Crop environmental quantification and decision system based on growth stage adaptation

Abstract

The application discloses a crop environment quantitative evaluation and decision system based on growth stage self-adaptation and relates to the technical field of agricultural wisdom decision-making.The data acquisition module collects original environmental data sequences of multiple types of sensors in the field;the dynamic identification module analyzes the sequences by using a preset crop growth stage identification model, identifies the current growth stage, and outputs a corresponding key environmental parameter weight template;the feature fusion module fuses the original data according to the template to generate a stage-adaptive environmental feature vector;the state evaluation module converts the vector into a growth state quantitative evaluation value through a growth state evaluation model;and the decision trigger judgment module compares the preset dynamic threshold interval corresponding to the growth stage according to the growth state quantitative evaluation value, and generates a decision trigger instruction when the growth state quantitative evaluation value deviates.The application realizes stage self-adaptive intelligence of crop growth monitoring and decision-making, and improves the accuracy of environmental state evaluation and the timeliness of management decision-making.

Description

Technical Field

[0001] This invention relates to the field of agricultural intelligent decision-making technology, specifically to a crop environment quantitative assessment and decision-making system based on growth stage adaptation. Background Technology

[0002] Precision agriculture is an important direction for the development of modern agriculture. Its core lies in achieving refined crop production management and efficient resource utilization through real-time and accurate environmental perception and intelligent decision-making. With the advancement of sensor technology, the Internet of Things, and data algorithms, field crop growth environment dynamic monitoring systems can now integrate multiple types of sensors to acquire multi-dimensional environmental data such as soil temperature and humidity, light intensity, and carbon dioxide concentration in real time. Based on fixed thresholds or multi-parameter fusion models, these systems can perform status assessments and issue alarms, thereby improving the automation level of environmental perception to a certain extent.

[0003] However, throughout the complete growth cycle of crops, from germination and vegetative growth to reproductive growth, their sensitivity to different environmental factors and their required weights vary. For example, the influence weights of different environmental parameters on crop growth status are not constant during the seedling, vegetative, and reproductive growth stages. Existing technologies commonly employ unified parameter weighting or static evaluation models, which typically implicitly assume that the crop's environmental response relationship is relatively stable over time. When crop growth stages change, these assumptions easily fail, leading to inaccurate reflection of the actual contribution of environmental parameters to growth status. In this situation, even if the raw environmental data is accurately collected, the lack of dynamic identification of crop growth stages and modeling mechanisms for stage-differentiated environmental characteristics means that existing technologies may still result in evaluations that deviate from the actual growth conditions, thus affecting the timing and direction of subsequent management decisions. This problem is particularly prominent during rapid crop growth transitions, easily causing delays or mismatches in environmental regulation, thereby reducing the overall effectiveness of precision agriculture systems. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a crop environment quantitative assessment and decision-making system based on growth stage adaptation.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] This invention discloses a crop environment quantitative assessment and decision-making system based on growth stage adaptation, comprising:

[0007] The data acquisition module is used to acquire raw environmental data sequences collected by multiple types of sensors;

[0008] The dynamic recognition module is used for:

[0009] The system receives the raw environmental data sequence and inputs it into a preset crop growth stage discrimination model for matching analysis to identify the current growth stage of the crop.

[0010] Dynamically output the current growth stage identifier of the crop and the corresponding key environmental parameter weight template;

[0011] The feature fusion module is used to perform weighted fusion calculation on each parameter data in the original environmental data sequence based on the key environmental parameter weight template, and generate an environmental feature vector for stage adaptability.

[0012] The status assessment module is used to input the environmental feature vector into a preset growth status assessment model, analyze and output the quantitative assessment value of the current crop's growth status.

[0013] The decision-triggered judgment module is used for:

[0014] Receive the quantitative evaluation value of the growth status and the growth stage identifier, and call the corresponding preset dynamic threshold range according to the growth stage identifier;

[0015] The quantitative evaluation value of the growth status is compared with the preset dynamic threshold range;

[0016] When the quantitative evaluation value of the growth status deviates from the preset dynamic threshold range, a decision trigger instruction is generated and output.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0018] 1. This invention introduces a crop growth stage discrimination model, enabling the system to automatically identify the specific growth stage of the crop based on real-time environmental data. This solves the problem of the disconnect between environmental perception and actual crop needs caused by the traditional system using a single evaluation standard throughout the entire growth cycle, and makes the collection and interpretation of environmental parameters guided by agronomic time-series laws.

[0019] 2. This invention enables the system to highlight the influence of dominant environmental factors at the current stage and weaken the interference of secondary factors when generating environmental feature vectors. The resulting environmental feature vectors better reflect the actual needs of crops under their current physiological state, improving the accuracy of feature representation in describing crop growth status.

[0020] 3. The state assessment standards and decision triggering conditions of the system of this invention can evolve synchronously with the crop growth process. This allows the system to adopt differentiated management strategies at different stages such as the germination stage and the vegetative growth stage, thereby achieving more targeted and precise regulation in complex and ever-changing field environments, and effectively avoiding decision-making deviations caused by misjudgment of growth stages or mismatch of standards. Attached Figure Description

[0021] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0022] Figure 1 This is a system module connection diagram of the present invention;

[0023] Figure 2 This is a flowchart of the system modules of the present invention;

[0024] Figure 3 This is a flowchart illustrating the construction process of the crop growth stage discrimination model of the present invention. Detailed Implementation

[0025] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0026] Application Overview:

[0027] In existing technologies, dynamic monitoring of crop growth environment in the field mostly relies on fixed threshold alarms or static data fusion, failing to consider the dynamic changes in environmental factor requirements throughout the entire crop growth period. Traditional systems use the same set of evaluation criteria, which cannot adaptively adjust the weights and decision thresholds of key parameters at different stages such as germination, vegetative growth, and reproductive growth, leading to distorted state assessments and misjudgments of decision timing, making it difficult to meet the differentiated management needs of precision agriculture.

[0028] To address the aforementioned issues, an adaptive optimization of the evaluation model can be achieved by establishing a dynamic mapping relationship between growth stages and the weights of key environmental parameters. Furthermore, by identifying growth stages in real time and switching evaluation strategies accordingly, the agronomic rationality of the system's perception can be effectively improved. Therefore, a concept for constructing a closed-loop intelligent monitoring system of "stage perception - weight adaptation - decision triggering" is proposed.

[0029] After introducing the basic concept of the present invention, the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0030] like Figure 1 and Figure 2 As shown, the crop environment quantitative assessment and decision-making system based on growth stage adaptation has its functional modules connected via wired or wireless communication, including:

[0031] The data acquisition module is used to acquire raw environmental data sequences collected by various types of sensors. These environmental sensors may include, but are not limited to, nanosensors or microsensors used to collect environmental parameters such as soil moisture, soil temperature, air temperature and humidity, and light intensity. Each sensor node converts the sensed environmental signals into digital signals through an analog-to-digital conversion circuit and generates raw environmental data sequences according to a preset sampling period.

[0032] The dynamic identification module, upon receiving the raw environmental data sequence, first performs necessary formatting and time alignment processing on the data. Then, it inputs the processed data into a pre-defined crop growth stage discrimination model for matching analysis, identifying the current growth stage of the crop and dynamically outputting the current growth stage identifier. Simultaneously, based on the identified growth stage, the dynamic identification module calls the corresponding key environmental parameter weight template and outputs this weight template as the basis for subsequent feature fusion.

[0033] In this embodiment, the crop growth stage discrimination model is a time-series data classification model based on supervised learning, used to divide the crop growth process into several discrete stages (e.g., seedling stage, vegetative growth stage, reproductive growth stage). This model can employ a structure suitable for processing time-series features, including but not limited to recurrent neural networks, long short-term memory networks, or hybrid classification models combining time-series feature extractors.

[0034] The crop growth stage discrimination model employs a two-layer LSTM structure. Inputs are time-series environmental parameters (such as soil temperature and humidity, light intensity, etc.), with 64 units per layer and a dropout rate of 0.3. The output layer uses Softmax to correspond to the four growth stages (seedling stage, vegetative growth stage, reproductive growth stage, and maturity stage). Training data comes from historical monitoring data from three growing seasons and ten fields, with no fewer than 500 samples per stage, divided into training and validation sets in an 8:2 ratio. The loss function is cross-entropy, the optimizer is Adam, the initial learning rate is 0.001, and the learning rate is halved every 10 rounds if the validation loss does not decrease. The early stopping round count is 20.

[0035] The feature fusion module receives the raw environmental data sequence and key environmental parameter weight templates. Based on the importance distribution of each environmental parameter according to the key environmental parameter weight templates, it performs weighted fusion calculations on the parameter data in the raw environmental data sequence, integrating the multi-dimensional, multi-source environmental parameter data into a stage-adaptive environmental feature vector that reflects the characteristics of the current growth stage. The environmental feature vector numerically reflects the comprehensive impact of various environmental factors on the crop growth status at the current stage.

[0036] The status assessment module is used to input environmental feature vectors into a preset growth status assessment model, analyze them, and output a quantitative assessment value of the current crop's growth status. The growth status assessment model is used to characterize the mapping relationship between environmental features and crop growth status. Through analysis and calculation of the environmental feature vectors, it outputs a quantitative assessment value of the current crop's growth status.

[0037] The growth status assessment model is used to quantitatively evaluate the current growth status of crops based on environmental feature vectors of stage adaptability. The output reflects the growth status level of the crop at the current growth stage and can be expressed in continuous numerical form or hierarchical numerical form.

[0038] Growth status assessment models can employ neural network-based regression models to map environmental feature vectors into continuous assessment values ​​representing growth status; or they can use structures such as multilayer perceptrons and support vector regression models to adapt to growth status assessment needs of varying complexity. Models can also be constructed by associating them with crop growth stage information, enabling them to exhibit different parameter response characteristics at different stages.

[0039] The training samples for the growth status assessment model can include: an environmental feature vector constructed by the feature fusion module, and a crop growth status reference value corresponding to that environmental feature vector. The growth status reference value can be derived from indicators reflecting crop growth status, such as crop growth amount, biomass, leaf area index, yield statistics, or manual assessment results, and is used as a supervision signal for the model after standardization. By introducing data from multiple growing seasons and multiple plots, the model's adaptability to differences in growth status under different environmental conditions can be enhanced. The growth status assessment model uses a three-layer fully connected neural network with hidden layer nodes in the range of [64, 32], an activation function of ReLU, and an output of growth status score. The training samples consist of an environmental feature vector and the corresponding standardized leaf area index value, using mean squared error loss, a batch size of 32, and a training duration of no more than 200 epochs.

[0040] The decision-triggered judgment module is used to further judge and process the growth status assessment results. After receiving the quantitative assessment value of the growth status and the growth stage identifier, it calls the preset dynamic threshold range corresponding to the current growth stage from the preset management strategy database according to the growth stage identifier.

[0041] The system compares and analyzes the quantitative assessment value of growth status with a preset dynamic threshold range. When the quantitative assessment value deviates from the preset dynamic threshold range, indicating that the current growth status is deviating from the management target for that stage, a decision trigger command is generated and output. This command is then sent to the corresponding field execution unit controller (such as irrigation valve controller, film rolling motor controller, fertilizer applicator control terminal) or the smart terminal of the farm manager via the system communication interface. When the command is sent to the execution unit controller, the system can automatically execute the corresponding control operation based on a preset strategy or after manual confirmation.

[0042] The crop growth stage discrimination model and the growth status assessment model can work collaboratively on the same computing platform, or they can be deployed separately on edge computing devices and central servers. Through the former's dynamic identification of crop growth stages and the latter's quantitative assessment of growth status, the system can continuously update the environmental feature construction method and status assessment results as the crop growth process changes, thereby providing a reliable data foundation for subsequent threshold comparison and decision triggering.

[0043] Through the above technical solution, this application can dynamically identify crop growth stages based on multi-source environmental data, and perform stage-adaptive fusion and analysis of environmental data according to the differentiated needs of environmental factors at different growth stages, thereby achieving quantitative assessment and decision-making triggers for crop growth status. This solution avoids the problem of insufficient adaptability caused by using fixed parameter weights or uniform evaluation standards, improves the consistency between crop growth environment monitoring results and actual growth status, and is conducive to improving the accuracy and reliability of dynamic monitoring and management decisions of field crop growth environment.

[0044] like Figure 3 The diagram shown is a flowchart of the construction process for the crop growth stage discrimination model. This application further proposes that the crop growth stage discrimination model be constructed through the following steps:

[0045] During the model building phase, historical environmental data sequences of the target crop over multiple growth cycles are first collected. These historical environmental data sequences originate from various types of sensors deployed in the field, covering environmental parameters including at least one or more of temperature, moisture, light, and nutrients. The aforementioned environmental data is then time-aligned with agronomic records for the corresponding cycles. The agronomic records clearly label the crop's growth stage for each time period, such as seedling stage, vegetative growth stage, reproductive growth stage, or maturity stage.

[0046] After obtaining the labeled data, the historical environmental data series undergoes temporal feature extraction processing. Specifically, each environmental parameter is analyzed using a sliding time window approach to extract the temporal features of multiple parameters from the historical environmental data series. These temporal features include, but are not limited to, the sliding window mean, the slope of the trend, and the periodicity intensity. The sliding window mean characterizes the average level of the environmental parameter within that period; the slope of the trend reflects the parameter's upward or downward trend over time; and the periodicity intensity characterizes the impact of diurnal or irrigation cycles on parameter fluctuations. The extraction of periodicity intensity can be based on Fourier transform or autocorrelation analysis.

[0047] The time-series features corresponding to each environmental parameter are combined to form a feature vector. This feature vector is used as the model input, and the corresponding growth stage label is used as the output to train a machine learning classifier, thus obtaining a crop growth stage discrimination model.

[0048] In one implementation, the Pearson correlation coefficient can be used to measure the correlation between the i-th environmental parameter and the target indicator:

[0049] ;

[0050] in: This represents the correlation coefficient of the i-th environmental parameter under growth stage G;

[0051] This represents the statistical value of the environmental parameter in the j-th sample;

[0052] This indicates the yield or physiological target value of the corresponding sample;

[0053] This represents the mean of the i-th environmental parameter across all N samples;

[0054] This represents the mean of the target indicator y across all N samples;

[0055] N is the number of samples.

[0056] Subsequently, the correlation coefficients of each environmental parameter were normalized to generate a weight template for the key environmental parameters corresponding to this growth stage. (Key Environmental Parameter Weight Template) The correlation coefficients between various environmental parameters and final yield or key physiological indicators at each growth stage are normalized. Key environmental parameters are environmental variables related to crop growth status, including at least one or more of soil temperature, soil moisture, light intensity, air temperature and humidity, and nutrient concentration. For example, the following normalization method can be used:

[0057] ;

[0058] Where n is the total number of environmental parameters participating in the weight allocation, and all weight coefficients satisfy the normalization condition.

[0059] when When this occurs, the weight of this environmental parameter is set to zero or it is not included in the weight template construction.

[0060] Training data for the crop growth stage discrimination model can be derived from historical field monitoring data. Each training sample set should include at least: a sequence of original multi-parameter environmental data collected over a period of time, and corresponding crop growth stage annotation information. The historical field monitoring data can be set according to crop type and monitoring density, for example, no fewer than 100 sample data sets per growth stage to ensure the stability of model training. The training set and validation set can be divided in a 7:3 or 8:2 ratio to prevent model overfitting. Growth stage annotation information can be obtained based on manual field surveys, agricultural records, or statistical results of crop growth history.

[0061] By constructing the crop growth stage discrimination model and key environmental parameter weight templates described above, this application can accurately identify the current growth stage of a crop based on multi-parameter time-series characteristics and dynamically assign importance weights to environmental parameters for different stages. This provides a clear stage-appropriate basis for the subsequent construction of environmental feature vectors and the assessment of growth status, thereby improving the accuracy and stability of the overall system in judging crop growth status under complex field environmental conditions, and enhancing the technical rationality and practical value of the field crop growth environment dynamic monitoring system.

[0062] This application further proposes that the feature fusion module extracts the measured values ​​of various parameters (such as soil moisture, temperature, light intensity, etc.) from the original environmental data sequence based on the current growth stage identifier, and calls the pre-statistical historical mean and standard deviation for that stage (sourced from historical data of multiple growth cycles, such as 3 growing seasons, 10 plots, ≥100 sets per season). Subsequently, a weighted fusion calculation is performed based on the weight template of key environmental parameters.

[0063] By performing standardization on each environmental parameter and combining it with the weight coefficients assigned to that parameter in the key environmental parameter weight template, the corresponding scalar feature value is calculated. In one specific implementation, the specific steps for generating the environmental feature vector for stage adaptation include:

[0064] Calculate scalar eigenvalues Through one or more scalar eigenvalues The set constitutes the environmental feature vector;

[0065] Among them, scalar eigenvalues The calculation formula is as follows:

[0066] ;

[0067] Where n represents the total number of environmental parameters involved in the fusion calculation;

[0068] i represents the sequence number of the environmental parameter. ;

[0069] This represents the current measurement value of the i-th environmental parameter extracted from the original environmental data sequence;

[0070] and Let represent the mean and standard deviation of the i-th environmental parameter in the historical dataset for the corresponding growth stage, respectively. The historical dataset should cover at least one complete growth cycle, preferably multiple growth cycles, to improve the stability of the statistical results. For cases with insufficient sample size, a sliding time window or data from adjacent growth stages can be used to supplement the statistics.

[0071] This represents the weight coefficient assigned to the i-th environmental parameter from the key environmental parameter weight template, and all weight coefficients satisfy the normalization condition: This normalization condition is used to ensure that the relative contributions of different environmental parameter weights in the fusion calculation are controllable, avoiding the excessive dominance of a single parameter on the scalar characteristic value. The weight coefficients typically range from [-1, 1] to [0, 1], and the specific value can be determined based on correlation analysis methods.

[0072] When the standard deviation of a certain environmental parameter in historical data at the current growth stage When the value is below a preset threshold, this parameter can be considered a stable parameter, and its standardized result has a relatively small impact on the scalar characteristic value. The standard deviation threshold can be set empirically based on the fluctuation range of historical data, for example, set to 5% to 15% of the historical mean of this parameter.

[0073] By employing the aforementioned environmental feature vector generation method, this application can transform multi-source environmental parameters into statistically significant and stage-adaptive feature expressions based on crop growth stage identification. This allows environmental features to reflect not only the current measurement value but also its deviation from the historical distribution of that growth stage. Combined with key environmental parameter weight templates, the differences in the importance of environmental parameters at different growth stages are reflected in the feature fusion results. This provides more stable and discriminative input features for subsequent growth status assessment models, helping to reduce the impact of differences in environmental parameter dimensions and random fluctuations on the assessment results, and improving the reliability and consistency of crop growth status assessment results.

[0074] This application further proposes that the system also includes a health compensation module, connected between the data acquisition module and the feature fusion module. During system operation, this module first receives and analyzes the data drift rate, noise level, or response consistency of individual sensors in the raw environmental data sequence to assess the sensor's health status. Specifically, the data drift rate describes the slow shift trend of the sensor's output value relative to a historical benchmark; the noise level describes the short-term fluctuation amplitude of the sensor's output signal; and the response consistency describes the degree of consistency in the output results of multiple similar sensors or the same sensor within adjacent time windows. The data drift rate can be calculated by comparing the mean of environmental parameters within the current time window with the historical reference mean; the noise level can be obtained by statistically analyzing the standard deviation or variance of the data within the current time window; and the response consistency can be evaluated through correlation analysis between similar sensors or between adjacent time windows. By comprehensively analyzing the above indicators, a health coefficient for the corresponding sensor is generated.

[0075] In one implementation, the sensor's health coefficient The following weighted formula can be used for calculation:

[0076] ;

[0077] in, This represents the normalized data drift rate.

[0078] This represents the normalized noise level.

[0079] In response to consistency metrics;

[0080] α, β, and γ are preset weighting coefficients and satisfy α+β+γ=1.

[0081] The above formula allows the sensor's health coefficient to be between 0 and 1, with the value closer to 1 indicating a better sensor health status.

[0082] The sensor health coefficients are fused with the key environmental parameter weight templates in a secondary process to generate compensated weight coefficients. This fusion process is used to suppress the weight contribution of sensors with poor health status or enhance the weight contribution of sensors with good health status while preserving the importance distribution relationship of environmental parameters during the growth stage.

[0083] During the weight compensation process, the compensated weight coefficients This can be achieved by multiplying or weighting the original weight coefficients with the corresponding sensor health coefficients, and then normalizing them to meet the constraints on the total weight amount during feature fusion. As an example, ,in The health coefficient is then re-normalized.

[0084] The feature fusion module updates the weight template of key environmental parameters using the compensated weight coefficients. When generating environmental feature vectors, the feature fusion module performs weighted fusion calculations on the parameter data in the original environmental data sequence based on the updated weight coefficients, so that the environmental feature vectors can simultaneously reflect the characteristics of the growth stage and the health status of the sensor.

[0085] By introducing a health compensation module and incorporating sensor health status into the calculation of environmental parameter weights, this application effectively reduces the uncertainty caused by sensor performance degradation or abnormal fluctuations during the environmental feature fusion stage. This mechanism enables the environmental feature vector to not only reflect the environmental adaptability during the growth stage but also to further reflect the reliability differences of environmental data sources, thereby improving the stability and reliability of the input features of the subsequent growth status assessment model.

[0086] This application further proposes that, during system operation, when the decision trigger judgment module receives the quantitative evaluation value of the growth status output by the growth status evaluation module, and completes the preliminary threshold interval comparison in combination with the growth stage identifier output by the dynamic identification module, it further executes the judgment process based on future environmental evolution.

[0087] Before generating and outputting the decision trigger command, the decision trigger judgment module is also used for:

[0088] Acquire weather forecast data for a predetermined time period. Specifically, weather forecast data can be obtained through an external meteorological service interface and must include meteorological parameters closely related to changes in the field environment, such as temperature, probability or amount of rainfall, light conditions, and air humidity, for at least several hours or days in the future. After acquisition, the weather forecast data can undergo necessary format conversion and time alignment processing to adapt to the system's internal analysis models. Regarding the setting of the predetermined time period, a reasonable forecast window can be selected based on crop type and management needs, such as the next 12 to 72 hours, to cover the timescales at which common meteorological changes affect crop growth. The length of this time period can be set during the system deployment phase based on historical meteorological data and crop response characteristics.

[0089] Based on growth stage identifiers and quantitative assessment values ​​of growth status, a pre-defined environmental evolution simulation model is invoked to simulate the environmental evolution trend under the influence of weather forecast data. This model characterizes the potential impact of meteorological changes on field environmental parameters, such as the effect of rainfall on soil moisture and the impact of temperature changes on evapotranspiration rates. By inputting weather forecast data into the model, a predicted sequence of environmental parameters for a preset future time period is generated. The decision-triggered judgment module further calculates the corresponding growth status change trend based on the environmental parameter prediction sequence, obtaining the growth status evolution results under simulated conditions. By comprehensively analyzing this simulation result with the current quantitative assessment value of growth status and the corresponding threshold range, it is determined whether the deviation of the current growth status will persist or show a worsening trend in the future.

[0090] Environmental evolution simulation models can be trained based on historical environmental and meteorological data. Training data can include historical weather forecasts, corresponding actual meteorological observations, and records of environmental parameter changes during the same period. By performing regression analysis or machine learning modeling on this data, a mapping relationship between weather parameters and changes in field environmental parameters can be established. The learning rate ranges from 0.001 to 0.01, the number of training epochs is 100-500, and the batch size is 16-64. After the model is trained, it can quickly predict environmental changes under future weather conditions during system operation.

[0091] Subsequently, a decision trigger command is generated only when the simulation results meet the preset conditions for continuous deterioration. These conditions include: the quantitative assessment value of the growth status falling outside a preset dynamic threshold range at multiple consecutive time points in the future. In practice, the number of consecutive time points can be set according to the crop type and environmental response characteristics, for example, 2 to 5 consecutive time points, to avoid misjudgments of the decision trigger results due to single-point prediction errors.

[0092] Furthermore, conditions for continued deterioration may also include: the deviation of the quantitative assessment value of growth status further increases compared to the current assessment value within a future time period. The threshold for determining the increase in deviation can be set at 10% to 30% of the deviation of the current quantitative assessment value of growth status, in order to distinguish between normal fluctuations and actual deterioration trends.

[0093] By introducing an environmental evolution simulation analysis mechanism based on weather forecasts, this application can proactively assess the necessity of management decisions by considering the current crop growth status and future trends in natural environmental changes. This effectively reduces unnecessary decision triggers caused by short-term environmental fluctuations, making the generated decision trigger instructions more targeted and timely, and contributing to improving the stability and management efficiency of the field crop growth environment dynamic monitoring system in practical applications.

[0094] This application further proposes that, in the step of comparing the quantitative evaluation value of growth status with a preset dynamic threshold range, the preset dynamic threshold range is a dynamic function based on the collaborative relationship of multiple parameters in the original environmental data sequence, specifically including:

[0095] During system operation, after receiving the quantitative evaluation value of growth status and the corresponding growth stage identifier, the decision trigger judgment module no longer directly calls the fixed threshold range for comparison, but first calculates the dynamic decision threshold based on the current growth stage and the original environmental data sequence.

[0096] Specifically, the decision-triggered judgment module retrieves the basic threshold corresponding to the growth stage from the preset threshold library based on the growth stage identifier G. This baseline threshold is used to characterize the reasonable evaluation boundary of crop growth status within this growth stage under standard or reference environmental conditions.

[0097] Subsequently, several co-environmental parameters that showed significant correlation with changes in growth state were selected from the original environmental data sequence. Factors such as temperature, soil moisture, light intensity, and air humidity are considered, and dynamic correction terms are constructed based on these parameters. By substituting the current measurements of each co-environmental parameter into the standardized influence function, the degree of deviation of each parameter from the historical mean of this growth stage is obtained.

[0098] Based on this, and according to the growth stage identifiers and relevant environmental parameters in the original environmental data sequence, a dynamic decision threshold is calculated using a dynamic function. The formula is as follows:

[0099] ;

[0100] in, The basic threshold corresponding to growth stage G can be obtained based on historical planting trial data, agronomic standards, or long-term monitoring data. For example, by analyzing the distribution of quantitative evaluation values ​​of the growth status of healthy growth samples at each growth stage, the mean or confidence interval boundary can be selected as the basic threshold reference.

[0101] Let G be the k-th cooperative environmental parameter during growth stage G. The influence coefficient on the threshold is obtained through historical data regression analysis. Specifically, during the model training phase, multiple regression or regularized regression methods can be used to analyze the sensitivity of each collaborative environment parameter to the growth status assessment results, thereby determining its weight at different growth stages. ∈(-1,1), representing the positive or negative influence of this parameter on the threshold;

[0102] For collaborative environment parameters The standardized influence function is defined as:

[0103] ;

[0104] in and These are the parameters of the collaborative environment. Mean and standard deviation of G in historical data during the growth phase;

[0105] m represents the number of collaborative environmental parameters, which can be set according to crop type and sensor configuration, for example, 2 to 6.

[0106] The decision-triggered judgment module combines the quantitative evaluation value of the growth status with the dynamic decision threshold. The comparison is performed, and a decision-making instruction is triggered based on preset comparison rules. Regarding the comparison rules, a dynamic decision threshold can be set for the quantitative evaluation value of the growth status to be higher or lower than the threshold value. A certain percentage (e.g., 5% to 20%) is considered a valid deviation to avoid false triggering caused by minor fluctuations.

[0107] By constructing a preset dynamic threshold range as a dynamic function based on the synergistic relationship of multiple environmental parameters, this application enables the decision-triggered judgment process to fully consider the comprehensive impact of environmental condition changes at different growth stages. This avoids the problem of insufficient adaptability of static thresholds in complex field environments, making the comparison results of quantitative assessment values ​​of growth status more consistent with actual physiological response characteristics. This improves the accuracy and environmental adaptability of the decision-triggered results, and overall enhances the practicality and reliability of the field crop growth environment dynamic monitoring system.

[0108] This application further proposes that after generating and outputting decision triggering instructions, the system also includes an instruction feedback learning module. During system operation, whenever the decision triggering judgment module generates and outputs a decision triggering instruction, the instruction feedback learning module fully records that decision. It records each output decision triggering instruction and its corresponding original environmental data sequence, growth stage identifier, and changes in the quantitative evaluation value of the growth state before and after the decision.

[0109] As the system operates over a long period, the instruction feedback learning module will accumulate data from multiple complete growth cycles. After a preset number of M growth cycles, the module performs batch analysis of historical records, grouping decision samples with the same or similar instruction types, the same growth stage identifiers, and environmental feature vector similarity exceeding a preset threshold into the same scenario set. Within this set, the instruction feedback learning module analyzes and generates effectiveness indicators for similar instructions in similar scenarios by comparing the trends in quantitative evaluation values ​​of growth status before and after decision execution. The number of growth cycles M can be set according to the crop type and growth cycle length. For example, for crops that are harvested once a year, M can be set to 2 to 5 complete growth cycles to ensure sufficient data support for feedback learning while avoiding system instability caused by overly frequent updates.

[0110] Effectiveness indicators can be used to characterize whether an instruction prompts the growth status to revert to the expected range, as well as the rate and stability of this regression. It can be defined as a combination of the magnitude and rate at which the quantitative assessment value of the growth status reverts to the target threshold range within a preset time window after the decision. The specific calculation method can be adjusted according to the crop type, but is not limited to a fixed form.

[0111] Based on the effectiveness index, the instruction feedback learning module further determines whether adaptive adjustments to system parameters are needed. When a certain type of instruction consistently demonstrates insufficient effectiveness in a specific growth stage and environmental scenario, the module will adaptively adjust and update the preset dynamic threshold range or key environmental parameter weight template for the corresponding growth stage. To ensure system stability, the single adjustment ratio of the threshold range or weight template can be limited to ±5% to ±15% of the original parameter value to achieve gradual optimization. When the effectiveness remains at a relatively good level, the existing parameter settings are maintained unchanged.

[0112] Through the above technical solution, this application enables the system to gradually accumulate decision-making experience during long-term operation and continuously revise internal parameter settings based on actual growth feedback. This allows the preset dynamic threshold range and key environmental parameter weight templates to no longer rely on one-time manual settings, but rather to adaptively evolve with changes in crop growth characteristics and environmental conditions. This improves the adaptability and continuous effectiveness of decision-triggered commands in complex field environments, enhancing the stability and intelligence level of the overall field crop growth environment dynamic monitoring system.

[0113] This application further proposes that the system also includes a spatial heterogeneity correction module, which receives raw environmental data sequences from monitoring points at different locations in the field during system operation. The spatial location of each monitoring point can be distinguished by pre-set geographic coordinates, field grid numbers, or relative location identifiers.

[0114] Based on each monitoring point, the coefficient of variation of historical data and soil texture auxiliary data are calculated. The coefficient of variation of historical data is used to characterize the relative stability of environmental parameter fluctuations at the monitoring point within a preset time scale. A smaller coefficient of variation indicates that the environmental conditions at that location are relatively stable and are more likely to represent the long-term average state of the area.

[0115] The statistical time window for the coefficient of variation of historical data can be set according to the crop growth rhythm, such as selecting 7 days, 14 days or a complete growth stage as the calculation period, so as to reflect spatial stability while avoiding interference from short-term abnormal fluctuations.

[0116] In one implementation method, for the j-th monitoring point, its historical data variation coefficient is... It can be calculated as follows:

[0117] ;

[0118] in, This represents the standard deviation of a certain environmental parameter at that monitoring point within the statistical time window. This represents the corresponding mean. When multiple environmental parameters are involved, the coefficients of variation of each parameter can be weighted and averaged or the maximum value can be taken as the comprehensive coefficient of variation.

[0119] Soil texture auxiliary data may include information such as soil type, texture grade, sand-clay ratio, or organic matter content range. This type of data can be derived from soil survey results, database data, or previous manual sampling analysis results, and is used to characterize the differences in prior conditions at different spatial locations. Soil texture auxiliary data can be mapped into numerical form through discrete levels, for example, mapping sandy soil, loam, and clay to different texture coefficient ranges to participate in the calculation of spatial representation weights.

[0120] Based on this, the module integrates historical data variation coefficients with soil texture auxiliary data to calculate the spatial representative weight of each monitoring point. This reflects the contribution of each monitoring point to the integration of overall field environmental characteristics. To avoid bias in the overall data due to excessively high weights for individual monitoring points, an upper limit constraint can be set on the spatial representative weight, for example, limiting the weight of a single monitoring point to no more than 40% to 50% of the total weight.

[0121] In one possible implementation, space represents weight. It can be calculated in the following form:

[0122] ;

[0123] in: This represents the soil texture characteristic value corresponding to the i-th monitoring point;

[0124] and This is a normalization function used to map data of different dimensions to a uniform scale;

[0125] To adjust the fusion coefficient between historical stability and soil prior information relative weights, it can be set between 0.5 and 0.8.

[0126] N represents the total number of monitoring points.

[0127] Data from different monitoring points are weighted and fused according to spatial representation weights to generate a unified, standardized environmental data sequence representing the field scale. The original environmental data sequences received and processed by the dynamic identification module and subsequent modules are standardized environmental data sequences processed by the spatial heterogeneity correction module.

[0128] By introducing a spatial heterogeneity correction module and performing weighted fusion processing on environmental data from multiple monitoring points, this application can effectively reduce the interference of local abnormal environments or highly fluctuating areas on the overall analysis results at the field scale, making the environmental data used by subsequent modules more stable and spatially representative. This improves the consistency and reliability of crop growth stage discrimination, growth status assessment, and decision trigger judgment under complex spatial environmental conditions, thereby enhancing the adaptability of the entire field crop growth environment dynamic monitoring system to practical application scenarios.

[0129] This application further proposes that the system also includes a dynamic nutrient compensation module, connected between the data acquisition module and the feature fusion module. The dynamic nutrient compensation module first receives real-time nutrient concentration data collected by nanosensors and queries a pre-set crop stage nutrient requirement map corresponding to the growth stage identifier. The nutrient concentration data may include one or more of available nitrogen, phosphorus, and potassium in the soil, and its collection frequency is consistent with the original environmental data sequence or synchronized through time alignment. The crop stage nutrient requirement map is used to characterize the target requirement range or target value of various nutrients for crops at different growth stages. The crop stage nutrient requirement map can be established separately according to different crop types. For example, for food crops, the growth cycle can be divided into seedling stage, vegetative growth stage, reproductive growth stage, etc., and target nutrient ranges can be set for each stage. The target value can be selected as the median of the range or an empirically recommended value as the calculation benchmark.

[0130] After obtaining real-time nutrient concentrations and staged nutrient requirement targets, the deviation between the real-time nutrient concentrations and the target values ​​in the crop stage nutrient requirement map is calculated. The deviation is used to quantify the degree of insufficiency or surplus of current nutrient supply relative to target demand. Based on the deviation and a preset nutrient supplementation efficiency coefficient, the deviation is corrected to generate a dynamic nutrient compensation factor, reflecting the actual impact of nutrient deviation on crop growth under the current growth stage and environmental conditions.

[0131] When generating environmental feature vectors, the feature fusion module treats the dynamic nutrient compensation factor as an independent environmental stress correction term. After weighting according to the key environmental parameter weight template, it is superimposed on the weighted fusion calculation result to generate an environmental feature vector that includes the influence of nutrient stress.

[0132] In one possible implementation, the calculation process of the dynamic nutrient compensation factor can be expressed in the following form:

[0133] First, define the nutrient deviation ΔN(G):

[0134] ;

[0135] in: This indicates the real-time nutrient concentration collected by the nanosensor;

[0136] This represents the target nutrient value in the nutrient requirement map corresponding to the crop stage G.

[0137] Subsequently, the nutrient replenishment efficiency coefficient η(G) is introduced to generate a dynamic nutrient compensation factor. :

[0138] ;

[0139] in, This value, ranging from 0.3 to 1.2, is used to characterize the crop's response efficiency to nutrient changes at growth stage G, and is determined by historical experimental data or model regression. For example, during the rapid growth phase... It can be set to 0.8 to 1.2, and can be reduced to 0.3 to 0.6 during the mature stage.

[0140] To avoid the excessive influence of nutrient outliers on feature fusion, a dynamic nutrient compensation factor is used. The absolute value can be set with an upper limit threshold, for example, limited to the range of -1 to 1.

[0141] In the feature fusion stage, environmental feature vector components that include the effects of nutrient stress are included. It can be represented as:

[0142] ;

[0143] in: The feature values ​​are obtained by weighted fusion of the original environmental parameters;

[0144] This represents the corresponding weight of the nutrient compensation factor in the key environmental parameter weight template.

[0145] By introducing a dynamic nutrient compensation module during feature fusion, this application quantitatively integrates real-time nutrient supply-demand deviations into the environmental feature vector, enabling the growth status assessment model to explicitly consider nutrient stress factors when judging crop growth status. This improves the system's sensitivity to nutrient deficiency or excess, making the generated quantitative growth status assessment values ​​more consistent with actual physiological response characteristics, thereby enhancing the accuracy and practicality of the field crop growth environment dynamic monitoring system in nutrient management-related decision-making.

[0146] This application further proposes that the system also includes a growth trend consistency verification module, connected between the state assessment module and the decision trigger judgment module. The growth trend consistency verification module first receives a sequence of quantitative evaluation values ​​of growth status within a continuous time window. Simultaneously, the module receives a growth stage identifier sequence that is time-aligned with the quantitative evaluation value sequence of growth status, used to clarify the growth stage or transition interval of the crop within that time window. The time window can cover several continuous sampling periods, and its length can be set according to the crop growth rhythm to ensure that it reflects the stage-specific change trend. Based on the growth stage identifier sequence, it determines whether the actual change trajectory of the crop growth status conforms to the expected growth model of the corresponding stage. This expected growth model is used to describe the typical trend of the quantitative evaluation value of growth status changing over time within that growth stage under normal conditions. The expected growth model can be trained based on historical health sample data, for example, by performing regression fitting or time series modeling on the quantitative evaluation values ​​of growth status within different growth stages to form a typical growth curve for each stage. The model form can be a linear trend model, an exponential model, or a smoothed time series model, without specific limitations.

[0147] Subsequently, the growth trend consistency verification module aligns the actual growth state quantitative evaluation value sequence with the expected growth model output sequence and calculates the similarity index between the two. This similarity index is used to quantify the degree of consistency between the actual growth trajectory and the theoretical expected trajectory.

[0148] When generating the decision trigger command, the decision trigger judgment module further incorporates this similarity index as a confidence weight for judgment:

[0149] When the similarity index is greater than the preset similarity threshold, a decision trigger instruction is executed;

[0150] When the similarity index is less than the preset similarity threshold, the system is considered to be in an abnormal state or there is unmonitored interference. This triggers the system self-check process and suspends the execution of decisions, while outputting a prompt message indicating that manual review is required.

[0151] In one possible implementation, the similarity index S can be calculated based on the distance or correlation between the actual growth state quantitative assessment value sequence and the expected growth model sequence, for example, expressed in the form of a normalized correlation coefficient:

[0152] ;

[0153] in: This represents the quantitative assessment value of the actual growth status at time point t.

[0154] This represents the expected output value of the growth model at time point t;

[0155] and These represent the mean of the corresponding sequences within the time window;

[0156] T represents the number of sampling points within the time window. The length of the time window T can be set according to the crop type and monitoring frequency. For example, under daily sampling conditions, 5 to 15 consecutive sampling points can be selected to balance trend stability and response sensitivity.

[0157] The similarity index S ranges from -1 to 1, with a value closer to 1 indicating a closer match between the actual growth trend and the expected model. A preset similarity threshold is set. The value can be set based on historical verification results, and can be set to 0.6~0.8. When the similarity index is lower than the preset similarity threshold... When this occurs, the exception handling logic is triggered.

[0158] The system self-test process may include operations such as sensor health verification, data integrity verification, or communication status detection. Its specific implementation does not affect the core functional description of this module.

[0159] By introducing a growth trend consistency verification mechanism, this application can effectively distinguish between normal growth fluctuations and abnormal changes, enabling the system to operate more robustly under complex field environments and multi-source data conditions, reducing the probability of misjudgment and false triggering, and improving the overall security and reliability of the field crop growth environment dynamic monitoring and management system.

[0160] When there are potential conflicts in the processing results or judgment conditions of different modules, this application uses the following rules to handle them uniformly:

[0161] When the spatial heterogeneity correction module or the health compensation module determines that the reliability of the data of a certain environmental parameter or monitoring point is lower than the preset threshold, the data corresponding to that parameter or monitoring point will have its weight reduced or will not participate in the fusion process in the subsequent feature fusion process. Its result takes precedence over the input requirements of the dynamic identification module and the decision trigger judgment module.

[0162] When the growth stage identifier is updated, the system uses the latest identified growth stage identifier as the current valid stage, and accordingly calls up the corresponding key environmental parameter weight template, nutrient requirement map and threshold calculation rules again. The intermediate results generated in the previous stage are no longer used for decision-making in the current cycle.

[0163] When the quantitative evaluation value of growth status meets the threshold triggering condition, but the similarity index output by the growth trend consistency verification module is lower than the preset similarity threshold. When this happens, the system prioritizes executing the trend consistency verification results, postpones or cancels the execution of decision trigger instructions, and enters the anomaly detection or manual review process.

[0164] When the health compensation factor, nutrient dynamic compensation factor, and spatial representation weight are applied simultaneously to the construction process of the environmental feature vector, the effect of each compensation factor must meet the preset amplitude constraint conditions to prevent a single compensation factor from having an excessive dominant influence on the fusion result; when the superposition result exceeds the preset upper limit, the system normalizes or truncates the compensation result.

[0165] When the dynamic threshold calculation results are inconsistent with the environmental evolution simulation results based on weather forecasts, the system prioritizes the judgment results based on the environmental evolution trend in order to avoid unnecessary decision-making triggers caused by short-term fluctuations or prediction errors.

[0166] The following is a specific example of a crop environment quantitative assessment and decision-making system based on growth stage adaptation:

[0167] A smart agriculture demonstration zone conducted dynamic monitoring of the entire growth cycle of winter wheat planted in contiguous areas. The system evenly deployed five monitoring nodes in the field, each integrating multiple types of nanosensors: soil temperature and humidity were monitored using mesoporous silica nanosensors (sampling frequency 1 time / hour); soil nitrogen, phosphorus, and potassium nutrients were monitored using quantum dot fluorescent nanosensors (detection limit down to 0.1 mg / kg); and air temperature, humidity, and light intensity were monitored using carbon nanotube composite sensors (response time <2 seconds). The data acquisition module collected raw environmental data sequences in real time via a LoRa wireless communication network (transmission distance 5 km), covering soil moisture (%), soil temperature (°C), air humidity (%), and light intensity (%). Five parameters were included: soil available nitrogen (mg / kg).

[0168] During the jointing stage of winter wheat (identified by a dynamic identification module using a crop growth stage discrimination model, with input time-series features including the sliding window mean, trend slope, and periodic intensity), the current growth stage is dynamically identified as "jointing stage," and the corresponding key environmental parameter weight template is invoked. This template is obtained by normalizing the correlation coefficients between various environmental parameters during the jointing stage and the final yield: soil moisture weight. Soil temperature air humidity Light intensity available nitrogen in soil The feature fusion module uses this template to perform weighted fusion of the original data, generating a stage-adaptive environment feature vector. Taking the original data at a certain moment as an example: soil moisture... (historical average) Standard deviation ), soil temperature ℃ ℃, ℃), air humidity ( , ), light intensity ( , available nitrogen in soil mg / kg ( mg / kg, (mg / kg). Scalar eigenvalue V was calculated. .

[0169] The state assessment module inputs environmental feature vectors into the growth state assessment model (trained based on a multilayer perceptron), outputting a quantitative assessment value of 1.417 for the current growth state. The decision-triggered judgment module calls the preset dynamic threshold range (1.0~1.3) for the jointing stage and finds that the quantitative assessment value of 1.417 deviates from the upper limit of the threshold. Further acquisition of 72-hour weather forecast data (temperature 16~18℃, no rainfall, sufficient sunshine) and simulation of environmental evolution trends show that soil moisture will continue to increase (predicted value 30%), and the deviation in growth state will worsen. Simultaneously, the system calculates the dynamic threshold using the dynamic decision threshold formula: (Basic threshold at the jointing stage), select soil moisture ( ) and light intensity ( ( ) represents the collaborative environment parameter, and its influence coefficient. , Standardized influence function , Obtain the dynamic decision threshold .

[0170] Given that the assessed value of 1.417 is still below the dynamic threshold The initial assessment was that the soil was safe. However, combined with the weather forecast simulation results, which indicated that soil moisture would continue to rise abnormally, potentially leading to risks not covered by the static model, such as root hypoxia, the system ultimately determined, based on the 'weather forecast simulation results' and the 'expert rule base,' that a decision trigger command to 'reduce soil moisture' still needed to be generated. If the risk level predicted by the weather forecast simulation exceeds a preset risk threshold, the decision trigger command will be generated primarily based on the weather forecast simulation results.

[0171] During system operation, the health compensation module monitors sensor status in real time: it detected that the data drift rate of one group of soil moisture sensors reached 8%, generating a health coefficient of 0.85. Its weight was adjusted from 0.35 to 0.35 × 0.85 = 0.2975, and the weight template was updated after re-normalization. The spatial heterogeneity correction module calculates the spatial representativeness weight (0.18–0.22) of each point based on the historical data variation coefficients (range 0.05–0.12) and soil texture (mainly loam) of five monitoring points. After weighted fusion, a standardized environmental data sequence is generated to ensure data representativeness. The nutrient dynamic compensation module detects that the available nitrogen in the soil deviates from the target value by 10 mg / kg, generating a dynamic nutrient compensation factor of 0.07, which is superimposed on the environmental feature vector to make the assessment results more accurate.

[0172] The growth trend consistency verification module receives a 7-day sequence of quantitative evaluation values ​​of growth status and compares it with the expected growth model (linear growth trend) during the jointing stage. A similarity index of 0.82 (higher than the preset threshold of 0.7) confirms a normal growth trend, triggering the decision-making instruction. After multiple system cycles, the instruction feedback learning module adaptively optimizes the threshold or weight template based on historical decision-making results.

[0173] During its application in the demonstration area, this implementation achieved dynamic identification of winter wheat growth stages and stage-specific adaptability assessment of environmental parameters. By dynamically adjusting weights and decision thresholds, the system promptly triggered irrigation control commands when soil moisture abnormally increased, thus maintaining stable growth during the winter wheat jointing stage and laying the foundation for high yields during the subsequent heading stage.

[0174] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A crop environment quantitative assessment and decision-making system based on growth stage adaptation, characterized in that: include: The data acquisition module is used to acquire raw environmental data sequences collected by multiple types of sensors; The dynamic recognition module is used for: The system receives the raw environmental data sequence and inputs it into a preset crop growth stage discrimination model for matching analysis to identify the current growth stage of the crop. The system dynamically outputs the current growth stage identifier of the crop and the corresponding key environmental parameter weight template. The key environmental parameter weight template is obtained by normalizing the correlation coefficients between each environmental parameter and the final yield or key physiological indicators in each growth stage. The key environmental parameters are environmental variables related to the crop growth status, including at least one or more of soil temperature, soil moisture, light intensity, air temperature and humidity, and nutrient concentration. The feature fusion module is used to perform weighted fusion calculation on each parameter data in the original environmental data sequence based on the key environmental parameter weight template, and generate an environmental feature vector for stage adaptability. The status assessment module is used to input the environmental feature vector into a preset growth status assessment model, analyze and output the quantitative assessment value of the current crop's growth status. The decision-triggered judgment module is used for: Receive the quantitative evaluation value of the growth status and the growth stage identifier, and call the corresponding preset dynamic threshold range according to the growth stage identifier; The quantitative evaluation value of the growth status is compared with the preset dynamic threshold range; When the quantitative evaluation value of the growth status deviates from the preset dynamic threshold range, a decision trigger instruction is generated and output. The preset dynamic threshold range is a dynamic function based on the collaborative relationship of multiple parameters in the original environmental data sequence, specifically including: Based on the growth stage identifier and relevant environmental parameters in the original environmental data sequence, the dynamic decision threshold is calculated using the dynamic function. The formula is as follows: in, This is the base threshold corresponding to the growth stage G; For the k-th cooperative environmental parameter under the growth stage G, The influence coefficient on the threshold was obtained through regression analysis of historical data; the collaborative environment parameters Selected from the original environmental data sequence; For collaborative environment parameters The standardized influence function is defined as: in and These are the parameters of the collaborative environment. Mean and standard deviation of G in historical data during the growth phase; m is the number of collaborative environment parameters; The decision triggering judgment module compares the quantitative evaluation value of the growth state with the dynamic decision threshold. The comparison is performed, and a decision instruction is triggered based on the preset comparison rules.

2. The crop environment quantitative assessment and decision-making system based on growth stage adaptation according to claim 1, characterized in that: The crop growth stage discrimination model is constructed through the following steps: Collect historical environmental data sequences and corresponding agronomic record growth stage labels for the target crop throughout its complete growth cycle; The temporal features of multiple parameters in the historical environmental data sequence are extracted; the temporal features include the sliding window mean, the slope of the change trend, and the intensity of periodicity. Using the time-series features as input and the growth stage labels as output, a machine learning classifier is trained to obtain the crop growth stage discrimination model.

3. The crop environment quantitative assessment and decision-making system based on growth stage adaptation according to claim 1, characterized in that: The specific steps for generating environmental feature vectors for adaptation during the generation phase include: Calculate scalar eigenvalues Through one or more of the scalar feature values The set constitutes the environmental feature vector; Wherein, the scalar eigenvalue The calculation formula is as follows: Where n represents the total number of environmental parameters involved in the fusion calculation; i represents the sequence number of the environmental parameter. ; This represents the current measurement value of the i-th environmental parameter extracted from the original environmental data sequence; and Let represent the mean and standard deviation of the i-th environmental parameter in the historical dataset of the corresponding growth stage, respectively. This represents the weight coefficient assigned to the i-th environmental parameter from the key environmental parameter weight template, and all weight coefficients satisfy the normalization condition: .

4. The crop environment quantitative assessment and decision-making system based on growth stage adaptation according to claim 1, characterized in that: The system further includes a health compensation module, connected between the data acquisition module and the feature fusion module, for: Receive and analyze the data drift rate, noise level, or response consistency of individual sensors in the raw environmental data sequence to generate sensor health coefficients; The sensor health coefficient is fused with the key environmental parameter weight template in a secondary manner to generate a compensated weight coefficient. The feature fusion module updates the key environmental parameter weight template using the compensated weight coefficients and performs the weighted fusion calculation.

5. The crop environment quantitative assessment and decision-making system based on growth stage adaptation according to claim 1, characterized in that: Before generating and outputting the decision triggering instruction, the decision triggering judgment module is also used for: Obtain weather forecast data for a preset time period in the future; Based on the growth stage identifier and the quantitative evaluation value of the growth status, the environmental evolution trend under the influence of the weather forecast data is simulated. The decision trigger instruction is generated only when the simulation results meet the preset continuous deterioration conditions; wherein, the continuous deterioration conditions include: the quantitative evaluation value of the growth state is outside the preset dynamic threshold range at multiple consecutive time points in the future.

6. The crop environment quantitative assessment and decision-making system based on growth stage adaptation according to claim 1, characterized in that: After generating and outputting the decision trigger command, the system further includes an command feedback learning module, used for: Record each output decision trigger command and its corresponding original environmental data sequence, growth stage identifier, and changes in the quantitative evaluation value of growth status before and after the decision; After a preset number of M growth cycles, the effectiveness indicators of similar instructions in similar scenarios are analyzed and generated. Based on the effectiveness index, the preset dynamic threshold range or the key environmental parameter weight template is adaptively adjusted.

7. The crop environment quantitative assessment and decision-making system based on growth stage adaptation according to claim 1, characterized in that: The system also includes a spatial heterogeneity correction module, used for: Receive the raw environmental data sequences from monitoring points at different locations in the field; Based on the historical data variation coefficients and soil texture auxiliary data of each monitoring point, the spatial representative weight of each monitoring point is calculated. Data from different monitoring points are weighted and fused according to the spatial representation weights to generate a standardized environmental data sequence representing the field scale. The original environmental data sequence received and processed by the dynamic identification module and subsequent modules is the standardized environmental data sequence after being processed by the spatial heterogeneity correction module.

8. The crop environment quantitative assessment and decision-making system based on growth stage adaptation according to claim 1, characterized in that: The system further includes a nutrient dynamic compensation module, connected between the data acquisition module and the feature fusion module; used for: Receive real-time nutrient concentration data collected by nanosensors and query a preset crop stage nutrient requirement map corresponding to the growth stage identifier; Calculate the deviation between the real-time nutrient concentration and the target value in the nutrient requirement map of the crop stage, and generate a dynamic nutrient compensation factor based on the deviation and a preset nutrient supplementation efficiency coefficient. When generating the environmental feature vector, the feature fusion module treats the dynamic nutrient compensation factor as an independent environmental stress correction term, weights it according to the key environmental parameter weight template, and then superimposes it into the weighted fusion calculation result to generate an environmental feature vector that includes the influence of nutrient stress.

9. The crop environment quantitative assessment and decision-making system based on growth stage adaptation according to claim 1, characterized in that: The system also includes a growth trend consistency verification module, connected between the state evaluation module and the decision trigger judgment module, used for: Receive the sequence of quantitative evaluation values ​​of the growth status within a continuous time window, and based on the sequence of growth stage identifiers, determine whether the actual change trajectory of the crop growth status conforms to the expected growth model of the corresponding stage and calculate the similarity index. When generating the decision trigger instruction, the similarity index is used as a confidence weight for judgment; When the similarity index is greater than a preset similarity threshold, the decision trigger instruction is executed.

Images