A method for predicting dry matter accumulation of corn based on modeling of root bleeding characteristics

Abstract

The present application relates to the field of agricultural prediction, and more particularly to a corn dry matter accumulation prediction method based on root bleeding characteristics modeling, which comprises: obtaining complete multi-dimensional time series data set by synchronizing preprocessing of multi-source time series data; obtaining adaptive input damping factor by dynamic baseline and local variability analysis of original bleeding data, and performing preliminary correction of bleeding data through the adaptive input damping factor; obtaining posterior compensation factor by peak event and environmental constraint analysis of smoothed bleeding data, and performing secondary correction of smoothed bleeding data through the posterior compensation factor; modeling and predicting the corrected bleeding characteristics, and obtaining the corn dry matter accumulation rate through the prediction model; and obtaining the corn yield prediction result by accumulating and integrating the corn dry matter accumulation rate, thereby solving the problem of distorted prediction results caused by transient peak of bleeding data in the post-drought rehydration scenario.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of agricultural prediction, and particularly relates to a corn dry matter accumulation prediction method based on root bleeding characteristic modeling. BACKGROUND

[0002] As an important food crop worldwide, the yield level of corn is directly related to the agricultural production efficiency and food security. In the growth process of corn, the accumulation amount of dry matter is a core physiological index for measuring the growth of crops and predicting the final yield. In order to realize the precision and scientific management of field work, it has significant application value and popularization significance to develop a method for continuously and accurately predicting the accumulation of corn dry matter. Existing research and practice show that establishing a prediction model based on the physiological activity characteristics of crop roots is an important technical path to realize the prediction of dry matter accumulation. Root is the key organ for crop to absorb water and nutrients, and its physiological activity is closely related to the accumulation of aboveground biomass. Root bleeding fluid, as a comprehensive index reflecting root activity, can effectively represent the water absorption and nutrient supply capacity of roots. Therefore, in the existing technology, the bleeding fluid related characteristic data is usually collected, and combined with the actual dry matter accumulation of corn, a prediction model from bleeding characteristics to dry matter accumulation is established by using regression analysis, support vector machine or random forest and other machine learning algorithms.

[0003] However, in practical applications, the sap flow data collected by sensors is a time series data with much noise and frequent fluctuations, so data preprocessing must be performed before modeling. In order to filter out noise and extract the baseline signal that can reflect the stable growth trend of plants, the Savitzky-Golay (SG) filter is widely used in the processing of crop physiological time series because it can smooth the data while maintaining the signal form characteristics. The principle of the SG filter is to smooth the data points in the sliding window by local polynomial least squares fitting, which has become a common and mature technical means in this field. However, the inventors found in research and application that the SG filter has inherent defects when processing some typical scenarios in corn fields. In particular, after the corn is subjected to drought stress for a period of time, when sudden heavy rain or concentrated irrigation occurs, the root system will exhibit a compensatory water absorption phenomenon, resulting in a transient peak in the sap flow time series with a short duration and a high amplitude. The SG filter cannot distinguish whether the peak is a physiological artifact caused by short-term stress response or represents the real dry matter growth trend of the aboveground part, so it will mechanically include the extreme value point in the fitting process, causing the smoothed result to be significantly raised in this interval. When the subsequent prediction model uses the distorted smoothed data for inference, it will misjudge as a pulse-like increase in dry matter accumulation rate, leading to systematic overestimation of the stage prediction result and gradually accumulating errors in the growth cycle, thereby seriously affecting the accuracy of yield prediction. Therefore, the existing technology cannot effectively distinguish between physiological artifacts caused by compensatory water absorption and real growth trends in typical scenarios such as rehydration after drought, resulting in distorted and accumulated errors in the prediction results of corn dry matter accumulation. Therefore, it is urgent to identify and correct such abnormal peak interference. SUMMARY

[0004] Therefore, the present application aims to provide a corn dry matter accumulation prediction method based on root bleeding feature modeling to solve the problem of distorted prediction results caused by transient peaks in bleeding data in the rehydration scenario after drought.

[0005] To achieve the above-mentioned purpose, the technical scheme of the present application is as follows:

[0006] A corn dry matter accumulation prediction method based on root bleeding feature modeling, the method comprising the following steps:

[0007] Step S1: obtaining complete multi-dimensional time series data set by synchronizing preprocessing of multi-source time series data;

[0008] Step S2: obtaining adaptive input damping factor by dynamic baseline and local variability analysis of original bleeding data, and performing preliminary correction of bleeding data by adaptive input damping factor;

[0009] Step S3: Obtain a posterior compensation factor by performing peak event and environmental constraint analysis on the smoothed sap flow data, and perform secondary correction on the smoothed sap flow data by the posterior compensation factor;

[0010] Step S4: Obtain the corn dry matter accumulation rate by modeling and predicting the modified sap flow characteristics, and obtain the corn dry matter accumulation rate by the prediction model;

[0011] Step S5: Obtain the corn yield prediction result by cumulatively integrating the corn dry matter accumulation rate.

[0012] Further, the complete multi-dimensional time series data set obtained by synchronizing and preprocessing the multi-source time series data includes:

[0013] After the corn enters the jointing stage, a representative plant is selected to install a sap flow meter, and the sap flow data sequence of the stem is continuously collected and recorded at a sampling frequency of every 30 minutes; soil moisture sensors are arranged in the main distribution area of the plant roots, and the soil moisture data sequence is collected and recorded at the same sampling frequency as the sap flow rate collection; a photosynthetically active radiation sensor is installed above the corn canopy or in the field weather station, and the photosynthetically active radiation data sequence is collected and recorded at the same sampling frequency as the sap flow rate collection; and after the above data collection is completed, the sap flow rate data sequence, the soil moisture data sequence and the photosynthetically active radiation data sequence are time-stamped and aligned, and the missing data points are filled by the nearest point interpolation or linear interpolation method, to obtain a complete and synchronized multi-dimensional time series data set.

[0014] Further, the adaptive input damping factor is obtained by analyzing the dynamic baseline and local variation degree of the original sap flow data, and the original sap flow data is preliminarily corrected by the adaptive input damping factor, including:

[0015] The number of evaluations for dynamic growth baseline evaluation and local data fluctuation evaluation is set; for any target time in the original sap flow data, the median of the selected continuous sap flow data points before the target time is selected according to the number of evaluations, and the median of the selected continuous sap flow data points is taken as the dynamic growth baseline of the target time; the standard deviation of the selected continuous sap flow data is taken as the local fluctuation degree of the target time;

[0016] The adaptive input damping factor is obtained by jointly analyzing the deviation degree of the original sap flow data according to the dynamic growth baseline and the local fluctuation degree, and the original sap flow data is preliminarily corrected.

[0017] Further, the adaptive input damping factor is obtained by jointly analyzing the deviation degree of the original sap flow data according to the dynamic growth baseline and the local fluctuation degree, and the original sap flow data is preliminarily corrected, including:

[0018] For any target moment in the original sap flow data, the calculation result of subtracting the dynamic growth baseline of the target moment from the sap flow data of the target moment is taken as the first baseline deviation evaluation of the target moment, the calculation result of adding the local fluctuation degree of the target moment to a very small positive number is taken as the denominator, and the corresponding fraction is taken as the first input damping evaluation of the target moment; the inverse of the square of the first input damping evaluation of the target moment is exponentially mapped with the natural constant as the base, and the corresponding mapping result is taken as the adaptive input damping factor of the target moment;

[0019] The original sap flow data is modified based on the baseline deviation through the adaptive input damping factor to obtain the preliminary modified sap flow data.

[0020] Further, the original sap flow data is modified based on the baseline deviation through the adaptive input damping factor to obtain the preliminary modified sap flow data, comprising:

[0021] For any target moment in the original sap flow data, the calculation result of subtracting the dynamic growth baseline from the sap flow data of the target moment is taken as the first baseline deviation evaluation of the target moment, the calculation result of multiplying the adaptive input damping factor of the target moment by the first baseline deviation evaluation of the target moment is taken as the first modification evaluation of the target moment, and the calculation result of adding the dynamic growth baseline of the target moment to the first modification evaluation of the target moment is taken as the preliminary modified sap flow data of the target moment.

[0022] Further, the original sap flow data is modified based on the baseline deviation through the adaptive input damping factor to obtain the preliminary modified sap flow data, comprising:

[0023] The sliding window length and the polynomial fitting order of the SG filter are set, and the preliminary modified sap flow data is smoothed by SG filtering to obtain the smoothed sap flow data.

[0024] The peak value compensation reference quantity is obtained by identifying and calculating the peak value event of the smoothed sap flow data.

[0025] The environmental adjustment factor is obtained by analyzing and processing the soil humidity data and the photosynthetically active radiation data.

[0026] The posterior compensation factor is obtained by jointly processing the peak value compensation reference quantity and the environmental adjustment factor, and the smoothed sap flow data is modified again.

[0027] Further, the peak value compensation reference quantity is obtained by identifying and calculating the peak value event of the smoothed sap flow data, comprising:

[0028] set a peak evaluation coefficient; for any target moment in the original sap flow data, multiply the peak evaluation coefficient by the local fluctuation degree of the target moment and add the dynamic growth baseline of the target moment to obtain a peak event starting evaluation threshold of the target moment; for any target moment, if the value of the smoothed sap flow data of the target moment is greater than the peak event starting evaluation threshold of the target moment, the target moment is determined as the starting moment of the peak event; the first moment smaller than the peak event starting evaluation threshold after the starting moment of the peak event is taken as the ending moment of the peak time corresponding to the starting moment of the peak event; all peak events in the original sap flow data are obtained through the evaluation of the starting moment and the ending moment of the peak event;

[0029] subtract the sap flow data of any target moment in the original sap flow data from the preliminary corrected sap flow data of the target moment to obtain a first peak flux evaluation of the target moment;

[0030] for any target peak event in all peak events in the original sap flow data, add the first peak flux evaluation of all moments in the target peak event to obtain a peak compensation reference quantity of the target peak event.

[0031] Further, the environmental adjustment factor is obtained by performing environmental constraint analysis and processing on the soil humidity data and the photosynthetically active radiation data, and the environmental constraint analysis and processing comprises:

[0032] for any target peak event in all peak events in the original sap flow data, the moment of the maximum value of the original sap flow data in the target peak event is taken as the transient peak moment of the target peak event;

[0033] for the transient peak moment of any target peak event, the average value of the soil humidity data in the interval formed by the ten days before the transient peak moment to the four days before the transient peak moment is taken as the reference soil humidity of the target peak event;

[0034] for the transient peak moment of any target peak event, the maximum photosynthetically active radiation data within the seven days before the transient peak moment is taken as the reference photosynthetically active radiation value of the target peak event;

[0035] for any target moment in any target peak event, the calculation result of dividing the soil humidity data of the target moment by the reference soil humidity of the target peak event is taken as the first soil humidity evaluation of the target moment; the calculation result of dividing the photosynthetically active radiation data of the target moment by the reference photosynthetically active radiation value of the target peak time is taken as the first photosynthetically active radiation evaluation of the target moment;

[0036] The calculation result of multiplying the first soil humidity evaluation of the target moment in the target peak event and the first photosynthetically active radiation evaluation is taken as the environmental adjustment factor of the target peak event at the target moment.

[0037] Further, the posterior compensation factor is obtained by jointly processing the peak compensation reference quantity and the environmental adjustment factor, and the smoothed girdling data is secondarily revised, including:

[0038] A physiological response time constant is set; for any target moment in any target peak event, the calculation result of subtracting the target moment from the moment of the instantaneous peak of the target peak event is taken as the numerator, the square of the physiological response time constant is taken as the denominator, and the corresponding fraction is taken as the first pulse evaluation of the target moment; the calculation result of subtracting the target moment from the moment of the instantaneous peak of the target peak event is taken as the numerator, the physiological response time constant is taken as the denominator, the corresponding fraction is subjected to exponential mapping with the natural constant as the base number, and the corresponding fraction is taken as the second pulse evaluation of the target moment; the calculation result of multiplying the first pulse evaluation and the second pulse evaluation of the target moment is taken as the pulse response evaluation of the target moment;

[0039] The calculation result of multiplying the environmental adjustment factor of the target peak event at the target moment, the peak compensation reference quantity of the target peak event and the pulse response evaluation of the target peak event at the target moment is taken as the posterior compensation factor of the target peak event at the target moment;

[0040] The calculation result of adding the posterior compensation factor of the target peak event at the target moment and the corresponding target moment data in the smoothed girdling data is taken as the girdling data after secondary revision.

[0041] Compared with the prior art, the present application has the following advantages:

[0042] The corn dry matter accumulation prediction method based on root bleeding characteristics modeling provided in the present application realizes effective suppression of the instantaneous peak value of bleeding liquid flow under the post-drought rehydration scenario and dynamic recovery of physiological compensation information by introducing a double optimization mechanism of an adaptive input damping factor and a posterior compensation factor in the bleeding liquid flow data processing process. In the actual field environment, the corn root system is affected by water stress and environmental mutations, and the data collected by the sensor often has short-time high-amplitude fluctuations. The traditional filtering method is easy to misjudge such peak values as real growth trends, resulting in systematic deviation of the prediction model. The present application can automatically identify and suppress abnormal fluctuations by joint modeling of dynamic baseline and local variability, so that the processed data is closer to the real law of corn growth, thereby improving the stability and reliability of data input while maintaining data fidelity. Further, the present application not only suppresses the disturbance of abnormal peak values to data smoothing, but also injects the potential physiological gain brought by environmental improvement into the prediction model through the construction of a compensation factor. The compensation process not only considers the peak event intensity, but also fuses environmental constraint factors such as soil humidity and photosynthetically active radiation, and dynamically adjusts the physiological response rhythm of crops, so as to more truly reflect the compensatory growth trend of corn after water recovery. The dry matter accumulation prediction results obtained thereby show higher accuracy in complex and variable field environments, and can provide reliable support for growth assessment, field management optimization and yield prediction in agricultural production. BRIEF DESCRIPTION OF DRAWINGS

[0043] The accompanying drawings, which form a part of this application, are included to provide a further understanding of the application and are incorporated in and constitute a part of this application. The embodiments of the present application and its description are used to explain the present application and are not used to limit the present application. In the drawings:

[0044] Figure 1 The method flowchart of the corn dry matter accumulation prediction method based on root bleeding characteristics modeling provided in the present application. DETAILED DESCRIPTION

[0045] The present application will be described in detail below with reference to the accompanying drawings and embodiments.

[0046] Reference Figure 1 The method flowchart of the corn dry matter accumulation prediction method based on root bleeding characteristics modeling provided in the present application. Figure 1 The corn dry matter accumulation prediction method based on root bleeding characteristics modeling can include:

[0047] Step S1, obtaining a complete multi-dimensional time series data set by synchronizing and preprocessing multi-source time series data.

[0048] In the embodiments of the present application, first, a multi-source time series data set reflecting the growth state and environmental conditions of corn is needed to be acquired. Specifically, after the corn enters the jointing stage, a representative plant is selected to install a stem flow meter to continuously collect and record the sap flow rate data sequence of the stem at a sampling frequency of every 30 minutes; soil humidity sensors are arranged in the main distribution area of the plant roots to collect and record the soil humidity data sequence at the same sampling frequency as the sap flow rate collection; a photosynthetically active radiation sensor is installed above the corn canopy or in the field weather station to collect and record the photosynthetically active radiation data sequence at the same sampling frequency as the sap flow rate collection; and after the above data collection is completed, the sap flow rate data sequence, the soil humidity data sequence, and the photosynthetically active radiation data sequence are time-stamped aligned, and missing data points are filled in by the nearest point interpolation or linear interpolation method to acquire a complete multi-dimensional time series data set.

[0049] At this point, the complete multi-dimensional time series data set is acquired by synchronizing and preprocessing the multi-source time series data.

[0050] In step S2, the adaptive input damping factor is acquired by performing dynamic baseline and local variability analysis on the original sap flow data, and the sap flow data is preliminarily corrected by the adaptive input damping factor.

[0051] The technical problem to be solved by the present application is that the standard Savitzky-Golay (SG) filter regards each data point in the sliding window as a signal with equal significance and gives it a fixed weight determined by the polynomial order and window position when performing smoothing operation. However, in the case of post-drought rehydration, the instantaneous peak value appearing in the collected sap flow data has an essential difference in physiological significance from the data points in the stable period. The peak value mainly reflects the short-term hydraulic adjustment behavior of the root system under the dramatic change of water potential, rather than the synchronous transition of the aboveground dry matter accumulation rate. Therefore, directly inputting the original data sequence containing this instantaneous peak value into the SG filter will cause the filter to produce a severely distorted output in order to fit this extreme value point which is statistically and physiologically abnormal. To solve this problem, the peak value cannot be removed or flattened, because the occurrence of the peak value itself marks a key transition in the physiological state of the crop. The solution of the present application is to introduce a dynamic preprocessing step before the data enters the SG filter. A mechanism is constructed to evaluate the abnormality degree of each data point. Specifically, the mechanism can compare the sap flow data point at the current time with its recent historical background. The abnormality degree of a data point is not determined by its absolute value, but by its deviation from the recent stable growth trend, combined with the volatility of the trend itself. Therefore, two key features are extracted from the historical data: a dynamic baseline that can represent the current potential growth rate, and a local variability that can quantify the normal fluctuation range of recent data. Based on these two features, an adaptive damping factor is constructed. When a data point deviates greatly from the dynamic baseline and local variability, the factor should automatically reduce the influence of the data point; on the contrary, when the data point is within the normal fluctuation range, the factor should not have an impact to preserve the original information of the data. By applying this damping factor to the original data, the excessive interference of the peak value on the subsequent SG filter fitting process can be effectively suppressed without completely removing the peak value information, thereby creating conditions for extracting the true growth trend baseline.

[0052] In summary, first, the evaluation number for dynamic growth baseline evaluation and local data fluctuation evaluation is set, and in the embodiment of the present application, the evaluation number is set to 48, corresponding to a time window of 24 hours; for any target time in the original sap flow data, the median of the selected continuous sap flow data points before the target time is taken as the dynamic growth baseline of the target time; and the standard deviation of the selected continuous sap flow data is taken as the local fluctuation degree of the target time;

[0053] After obtaining the dynamic growth baseline and the local fluctuation degree at the target time, the adaptive input damping factor is obtained by performing a joint deviation analysis on the original wound flow data based on the dynamic growth baseline and the local fluctuation degree. The original wound flow data is then preliminarily corrected. Specifically, for any target time in the original wound flow data, the result of subtracting the wound flow rate data at the target time from the dynamic growth baseline at the target time is used as the numerator, and the result of adding the local fluctuation degree at the target time to the smallest positive number is used as the denominator. The corresponding fraction is used as the first input damping assessment at the target time. The negative of the square of the first input damping assessment at the target time is then subjected to an exponential mapping with the natural constant as the base, and the corresponding mapping result is used as the adaptive input damping factor at the target time.

[0054] In one implementation, assume the first The original injury flow data at each moment is ;No. The dynamic growth baseline of the raw wound flow data at each time point is: ;No. The degree of local fluctuation at each moment is Then the first The formula for calculating the adaptive input damping factor at time n is:

[0055]

[0056] in, Indicates the first The adaptive input damping factor at each moment; Indicates the first Raw wound flow data at each moment; Indicates the first The dynamic growth baseline of the raw wound flow data at each moment; Indicates the first The degree of local fluctuation at a given moment; Denotes the natural constant e; The use of extremely small positive numbers to prevent the denominator from being 0 is specified in this embodiment of the invention. .

[0057] After obtaining the adaptive input damping factor, the original wound flow data is further corrected based on baseline deviation using the adaptive input damping factor to obtain preliminary corrected wound flow data. Specifically, for any target time in the original wound flow data, the result of subtracting the wound flow rate data at the target time from the dynamic growth baseline is used as the first baseline deviation assessment at the target time; the result of multiplying the adaptive input damping factor at the target time with the first baseline deviation assessment at the target time is used as the first correction assessment at the target time; and the result of adding the dynamic growth baseline at the target time to the first correction assessment at the target time is used as the preliminary corrected wound flow data at the target time.

[0058] In one embodiment, the first The calculation expression for the preliminary corrected injury flow data at each time point is:

[0059]

[0060] in, Indicates the first Preliminary corrected wound flow data at a given moment; Indicates the first The adaptive input damping factor at each moment; Indicates the first Raw wound flow data at each moment; Indicates the first The dynamic growth baseline of the raw wound flow data at each time point.

[0061] It should be noted that the formula is obtained through... This item will take the raw data points at the current moment. Transform its absolute value into a standardized deviation metric. Specifically, this step will... Its dynamic growth baseline The difference is normalized by the local data volatility within its time window. This allows the deviation measure to objectively reflect... The statistical significance under the current physiological context avoids misjudgment that may be caused by the inherent fluctuations in crop growth at different stages.

[0062] Subsequently, this standardized deviation measure is input as an independent variable into a Gaussian function to generate the damping factor. The nonlinear characteristics of this function are the core of achieving selective correction. When When the data values ​​are within the normal physiological fluctuation range, their standardized deviation metric is small, resulting in a smaller calculated value. the value approaches 1. This property guarantees that for regular data representing true growth dynamics, the correction process of the present invention does not have substantial impact, thus ensuring data fidelity. On the contrary, when a compensatory water uptake phenomenon is triggered by a post-drought rehydration event, resulting in a statistically highly unlikely transient peak, the normalized deviation metric value will become extremely large. In this case, the exponential decay property of the Gaussian function will cause the value to rapidly and nonlinearly approach 0. This property enables the present invention to accurately identify such physiological artifact signals triggered by short-term environmental changes. Finally, the damping factor is applied through the correction formula which structurally decomposes the original data point into a baseline trend component and a transient deviation component . The damping factor only acts on the transient deviation component. Therefore, when approaches 0 (i.e. a transient peak is identified), the transient deviation component is effectively eliminated, making the corrected value converge to a stable dynamic baseline , thus preventing the abnormal peak from distorting the subsequent SG filtering process. When approaches 1 (i.e. regular data is processed), the transient deviation component is completely preserved, essentially equivalent to the original value .

[0063] So far, the adaptive input damping factor is obtained by performing dynamic baseline and local variability analysis on the original sap flow data, and the sap flow data is preliminarily corrected by the adaptive input damping factor.

[0064] Step S3, the posterior compensation factor is obtained by performing peak event and environmental constraint analysis on the smoothed sap flow data, and the smoothed sap flow data is secondarily corrected by the posterior compensation factor.

[0065] After the processing of step S2, the instantaneous peak artifacts in the original sap flow data have been effectively suppressed, thus providing a clean input for the SG filter to generate a growth baseline free from outliers. However, the aforementioned instantaneous peak, although its amplitude does not represent a simultaneous dry matter accumulation, its existence itself is a critical physiological information indicator, which indicates that the growth environment of the crop has fundamentally improved, i.e., from water stress to water abundance, its future growth potential has been activated and significantly improved. The damping operation in step S2 essentially erases this peak information from the data sequence, resulting in a smoothed baseline that cannot reflect this future enhanced growth trend. Specifically, the response of photosynthesis and biomass synthesis of the plant aboveground part to the improvement of root water status has inherent physiological lag. Therefore, within a period of time after the peak event, the true dry matter accumulation rate will gradually climb to a higher steady state level. The smoothed curve obtained only by step S2 will underestimate this lag recovery and compensatory growth process. To solve this problem, the present application further constructs an additional compensation factor, the function of which is to inject a compensation signal into the smoothed baseline after the peak event, which can simulate and quantify the lagging growth potential improvement. Its trigger should depend on the identification of the peak event; its overall magnitude should be positively correlated with the intensity of the suppressed peak (representing the degree of stress relief); its instantaneous size should also be subject to the current environmental conditions (such as light and soil moisture), because the realization of growth potential cannot be separated from good external environment; its time-varying morphology should also exhibit a dynamic process of starting from zero, gradually increasing, reaching a peak, and then slowly stabilizing, to truly reflect the response rhythm of the physiological activity of the aboveground part.

[0066] In summary, first, the length of the sliding window and the order of polynomial fitting of the SG filter are set. In the embodiment of the present application, the length of the sliding window is set to 11, and the order of polynomial fitting is set to 3. The length of the sliding window and the order of polynomial fitting can be adjusted according to the actual scene, and are not required.

[0067] Afterwards, the peak value compensation reference quantity is obtained by performing peak event identification and intensity calculation processing on the smoothed sap flow data. Specifically, a peak evaluation coefficient is set, and in the embodiment of the present application, the peak evaluation coefficient is set to 3. For any target moment in the original sap flow data, the calculation result of multiplying the peak evaluation coefficient by the local fluctuation degree of the target moment and adding the dynamic growth baseline of the target moment is taken as the peak event starting evaluation threshold of the target moment. For any target moment, if the value of the smoothed sap flow data of the target moment is greater than the peak event starting evaluation threshold of the target moment, the target moment is determined as the starting moment of the peak event. The first moment smaller than the peak event starting evaluation threshold after the starting moment of the peak event is taken as the ending moment of the peak time corresponding to the starting moment of the peak event. The starting moment and the ending moment of the peak event are evaluated to obtain all peak events in the original sap flow data.

[0068] The calculation result of subtracting the sap flow data of any target moment in the original sap flow data from the preliminary corrected sap flow data of the target moment is taken as the first peak flux evaluation of the target moment. For any target peak event in all peak events in the original sap flow data, the calculation result of adding the first peak flux evaluation of all moments in the target peak event is taken as the peak compensation reference quantity of the target peak event.

[0069] After obtaining the peak compensation reference quantity, the environmental adjustment factor is obtained by continuing to perform environmental constraint analysis processing on the soil humidity data and the photosynthetically active radiation data. Specifically, for any target peak event in all peak events in the original sap flow data, the moment of the maximum value of the original sap flow data in the target peak event is taken as the transient peak moment of the target peak event. For the transient peak moment of any target peak event, the average value of the soil humidity data in the interval formed by the ten days before the target peak event to the four days before the target peak event is taken as the reference soil humidity of the target peak event.

[0070] For the transient peak moment of any target peak event, the maximum photosynthetically active radiation data within the seven days before the transient peak moment of the target peak event is taken as the reference photosynthetically active radiation value of the target peak event.

[0071] For any target moment in any target peak event, the calculation result of dividing the soil humidity data of the target moment by the reference soil humidity of the target peak event is taken as the first soil humidity evaluation of the target moment. The calculation result of dividing the photosynthetically active radiation data of the target moment by the reference photosynthetically active radiation value of the target peak time is taken as the first photosynthetically active radiation evaluation of the target moment.

[0072] The calculation result of multiplying the first soil humidity evaluation of the target moment in the target peak value event and the first photosynthetically active radiation evaluation is taken as the environmental adjustment factor of the target peak value event at the target moment.

[0073] After the environmental adjustment factor of the target moment is acquired, the posterior compensation factor is acquired by jointly processing the peak compensation reference quantity and the environmental adjustment factor, and the smoothed girdling data is secondarily revised. Specifically, a physiological response time constant is set, which is 24 in the embodiment of the present application. For any target moment in any target peak value event, the calculation result of subtracting the target moment from the instantaneous peak moment of the target peak value event is taken as the numerator, the square of the physiological response time constant is taken as the denominator, and the corresponding fraction is taken as the first pulse evaluation of the target moment. The calculation result of subtracting the target moment from the instantaneous peak moment of the target peak value event is taken as the numerator, the physiological response time constant is taken as the denominator, the corresponding fraction is subjected to exponential mapping with the natural constant as the base number, and the corresponding fraction is taken as the second pulse evaluation of the target moment. The calculation result of multiplying the first pulse evaluation and the second pulse evaluation of the target moment is taken as the pulse response evaluation of the target moment.

[0074] The calculation result of multiplying the environmental adjustment factor of the target peak value event at the target moment, the peak compensation reference quantity of the target peak value event and the pulse response evaluation of the target peak value event at the target moment is taken as the posterior compensation factor of the target peak value event at the target moment.

[0075] In an embodiment, it is assumed that the soil humidity data of the first moment is , the reference soil humidity of the peak value event corresponding to the first moment is ; the photosynthetically active radiation value of the first moment is , the reference photosynthetically active radiation value of the peak value event corresponding to the first moment is ; the peak compensation reference quantity of the peak value event corresponding to the first moment is ; the instantaneous peak moment of the peak value event corresponding to the first moment is ; and the physiological response time constant is . The calculation expression of the posterior compensation factor of the first moment is:

[0076]

[0077] Wherein, represents the posterior compensation factor of the first moment; soil moisture data at the first time point, reference soil moisture corresponding to the peak event at the first time point; photosynthetically active radiation value at the first time point, reference photosynthetically active radiation value corresponding to the peak event at the first time point; peak compensation reference amount corresponding to the peak event at the first time point; instantaneous peak time corresponding to the peak event at the first time point; physiological response time constant.

[0078] After obtaining the posterior compensation factor at the target time point, the calculation result of adding the posterior compensation factor of the target peak event at the target time point to the corresponding target time point data in the smoothed sap flow data is taken as the sap flow data after secondary correction.

[0079] It should be noted that in the present application, the purpose of the posterior compensation factor is to quantitatively recover the physiological information lost after step S2 processing, so as to solve the problem of underestimating the compensatory growth after stress of the smoothed baseline. The core term of the formula quantifies the total flux of root compensatory water absorption in the rehydration event after drought, which represents the intensity of the state transition. Taking as the basic scale of the total compensation amount ensures that the intensity of compensation is directly related to the intensity of the previous stress relief event, so that the compensation has a solid data basis. Secondly, the formula introduces this environmental potential coefficient to realistically constrain the implementation of compensation. The realization of crop growth potential cannot be separated from suitable water and light conditions. This coefficient compares and normalizes the real-time measured soil moisture and photosynthetically active radiation with their respective reference values. This allows the calculated value of the posterior compensation factor to be dynamically adjusted according to the adequacy of the current environmental resources. In conditions of sufficient water and good light, the coefficient value is higher, allowing stronger compensation to occur; conversely, if it encounters rainy weather later, the coefficient value will be reduced, thereby reasonably inhibiting the intensity of compensation. Through this impulse response function, the dynamic process of the evolution of the compensation effect over time is described. This function simulates the universal law of lag response in biological systems: at the peak value time point, the compensation value is zero; with the passage of time, the compensation effect gradually increases, and in a physiological time constant ​​​​​​The peak value of the determined moment reaches the peak; then, the compensation effect is smoothly attenuated as the plant gradually adapts to the new superior environment and enters a new growth steady state. This design ensures that the application of the compensation signal is not a sudden step, but a dynamic process that meets the physiological rhythm of the plant. Finally, by adding the smoothed sap flow data to the posterior compensation factor, the secondary revised sap flow data is obtained, which eliminates the distortion and artifacts caused by short-term stress response in the original data, and restores the lagging growth gain brought by environmental improvement, and finally obtains a characteristic sequence that can accurately reflect the real growth trend of corn after experiencing environmental changes.

[0080] At this point, the posterior compensation factor is obtained by analyzing the peak value event and environmental constraints of the smoothed sap flow data, and the smoothed sap flow data is revised again by the posterior compensation factor.

[0081] Step S4, the corn dry matter accumulation rate is obtained by modeling and predicting the revised sap flow characteristics and through the prediction model.

[0082] After obtaining the secondary revised sap flow data through the aforementioned steps S2 and S3, the sap flow data will be sent as the core input characteristics into a pre-trained corn dry matter accumulation prediction model. The prediction model is constructed by using the machine learning algorithm in the prior art, and preferably, the pre-trained corn dry matter accumulation prediction model is obtained by pre-training the corn dry matter accumulation prediction model through the support vector regression model. The model has learned the nonlinear mapping relationship between the input of multiple physiological and environmental factors including sap flow characteristics and the dry matter accumulation rate of corn per unit time in the training stage. Therefore, the secondary revised sap flow data revised by the dynamic correction method is input into the model to obtain the corn dry matter accumulation prediction rate.

[0083] At this point, the corn dry matter accumulation rate is obtained by modeling and predicting the revised sap flow characteristics and through the prediction model.

[0084] Step S5, the corn yield prediction result is obtained by accumulating and integrating the corn dry matter accumulation rate.

[0085] By steps S2 to S4, the present application successfully solves the core technical problem in the prior art that the prediction is inaccurate due to the inability to distinguish between physiological artifacts and real growth trends in special scenarios such as post-drought rehydration. Specifically, the present application realizes automatic identification and interference suppression of the instantaneous peak value caused by compensatory water uptake effect through an adaptive input damping factor; and restores the lagging compensation growth information brought about by environmental improvement through a posterior compensation factor. This double correction mechanism ensures that the twice-corrected phytotoxicity data obtained can truly reflect the root physiological activity related to dry matter accumulation. Therefore, the obtained prediction result of the dry matter accumulation rate has significantly higher accuracy compared to the prior art. By integrating the predicted accumulation rates at each time during the entire growth period, the total dry matter accumulation amount in any time period can be obtained, thereby providing reliable data support for accurate assessment of crop growth status, optimization of field management decisions, and prediction of final yield.

[0086] At this point, the corn yield prediction result is obtained by integrating the corn dry matter accumulation rate.

[0087] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting dry matter accumulation in maize based on root sap flow characteristics, characterized in that, The method includes: Step S1: Obtain a complete multidimensional time series dataset by performing synchronization preprocessing on multi-source time series data; Step S2: By performing dynamic baseline and local variability analysis on the original wound flow data, the adaptive input damping factor is obtained, and the wound flow data is initially corrected using the adaptive input damping factor; Step S3: By performing peak event and environmental constraint analysis on the smoothed wound flow data, the posterior compensation factor is obtained, and the smoothed wound flow data is then corrected a second time using the posterior compensation factor. Step S4: Model and predict the corrected wound flow characteristics, and obtain the corn dry matter accumulation rate through the prediction model; Step S5: Obtain the corn yield prediction result by cumulatively integrating the corn dry matter accumulation rate; The process of obtaining a complete multidimensional time series dataset by synchronizing and preprocessing multi-source time series data includes: collecting and recording stem sap flow rate data sequence using a stem flow meter; collecting and recording soil moisture data sequence using a soil moisture sensor in the main distribution area of ​​plant roots; and collecting and recording photosynthetically active radiation data sequence using an effective radiation sensor. The process involves performing dynamic baseline and local variability analysis on the original wound flow data to obtain an adaptive input damping factor, and then using this adaptive input damping factor to perform preliminary correction of the wound flow data. This includes: setting the number of assessments for dynamic growth baseline evaluation and local data fluctuation evaluation; for any target time in the original wound flow data, selecting continuous wound flow data points before the target time based on the number of assessments, and using the median of the selected continuous wound flow data points as the dynamic growth baseline for the target time; using the standard deviation of the selected continuous wound flow data as the local fluctuation degree for the target time; and obtaining the adaptive input damping factor by performing joint deviation analysis on the original wound flow data based on the dynamic growth baseline and local fluctuation degree, and then performing preliminary correction of the original wound flow data. The process involves analyzing peak events and environmental constraints on smoothed wound flow data to obtain a posterior compensation factor, and then using this posterior compensation factor to perform a secondary correction on the smoothed wound flow data. This includes: setting the sliding window length and polynomial fitting order of the SG filter; smoothing the initially corrected wound flow data using SG filtering to obtain smoothed wound flow data; identifying and calculating peak events in the smoothed wound flow data to obtain a peak compensation baseline; analyzing soil moisture data and photosynthetically active radiation data for environmental constraints to obtain an environmental adjustment factor; and jointly processing the peak compensation baseline and the environmental adjustment factor to obtain a posterior compensation factor, which is then used to perform a secondary correction on the smoothed wound flow data.

2. The method for predicting maize dry matter accumulation based on root sap flow characteristics according to claim 1, characterized in that, The process of obtaining a complete multidimensional time series dataset by synchronizing and preprocessing multi-source time series data includes: After maize enters the jointing stage, representative plants are selected and equipped with stem flow meters to continuously collect and record the sap flow rate data sequence of the stem at a sampling frequency of once every 30 minutes. Soil moisture sensors are deployed in the main distribution area of ​​the plant roots to collect and record soil moisture data sequences at the same sampling frequency as the sap flow rate data. Photosynthetically active radiation sensors are installed above the maize canopy or in field weather stations to collect and record photosynthetically active radiation data sequences at the same sampling frequency as the sap flow rate data. After completing the above data collection, the sap flow rate data sequence, soil moisture data sequence, and photosynthetically active radiation data sequence are timestamped and aligned. Missing data points are filled in using nearest neighbor interpolation or linear interpolation methods to obtain a complete and synchronized multidimensional time series dataset.

3. The method for predicting maize dry matter accumulation based on root sap flow characteristics according to claim 1, characterized in that, The process involves performing a joint deviation analysis on the original wound flow data based on the dynamic growth baseline and the degree of local fluctuations to obtain an adaptive input damping factor, and then performing preliminary corrections on the original wound flow data, including: For any target time in the original wound flow data, the result of subtracting the wound flow rate data at the target time from the dynamic growth baseline at the target time is used as the numerator, and the result of adding the local fluctuation degree at the target time to the minimum positive number is used as the denominator. The corresponding fraction is used as the first input damping assessment at the target time. The negative of the square of the first input damping assessment at the target time is subjected to an exponential mapping with the natural constant as the base, and the corresponding mapping result is used as the adaptive input damping factor at the target time. The original wound flow data is corrected based on baseline deviation by using an adaptive input damping factor to obtain preliminary corrected wound flow data.

4. The method for predicting maize dry matter accumulation based on root sap flow characteristics according to claim 3, characterized in that, The process of correcting the original wound flow data based on baseline deviation using an adaptive input damping factor to obtain preliminarily corrected wound flow data includes: For any target time in the original wound flow data, the result of subtracting the wound flow rate data at the target time from the dynamic growth baseline is used as the first baseline deviation assessment for the target time; the result of multiplying the adaptive input damping factor at the target time with the first baseline deviation assessment for the target time is used as the first correction assessment for the target time; and the result of adding the dynamic growth baseline at the target time to the first correction assessment for the target time is used as the preliminary corrected wound flow data for the target time.

5. The method for predicting maize dry matter accumulation based on root sap flow characteristics according to claim 1, characterized in that, The process of identifying peak events and calculating the intensity of the smoothed wound flow data to obtain the peak compensation benchmark includes: A peak evaluation coefficient is set. For any target time in the original wound flow data, the peak evaluation coefficient is multiplied by the local fluctuation degree of the target time and added to the dynamic growth baseline of the target time. The result is used as the peak event start evaluation threshold for the target time. For any target time, if the value of the smoothed wound flow data at the target time is greater than the peak event start evaluation threshold, the target time is determined as the start time of the peak event. The first time after the start time of the peak event that is less than the peak time start evaluation threshold is used as the end time of the peak time corresponding to the start time of the peak event. All peak events in the original wound flow data are obtained through the evaluation of the start and end times of the peak events. The result of subtracting the initial corrected flow data at the target time from the flow data at any target time in the original flow data is used as the first peak flux assessment at the target time. For any target peak event among all peak events in the original wound flow data, the calculation result of summing the first peak flux assessments at all times in the target peak event is used as the peak compensation benchmark for the target peak event.

6. The method for predicting maize dry matter accumulation based on root sap flow characteristics according to claim 1, characterized in that, The process involves performing environmental constraint analysis on soil moisture data and photosynthetically active radiation data to obtain environmental regulation factors, including: For any target peak event among all peak events in the original wound flow data, the moment when the original wound flow data has the maximum value in the target peak event is taken as the instantaneous peak moment of the target peak event; For any instantaneous peak moment of a target peak event, the average value of the soil moisture data in the interval formed by the ten days to the four days prior to the instantaneous peak moment of the target peak event is taken as the reference soil moisture of the target peak event. For any instantaneous peak moment of a target peak event, the maximum photosynthetically active radiation data within the seven days preceding the instantaneous peak moment of the target peak event is used as the reference photosynthetically active radiation value of the target peak event; For any target time in any target peak event, the soil moisture data at the target time is divided by the calculated result of the reference soil moisture of the target peak event as the first soil moisture assessment at the target time; the photosynthetically active radiation data at the target time is divided by the calculated result of the calculated result of the reference photosynthetically active radiation value at the target peak time as the first photosynthetically active radiation assessment at the target time. The result of multiplying the first soil moisture assessment and the first photosynthetically active radiation assessment at the target time during the target peak event is used as the environmental adjustment factor for the target peak event at the target time.

7. The method for predicting maize dry matter accumulation based on root sap flow characteristics according to claim 1, characterized in that, The process involves jointly processing the peak compensation baseline and the environmental adjustment factor to obtain the posterior compensation factor, and then performing a secondary correction on the smoothed wound flow data, including: Define a physiological response time constant; for any target moment in any target peak event, use the result of subtracting the instantaneous peak moment of the target peak event from the target moment as the numerator, and the square of the physiological response time constant as the denominator. Use the resulting fraction as the first pulse evaluation of the target moment; use the result of subtracting the instantaneous peak moment of the target peak event from the target moment as the numerator, and the physiological response time constant as the denominator. Perform an exponential mapping with the natural constant as the base on the resulting fraction. Use the resulting fraction as the second pulse evaluation of the target moment. Multiply the first pulse evaluation and the second pulse evaluation of the target moment as the pulse response evaluation of the target moment. The result of multiplying the environmental adjustment factor of the target peak event at the target time, the peak compensation benchmark of the target peak event, and the impulse response evaluation of the target peak event at the target time is used as the posterior compensation factor of the target peak event at the target time. The result of adding the posterior compensation factor of the target peak event at the target time to the corresponding target time data in the smoothed wound flow data is used as the wound flow data for secondary correction.

Images