Agricultural disaster early warning decision method and system

Abstract

The present application relates to early warning decision-making technical field, specifically to a kind of agricultural disaster early warning decision-making method and system, comprising the following steps: obtain structural parameter set by multispectral sensor, sliding window smoothing processing and marking direction consistency, input hidden markov model to calculate transition probability, identify stage trigger point and output label, generate response mapping table by matching physiology and meteorological time series, extract trend overlap section label impact window, combine multi-factor weighted ordering to determine disaster grade, output early warning instruction.In the present application, by continuously observing growth parameters and smoothing time series data, analyzing structural index dynamic consistency, identifying growth mutation, aligning physiological and meteorological factors, extracting trend overlap sensitive area, superimposing residual to lock risk time window, the sensitivity and pertinence of disaster early warning are improved, the false alarm and missed alarm risk is reduced, fine data support is provided for management, and the scientificity and initiative of risk control are enhanced.

Description

Technical Field

[0001] This invention relates to the field of early warning decision-making technology, and in particular to an agricultural disaster early warning decision-making method and system. Background Technology

[0002] The field of early warning and decision-making technology refers to the use of various technologies to predict, monitor, and issue early warnings for disasters occurring in agricultural production, in order to reduce or avoid losses caused by disasters. This field includes the collection, processing, and analysis of information from meteorological, geographical, and environmental sources, involving disaster type identification, early warning model construction, and the implementation of decision support systems. Core aspects include data acquisition and transmission, disaster identification and classification, risk assessment, and decision support. Through the comprehensive analysis of multi-source information, it provides scientific early warning services for agricultural production, reducing the uncertainties caused by disasters.

[0003] Traditional agricultural disaster early warning decision-making methods refer to approaches that rely on experience or basic models for disaster early warning, employing traditional means such as meteorological data analysis and analysis of original disaster records for disaster prediction and warning. These traditional methods, based on statistical analysis of weather parameters and using fixed disaster models for early warning, neglect real-time changing environmental factors, resulting in low timeliness and accuracy of warnings, making them ill-suited for complex and ever-changing disaster situations.

[0004] Traditional methods rely mainly on fixed meteorological models and raw data in application, which cannot dynamically perceive the real-time growth status of crops. They lack direct observation means for the physiological responses of crops under meteorological changes, which can easily lead to delayed disaster risk identification, insufficiently refined risk window division, and discrepancies between early warning results and the actual time of disaster occurrence. This makes it difficult to provide managers with timely and accurate risk response suggestions, resulting in decision-making delays and increased agricultural losses. Summary of the Invention

[0005] To address the technical problems of traditional methods that rely primarily on fixed meteorological models and raw data, failing to dynamically perceive the real-time growth status of crops and lacking direct observation methods for crop physiological responses to meteorological changes, leading to delayed disaster risk identification, insufficiently refined risk window segmentation, discrepancies between early warning results and actual disaster occurrence times, and difficulty in providing timely and accurate risk response suggestions to managers, resulting in decision-making delays and increased agricultural losses, this invention provides an agricultural disaster early warning decision-making method and system. The technical solution is as follows:

[0006] On the one hand, an agricultural disaster early warning decision-making method is provided, which includes:

[0007] S1: The daily increment of crop leaf age, daily growth of stem height, and daily variation of internode length are obtained by multispectral sensors. The time-series smoothing is performed by a sliding window algorithm. The daily increment of leaf age and daily growth of stem height are marked with directional consistency to generate a set of structural parameters.

[0008] S2: Input the set of structural parameters into the Hidden Markov Model to calculate the stage transition probability, determine the critical point based on the sliding mean of stem height and the second difference of internode length, accumulate the daily increment of leaf age, and mark the stage transition trigger point when the cumulative value exceeds the stage threshold for two consecutive days, and output the stage label.

[0009] S3: Call the transpiration rate threshold, root water potential response lag time, and organ morphological integrity rate corresponding to the stage label, and use the dynamic time warping algorithm to align the meteorological data with the time series of physiological parameters to generate a response mapping table;

[0010] S4: Based on the regression residual sequence in the response mapping table, the sliding trend overlap method is used to determine the trend direction of the temperature change amplitude, wind speed change slope and radiation intensity range of meteorological parameters, and to extract the segments that overlap with the evaporation rate trend direction in the response mapping table and mark them as impact window segments.

[0011] As a further embodiment of the present invention, the structural parameter set includes directional consistency labels, time-smoothed parameter values, and leaf age growth rate; the stage labels include stage identification identifiers, transition judgment conditions, and stage transition cumulative thresholds; the response mapping table includes meteorological variable time alignment results, plant physiological parameter matching dimensions, and trend residual parameters; and the impact window segment includes temperature disturbance segment, wind speed anomaly segment, and radiation intensity variation segment.

[0012] As a further aspect of the present invention, the specific steps of S1 include:

[0013] S101: Acquire crop leaf reflectance, stem reflectance and internode reflectance data for consecutive days using a multispectral sensor, extract the daily variation of leaf number, stem height and internode height for adjacent days, convert them into daily variations of leaf age, stem height and internode length, and generate a crop growth variation dataset.

[0014] S102: Call the three sequence data in the crop growth change dataset, set three dates as a sliding window, calculate the average value of leaf age increment, stem height growth and internode length change within the window, shift the window and repeat the calculation of the processing interval to generate a smoothed value set of growth sequence;

[0015] S103: Call the daily increment of leaf age and daily increase of stem height in the growth sequence smoothing value set, determine whether the change direction of adjacent dates is consistent, combine the consistent direction mark with the corresponding internode length change value, and generate a structural parameter set.

[0016] As a further aspect of the present invention, the specific steps of S2 include:

[0017] S201: Input the set of structural parameters into the Hidden Markov Model, calculate the transition probability between the current state and the original state based on the state data sequence, filter the channels with a difference greater than the state jump threshold, record the state channel information, calculate the jump trend value, and generate the stage jump tendency coefficient.

[0018] S202: Based on the stage jump tendency coefficient, extract the internode length sequence and the stem height sliding mean, perform second-order difference operation, screen the segments that fluctuate synchronously with the stem height sliding mean, calculate the segment variation slope value, cross-compare, and obtain the synchronous fluctuation discrimination value;

[0019] S203: Based on the synchronous fluctuation discrimination quantity, construct the daily incremental sequence of leaf age, screen the time nodes where the increment is greater than the critical threshold of stage jump for two consecutive days, identify the monitoring time window, and obtain the stage label.

[0020] As a further aspect of the present invention, the state transition threshold is a critical value for determining whether the change in the channel state transition probability has reached a significant level.

[0021] The stage jump tendency coefficient measures the overall jump trend strength of multiple channel states within the current stage and predicts the probability of state reversal.

[0022] The synchronous fluctuation discrimination quantity quantifies the numerical index of the consistency of fluctuations between the internode length and the moving average of stem height through the second-order difference synchronicity.

[0023] The stage transition critical threshold is used to screen time points where leaf age increases significantly over two consecutive days, marking growth stage mutation events.

[0024] As a further aspect of the present invention, the specific steps of S3 include:

[0025] S301: Call the transpiration rate threshold, root water potential response lag time and organ morphology integrity rate corresponding to the stage label, sort the stage labels by time, classify the transpiration rate by threshold, match the root water potential delay response and extract the curve, calculate the organ morphology integrity rate and align the parameter time sequence, and generate the time sequence parameter integrated value.

[0026] The integrated value of the time series parameters is a comprehensive result of multiple physiological and meteorological time series data, reflecting the indicators of changes in physiological state and environmental factors;

[0027] S302: Based on the integrated value of the time series parameters, extract the temperature, sunshine duration and wind speed sequences from the meteorological data, compare them with the trajectory of physiological parameter changes, apply the dynamic time warping algorithm to calculate their offset and mapping path, and obtain the meteorological alignment offset metric value.

[0028] The meteorological alignment offset metric is calculated by using a dynamic time warping algorithm to determine the time offset and mapping path between meteorological factors and physiological parameters.

[0029] S303: Based on the meteorological alignment offset metric, and according to the transpiration rate threshold range, root water potential response lag range, and organ morphology integrity rate variation range, perform secondary alignment of meteorological data and physiological parameters, perform matching index mapping, and generate a response mapping table.

[0030] As a further aspect of the present invention, the specific steps of S4 include:

[0031] S401: Based on the regression residual sequence in the response mapping table, obtain the temperature sequence, wind speed sequence and radiation intensity sequence, extract the temperature difference, wind speed slope and radiation range by sliding window, and normalize them to determine the trend direction, and generate a meteorological change trend direction sequence.

[0032] The normalization process uses max-min normalization to standardize the sequence.

[0033] S402: Based on the meteorological change trend direction sequence, the meteorological parameter trend direction is symbolically encoded, and combined with the evaporation rate trend direction sequence, the consistency of the two sequences in the same time period is compared. Time window segments are selected according to the consistency standard to obtain the set of trend direction consistent segment numbers.

[0034] S403: Call the set of segment numbers with consistent trend direction, mark the start and end indices of the corresponding windows in the original sequence, extract the time index and establish a trend overlap mapping table, classify and organize the continuous segments under the time axis, and obtain the impact window segments.

[0035] As a further aspect of the present invention, the method includes step S5:

[0036] S5: Based on the relationship between the absolute value of the temperature change amplitude and the evaporation rate threshold in the response mapping table within the impact window segment, the linear relationship between the wind speed change slope and the organ integrity rate corresponding to the stage label, and the joint relationship between the soil water potential time difference matching result and the radiation intensity range within the impact window segment, the weighted sorting method is invoked to prioritize the disaster level and output the early warning decision instruction.

[0037] The early warning decision instructions include disaster level assessment factors, priority ranking weights, and response strategy recommendations.

[0038] As a further aspect of the present invention, the specific steps of S5 include:

[0039] S501: Based on the relationship between the absolute value of the temperature change amplitude within the impact window segment and the evaporation rate threshold in the response mapping table, extract the temperature change sequence and match the threshold interval, determine the location of the change signal node, calculate the distribution range of the dense response segment on the time axis, and generate the temperature change response amplitude value.

[0040] The temperature change response amplitude value refers to the degree of influence of drastic temperature changes on plant transpiration, which is calculated by comprehensively considering factors such as temperature change slope, transpiration threshold matching and response density.

[0041] S502: Based on the temperature change response amplitude value, match the wind speed change slope and organ integrity rate coefficient sequence, construct a response comparison chain, compare the change ratio by continuous segment and extract the offset difference amplitude to obtain the wind speed-organ correlation difference degree.

[0042] S503: Call the time difference matching results of the wind speed organ correlation difference and soil water potential, calculate the response index by combining the fluctuation range of radiation intensity within the section, construct a weighted structure vector for sorting and extract the corresponding disaster impact level, match the intervention instruction set bound to the response level, and output the early warning decision instruction.

[0043] On the other hand, an agricultural disaster early warning and decision-making system is provided, which is used to execute the above-mentioned agricultural disaster early warning and decision-making method, and the system includes:

[0044] The temporal smoothing module is used to acquire the daily increment of leaf age, daily increase of stem height, and daily change of internode length through multispectral sensors. It uses a sliding window algorithm to perform temporal smoothing processing, performs directional consistency marking operation to generate a set of structural parameters, and passes them to the step identification module.

[0045] The stage transition identification module is used to input the set of structural parameters into a hidden Markov model to calculate the stage transition probability, determine the critical point based on the sliding mean of stem height and the second difference of internode length, accumulate the daily increment of leaf age and compare it with a preset threshold, generate stage labels, and pass them to the time alignment module.

[0046] The time-series alignment module is used to call the transpiration rate threshold, root water potential response lag time, and organ morphological integrity rate corresponding to the stage label, and uses a dynamic time warping algorithm to align meteorological and physiological time-series data, generate a response mapping table, and pass it to the trend overlap module.

[0047] The trend overlap module is used to call the regression residual sequence in the response mapping table, use the sliding trend overlap method to determine the trend direction of temperature change amplitude, wind speed change slope and radiation intensity range, extract the overlapping segment with the evaporation rate trend direction, mark the impact window segment, and pass it to the disaster assessment module.

[0048] The disaster assessment module is used to prioritize disaster levels by comparing the magnitude of temperature change with the evaporation rate threshold within the impact window segment, performing linear regression calculations of wind speed change slope and organ integrity rate, and combining the soil water potential time difference matching results with the radiation intensity range, and outputs early warning decision instructions by calling a weighted sorting method.

[0049] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0050] By continuously observing crop growth parameters and smoothing time-series data, combined with dynamic consistency analysis among growth structure indicators, we can achieve detailed identification of abrupt changes in growth stages. The synchronous alignment and trend overlap analysis of crop physiological parameters and meteorological factors can effectively reveal sensitive segments of physiological responses under meteorological fluctuations. By judging the superposition of trend residuals, we can accurately lock the risk impact window, improve the sensitivity and pertinence of disaster early warning, reduce the risk of false alarms and missed alarms, provide timely and detailed data support for subsequent management measures, and enhance the scientific and proactive nature of agricultural production risk management. Attached Figure Description

[0051] Figure 1 This is a schematic diagram of the workflow of the present invention;

[0052] Figure 2 This is a detailed flowchart of S1 of the present invention;

[0053] Figure 3 This is a detailed flowchart of the S2 process of the present invention;

[0054] Figure 4 This is a detailed flowchart of the S3 process of the present invention;

[0055] Figure 5 This is a detailed flowchart of the S4 process of the present invention;

[0056] Figure 6 This is a detailed flowchart of S5 of the present invention;

[0057] Figure 7 This is a system flowchart of the present invention. Detailed Implementation

[0058] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0059] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0060] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0061] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0062] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0063] Please see Figure 1 This invention provides an agricultural disaster early warning decision-making method, the processing flow of which may include the following steps:

[0064] S1: The daily increment of crop leaf age, daily growth of stem height, and daily variation of internode length are obtained by multispectral sensors. The time-series smoothing is performed by a sliding window algorithm. The daily increment of leaf age and daily growth of stem height are marked with directional consistency to generate a set of structural parameters.

[0065] S2: Input the set of structural parameters into the Hidden Markov Model to calculate the stage transition probability. Determine the critical point based on the sliding mean of stem height and the second difference of internode length. Accumulate the daily increment of leaf age. When the cumulative value exceeds the stage threshold for two consecutive days, mark the stage transition trigger point and output the stage label.

[0066] S3: Call the transpiration rate threshold, root water potential response lag time, and organ morphological integrity rate corresponding to the stage label, and use the dynamic time warping algorithm to align the meteorological data with the time series of physiological parameters to generate a response mapping table;

[0067] S4: Based on the regression residual sequence in the response mapping table, the sliding trend overlap method is used to determine the trend direction of the temperature change amplitude, wind speed change slope and radiation intensity range of meteorological parameters, and to extract the segments that overlap with the evaporation rate trend direction in the response mapping table and mark them as impact window segments.

[0068] S5: Based on the relationship between the absolute value of the temperature change amplitude and the evaporation rate threshold in the response mapping table within the impact window segment, the linear relationship between the wind speed change slope and the organ integrity rate corresponding to the stage label, and the joint relationship between the soil water potential time difference matching result and the radiation intensity range within the impact window segment, the weighted sorting method is called to prioritize the disaster level and output the early warning decision instruction.

[0069] The structural parameter set includes directional consistency labels, time-smoothed parameter values, and leaf age growth rate; the stage labels include stage identification markers, transition judgment conditions, and cumulative thresholds for stage transitions; the response mapping table includes meteorological variable time alignment results, plant physiological parameter matching dimensions, and trend residual parameters; the impact window segments include temperature disturbance segments, wind speed anomaly segments, and radiation intensity variation segments; and the early warning decision instructions include disaster level assessment factors, priority ranking weights, and response strategy recommendation results.

[0070] Specifically, such as Figure 2 As shown, the specific steps of S1 are as follows:

[0071] S101: Acquire crop leaf reflectance, stem reflectance and internode reflectance data for consecutive days using a multispectral sensor, extract the daily variation of leaf number, stem height and internode height for adjacent days, convert them into daily variations of leaf age, stem height and internode length, and generate a crop growth variation dataset.

[0072] To acquire crop leaf, stem, and internode reflectance data over consecutive days using multispectral sensors, it is necessary to first define the spectral band settings within the monitoring area. For example, fine sampling should be performed in the blue, green, and red bands within the 450–750 nm visible light region. The plant tissue reflectance data collected daily at 10:00 AM under the same illumination conditions should be recorded, and the daily reflectance values ​​for each spectral band should be recorded in a two-dimensional matrix. For a cornfield sample of plants at the 5th–6th leaf stage, the reflectance data for leaves, stems, and internodes should be acquired daily. For instance, if the red band reflectance of the leaves is 0.34 on July 1, 2025, and 0.38 the following day, the reflectance between the two days is... The increment is 0.04. The reflectance of the stem and internodes is processed in the same way. Then, the changes in leaf number, stem height, and internode height caused by the reflectance change need to be extracted sequentially. Specifically, this can be achieved by setting the leaf number change rate corresponding to each unit change in reflectance. For example, setting a change of 0.1 in red light reflectance to correspond to a change of 1 leaf. If the reflectance increases by 0.04, the leaf increment is 0.4, retaining decimal precision in daily statistics. Stem height is determined based on changes in the red-edge band (700–740nm). Setting a change of 0.05 in reflectance to correspond to a height change of 0.6cm, if a plant's red-edge reflectance increases from 0.45 to 0.51 over three days, the stem height change is... The internode height is set according to the near-infrared band (750–900nm) variation range. Every 0.1% change in reflectance corresponds to a 1.2cm change in internode height. If the change is 0.08, the internode height is 0.96cm. After processing the daily difference, a daily growth change dataset is generated. The data structure includes fields such as date, leaf number increment, stem height increment, and internode length increment, which ultimately form a continuous growth dataset.

[0073] S102: Call the three sequence data in the crop growth change dataset, set three dates as a sliding window, calculate the average value of leaf age increment, stem height growth and internode length change within the window, shift the window and repeat the calculation of the processing interval to generate a smoothed value set of growth sequence;

[0074] The crop growth change dataset uses three time series data points: daily increments of leaf age, stem height, and internode length. A sliding window processing flow is constructed in chronological order, with each window covering three days of data. For example, July 2nd to July 4th, 2025, is selected as the first window. The daily increment vectors of leaf age are extracted as [0.3, 0.5, 0.6] leaves / day, stem height as [0.4, 0.6, 0.7] cm / day, and internode length as [0.5, 0.8, 0.9] cm / day. The average values ​​of these terms within each three-day window are calculated, where the average increment of leaf age is... Shoots / day, average stem height increase is cm / day, with an average intersegmental length variation of [value missing]. cm / day, record the average value vector as the smoothed value of the current window; then move the window to the right by one day and repeat the same calculation process. For example, if the window changes to July 3 to July 5, take three sets of daily increment vectors again and perform the above average calculation. The sliding process of the window needs to cover the entire monitoring period. For example, taking 12 consecutive days of data as an example, 10 windows will be constructed, and a set of smoothed data vectors will be generated for each window. Finally, the smoothed value set of the growth sequence will be formed, and the recorded fields are the window start date, the average leaf age increment, the average stem height, and the average internode length. This structure is used for subsequent structural parameter analysis and processing.

[0075] S103: Call the daily increment of leaf age and daily increase of stem height from the growth sequence smoothing value set, determine whether the change direction of adjacent dates is consistent, combine the consistent direction mark with the corresponding internode length change value, and generate a set of structural parameters.

[0076] The algorithm retrieves the daily increments of leaf age and stem height from the smoothed growth sequence set. It then compares the direction of change between any two consecutive days, defining the direction as "increase" or "decrease." The criterion is the sign of the difference between the current date value and the previous date value. If the signs match, the direction is the same. For example, if the leaf age increment is 0.45 on July 3rd and 0.52 on July 4th, the direction is continuous increase. Stem height increments of 0.62 and 0.65 respectively also indicate an increase. If the directions match, the algorithm is considered true, and the judgment value is set to 1; otherwise, it is considered false. 0; The directional consistency label is combined with the internode length change value on the same day to construct a structural parameter set. The structural parameters consist of fields such as date, directional consistency, and internode length change value. For example, on July 4, 2025, the directional consistency is 1, and the internode length change is 0.9 cm, so the structural parameter entry is [2025-07-04, 1, 0.9]. The specific judgment process requires calculating the change trend of the two sequences in the daily data separately, taking the difference between the two averages and performing a sign operation. The leaf age direction is... The direction of stem height is If the daily increment of leaf age and the daily increment of stem height have the same sign (i.e., both are positive or both are negative), then they are considered to be in the same direction; otherwise, if the change in leaf age is positive and the change in stem height is negative on a certain day, then the consistency of direction is 0. This operation is repeated for the entire sequence to construct a complete set of structural parameters.

[0077] Table 1: Daily Growth Increment and Structural Parameters for Growth Monitoring

[0078]

[0079] Table 1 shows the daily incremental values ​​of leaf age, stem height, and internode length extracted from multispectral sensor data analysis, and further logically determines whether the growth direction is consistent over two consecutive days.

[0080] The results indicate that during the period from July 2 to July 4, 2025, the plant leaf age and stem height were growing synchronously, and the internode length showed a continuous growth trend with a directional consistency of 1. This can be used to judge the trend of subsequent growth structure and as a basis for triggering disaster early warning.

[0081] Specifically, such as Figure 3 As shown, the specific steps of S2 are as follows:

[0082] S201: Input the structural parameter set into the Hidden Markov Model, calculate the transition probability between the current state and the original state based on the state data sequence, filter the channels with a difference greater than the state jump threshold, record the state channel information, calculate the jump trend value, and generate the stage jump tendency coefficient.

[0083] To input the structural parameter set into a Hidden Markov Model, the structural parameter set must first be decomposed. Assume the structural parameter set consists of data collected from a single maize plant during consecutive growth stages, including the number of leaves, internode length, stem diameter, leaf area, and leaf color index. Each parameter corresponds to a time series within a growth stage. For example, if the internode length of a maize plant on days 5 to 10 is 3.2cm, 3.5cm, 3.9cm, 4.3cm, 4.7cm, and 5.0cm, and the leaf color index is 0.72, 0.74, 0.77, 0.80, 0.84, and 0.86, then... First, the structural parameters need to be normalized to fit the state-space input format of the Hidden Markov Model. The number of states is set to three, corresponding to the early, middle, and late reproductive states. For each normalized time series, the state transition probability matrix is ​​calculated sequentially with the initial state series. This involves comparing the probability changes of the daily structural parameters in the current and previous states. The state series is then derived using the Viterbi algorithm, forming the state transition path between the current and original states. The positions where state transitions occur along the transition path are recorded and filtered. For channels with large fluctuations, the transition amplitude is defined as the absolute difference in state transition probabilities. For example, if the transition probability of a channel jumps from 0.2 to 0.65 on day 6, the transition amplitude is 0.45. When this value exceeds a preset state transition threshold, it is recorded as a transition channel. This threshold is set by adding a standard deviation to the average of the maximum transition amplitudes of similar channels in the first 100 samples. For example, if the average is 0.31 and the standard deviation is 0.06, the transition threshold is set to 0.37. If the transition amplitude of a channel is greater than 0.37, the channel number is recorded and matched with the corresponding value. Based on the information on state changes, the direction of change in the time series is extracted from the transition channels, i.e., it is determined whether the transition is from a low state to a high state or the opposite direction. By constructing a transition direction vector and statistically analyzing the proportion of consistent directions, a transition trend value is obtained. For example, if 7 out of 10 transition channels transition from a mid-stage state to a late-stage state, the transition trend value is 0.7. Then, combining the number of transition channels and the transition trend direction, a stage transition tendency coefficient is calculated through linear combination weighting. Assuming the proportion of transitions is 0.6 and the trend value is 0.7, the stage transition tendency coefficient is set as its weighted average.

[0084] ;

[0085] The results indicate that the parameters during this period show a trend toward a later state across multiple channels. This coefficient serves as the basis for subsequent analysis to determine synchronous fluctuations.

[0086] S202: Based on the stage jump tendency coefficient, extract the internode length sequence and the stem height sliding mean, perform second-order difference operation, screen the segments that fluctuate synchronously with the stem height sliding mean, calculate the segment variation slope value, cross-compare, and obtain the synchronous fluctuation discriminant.

[0087] Based on the stage jump tendency coefficient, the synchronicity between the internode length sequence and the stem height moving average, which are the structural parameters most closely related to the plant's growth status, was compared. The internode length sequence was represented by daily internode length data, and the stem height moving average was calculated using a 3-day moving average method. Example data is as follows: The stem height of a plant from day 5 to day 10 was 34.2, 35.5, 37.8, 40.6, 44.3, and 47.9 cm, respectively. Therefore, the moving average on day 7 was:

[0088] ;

[0089] This process is repeated to obtain the moving mean sequence. The second difference between the internode length and the moving mean is then calculated, i.e., for each day... The difference is calculated as follows:

[0090] ;

[0091] Taking the intersegmental length on day 7 as 3.9 cm, and the lengths on the days before and after as 3.5 cm and 4.3 cm respectively, the difference is:

[0092] ;

[0093] Applying the above calculations to the daily data, segments that fluctuate synchronously with the moving average of stem height are selected, i.e., time point sequences with the same direction of the difference sequence. For example, if the internode length and the second difference direction of the moving average are consistent from day 6 to day 8, it is determined to be a synchronous segment. Based on this, the slope value of the segment variation is calculated, that is, the slope value is obtained by fitting each segment of data through linear regression. Assuming that the moving averages from day 6 to day 8 are 36.5, 37.9, and 39.6 cm, respectively, corresponding to days 1, 2, and 3, the slope is:

[0094] ;

[0095] The same operation is performed on the intersegment length to obtain its variation slope value. By comparing whether the slope direction (positive or negative sign) and numerical amplitude of the two segments are on the same order of magnitude, if the absolute difference between the two slopes is less than 0.5 cm / day and the signs are consistent, it is marked as synchronous fluctuation. Finally, the ratio of the number of synchronous judgments in the synchronous segments to the total number of segments is calculated to generate the synchronous fluctuation discriminant. If there are 10 time periods in the sample and 8 segments meet the synchronization condition, the synchronous fluctuation discriminant is 0.8. This value is used to determine the level of synchronization after the stage jump.

[0096] S203: Based on the synchronous fluctuation discriminant, construct the cumulative sequence of daily leaf age increments, screen time nodes where the increments for two consecutive days are greater than the critical threshold of stage jump, identify the monitoring time window, and obtain stage labels;

[0097] To construct a cumulative sequence of daily leaf age increments based on synchronous fluctuation discriminant, it is necessary to first obtain the daily leaf age value, i.e., the actual number of unfolded leaves on the plant. The sampling period is once a day. Taking a corn plant as an example, the leaf ages from day 5 to day 10 are 4.1, 4.3, 4.5, 4.9, 5.2, and 5.5, respectively. The daily increment is a difference sequence: 0.2, 0.2, 0.4, 0.3, and 0.3. The cumulative sequence is constructed as 0.2, 0.4, 0.8, 1.1, and 1.4. Calculating the increments for two consecutive days yields the sequence: 0.4, 0.6, 0.7, and 0.6. The critical threshold for stage transition is set to 0.5, which is obtained by using the 75th percentile of the sample distribution. For example, the distribution of daily leaf age increments in the statistical sample is as follows:

[0098] Table 2: Distribution of Leaf Age Increment

[0099]

[0100] As shown in Table 2, the critical threshold is set as the mean between the median and the highest value, that is:

[0101] ;

[0102] If the actual increase in the number of plants monitored for two consecutive days exceeds this value, it is determined to be a jump node. For example, if the increase from day 8 to day 9 is 0.7 pieces, it meets the jump condition, and day 9 is marked as a jump node. The time period of the jump node constitutes the monitoring time window. The synchronous status of the structural parameters within the jump time window is marked as a stage label, which serves as the state sample label during structural analysis and model training.

[0103] Specifically, such as Figure 4 As shown, the specific steps of S3 are as follows:

[0104] S301: Call the transpiration rate threshold, root water potential response lag time and organ morphology integrity rate corresponding to the stage label, sort the stage labels by time, classify the transpiration rate by threshold, match the root water potential delay response and extract the curve, calculate the organ morphology integrity rate and align the parameter time series, and generate the time series parameter integrated value.

[0105] The integrated value of time series parameters is a comprehensive result of multiple physiological and meteorological time series data, reflecting the changes in physiological state and environmental factors;

[0106] The process involves retrieving the transpiration rate threshold, root water potential response lag time, and organ morphology integrity corresponding to each stage label. First, the plant growth stage identifiers are extracted from the time series data. For example, based on typical dates of different growth stages (such as vegetative growth, flowering, and grain-filling stages), stage labels are marked as T1 to T5. Based on this timeline, three parameters corresponding to each stage label are matched sequentially. Specifically, the transpiration rate threshold is set as the upper and lower limits of the transpiration rate per unit leaf area, such as an average daytime transpiration rate of 6.3 mmol·m³. -2 ·s -1 Based on this, the classification intervals can be divided as follows: below 4.0 is low transpiration (L), 4.0 to 7.5 is medium transpiration (M), and above 7.5 is high transpiration (H). Then, the root water potential response lag time is read and loaded. This time reflects the time delay between stomatal closure or opening and root pressure change. For example, if the grain-filling period sample is set to a 6-hour lag, then the transpiration change in the previous time window needs to correspond to the root water potential response 6 hours later. The curves are mapped using a time-shifting operation. Simultaneously, the organ morphology integrity rate is extracted. The morphology integrity rate is calculated using the area ratio after image segmentation or the vascular bundle connectivity index. Assuming the intact area of ​​the control plant's organs is 100 cm², and the area of ​​the current sample's leaf damaged by insects is 28 cm², then the integrity rate is... Next, the three parameters corresponding to the stage labels are constructed into a structure or matrix and arranged in time series. The transpiration rate is then classified and judged. During classification, the transpiration rate data is read point by point and classified into the L / M / H category array according to the above interval range. Then, based on the known lag time value, the root water potential value of the corresponding time period is shifted backward. The missing points in the shifted curve are filled in by interpolation. For example, when processing a sequence recorded once per hour with a 6-hour lag, 6 empty positions need to be filled in the head of the new curve and filled in by linear interpolation. Finally, the organ integrity rate of the multi-stage is aligned with the above two parameters in time order to ensure that the three parameters have a unified correspondence at the same time node and are integrated into a time series parameter matrix. The specific data is shown in Table 3.

[0107] Table 3: Integrated Values ​​of Timing Parameters

[0108]

[0109] As shown in Table 3, the lag treatment allows the root water potential value to be synchronized with the transpiration rate and organ integrity rate in the time dimension. In this way, a time-series parameter integration value of the transpiration intensity level, the root response delay-adjusted value, and the organ health status over a continuous time period is finally obtained.

[0110] S302: Based on the integrated value of time series parameters, extract the temperature, sunshine duration and wind speed sequences from meteorological data, compare them with the change trajectory of physiological parameters, apply the dynamic time warping algorithm to calculate their offset and mapping path, and obtain the meteorological alignment offset metric value.

[0111] The meteorological alignment offset metric is calculated using a dynamic time warping algorithm to determine the time offset and mapping path between meteorological factors and physiological parameters.

[0112] Based on the acquired integrated time-series parameters, meteorological data for the corresponding time period is extracted. Temperature, sunshine duration, and wind speed are converted into time-series arrays. For example, if recorded hourly throughout the day, the temperature sequence would be [28.1, 29.3, 30.5, 31.2]. Sunshine duration can be converted to hours based on radiation intensity; if the average radiation intensity is higher than 1200 W / m² during a certain period, it is considered an effective sunshine duration of 1 hour, resulting in a sunshine duration sequence such as [0.8, 1.0, 0.9, 0.6]. The wind speed sequence can be set using field weather station records as [1.5, 2.3, 3.1, 2.8]. After forming a meteorological parameter time-series matrix, it is compared with a physiological parameter time-series matrix. To identify the offset trends of meteorological parameters and physiological responses on the time axis, both multidimensional time series are normalized. Temperature is normalized to the 0-1 range using a linear normalization formula:

[0113] ;

[0114] Taking temperature as an example, if the minimum temperature in a day is 26.5℃ and the maximum temperature is 35.2℃, then the normalized value of 31.2℃ is:

[0115] ;

[0116] After normalization, the distance between nodes in the two sets of multidimensional time series is calculated by constructing an Euclidean distance matrix. Then, a path search is performed, using a greedy algorithm to compare the minimum distance points one by one to construct the offset path. The overall path offset is then calculated. For example, if the normalized meteorological path is P1=[0.2, 0.3, 0.5, 0.7] and the physiological path is P2=[0.1, 0.4, 0.6, 0.75], then the offset of node 1 is 0.1, the offset of node 2 is 0.1, the offset of node 3 is 0.1, and the offset of node 4 is 0.05. Finally, the average offset is taken as follows:

[0117] ;

[0118] This offset is the meteorological alignment offset metric, which indicates the degree of difference in the overall time series of the two trajectories.

[0119] S303: Based on the meteorological alignment offset metric, and based on the transpiration rate threshold range, root water potential response lag range, and organ morphology integrity rate variation range, perform secondary alignment of meteorological data and physiological parameters, perform matching index mapping, and generate a response mapping table.

[0120] Based on the calculated offset metric value, for example, 0.0875, it is compared with the set offset reference interval to determine whether to perform a secondary alignment operation. If the offset value is higher than the upper limit of the preset interval (set to 0.05), the subsequent processing flow is initiated. First, based on the transpiration rate interval, the real-time values ​​of transpiration rate at multiple times are reclassified, with the interval values ​​referring to the L / M / H classification rules. Then, based on the root water potential response lag interval, the current time point is pushed forward. For example, if the lag time is set to 6 hours during the grouting period, the meteorological data is aligned forward by 6 hours. Finally, based on the range of variation in organ morphology integrity rate, a distinction threshold is set. For example, an integrity rate greater than 0.85 is considered intact, less than 0.65 is considered severe damage, and between is considered moderate damage. If the integrity rate is continuously lower than 0.65 for a certain period, it needs to be isolated from the data of the high integrity rate stage through compensation matching. Subsequently, for the above three types of mapping conditions, physiological parameters and meteorological parameters are scanned time by time, index points are matched, their mapping numbers are recorded, and a response mapping table is generated. The original meteorological time nodes are adjusted to be consistent with the physiological nodes to form the final response mapping table, which serves as the input source for subsequent simulation or regulation.

[0121] Specifically, such as Figure 5 As shown, the specific steps of S4 are as follows:

[0122] S401: Based on the regression residual sequence in the response mapping table, obtain the temperature sequence, wind speed sequence and radiation intensity sequence, extract the temperature difference, wind speed slope and radiation range by sliding window, and normalize them to determine the trend direction, and generate a meteorological change trend direction sequence.

[0123] Normalization uses min-max normalization to standardize the sequence;

[0124] Based on the regression residual sequence in the response mapping table, the residual sequence is first divided chronologically to correspond to environmental data items at each sampling time point. The residual value at each time point is calculated as the difference between the predicted and measured evaporation rates. Then, raw meteorological data records are retrieved for each time point, extracting three types of raw environmental parameters: temperature, wind speed, and radiation intensity. These are used to form a meteorological parameter triplet sequence. Assuming an initial sampling frequency of 10 minutes, six triplet data sets can be obtained per hour. Based on this, a sliding window of length 6 is used for local analysis of each parameter type. Specifically, the sixth value within each window of the temperature sequence is subtracted from the first value, resulting in the value for that window. The temperature difference within the mouth; linear fitting is performed on each window in the wind speed sequence, and the slope of the fitted line is extracted as the wind speed slope value within that window; the maximum and minimum values ​​are extracted from the radiation intensity sequence, and their difference is calculated as the radiation range of that window. The temperature difference is expressed in degrees Celsius, the wind speed slope in m / s per minute, and the radiation range in W / m². After obtaining the local variation indices of the three types of parameters within the above windows, a maximum-minimum normalization method is uniformly adopted to map the values ​​of multiple parameters to the interval [−1, 1]. The direction of the change trend is determined by judging the sign of the normalized values, where positive values ​​indicate an upward trend and negative values ​​indicate a downward trend. The normalization formula is:

[0125] ;

[0126] in These are the original eigenvalues. These represent the minimum and maximum values ​​of the parameter over the entire period. For example, in a certain sampling window, the temperature is 31.2℃ in the 6th minute and 29.7℃ in the 1st minute, so the difference is 1.5℃. Assuming the original maximum temperature is 35℃ and the minimum temperature is 25℃, the normalized value is:

[0127] ;

[0128] The trend direction is marked as downward; the wind speed slope is set to 0.06 m / s / min after fitting, with a minimum slope of -0.1 and a maximum of 0.2. The normalized value is then:

[0129] ;

[0130] Marked as rising; radiation range is 920W / m² − 610W / m² = 310W / m², original range 610~1200W / m², normalized to:

[0131] ;

[0132] Marked as rising, the process iterates through the time windows sequentially, constructing a combination of three trend directions under each sliding window, and finally outputting a time series array of continuous trend direction combinations for subsequent consistency filtering.

[0133] S402: Based on the meteorological change trend direction sequence, the trend direction of meteorological parameters is symbolically encoded. Combined with the evaporation rate trend direction sequence, the consistency of the two sequences in the same period is compared. Time window segments are selected according to the consistency criteria to obtain the set of trend direction consistent segment numbers.

[0134] Based on the meteorological trend direction sequence, the normalized results of temperature difference, wind speed slope, and radiation range within the window are first labeled with symbols. "1" represents an upward trend, "−1" represents a downward trend, and "0" represents a stable trend when the change amplitude is less than a threshold. The criterion for trend stability is that the absolute value of the normalized value is less than 0.1. For example, when the normalized temperature value is −0.08, the wind speed is 0.05, and the radiation is −0.03, the corresponding three trend directions are all assigned the value "0", forming a coding group such as [0, 0, 0]. This process generates directional coding sequences for each of the three trend items. Subsequently, the evaporation rate trend direction sequence is extracted simultaneously, using the same sliding window strategy to form a one-to-one correspondence with the meteorological trend direction. For each time segment sequence, the signs of the three meteorological direction codes and the evaporation rate trend direction are compared item by item. For example, if the evaporation direction is 1 and the meteorological direction group is [1, −1, 1], it is considered partially consistent. The consistency judgment criterion is further set as follows: if at least two of the three trend items are consistent with the evaporation direction, it is judged as "trend consistent", otherwise it is "inconsistent". After executing this criterion, the window segments are traversed, and the windows that meet the consistency criterion are recorded and numbered. Finally, the set of trend direction consistent segment numbers is output. For example, if the window number [12, 13, 14, 18, 19, 20] corresponds to a trend consistent window, then the number set is {12, 13, 14, 18, 19, 20}.

[0135] S403: Call the set of segment numbers with consistent trend direction, mark the start and end indices of the corresponding windows in the original sequence, extract the time index and establish a trend overlap mapping table, classify and organize the continuous segments under the time axis, and obtain the impact window segments.

[0136] The set of trend-oriented consistent segment numbers is called, and the corresponding start and end time indices are extracted from the original response mapping table one by one. The start index of each number corresponding to the window is (number-1)×window step size, and the end index is the start index+window length−1. For example, when the number is 12, the window step size is 2, so the start index is 22, and the window length is 6, so the end index is 27, resulting in the window start and end index [22, 27]. Then, a trend overlap mapping table is established, and trend-oriented consistent windows are classified and merged according to time index. For consecutive numbered windows such as [12, 13, 14], the start index is the start of window 12, and the end index is the end of window 14, which is organized into a continuous trend segment [22, 33]. Non-continuous numbers are recorded as multiple independent segments. After the mapping table is completed, the overlapping trend segments and corresponding index information are output. Further, through classification and organization, each type of continuous segment number is marked separately, and finally, the impact window segment set is output.

[0137] Table 4: Impact Window Segment Index Table

[0138]

[0139] As shown in Table 4, three consecutive time segments with consistent trend directions and forming impact windows are listed. The segment length is the total coverage length of multiple consecutive sliding windows. Combined with the start and end indices, these can be used for subsequent sequence feature extraction and response adjustment operations.

[0140] Specifically, such as Figure 6 As shown, the specific steps of S5 are as follows:

[0141] S501: Based on the relationship between the absolute value of the temperature change amplitude within the impact window segment and the evaporation rate threshold in the response mapping table, extract the temperature change sequence and match the threshold interval, determine the location of the change signal node, calculate the distribution range of the dense response segment on the time axis, and generate the temperature change response amplitude value.

[0142] The temperature abrupt change response amplitude value refers to the degree of influence of drastic temperature changes on plant transpiration, which is calculated by comprehensively considering factors such as temperature change slope, transpiration threshold matching, and response density.

[0143] Based on the relationship between the absolute value of the temperature abrupt change amplitude within the impact window segment and the evaporation rate threshold in the response mapping table, the temperature change sequence recorded on the monitoring time axis is segmented and extracted with a sampling interval of 1 minute. During the extraction process, a sliding window of 5 minutes is used to calculate the maximum temperature difference within each window. When this difference exceeds the set abrupt change identification benchmark (e.g., 5℃ / 5min), the start time of the window is recorded as a candidate abrupt change node. Then, the slope value of the temperature change at the candidate node is calculated using the slope formula. The rate of change was calculated segment by segment based on the temperature difference within a 1-minute time span, and the transpiration rate response threshold in the response mapping table was set to ±0.2 mmol·m. -2 ·s -1 These correspond to the mild and strong transpiration response ranges, respectively. By matching the slope values ​​one by one with the relationship between the transpiration rate change and the response threshold, if the rate of change is within the range of ±0.5℃ / min to ±1.5℃ / min, the mapped transpiration change is between ±0.15 mmol·m⁻². -2 ·s -1 With ±0.3 mmol·m -2 ·s -1 Between these points, the response interval is considered. If the slope value is greater than ±1.5℃ / min and the fluctuation trend is consistent within 2 consecutive minutes, it is identified as a sudden change signal node. For the sudden change signal node, the number of response segments within the 2 minutes before and after it is further counted. If the number of dense segments exceeds 3, that is, if there are more than 3 rate sudden change responses within 5 minutes, the segment is determined to be a dense response segment. At the same time, the start and end points of the time interval are recorded. The slope value within each dense response segment is weighted and averaged. Combined with the mapping table level of each segment in the response interval, the impact of the sudden change is weighted, accumulated, and averaged. Finally, the amplitude value of the temperature sudden change response is output. For example, in a temperature change sequence, the sudden change starts at 08:10 and ends at 08:16. The slopes recorded in this segment are 1.2, 1.5, 1.8, and 1.6℃ / min, respectively, corresponding to transpiration response intensities of 0.21, 0.27, 0.31, and 0.29 mmol·m. -2 ·s -1 The amplitude of the temperature change response can then be calculated as 0.27 mmol·m⁻². -2 ·s -1 .

[0144] S502: Based on the amplitude value of the temperature change response, the wind speed change slope and organ integrity rate coefficient sequence are matched to construct a response comparison chain. The change ratio is compared by continuous segment and the offset difference amplitude is extracted to obtain the wind speed-organ correlation difference degree.

[0145] Based on the temperature change response amplitude value obtained in the previous step, it is matched segment by segment with the wind speed change slope sequence and the plant organ integrity coefficient array at the corresponding time. First, the time axis is aligned, and the wind speed monitoring data is differentially processed every minute to calculate the wind speed change slope. For example, if the wind speed increases from 1.8 m / s to 3.0 m / s between 08:12 and 08:13, then:

[0146] ;

[0147] Subsequently, the standard deviation of the wind speed change slope within each consecutive 3-minute interval was calculated as the fluctuation coefficient. Then, organ integrity coefficients are extracted from the monitoring results corresponding to different plant organs (such as leaves, stems, calyxes, etc.). and The parameter was measured in a dry hot air experiment and is shown in Table 5. The value was assigned when the blade integrity was poor. The stem is This reflects the degree of response of different organs to changes in wind speed, and the product of the slope of wind speed change and its proportion in the response chain is calculated.

[0148] Table 5: Wind Speed ​​Organ Response Factors

[0149]

[0150] Table 5 shows the values ​​of wind speed change slope, wind speed fluctuation coefficient, wind speed weighting factor, and fluctuation weighting factor in the differentiated time periods.

[0151] The formula for calculating the wind speed organ correlation difference is as follows:

[0152] ;

[0153] in, Represents the degree of difference in wind speed among organs, in m / s². Representing the The slope of wind speed change over a continuous period of time, expressed in m / s². Represents all The average slope of the wind speed change in the segment, in m / s². Representing the The fluctuation coefficient of the slope of wind speed change per unit time within a segment, expressed in m / s². Representing the The wind speed slope weighting factor in the segment response chain is a dimensionless parameter. Representing the The weighting factor for wind speed fluctuations in the segment response chain is a dimensionless parameter. This represents the total number of consecutive comparison segments divided within the response chain; it is an integer count and has no unit.

[0154] Calculate using the above data and the formula:

[0155] ;

[0156] ;

[0157] ;

[0158] ;

[0159] ;

[0160] The results indicate that there is a moderate degree of response difference between plant organ integrity and wind speed change in this response chain, with a specific wind speed-organ correlation difference of 0.271 m / s².

[0161] S503: Call the time difference matching results of wind speed organ correlation difference and soil water potential, calculate the response index by combining the fluctuation range of radiation intensity within the section, construct a weighted structure vector for sorting and extract the corresponding disaster impact level, match the intervention instruction set bound to the response level, and output the early warning decision instruction.

[0162] The response time difference between wind speed organ correlation difference (0.271 m / s²) and soil water potential changes within the same segment was used. Based on the transpiration delay principle, the disturbance response time threshold was extracted. For example, if the soil water potential decreases at 08:20 after a sudden change at 08:12, the response time difference is set to 8 minutes, which is considered a significantly delayed response (>6 minutes as the criterion). Furthermore, the radiation intensity variation range within this time period was retrieved, and the difference between maximum and minimum radiation values ​​was calculated in 2-minute windows. If the radiation intensity fluctuated from 220 W / m² to 330 W / m² between 08:10 and 08:20, an amplitude of 110 W / m², it was classified as a high-amplitude fluctuation. This was combined with the temperature sudden change response amplitude of 0.27 mmol·m². -2 ·s -1 The correlation between wind speed and organ is 0.271 m / s². Using the following weighted vectors: temperature weight 0.4, wind speed weight 0.35, and radiation weight 0.25, the constructed structure vector is [0.27×0.4, 0.271×0.35, 110×0.25 / 1000] = [0.108, 0.09485, 0.0275]. The weighted summation yields a response index value of 0.23035. Based on the preset interference level table, if this value is greater than 0.2, it is classified as a Level 3 interference level. The corresponding early warning response instruction set is: "Activate spray cooling + shading control + segmental water control," thus outputting the early warning decision instruction.

[0163] like Figure 7 As shown, an agricultural disaster early warning and decision-making system includes:

[0164] The temporal smoothing module is used to acquire the daily increment of leaf age, daily increase of stem height, and daily change of internode length through multispectral sensors. It uses a sliding window algorithm to perform temporal smoothing processing, performs directional consistency marking operation to generate a set of structural parameters, and passes them to the step identification module.

[0165] The stage transition identification module is used to input the set of structural parameters into the Hidden Markov Model to calculate the stage transition probability, determine the critical point based on the sliding mean of stem height and the second difference of internode length, accumulate the daily increment of leaf age and compare it with the preset threshold, generate stage labels, and pass them to the time sequence alignment module.

[0166] The time-series alignment module is used to call the transpiration rate threshold, root water potential response lag time, and organ morphological integrity rate corresponding to the stage label. It uses a dynamic time warping algorithm to align meteorological and physiological time-series data, generate a response mapping table, and pass it to the trend overlap module.

[0167] The trend overlap module is used to call the regression residual sequence in the response mapping table, and use the sliding trend overlap method to determine the trend direction of the temperature change amplitude, wind speed change slope, and radiation intensity range. It extracts the overlapping segment with the evaporation rate trend direction, marks the impact window segment, and passes it to the disaster assessment module.

[0168] The disaster assessment module is used to prioritize disaster levels by comparing the magnitude of temperature change within the impact window segment with the evaporation rate threshold, performing linear regression calculations of wind speed change slope and organ integrity rate, and combining the soil water potential time difference matching results with the radiation intensity range. It then calls a weighted sorting method to output early warning decision instructions.

[0169] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A decision-making method for early warning of agricultural disasters, characterized in that, Includes the following steps: S1: The daily increment of crop leaf age, daily growth of stem height, and daily variation of internode length are obtained by multispectral sensors. The time-series smoothing is performed by a sliding window algorithm. The daily increment of leaf age and daily growth of stem height are marked with directional consistency to generate a set of structural parameters. S2: Input the set of structural parameters into the Hidden Markov Model to calculate the stage transition probability, determine the critical point based on the sliding mean of stem height and the second difference of internode length, accumulate the daily increment of leaf age, and mark the stage transition trigger point when the cumulative value exceeds the stage threshold for two consecutive days, and output the stage label. S3: Call the transpiration rate threshold, root water potential response lag time, and organ morphological integrity rate corresponding to the stage label, and use the dynamic time warping algorithm to align the meteorological data with the time series of physiological parameters to generate a response mapping table; S4: Based on the regression residual sequence in the response mapping table, the sliding trend overlap method is used to determine the trend direction of the temperature change amplitude, wind speed change slope and radiation intensity range of meteorological parameters, and to extract the segments that overlap with the evaporation rate trend direction in the response mapping table and mark them as impact window segments. S5: Based on the relationship between the temperature change amplitude within the impact window segment and the absolute value of the evaporation rate threshold in the response mapping table, the linear relationship between the wind speed change slope and the organ integrity rate corresponding to the stage label, and the joint relationship between the soil water potential time difference matching result and the radiation intensity range within the impact window segment, the weighted sorting method is invoked to prioritize the disaster level and output the early warning decision instruction.

2. The agricultural disaster early warning decision-making method according to claim 1, characterized in that, The structural parameter set includes directional consistency labels, time-smoothed parameter values, and leaf age growth rate. The stage labels include stage identification identifiers, transition judgment conditions, and cumulative thresholds for stage transitions. The response mapping table includes meteorological variable time alignment results, plant physiological parameter matching dimensions, and trend residual parameters. The impact window segments include temperature disturbance segments, wind speed anomaly segments, and radiation intensity variation segments.

3. The agricultural disaster early warning decision-making method according to claim 1, characterized in that, The specific steps of S1 include: S101: Acquire crop leaf reflectance, stem reflectance and internode reflectance data for consecutive days using a multispectral sensor, extract the daily variation of leaf number, stem height and internode height for adjacent days, convert them into daily variations of leaf age, stem height and internode length, and generate a crop growth variation dataset. S102: Call the three sequence data in the crop growth change dataset, set three dates as a sliding window, calculate the average value of leaf age increment, stem height growth and internode length change within the window, shift the window and repeat the calculation of the processing interval to generate a smoothed value set of growth sequence; S103: Call the daily increment of leaf age and daily increase of stem height in the growth sequence smoothing value set, determine whether the change direction of adjacent dates is consistent, combine the consistent direction mark with the corresponding internode length change value, and generate a structural parameter set.

4. The agricultural disaster early warning decision-making method according to claim 3, characterized in that, The specific steps of S2 include: S201: Input the set of structural parameters into the Hidden Markov Model, calculate the transition probability between the current state and the original state based on the state data sequence, filter the channels with a difference greater than the state jump threshold, record the state channel information, calculate the jump trend value, and generate the stage jump tendency coefficient. S202: Based on the stage jump tendency coefficient, extract the internode length sequence and the stem height sliding mean, perform second-order difference operation, screen the segments that fluctuate synchronously with the stem height sliding mean, calculate the segment variation slope value, cross-compare, and obtain the synchronous fluctuation discrimination value; S203: Based on the synchronous fluctuation discrimination quantity, construct the daily incremental sequence of leaf age, screen the time nodes where the increment is greater than the critical threshold of stage jump for two consecutive days, identify the monitoring time window, and obtain the stage label.

5. The agricultural disaster early warning decision-making method according to claim 4, characterized in that, The state transition threshold is a critical value for determining whether the change in the channel state transition probability has reached a significant level of difference. The stage jump tendency coefficient measures the overall jump trend strength of multiple channel states within the current stage and predicts the probability of state reversal. The synchronous fluctuation discrimination quantity quantifies the numerical index of the consistency of fluctuations between the internode length and the moving average of stem height through the second-order difference synchronicity. The stage transition critical threshold is used to screen time points where leaf age increases significantly over two consecutive days, marking growth stage mutation events.

6. The agricultural disaster early warning decision-making method according to claim 4, characterized in that, The specific steps of S3 include: S301: Call the transpiration rate threshold, root water potential response lag time and organ morphology integrity rate corresponding to the stage label, sort the stage labels by time, classify the transpiration rate by threshold, match the root water potential delay response and extract the curve, calculate the organ morphology integrity rate and align the parameter time sequence, and generate the time sequence parameter integrated value. S302: Based on the integrated value of the time series parameters, extract the temperature, sunshine duration and wind speed sequences from the meteorological data, compare them with the trajectory of physiological parameter changes, apply the dynamic time warping algorithm to calculate their offset and mapping path, and obtain the meteorological alignment offset metric value. S303: Based on the meteorological alignment offset metric, and according to the transpiration rate threshold range, root water potential response lag range, and organ morphology integrity rate variation range, perform secondary alignment of meteorological data and physiological parameters, perform matching index mapping, and generate a response mapping table.

7. The agricultural disaster early warning decision-making method according to claim 6, characterized in that, The specific steps of S4 include: S401: Based on the regression residual sequence in the response mapping table, obtain the temperature sequence, wind speed sequence and radiation intensity sequence, extract the temperature difference, wind speed slope and radiation range by sliding window, and normalize them to determine the trend direction, and generate a meteorological change trend direction sequence. S402: Based on the meteorological change trend direction sequence, the meteorological parameter trend direction is symbolically encoded, and combined with the evaporation rate trend direction sequence, the consistency of the two sequences in the same time period is compared. Time window segments are selected according to the consistency standard to obtain the set of trend direction consistent segment numbers. S403: Call the set of segment numbers with consistent trend direction, mark the start and end indices of the corresponding windows in the original sequence, extract the time index and establish a trend overlap mapping table, classify and organize the continuous segments under the time axis, and obtain the impact window segments.

8. The agricultural disaster early warning decision-making method according to claim 1, characterized in that, The early warning decision instructions include disaster level assessment factors, priority ranking weights, and response strategy recommendations.

9. The agricultural disaster early warning decision-making method according to claim 8, characterized in that, The specific steps of S5 include: S501: Based on the relationship between the absolute value of the temperature change amplitude within the impact window segment and the evaporation rate threshold in the response mapping table, extract the temperature change sequence and match the threshold interval, determine the location of the change signal node, calculate the distribution range of the dense response segment on the time axis, and generate the temperature change response amplitude value. S502: Based on the temperature change response amplitude value, match the wind speed change slope and organ integrity rate coefficient sequence, construct a response comparison chain, compare the change ratio by continuous segment and extract the offset difference amplitude to obtain the wind speed-organ correlation difference degree. S503: Call the time difference matching results of the wind speed organ correlation difference and soil water potential, calculate the response index by combining the fluctuation range of radiation intensity within the section, construct a weighted structure vector for sorting and extract the corresponding disaster impact level, match the intervention instruction set bound to the response level, and output the early warning decision instruction.

10. An agricultural disaster early warning and decision-making system, characterized in that, The system is used to implement the agricultural disaster early warning decision-making method according to any one of claims 1-9, and the system includes: The temporal smoothing module is used to acquire the daily increment of leaf age, daily increase of stem height, and daily change of internode length through multispectral sensors. It uses a sliding window algorithm to perform temporal smoothing processing, performs directional consistency marking operation to generate a set of structural parameters, and passes them to the step identification module. The stage transition identification module is used to input the set of structural parameters into a hidden Markov model to calculate the stage transition probability, determine the critical point based on the sliding mean of stem height and the second difference of internode length, accumulate the daily increment of leaf age and compare it with a preset threshold, generate stage labels, and pass them to the time alignment module. The time-series alignment module is used to call the transpiration rate threshold, root water potential response lag time, and organ morphological integrity rate corresponding to the stage label, and uses a dynamic time warping algorithm to align meteorological and physiological time-series data, generate a response mapping table, and pass it to the trend overlap module. The trend overlap module is used to call the regression residual sequence in the response mapping table, use the sliding trend overlap method to determine the trend direction of temperature change amplitude, wind speed change slope and radiation intensity range, extract the overlapping segment with the evaporation rate trend direction, mark the impact window segment, and pass it to the disaster assessment module. The disaster assessment module is used to prioritize disaster levels by comparing the magnitude of temperature change with the evaporation rate threshold within the impact window segment, performing linear regression calculations of wind speed change slope and organ integrity rate, and combining the soil water potential time difference matching results with the radiation intensity range, and outputs early warning decision instructions by calling a weighted sorting method.

Images