait set, including reorganizing the seedling stage agronomic trait data in the candidate salt tolerance trait set according to the seedling stage agronomic trait category, seedling growth stage, and response degree index value order, so that each seedling stage agronomic trait dimension forms a seedling stage salt tolerance feature that corresponds one-to-one with the response degree index.

[0061] The expression for calculating the rank correlation coefficient is:

[0062] ;

[0063] in, Represents the rank correlation coefficient. Indicates the quantity of rice plants. Indicates the first The position number of the response degree index corresponding to each rice object in the set of response degree indices of all rice objects within the same salt stress intensity range, after being sorted by numerical value; Indicates the first The position number of the change in agronomic traits during the seedling stage for each rice object within the set of changes in all rice objects in the same salt stress intensity range, sorted by numerical value.

[0064] S2.3 Extract the salt stress concentration parameter, salt stress duration parameter, and salt stress application method parameter from the experimental environment information to form environmental parameters.

[0065] It should be noted that salt stress concentration parameters, salt stress duration parameters, and salt stress application method parameters are extracted from the experimental environment information. Specifically, the salt stress concentration parameter is used to characterize the concentration level of soluble salts in the rice growth medium or irrigation water during the experiment; the salt stress duration parameter is used to characterize the time span of rice under the salt stress conditions; and the salt stress application method parameter is used to characterize the form in which the salt stress conditions are applied during the experiment. The salt stress concentration parameter, salt stress duration parameter, and salt stress application method parameter are then structured and summarized to form environmental parameters.

[0066] S2.4. Based on environmental parameters, perform partition mapping processing on the salt stress test conditions, divide the salt stress test conditions into multiple salt stress intensity intervals, and assign corresponding salt stress intensity interval identifiers.

[0067] It should be noted that the environmental parameters corresponding to each group of salt stress test conditions are treated as a combination of environmental parameters. Within the environmental parameter combination space, salt stress test conditions with similar distributions of environmental parameter combinations are mapped to the same partition, and different partitions are assigned to different salt stress intensity intervals. A unique salt stress intensity interval identifier is assigned to each salt stress intensity interval.

[0068] S2.5. Based on the seedling agronomic traits and corresponding seedling salt tolerance characteristics within each salt stress intensity interval, calculate the interval salt tolerance index, and store the interval salt tolerance index corresponding to each salt stress intensity interval in a structured manner to form the rice salt tolerance index.

[0069] It should be noted that rice species falling within the salt stress intensity range were selected from the seedling agronomic trait data, and the corresponding seedling salt tolerance trait values ​​were extracted to form a set of interval trait values. The ratio of the interval standard deviation to the interval mean in the set of interval trait values ​​was used as the interval dispersion. The salt tolerance trait weights were determined based on the dispersion of each seedling salt tolerance trait, with larger weights for trait characteristics with smaller dispersion. Based on the salt tolerance trait weights, the salt tolerance characteristics of each seedling in the set of interval trait values ​​were weighted and summed to obtain the interval salt tolerance index. The interval salt tolerance index was then associated with and stored with the corresponding salt stress intensity interval identifier to form the rice salt tolerance index.

[0070] S3. For each salt stress intensity range, multiple yield salt tolerance prediction sub-models are constructed using the corresponding seedling salt tolerance characteristics and rice salt tolerance index, and the environmental similarity index between experimental environmental information and historical experimental environmental information is calculated.

[0071] S3.1. Calculate the time-series derived indicators corresponding to each salt tolerance feature dimension of the seedling stage, and perform window aggregation and vector splicing processing to construct the seedling stage response path representation layer.

[0072] It should be noted that, for each seedling salt tolerance feature dimension, the feature value sequence corresponding to the seedling salt tolerance feature at different sampling times is extracted along the seedling growth time axis; the feature value sequence is divided into time windows according to the seedling growth stage, and the time-series derived index of the seedling salt tolerance feature is calculated in each time window, wherein the time-series derived index includes the average value, value change amplitude, and value change rate of the seedling salt tolerance feature in the time window; the time-series derived indexes obtained for the same seedling salt tolerance feature in different time windows are arranged in chronological order to form a window-level feature sequence; the window-level feature sequences of different seedling salt tolerance features are vector-concatenated according to the feature dimension order to form the seedling response path representation layer.

[0073] S3.2 Extract the interval statistics from the rice salt tolerance index corresponding to the salt stress intensity interval markers, perform vector splicing and conditional modulation processing on the salt stress intensity interval markers and interval statistics, and construct a salt intensity conditional gating layer.

[0074] It should be noted that, for each salt stress intensity interval, the corresponding interval salt tolerance index value is extracted from the rice salt tolerance index, and the interval mean, interval variance, and interval quantile statistics (median) are calculated as interval statistics. The salt stress intensity interval identifier is combined with the corresponding interval statistics to form a salt stress intensity interval condition vector. Using the salt stress intensity interval condition vector as the modulation basis, conditional modulation is performed on the response intensity corresponding to different time windows and different salt tolerance feature dimensions in the seedling response path representation layer. This ensures that the response features in the seedling response path representation layer that match the salt tolerance characteristics of the current salt stress intensity interval maintain their expression amplitude, while the response features that deviate from the salt tolerance characteristics of the current salt stress intensity interval are suppressed in terms of expression amplitude, thereby constructing a salt intensity conditional gating layer.

[0075] S3.3 Within each salt stress intensity range, by establishing and solidifying the mapping rules from time-series derived indices to rice salt tolerance indices, multiple yield salt tolerance prediction sub-models corresponding one-to-one with each salt stress intensity range are constructed.

[0076] It should be noted that, for each salt stress intensity interval, rice samples falling within the salt stress intensity interval are selected, and the corresponding seedling response path representation vector and the rice salt tolerance index value corresponding to the salt stress intensity interval are extracted. Within the current salt stress intensity interval, the seedling response path representation vector is used as input data, and the rice salt tolerance index value is used as target data. The correspondence between the two is fitted to establish a mapping rule for representing the relationship between the seedling response path and the salt tolerance result. The structural parameters and corresponding relationship parameters of the current mapping rule are fixedly stored to form a yield salt tolerance prediction sub-model that corresponds one-to-one with the salt stress intensity interval.

[0077] Based on rice samples falling within the corresponding salt stress intensity range, the seedling response path representation vector corresponding to the rice sample is selected as the training input data, and the rice salt tolerance index corresponding to the rice sample is selected as the training output data. During training, the correspondence parameters in the mapping rule are iteratively adjusted so that the predicted output result obtained based on the training input data gradually approaches the corresponding rice salt tolerance index value. When the deviation between the predicted output result and the corresponding rice salt tolerance index remains basically consistent during multiple rounds (e.g., 10 rounds) of parameter adjustment within the salt stress intensity range, and the predicted output result no longer changes significantly with the adjustment of the correspondence parameters, the mapping relationship is determined to have reached a stable state. After the mapping relationship reaches a stable state, the structural parameters and correspondence parameters corresponding to the mapping rule are fixed and saved to form the trained yield salt tolerance prediction sub-model.

[0078] It should also be noted that the structural parameters are parameters that are determined simultaneously when establishing the mapping rules between the seedling response path representation vector and the rice salt tolerance index, in order to limit the composition, dimensional order and output format of the input data. For example, when establishing the mapping rules, if the seedling response path representation vector containing multiple time window derived indicators is determined as the input and a single rice salt tolerance index is determined as the output, then the number of dimensions, the dimensional order and the single output format of the input vector constitute the structural parameters.

[0079] The correspondence parameter is formed synchronously during the mapping rule establishment process based on the sample correspondence between the seedling response path representation vector and the value of the rice salt tolerance index within the same salt stress intensity range. It is used to characterize the specific correlation between input changes and output changes. For example, when establishing a mapping relationship between the seedling response path representation vector and the corresponding salt tolerance index value of multiple rice objects within a certain salt stress intensity range, the influence of different time window features on the salt tolerance index is solidified into the correspondence parameter.

[0080] S3.4. Parametric processing is performed on the experimental environment information and the historical experimental environment information corresponding to each salt stress intensity range to form an environmental input feature vector and a historical environmental feature vector.

[0081] It should be noted that, based on the current experimental environment information, salt stress concentration parameters, salt stress duration parameters, and salt stress application method parameters are extracted and organized according to a unified data type and value scale, so that different parameters have a feature form that can be represented in parallel, thereby forming an environmental input feature vector. At the same time, for each salt stress intensity range, salt stress concentration parameters, salt stress duration parameters, and salt stress application method parameters corresponding to the salt stress intensity range are extracted from the historical experimental environment information, and organized according to the same parameterization method as the current experimental environment information, forming multiple historical environmental feature vectors corresponding to different salt stress intensity ranges.

[0082] S3.5. Calculate the corresponding environmental similarity index based on the degree of difference between the environmental input feature vector and the historical environmental feature vector.

[0083] The expression for calculating the environmental similarity index is:

[0084] ;

[0085] in, Indicators representing environmental similarity. This indicates the number of dimensions of the environmental parameters involved in the environmental similarity calculation. This indicates that under the current salt stress test conditions, the first... The parameter values ​​corresponding to each environmental parameter dimension include salt stress concentration parameter, salt stress duration parameter, and salt stress application method parameter; This represents a historical test environment compared to current salt stress test conditions, specifically the [number]th [test]. The parameter values ​​corresponding to each environmental parameter dimension; Indicates the first The global range of each environmental parameter dimension in the historical test environment is used to unify the scale of differences between different environmental parameter dimensions, so that parameters of different dimensions are comparable under the same similarity calculation framework.

[0086] S4. Based on the environmental similarity index, determine the applicability of the yield salt tolerance prediction sub-model, generate scenario control parameters, and input the seedling salt tolerance characteristics into the corresponding yield salt tolerance prediction sub-model according to the scenario control parameters, and output the predicted value of the yield salt tolerance index.

[0087] S4.1. Based on the environmental similarity index, sort the matching degree of each yield salt tolerance prediction sub-model to form a set of candidate prediction sub-models.

[0088] It should be noted that the environmental similarity index is associated with each yield salt tolerance prediction sub-model, wherein each yield salt tolerance prediction sub-model is associated with the historical environmental characteristics of its respective salt stress intensity range; the yield salt tolerance prediction sub-models are sorted according to the value order of the environmental similarity index to form a model sorting sequence based on environmental similarity; in the model sorting sequence, the value difference between the environmental similarity index corresponding to adjacent sorting positions is counted sequentially, and compared with the statistical distribution of all adjacent value differences in the model sorting sequence, and the yield salt tolerance prediction sub-models corresponding to consecutive sorting positions within the same value change range are selected to form a candidate prediction sub-model set.

[0089] It should also be noted that after sorting the environmental similarity indices corresponding to each yield salt tolerance prediction sub-model, the difference in values ​​of the environmental similarity indices corresponding to adjacent sorting positions in the sorting results is calculated to form an adjacent value difference sequence. In the adjacent value difference sequence, the analysis is performed according to the order of the value differences in the sorting structure. When several adjacent value differences appear consecutively in the adjacent value difference sequence without being interrupted by other value differences, it is determined that the corresponding adjacent sorting positions belong to the same value change segment. When the adjacent value differences are interrupted in the adjacent value difference sequence, it is determined that the value change segment has switched.

[0090] S4.2 Identify the similarity concentration interval and similarity decay interval in the candidate prediction sub-model set, and perform interval mapping processing to generate scene control parameters.

[0091] It should be noted that, according to the arrangement order of the candidate prediction sub-model set in the model sorting sequence, the adjacent value difference sequence corresponding to the candidate prediction sub-model set is extracted. In the adjacent value difference sequence, based on the continuous arrangement structure of the value difference in the sequence, the continuous segment where the adjacent value difference sequence is continuously arranged and there is no segment switching within the candidate prediction sub-model set is determined as the similarity concentration interval. The continuous segment where there is segment switching between the end of the candidate prediction sub-model set and the adjacent position of the non-candidate prediction sub-model is determined as the similarity decay interval. After completing the identification of the similarity concentration interval and the similarity decay interval, an interval mapping relationship of "segment identifier - control strategy identifier" is established. The similarity concentration interval is mapped to the scene control parameter used to enable the sub-models within the candidate prediction sub-model set, and the similarity decay interval is mapped to the scene control parameter used to restrict the participation of sub-models outside the candidate prediction sub-model set in the prediction, thereby generating the scene control parameter.

[0092] S4.3 The seedling response path representation layer performs temporal derivation processing on the salt tolerance characteristics of the seedling stage to generate the seedling response path representation vector.

[0093] It should be noted that the seedling response path representation layer takes the seedling salt tolerance feature corresponding to a single rice object as input, and organizes the seedling salt tolerance feature sequentially according to the seedling growth time, so that each feature dimension forms a continuous seedling feature sequence on the time axis; for each seedling salt tolerance feature dimension, the feature change of the seedling salt tolerance feature between adjacent sampling times in the seedling feature sequence is calculated, and stage summary processing is performed within the seedling growth stage to represent the evolution trajectory of the feature with the seedling growth process; the stage change results corresponding to each seedling salt tolerance feature dimension are combined in a unified order to form the seedling response path representation vector.

[0094] S4.4 The salt intensity condition gating layer performs interval mapping processing on the seedling response path representation vector based on the scenario control parameters to generate the controlled seedling response path vector.

[0095] It should be noted that, based on the salt stress intensity interval identifier indicated in the scenario control parameters, the salt intensity condition gating layer performs interval association on the path components (evolutionary trajectories) in the seedling response path representation vector that correspond to different salt tolerance feature dimensions at different seedling stages, so that each path component establishes a one-to-one interval association relationship with the corresponding salt stress intensity interval. According to the interval association relationship, the salt intensity condition gating layer performs interval mapping processing on the path components in the seedling response path representation vector, retaining only the path components that match the current salt stress intensity interval, and performing structural masking processing on the path components that do not belong to the current salt stress intensity interval. The salt intensity condition gating layer recombines the controlled path components according to the original structural order to generate a controlled seedling response path vector that is adapted to the current salt stress intensity interval.

[0096] S4.5 Calculate the predicted yield salt tolerance index of the controlled seedling stage response path vector under the current salt stress test conditions.

[0097] The expression for calculating the predicted value of the salt tolerance index for yield is:

[0098] ;

[0099] in, The salt stress intensity range is indicated by... The corresponding predicted value of salt tolerance index for yield, This represents the controlled seedling stage response path vector obtained by the current rice object under the current salt stress test conditions. The salt stress intensity range is indicated by... The number of training samples, The salt stress intensity range is indicated by... The The controlled seedling stage response path vector corresponding to each training sample. The salt stress intensity range is indicated by... The The values ​​of the rice salt tolerance index corresponding to each training sample. Indicates the range of salt stress intensity. One-to-one mapping parameter vector, Indicates the range of salt stress intensity. The one-to-one bias parameter represents the vector transpose operation.

[0100] It should also be noted that each component of the controlled seedling stage response path vector was subjected to scale normalization before calculation, so that the values ​​of each component are within a uniform numerical scale range, thus making the mapping parameter vector and the bias parameter dimensionless parameters, and the yield salt tolerance index prediction value is a dimensionless evaluation index.

[0101] S5. Assign corresponding prediction credibility tags to the predicted yield salt tolerance index value through credibility assessment, and perform output decision on the predicted yield salt tolerance index value according to the prediction credibility tags, and output salt tolerance prediction report.

[0102] S5.1. Perform stage consistency judgment processing on the time-series derived indicators corresponding to each seedling salt tolerance characteristic dimension according to the seedling stage time window, and generate the predicted consistency metric value.

[0103] It should be noted that the time-series derived indicators are divided into time windows according to the seedling growth stages, so that the same salt tolerance characteristic dimension at the seedling stage forms corresponding phased indicator sequences in different seedling stage time windows. The phased indicator sequences of the same salt tolerance characteristic dimension at the seedling stage in adjacent seedling stage time windows are aligned, and the direction, magnitude, and rate of change of the time-series derived indicators between adjacent seedling stage time windows are calculated to form the phased change description results. The direction of change is used to characterize the increase or decrease trend of the phased indicators, the magnitude of change is used to characterize the absolute degree of change of the phased indicators, and the rate of change is used to characterize the intensity of change per unit time.

[0104] For each seedling stage salt tolerance feature dimension, stage consistency determination is performed on the description results of stage changes within the entire seedling stage time window. Stage consistency determination includes directional consistency determination and amplitude stability determination. When the change direction between adjacent seedling stage time windows remains in the same direction and the fluctuation of the change amplitude does not exceed the amplitude stability constraint range, the current seedling stage salt tolerance feature dimension is determined to have stage consistency at the seedling stage time window scale. When the change direction changes in the opposite direction, or the fluctuation of the change amplitude exceeds the amplitude stability constraint range, the current seedling stage salt tolerance feature dimension is determined to lack stage consistency at the seedling stage time window scale. The stage consistency determination results corresponding to all seedling stage salt tolerance feature dimensions are structured and summarized, and the proportion of the number of seedling stage salt tolerance feature dimensions with stage consistency to the total number of seedling stage salt tolerance feature dimensions is used as the prediction consistency metric.

[0105] It should also be noted that, in the historical sample set corresponding to the salt stress intensity range, the variation amplitude values ​​of the same seedling salt tolerance feature dimension between adjacent seedling time windows in all historical samples are extracted to form a historical variation amplitude sample sequence; the historical variation amplitude sample sequence is sorted according to the numerical value, and the position segment where the sample points are arranged adjacently on the numerical axis and the spacing between adjacent historical samples remains continuous without any discontinuity is taken as the main cluster segment of the variation amplitude sample; the numerical range covered by the main cluster segment is taken as the amplitude stability constraint range.

[0106] S5.2 Extract the prediction stability benchmark interval of the predicted yield salt tolerance index and perform interval determination processing with the prediction consistency metric to generate a prediction confidence label.

[0107] It should be noted that, in the historical prediction records corresponding to the current salt stress intensity range, the predicted values ​​of yield salt tolerance index obtained during the historical prediction process are collected and sorted in order of magnitude; marginal values ​​located at both ends of the sorting sequence are identified and excluded from the sorting results, where marginal values ​​include a certain proportion of the predicted values ​​at the front and back ends of the sorting results (e.g., each accounting for 10% of the total number of sorting results); after completing the exclusion of marginal values, the range of values ​​covered by the continuously distributed historical prediction values ​​in the middle of the sorting results is used as the prediction stability benchmark interval; the current yield salt tolerance index prediction value is assigned to the prediction stability benchmark interval, and combined with the seedling stage evolution consistency results reflected by the prediction consistency metric, the current prediction result is jointly judged at both the numerical stability and evolutionary stability levels to generate a prediction confidence label.

[0108] It should also be noted that when performing interval assignment determination, the current production salt tolerance index prediction value is compared with the interval boundary of the prediction stability benchmark interval. When the production salt tolerance index prediction value is within the range of values ​​covered by the prediction stability benchmark interval, the current prediction value is determined to belong to the prediction stability benchmark interval.

[0109] After completing the interval assignment determination, the interval assignment determination results are combined with the stage consistency determination results corresponding to the prediction consistency metric. When the interval assignment determination results indicate that the predicted yield salt tolerance index belongs to the prediction stability benchmark interval, and the stage consistency determination results indicate that the seedling evolution process maintains consistency on the seedling time window scale, a prediction credibility label is generated to characterize that the prediction results are in a stable and credible state. When at least one of the interval assignment determination results and the stage consistency determination results does not meet the conditions, a prediction credibility label is generated to characterize that the prediction results are in an unstable state.

[0110] S5.3 Extract the predicted value field and the confidence status field of the predicted confidence marker and the predicted salt tolerance index, respectively, and perform consistency verification to generate the prediction result record.

[0111] It should be noted that, by parsing the predicted yield salt tolerance index, the predicted numerical field, which characterizes the numerical content of the predicted result, is extracted, and the rice object identifier, salt stress test condition identifier, and prediction time identifier associated with the predicted numerical field are extracted simultaneously. Field parsing is also performed on the prediction credibility marker to extract the credibility status field, which characterizes the credibility status of the predicted result, and the rice object identifier, salt stress test condition identifier, and prediction time identifier associated with the credibility status field are extracted simultaneously. The predicted numerical field and the credibility status field are compared and matched item by item according to the rice object identifier, salt stress test condition identifier, and prediction time identifier. When the three types of identifiers correspond one-to-one between the predicted numerical field and the credibility status field, the consistency check is considered passed. If the consistency check passes, the successfully matched predicted numerical field and the credibility status field are merged and encapsulated to generate a prediction result record containing both predicted numerical information and credibility status information.

[0112] S5.4 Perform output format adjudication and structured encapsulation processing on the prediction results record to generate a salt tolerance prediction report.

[0113] It should be noted that the predicted numerical field and the confidence status field contained in the prediction result record are read; based on the interval judgment result between the prediction consistency metric value corresponding to the confidence status field and the prediction stability benchmark interval, it is determined whether the prediction result record meets the reliable output condition. Specifically, when the prediction consistency metric value falls within the corresponding prediction stability benchmark interval, the prediction result record is determined to meet the reliable output condition; when the prediction consistency metric value does not fall within the corresponding prediction stability benchmark interval, the prediction result record is determined not to meet the reliable output condition. Standard output encapsulation processing is performed on the prediction result records that meet the reliable output condition, and confidence prompt information is generated for the prediction result records that do not meet the reliable output condition while retaining the predicted yield salt tolerance index value. The encapsulated prediction results are structured and organized according to the rice object identifier, salt stress test condition identifier, and prediction time sequence to generate a salt tolerance prediction report.

[0114] In summary, this invention achieves explicit modeling of the differences in rice response mechanisms under different salt stress intensities by constructing multiple prediction sub-models according to salt stress intensity ranges; it enables the selection of the most suitable prediction sub-model based on environmental similarity during the prediction stage, improving the prediction stability and applicability under different salt stress scenarios; and it upgrades the prediction results from "numerical output" to "numerical + credible semantic" decision information through a credibility assessment and output adjudication mechanism that combines prediction consistency and stability, reducing the risk of prediction misjudgment and improving the usability of prediction results in breeding decisions and production management.

[0115] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for predicting salt tolerance in rice, characterized in that: include, Data on the growth status of rice seedlings under salt stress were collected and analyzed over time to generate agronomic trait data for the seedling stage, while the corresponding experimental environment information was recorded. Extract the salt tolerance characteristics of seedling agronomic traits from the seedling stage data, divide the salt stress intensity range of the salt stress test conditions according to the experimental environment information, and calculate the rice salt tolerance index in each salt stress intensity range. For each salt stress intensity range, multiple yield salt tolerance prediction sub-models were constructed using the corresponding seedling salt tolerance characteristics and rice salt tolerance index, and the environmental similarity index between experimental environmental information and historical experimental environmental information was calculated. Based on the environmental similarity index, the applicability of the yield salt tolerance prediction sub-model is determined, scenario control parameters are generated, and the seedling salt tolerance characteristics are input into the corresponding yield salt tolerance prediction sub-model according to the scenario control parameters, and the predicted value of the yield salt tolerance index is output. The yield salt tolerance index prediction is assigned a corresponding prediction confidence label through confidence assessment, and an output decision is made on the yield salt tolerance index prediction based on the prediction confidence label, and a salt tolerance prediction report is output.

2. The method for predicting rice salt tolerance as described in claim 1, characterized in that: The process of generating seedling agronomic trait data and recording corresponding experimental environment information involves the following steps: Under salt stress test conditions, seedling growth status data were organized and divided into stages according to the seedling growth time sequence to form seedling time sequence data; The growth status of rice seedlings was structurally summarized based on the seedling time series data to form agronomic trait data of the seedling stage, while the corresponding experimental environment information was recorded.

3. The method for predicting rice salt tolerance as described in claim 1, characterized in that: The specific steps for extracting the salt tolerance characteristics of seedlings from agronomic traits during the seedling stage are as follows: The variation range, rate of change, and stage differences of agronomic traits during the seedling growth stage are statistically analyzed to generate response degree indicators. Based on the response degree index, correlation screening and structured mapping were performed on the agronomic trait data of the seedling stage to generate salt tolerance characteristics of the seedling stage.

4. The method for predicting rice salt tolerance as described in claim 1, characterized in that: The specific steps for dividing the salt stress test conditions into salt stress intensity ranges based on the experimental environment information and calculating the rice salt tolerance index in each salt stress intensity range are as follows. Extract salt stress concentration parameters, salt stress duration parameters, and salt stress application method parameters from the experimental environment information to form environmental parameters; Based on environmental parameters, a partitioning mapping process is performed on the salt stress test conditions to divide the salt stress test conditions into multiple salt stress intensity intervals and assign corresponding salt stress intensity interval identifiers. Based on the seedling agronomic traits and corresponding seedling salt tolerance characteristics within each salt stress intensity interval, the interval salt tolerance index is calculated. The interval salt tolerance indices corresponding to each salt stress intensity interval are then stored in a structured manner to form the rice salt tolerance index.

5. The method for predicting rice salt tolerance as described in claim 4, characterized in that: The specific steps for constructing multiple yield and salt tolerance prediction sub-models are as follows. We statistically analyzed the time-series derived indices corresponding to each salt tolerance feature dimension at each seedling stage, and performed window aggregation and vector concatenation to construct a seedling stage response path representation layer. Extract interval statistics from the rice salt tolerance index corresponding to the salt stress intensity interval markers, and perform vector splicing and conditional modulation processing on the salt stress intensity interval markers and interval statistics to construct a salt intensity conditional gating layer. Within each salt stress intensity range, by establishing and solidifying the mapping rules from time-derived indices to rice salt tolerance indices, multiple yield salt tolerance prediction sub-models corresponding one-to-one with each salt stress intensity range are constructed.

6. The method for predicting rice salt tolerance as described in claim 1, characterized in that: The specific steps for calculating the environmental similarity index between the experimental environment information and historical experimental environment information are as follows. The experimental environment information and the historical experimental environment information corresponding to each salt stress intensity range are parameterized to form an environmental input feature vector and a historical environmental feature vector. Based on the degree of difference between the environmental input feature vector and the historical environmental feature vector, the corresponding environmental similarity index is calculated.

7. The method for predicting rice salt tolerance as described in claim 6, characterized in that: The specific steps for generating the scene control parameters are as follows: Based on the environmental similarity index, the matching degree of each yield salt tolerance prediction sub-model is ranked to form a set of candidate prediction sub-models; Identify the similarity concentration interval and similarity decay interval in the candidate prediction sub-model set, perform interval mapping processing, and generate scene control parameters.

8. The method for predicting rice salt tolerance as described in claim 1, characterized in that: The specific steps for predicting the salt tolerance index of the output yield are as follows. The seedling response path representation layer performs temporal derivation processing on the salt tolerance characteristics of the seedling stage to generate the seedling response path representation vector. The salt intensity condition gating layer performs interval mapping processing on the seedling response path representation vector based on the scenario control parameters, generating a controllable seedling response path vector. Calculate the predicted yield salt tolerance index of the controlled seedling stage response path vector under the current salt stress test conditions.

9. The method for predicting rice salt tolerance as described in claim 8, characterized in that: The specific steps for assigning corresponding prediction confidence labels to the predicted yield salt tolerance index values ​​through confidence assessment are as follows: According to the seedling stage time window, the time-series derived indicators corresponding to each seedling stage salt tolerance characteristic dimension are subjected to stage consistency judgment processing to generate a prediction consistency metric. Extract the baseline interval for the predicted stability of the yield salt tolerance index and perform interval determination processing with the predicted consistency metric to generate a prediction confidence label.

10. The method for predicting rice salt tolerance as described in claim 1, characterized in that: The specific steps for outputting the salt tolerance prediction report are as follows. Extract the predicted value field and the confidence status field of the predicted value of the predicted confidence marker and the predicted value of the salt tolerance index, respectively, and perform consistency verification to generate the prediction result record; The predicted results are recorded and processed for output format adjudication and structured encapsulation to generate a salt tolerance prediction report.

Images