Crop demand dynamic index generation method based on multi-sensor data fusion

Abstract

This invention provides a method for generating a dynamic crop demand index based on multi-sensor data fusion, applicable to the field of precision agriculture. The method includes: comparing multi-source heterogeneous data with a pre-defined normal trend relationship based on historical data to determine a deviation term for at least one type of data in the multi-source heterogeneous data; adjusting the weights of each data type in the multi-source heterogeneous data based on the deviation term using a Bayesian probability update model to obtain an adaptive weight set; generating a time-weighted sequence based on the current growth stage information of the target crop, and generating a spatial weight matrix based on the spatial environment information of the field where the target crop is located; adjusting the adaptive weight set using the time-weighted sequence and the spatial weight matrix to obtain a spatiotemporal weight set; and performing weighted fusion calculation on the multi-source heterogeneous data based on the spatiotemporal weight set to generate a dynamic demand index for the target crop. This invention enables early, accurate, and quantitative assessment of crop water and fertilizer requirements.

Description

Technical Field

[0001] This invention relates to the field of precision agriculture technology, and in particular to a method for generating a dynamic crop demand index based on multi-sensor data fusion. Background Technology

[0002] In the field of precision agriculture, achieving efficient use of water and fertilizer resources is key to improving agricultural production efficiency. Currently, the assessment of crop demand status mainly relies on single indicators or experience-based judgments, which is insufficient to meet the needs of refined management.

[0003] Existing methods often rely on single physical or chemical parameters (such as soil moisture content and chlorophyll value), failing to comprehensively consider multi-dimensional factors such as soil supply, plant physiological response, and meteorological environment, resulting in one-sided assessment results and easy misjudgment.

[0004] This shows that the crop demand assessment methods in related technologies have significant limitations and technical problems. Summary of the Invention

[0005] This invention provides a method for generating a dynamic index of crop demand based on multi-sensor data fusion, which addresses the limitations of existing crop demand assessment methods and enables early, accurate, and quantitative assessment of crop water and fertilizer requirements.

[0006] This invention provides a method for generating a dynamic crop demand index based on multi-sensor data fusion, comprising the following steps: acquiring multi-source heterogeneous data of a target crop obtained from soil-plant-atmosphere continuum detection; comparing the multi-source heterogeneous data with a pre-defined normal trend relationship based on historical data to determine a deviation term for at least one type of data in the multi-source heterogeneous data; adjusting the weights of each data type in the multi-source heterogeneous data based on the deviation term using a Bayesian probability update model to obtain an adaptive weight set; generating a time weight sequence based on the current growth stage information of the target crop, and generating a spatial weight matrix based on the spatial environment information of the field where the target crop is located; adjusting the adaptive weight set using the time weight sequence and the spatial weight matrix to obtain a spatiotemporal weight set; and performing weighted fusion calculation on the multi-source heterogeneous data based on the spatiotemporal weight set to generate a dynamic demand index for the target crop.

[0007] According to the present invention, a method for generating a crop demand dynamic index based on multi-sensor data fusion, wherein acquiring multi-source heterogeneous data of a target crop obtained from soil-plant-atmosphere continuum detection includes: acquiring soil monitoring data from the soil level, wherein the soil monitoring data includes at least one of the following: soil moisture data measured by a soil moisture sensor and soil water potential data measured by a soil water potential sensor; acquiring plant monitoring data from the plant level, wherein the plant monitoring data includes at least one of the following: stem flow rate data measured by a stem flow sensor, canopy temperature data measured by a canopy infrared thermal imaging sensor, and stomatal conductance data measured by a stomatal conductance monitor; and acquiring atmospheric monitoring data from the atmospheric level, wherein the atmospheric monitoring data includes at least one of the following: carbon dioxide concentration data measured by a carbon dioxide concentration sensor and photosynthetically active radiation data measured by a photosynthetically active radiation sensor.

[0008] According to the present invention, a method for generating a crop demand dynamic index based on multi-sensor data fusion includes comparing the multi-source heterogeneous data with a preset normal trend relationship based on historical data to determine at least one type of deviation item in the multi-source heterogeneous data. This includes: establishing a normal trend relationship model among multiple target parameters based on historical data, and setting a deviation threshold for each target parameter, wherein the multiple target parameters include at least: soil water potential data, stem flow rate data, and canopy temperature data; comparing the multiple target parameters in the real-time collected multi-source heterogeneous data with the normal trend relationship model, and determining the target parameter with a deviation greater than the deviation threshold as a deviation item.

[0009] According to the present invention, a method for generating a dynamic index of crop demand based on multi-sensor data fusion is provided. The method further includes: determining the effective soil water range based on the soil moisture data and the soil water potential data; and determining the actual transpiration demand of the target crop based on the effective soil water range and the stem flow rate data.

[0010] According to the present invention, a method for generating a dynamic crop demand index based on multi-sensor data fusion, wherein adjusting the weights of various data types in the multi-source heterogeneous data based on the deviation term using a Bayesian probabilistic update model to obtain an adaptive weight set includes: identifying abnormal stress sources that cause data deviation based on the deviation term and determining the degree of correlation between the abnormal stress source and various data types in the multi-source heterogeneous data; in response to identifying the abnormal stress source, reducing the weights of data types whose correlation with the abnormal stress source is less than a correlation threshold, and increasing the weights of data types whose correlation with the abnormal stress source is greater than the correlation threshold, based on the Bayesian probabilistic update model, to obtain a redistributed weight set; and using the redistributed weight set as the adaptive weight set.

[0011] According to the present invention, a method for generating a dynamic crop demand index based on multi-sensor data fusion includes: generating a time-weighted sequence based on the current growth stage information of the target crop and generating a spatial weight matrix based on the spatial environment information of the field where the target crop is located; adjusting the adaptive weight set using the time-weighted sequence and the spatial weight matrix to obtain a spatiotemporal weight set, comprising: generating a time-weighted sequence by fitting a logistic growth function based on the current growth stage information of the target crop and the sensitivity differences of each data type in the multi-source heterogeneous data to each preset growth stage of the crop; generating the spatial weight matrix based on a spatial interpolation algorithm according to the spatial environment information of the field where the target crop is located, wherein the spatial environment information includes: crop planting density, vegetation coverage, and spatial distribution of vegetation index; and performing spatiotemporal dynamic adjustment of the adaptive weight set based on the time-weighted sequence and the spatial weight matrix to obtain a spatiotemporal weight set.

[0012] This invention also provides a crop demand dynamic index calculation device based on multi-sensor data fusion, comprising the following modules: an acquisition module for acquiring multi-source heterogeneous data of a target crop obtained from soil-plant-atmosphere continuum detection; a comparison module for comparing the multi-source heterogeneous data with a preset normal trend relationship based on historical data to determine a deviation term for at least one type of data in the multi-source heterogeneous data; an adaptive weighting module for adjusting the weights of each data type in the multi-source heterogeneous data based on the deviation term using a Bayesian probability update model to obtain an adaptive weight set; a spatiotemporal weighting module for generating a time weight sequence based on the current growth stage information of the target crop and generating a spatial weight matrix based on the spatial environment information of the field where the target crop is located; adjusting the adaptive weight set using the time weight sequence and the spatial weight matrix to obtain a spatiotemporal weight set; and a weighted fusion module for performing weighted fusion calculation on the multi-source heterogeneous data based on the spatiotemporal weight set to generate the demand dynamic index of the target crop.

[0013] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the crop demand dynamic index generation method based on multi-sensor data fusion as described above.

[0014] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the crop demand dynamic index generation method based on multi-sensor data fusion as described above.

[0015] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the crop demand dynamic index generation method based on multi-sensor data fusion as described above.

[0016] This invention provides a method for generating a dynamic crop demand index based on multi-sensor data fusion. First, it acquires multi-source heterogeneous data from a soil-plant-atmosphere continuum to provide a foundation for comprehensively understanding crop growth status. Next, it compares the multi-source heterogeneous data with a pre-defined normal trend relationship based on historical data to identify deviation terms and thus identify anomalies in the current data. Then, based on these deviation terms, it adjusts the weights of various data types in the multi-source heterogeneous data using a Bayesian probability update model to obtain an adaptive weight set that can adapt to data changes. Next, it generates a time-weighted sequence based on the current growth stage information of the target crop and combines it with field spatial environment information to generate a spatial weight matrix. These two components are used to adjust the adaptive weight set, resulting in a spatiotemporal weight set that integrates spatiotemporal features. Finally, it performs weighted fusion calculations on the multi-source heterogeneous data based on the spatiotemporal weight set to generate a dynamic demand index that dynamically and accurately reflects the actual needs of the crop. The entire process achieves adaptive and spatiotemporally optimized precise generation of a dynamic crop demand index from multi-source data. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced one by one below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating the crop demand dynamic index generation method based on multi-sensor data fusion provided by the present invention.

[0019] Figure 2 This is a schematic diagram of the detection of a portion of the sensors provided by the present invention.

[0020] Figure 3 This is a schematic diagram of another part of the sensor provided by the present invention.

[0021] Figure 4 This is a schematic diagram of the abnormal stress source identification process provided by the present invention.

[0022] Figure 5 This is a schematic diagram of the module of the crop demand dynamic index generation device based on multi-sensor data fusion provided by the present invention.

[0023] Figure 6 This is a schematic diagram of the physical structure of the electronic device provided by the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0025] In precision agriculture, efficient use of water and fertilizer resources is key to improving agricultural production efficiency. Currently, the assessment of crop needs mainly relies on single indicators or experience-based judgments, which are insufficient for the demands of refined management. The following is an introduction to existing crop needs assessment technologies: In related technologies, soil moisture sensors are deployed in the field to monitor soil moisture content in real time and compare it with a preset threshold. When the soil moisture falls below the set value, the irrigation system is automatically activated. This method automates irrigation, but it relies solely on soil moisture content and ignores the physiological responses of the crops themselves.

[0026] In related technologies, a handheld chlorophyll meter is used to measure the relative chlorophyll content of crop leaves, which serves as an indicator of the crop's nitrogen nutrition status and guides topdressing. This method is simple to operate, but it is a point measurement with limited representativeness and cannot reflect the overall water stress status of the crop.

[0027] In related technologies, drones equipped with multispectral cameras are used to acquire farmland images, calculate vegetation indices, and construct crop growth monitoring models to assess the growth status of crops over large areas. This method has the advantages of wide coverage and high efficiency, but the data update cycle is long, and it lacks direct perception of the crop root zone environment and real-time physiological processes.

[0028] The aforementioned technologies have promoted the intelligentization of agricultural production to some extent, but they still have significant shortcomings, mainly as follows: The assessment indicators are too simplistic and the information dimensions are insufficient: existing methods rely on single physical or chemical parameters (such as soil moisture content and chlorophyll value) and fail to comprehensively consider multi-dimensional factors such as soil supply, plant physiological response and meteorological environment, resulting in one-sided assessment results and easy misjudgment.

[0029] Delayed signal response and untimely early warning: Changes in soil moisture content or the appearance of visible leaf symptoms (such as wilting and yellowing) usually occur some time after crops have been subjected to stress. However, this invention, by monitoring early physiological indicators such as stomatal conductance and canopy temperature, can capture weak signals at the initial stage of stress, significantly advancing the early warning and intervention window.

[0030] Static and rigid models have poor adaptability: Traditional evaluation models typically have fixed parameters and weights, making them unable to adapt to the dynamic changes in water and fertilizer requirements at different crop growth stages (such as seedling, heading, and grain-filling stages). This invention employs an adaptive weighting algorithm, which can dynamically adjust according to the crop's growth stage and real-time environment, making the evaluation model more universal and accurate.

[0031] Insufficient spatial precision hinders homogenized management: Single sensor or remote sensing image analysis struggles to accurately depict the micro-scale differences in demand within a field caused by factors such as soil variations and uneven planting density. This invention combines spatial interpolation algorithms to generate a weighted distribution matrix, supporting variable irrigation and fertilization, moving from "homogeneous management" to "spatiotemporally differentiated decision-making."

[0032] Currently, agricultural production faces the dual challenges of increasingly tight resource constraints and growing environmental pressure, making traditional extensive water and fertilizer management models unsustainable. To achieve precise resource input and efficient utilization, an intelligent assessment method capable of real-time, comprehensive, and accurate perception of crop needs is urgently needed. This invention aims to propose a method for generating a dynamic crop demand index based on multi-sensor data fusion. By integrating multi-source heterogeneous data from the soil-plant-atmosphere continuum, a comprehensive assessment system with a physiological basis is constructed. This method utilizes an adaptive fusion algorithm to generate a "dynamic demand index" that dynamically reflects the intensity of crop water and fertilizer requirements, providing a scientific and quantitative basis for intelligent irrigation and precision fertilization, thereby effectively avoiding resource waste and ensuring healthy crop growth.

[0033] Figure 1 This is a flowchart illustrating the crop demand dynamic index generation method based on multi-sensor data fusion provided by the present invention, as shown below. Figure 1 As shown, the method includes the following steps.

[0034] Step 101: Obtain multi-source heterogeneous data of the target crop from soil-plant-atmosphere continuum detection.

[0035] In this embodiment of the invention, multi-source heterogeneous data from the soil-plant-atmosphere continuum are acquired in real time through a sensor network pre-deployed in the growth environment of the target crop.

[0036] Sensor networks include, but are not limited to: soil moisture sensors, soil water potential sensors, stem flow sensors, canopy infrared thermal imaging sensors, stomatal conductance monitors, atmospheric CO2 concentration sensors, and photosynthetically active radiation (PAR) sensors.

[0037] The aforementioned sensors continuously collect data to form a multi-source heterogeneous dataset. Among them, the soil moisture sensor and soil water potential sensor are used to monitor the volumetric water content and matrix potential in the soil to reflect the available water range of the soil; the stem flow sensor is used to measure the rate of water transport within the crop, directly indicating the actual transpiration demand of the plant; the canopy infrared thermal imaging sensor and stomatal conductance monitor are used to capture changes in crop canopy temperature and stomatal conductance to identify early physiological signals of water stress; and the atmospheric CO2 concentration sensor and PAR sensor are used to monitor carbon assimilation-related parameters in the environment.

[0038] All sensor data is uploaded in real time to a pre-associated data processor, ensuring data integrity and timeliness. The multi-source heterogeneous data covers multiple dimensions such as soil supply, plant physiological responses, and meteorological environment, providing a foundation for subsequent fusion computing.

[0039] Step 102: Compare the multi-source heterogeneous data with the normal trend relationship preset based on historical data to determine the deviation term of at least one type of data in the multi-source heterogeneous data.

[0040] In this embodiment of the invention, the normal trend relationship is established by analyzing historical data. For example, under normal conditions, soil water potential and stem flow rate should maintain a certain positive correlation, while canopy temperature and stomatal conductance should show a negative correlation. These relationship models are pre-stored, and deviation thresholds (such as standard deviation or percentage change) are set.

[0041] During the comparison process, the data processor calculates in real time the degree of deviation between the current value of each data type and the normal trend.

[0042] For example, when a soil moisture sensor indicates sufficient soil moisture, but a stem flow sensor detects a continuous decrease in stem flow rate, and canopy infrared thermal imaging shows an abnormally high canopy temperature, a bias is identified in the stem flow and canopy temperature data. The identification of this bias is based on multivariate co-analysis, ensuring that even if a single data point is normal, potential stress can be detected through abnormal patterns. Once the deviation exceeds a threshold, this type of data is marked as a bias, providing a basis for subsequent weight adjustments.

[0043] Step 103: Based on the bias term, adjust the weights of each data type in the multi-source heterogeneous data through a Bayesian probability update model to obtain an adaptive weight set.

[0044] In this embodiment of the invention, the Bayesian probability update model uses prior weights (such as initial uniform weights or weights set based on expert knowledge) as a basis and combines the current bias term as evidence to calculate the posterior probability.

[0045] Specifically, the Bayesian probability update model treats the bias term as a conditional event and updates the weight probability distribution of each data type.

[0046] For example, when deviations are detected in stem flow rate and canopy temperature while soil data are normal, the model will reduce the weights of soil moisture and soil water potential data, while increasing the weights of stem flow, canopy temperature, and meteorological data (such as CO2 concentration and PAR). The weight update formula is based on Bayes' theorem, adjusting the weight values ​​in real time to ensure that the weight set adaptively reflects the importance of the current data. The adaptive weight set is a vector, where each element corresponds to a weight for a data type, and the sum of the weights is 1.

[0047] Step 104: Generate a time weight sequence based on the current growth period information of the target crop, and generate a spatial weight matrix based on the spatial environment information of the field where the target crop is located; adjust the adaptive weight set using the time weight sequence and the spatial weight matrix to obtain the spatiotemporal weight set.

[0048] In this embodiment of the invention, the generation of the time-weighted sequence depends on the crop growth period model: the current growth period (such as seedling stage, heading stage, and grain-filling stage) is determined according to the crop type (such as wheat, corn) and growth calendar, and the response sensitivity of each data type at different growth stages is fitted by the logistic growth function.

[0049] For example, during the seedling stage, nitrogen demand is sensitive, so stomatal conductance and chlorophyll-related data have higher weights; during critical water periods (such as the jointing stage), stem flow rate and canopy temperature have increased weights. The time-weighted sequence is a weight vector that changes over time.

[0050] The spatial weight matrix is ​​generated based on the spatial heterogeneity of the field: by integrating planting density maps, vegetation cover data, and the spatial distribution map of the Normalized Difference Vegetation Index (NDVI), the weight of each spatial unit (e.g., a grid) is calculated using the Kriging interpolation algorithm to form a spatial weight matrix. The adaptive weight set is then adjusted using the temporal weight sequence and the spatial weight matrix to obtain a spatiotemporal weight set. This spatiotemporal weight set considers both temporal dynamics and spatial variability.

[0051] Step 105: Perform weighted fusion calculation on multi-source heterogeneous data based on spatiotemporal weight set to generate the dynamic demand index of the target crop.

[0052] In the embodiments of the invention, the weighted fusion calculation adopts a nonlinear weighted summation model. For example, firstly, the data of each data type is standardized to eliminate the difference in units; then, the standardized data is multiplied by the corresponding spatiotemporal weights; finally, all weighted data are summed and an activation function (such as the Sigmoid function) is applied to map to the range of 0-100 to generate a demand dynamic index.

[0053] The demand dynamic index is a scalar value; the higher the value, the more urgent the crop's need for water and fertilizer. For example, an index exceeding 70 can trigger an early warning mechanism. The entire calculation process is completed in real time within the data processor, ensuring that the demand dynamic index dynamically reflects the crop's actual needs. The demand dynamic index can be directly used to guide intelligent irrigation and precision fertilization decisions, improving resource utilization efficiency.

[0054] This invention first acquires multi-source heterogeneous data from the soil-plant-atmosphere continuum, providing a foundation for comprehensively understanding crop growth status. Next, the multi-source heterogeneous data is compared with a pre-defined normal trend relationship based on historical data to identify deviations and thus recognize anomalies in the current data. Then, based on these deviations, a Bayesian probability update model is used to adjust the weights of various data types within the multi-source heterogeneous data, resulting in an adaptive weight set that adapts to data changes. Next, a time-weighted sequence is generated based on the current growth stage information of the target crop, and a spatial weight matrix is ​​generated by combining this with field spatial environment information. These two components are then used to adjust the adaptive weight set, resulting in a spatiotemporal weight set that integrates spatiotemporal features. Finally, the spatiotemporal weight set is used to perform weighted fusion calculations on the multi-source heterogeneous data, generating a dynamic demand index that accurately reflects the actual needs of the crop. The entire process achieves the adaptive, spatiotemporally optimized, and precise generation of a crop demand dynamic index from multi-source data.

[0055] According to the present invention, a method for generating a crop demand dynamic index based on multi-sensor data fusion acquires multi-source heterogeneous data of a target crop obtained from soil-plant-atmosphere continuum detection, including: Obtain soil monitoring data from the soil layer, which includes at least one of the following: soil moisture data measured by a soil moisture sensor and soil water potential data measured by a soil water potential sensor. Acquire plant monitoring data from the plant level, which includes at least one of the following: stem flow rate data measured by a stem flow sensor, canopy temperature data measured by a canopy infrared thermal imaging sensor, and stomatal conductance data measured by a stomatal conductance monitor. Acquire atmospheric monitoring data from the atmospheric level, which includes at least one of the following: carbon dioxide concentration data measured by a carbon dioxide concentration sensor and photosynthetically active radiation data measured by a photosynthetically active radiation sensor.

[0056] In this embodiment of the invention, the limitation of relying solely on soil or leaf apparent parameters is overcome, and a higher-order index reflecting the real-time physiological state of crops is introduced.

[0057] refer to Figure 2 , Figure 2This is a schematic diagram of the detection of some sensors provided by the present invention, including: a photosynthetically active radiation sensor, a carbon dioxide concentration sensor, a stomatal conductance meter, an infrared thermal imaging sensor, and an edge computer (used to construct a "water-carbon-nitrogen" coupling model).

[0058] Based on the plant's transpiration regulation mechanism, in the early stages of water stress, crops close their stomata to reduce water loss, a process that directly leads to an increase in canopy temperature. By acquiring high-frequency canopy infrared thermal imaging data and stomatal conductance values, early signs of stress can be effectively captured before visible wilting symptoms appear on the leaves. Simultaneously, by combining atmospheric CO2 concentration and photosynthetically active radiation (PAR) data, a "water-carbon-nitrogen" coupled model is established, linking the crop's carbon assimilation capacity with its water and fertilizer requirements, making the assessment more physiologically meaningful.

[0059] refer to Figure 3 , Figure 3 This is a schematic diagram of another part of the sensors provided by the present invention, which includes: stem flow sensor, water potential sensor, moisture sensor and edge computer (used to determine the effective water range, the rate of water transport in the crop and the transpiration water requirement of the plant).

[0060] This invention recognizes that soil moisture content is not equivalent to the available water for crops. By combining soil moisture sensors and soil water potential sensors, based on the principles of plant water absorption kinetics, the range of "available water" in the soil that can be absorbed by crop roots can be accurately estimated. More importantly, the system integrates a stem flow sensor for continuous monitoring of the rate of water transport within the plant, which directly reflects the plant's actual transpiration requirements. This closed-loop sensing model of "soil supply capacity" and "actual plant needs" ensures the authenticity and reliability of the assessment results, effectively overcoming the limitation of a single soil sensor in distinguishing between "bound water" and "available water."

[0061] In some embodiments, an integrated sensor network deployed in the target crop field can simultaneously acquire monitoring data from three levels: soil, plants, and atmosphere, forming a multi-source heterogeneous dataset.

[0062] Soil environment is continuously monitored by deploying soil moisture sensors and soil water potential sensors at different depths in the root zone of the target crop. Soil moisture sensors measure the volumetric water content of the soil, expressed as a percentage or per unit volume. Soil water potential sensors measure the matric potential of the soil, expressed in kilopascals (kPa) or megapascals (MPa), reflecting the energy state of soil moisture and the ease with which crop roots can absorb water. By jointly analyzing soil moisture and soil water potential data, the system can accurately assess the range of "available water" in the soil that can be directly utilized by crops, overcoming the limitation of relying solely on moisture content to distinguish between bound water and available water. Data is uploaded to the data processor at set time intervals (e.g., every 15 minutes) via wired or wireless data acquisition and transmission units.

[0063] Physiological data of crops are directly acquired through stem flow sensors mounted on the stems of target crops, canopy infrared thermal imaging sensors facing the crop canopy, and contact or non-contact stomatal conductance monitors. Stem flow sensors continuously monitor and record the sap flow rate in the crop stems, directly reflecting the plant's actual transpiration intensity and water requirements. Canopy infrared thermal imaging sensors capture infrared radiation information from the crop canopy and convert it into canopy temperature data. Stomatal conductance monitors determine the degree of stomatal opening in leaves, obtaining stomatal conductance data, a key physiological indicator reflecting the early response of crops to water stress.

[0064] Carbon dioxide concentration data near the canopy of the target crop, measured by a carbon dioxide concentration sensor, and photosynthetically active radiation (PAR) data, measured by a photosynthetically active radiation (PAR) sensor. PAR data characterizes the intensity of light energy that can be used by plant leaves for photosynthesis. These atmospheric monitoring data are closely coupled with plant physiological processes to construct a water-carbon correlation model, helping to determine whether the crop's carbon assimilation capacity is affected by water or nutrient stress.

[0065] Through the embodiments of the present invention, by integrating multi-source heterogeneous sensor data from soil, plants and atmosphere, the environment and physiological state of crops can be comprehensively reflected, enabling accurate and dynamic assessment of crop water and nutrient requirements.

[0066] According to the present invention, a method for generating a crop demand dynamic index based on multi-sensor data fusion compares multi-source heterogeneous data with a pre-defined normal trend relationship based on historical data to determine at least one type of deviation term in the multi-source heterogeneous data, including: Based on historical data, a normal trend relationship model is established among multiple target parameters, and a deviation threshold is set for each target parameter. The multiple target parameters include at least: soil water potential data, stem flow rate data, and canopy temperature data. By comparing multiple target parameters in real-time acquired multi-source heterogeneous data with a normal trend relationship model, target parameters with deviations greater than the deviation threshold are identified as deviation items.

[0067] In some embodiments, extensive historical monitoring data (i.e., historical data) of the target crop under normal growth conditions (i.e., without significant water or nutrient stress) are collected. Based on this historical data, a model of the normal variation trend relationship between the target parameters is established through correlation analysis and regression modeling. The target parameters include, but are not limited to, soil water potential data, stem flow rate data, and canopy temperature data.

[0068] For example, a normal trend model would characterize the generally positive correlation between soil water potential and stem flow rate under normal conditions (i.e., higher soil water potential indicates stronger soil water supply capacity, and stem flow rate is usually higher), and the generally negative correlation between stem flow rate and canopy temperature (i.e., higher stem flow rate indicates stronger transpiration cooling effect, and canopy temperature is usually lower). A normal trend model can be expressed as a mathematical formula (such as a linear or nonlinear regression equation) or a set of rules describing the range of normal co-variables.

[0069] Simultaneously, for each target parameter, its standard deviation or percentile is calculated based on its historical data fluctuations under normal trends, and a deviation threshold is set accordingly. The deviation threshold is used to quantitatively determine whether real-time data significantly deviates from the normal trend. For example, the deviation threshold can be set to ±2 standard deviations of the historical normal value, or it can be set as the allowable percentage deviation relative to the predicted value of the normal trend line.

[0070] During the real-time operation of the sensor, the data processor continuously receives multi-source heterogeneous data acquired in real time. The current values ​​of the aforementioned target parameters (i.e., soil water potential data, stem flow rate data, canopy temperature data, etc.) are extracted from the received multi-source heterogeneous data.

[0071] Next, the current value of the target parameter is substituted into a pre-established normal trend relationship model for comparison. The comparison process includes: predicting the normal range of other related target parameters (such as stem flow rate data) from the real-time value of a certain target parameter (such as soil water potential data); then, calculating the degree of deviation between the actual measured value of the related target parameter and the predicted normal range.

[0072] If the deviation between the real-time value and the model's predicted value of one or more target parameters exceeds a predefined deviation threshold, the system identifies that target parameter as a deviation item. For example, in real-time monitoring, if soil water potential data shows sufficient soil water supply (its value is within the normal range), but the measured stem flow rate data is significantly lower than the lower limit predicted by the model based on the current soil water potential data, and the canopy temperature data is significantly higher than the upper limit predicted by the model, then the system will identify the stem flow rate data and the canopy temperature data as current deviation items. This co-deviation pattern often indicates possible crop physiological stress caused by factors such as atmospheric conditions.

[0073] Through the embodiments of the present invention, by establishing a multi-parameter normal change trend relationship model and setting a deviation threshold, abnormal deviation items in real-time monitoring data can be automatically identified, effectively distinguishing between the actual needs of crops and environmental disturbances, and improving the accuracy and robustness of crop demand judgment.

[0074] According to the present invention, a method for generating a crop demand dynamic index based on multi-sensor data fusion is provided, the method further includes: Based on soil moisture data and soil water potential data, the range of available soil water was determined; Based on soil available water range and stem flow rate data, the actual transpiration requirements of the target crop are determined.

[0075] In this embodiment of the invention, the effective water range of the soil is determined by combining soil moisture data (volume water content) and soil water potential data (matrix potential).

[0076] In this embodiment of the invention, soil water potential data is used as a key criterion. According to the principles of plant physiology, when the soil water potential is higher than the critical water absorption threshold of crop roots (usually about -0.03 MPa, corresponding to field capacity), although water can be absorbed by the crop, it is not the most effective range; when the soil water potential is lower than the permanent wilting point of the crop (usually about -1.5 MPa), the soil water is strongly adsorbed by soil particles and is difficult for the crop to utilize, which is considered ineffective water.

[0077] After obtaining the effective water range of the soil, the actual transpiration requirements of the target crop are comprehensively judged by combining stem flow rate data.

[0078] First, assess the current soil moisture status relative to the available soil water range. For example, if the current volumetric water content is close to or lower than the water content corresponding to the permanent wilting coefficient, it indicates a severe shortage of available soil water, and the crop may face drought stress.

[0079] However, soil moisture supply is only one aspect of demand; the actual transpiration demand of crops is more directly reflected by the plant's own physiological activities (i.e., the intensity of transpiration). Stem flow rate data directly characterizes the rate of water transport within the crop and is the most direct quantitative indicator of the crop's actual transpiration demand.

[0080] By analyzing the numerical level and diurnal variation of stem flow rate data (e.g., whether the stem flow rate can reach its potential peak under normal light conditions), and combining this with the current sufficiency of available soil water, a comprehensive judgment is made: If the available soil water range indicates sufficient moisture, but the stem flow rate data remains below its normal physiological expectation, it may indicate that the crop's transpiration is inhibited by non-soil factors (such as low atmospheric humidity, abnormally high temperature, etc.). In this case, the actual transpiration demand of the crop may not be met, and there is potential stress.

[0081] If the available soil water range is already insufficient, and the stem flow rate data also show suppressed characteristics, it strongly indicates that the actual transpiration demand of the crop cannot be met due to insufficient soil water supply, and irrigation intervention is urgently needed.

[0082] By performing a closed-loop comparative analysis of "soil supply capacity" (soil effective water range) and "actual plant consumption" (stem flow rate data) through the embodiments of the present invention, the actual transpiration demand of the target crop can be determined more realistically and reliably.

[0083] According to the present invention, a method for generating a dynamic crop demand index based on multi-sensor data fusion adjusts the weights of various data types in multi-source heterogeneous data based on a bias term using a Bayesian probability update model to obtain an adaptive weight set, including: Based on the deviation term, identify the abnormal stress sources that cause data deviation and determine the degree of correlation between the abnormal stress sources and various data types in multi-source heterogeneous data; In response to the identification of abnormal stress sources, based on the Bayesian probability update model, the weights of data types with a correlation degree less than the correlation threshold with the abnormal stress source are reduced, and the weights of data types with a correlation degree greater than the correlation threshold with the abnormal stress source are increased, resulting in a redistributed weight set; The redistributed weight set is used as the adaptive weight set.

[0084] refer to Figure 4 , Figure 4 This is a flowchart illustrating the process of identifying abnormal stress sources provided by the present invention, which includes: stem flow data, soil water potential data, canopy temperature data, ambient temperature data, adaptive adjustment of fusion weights, weight-based data fusion, and identification of abnormal stress sources.

[0085] In this embodiment of the invention, a normal trend relationship between soil water potential, stem flow rate, and canopy temperature was established based on historical data, and a deviation threshold was set. When the trend of real-time monitoring data deviates from the normal pattern (e.g., sufficient soil moisture but decreased stem flow rate and increased canopy temperature), it is determined that physiological inhibition may be caused by non-soil factors (such as high temperature and low humidity). At this time, a Bayesian probabilistic update model is activated (automatically reducing the weight of soil data while increasing the weight of meteorological data and canopy temperature) to identify abnormal stress sources and redistribute data weights, greatly improving the robustness and physiological interpretability of the assessment results.

[0086] In this embodiment of the invention, based on the determined deviation items, anomaly pattern recognition is performed to determine the type of abnormal stress source causing the data deviation. Common abnormal stress sources include, but are not limited to: atmospheric stress (such as high temperature, low humidity, strong radiation, etc.), soil stress (such as water stress, salinity stress, etc.), or biological stress. The recognition process is based on a preset abnormal pattern library. For example, when the deviation item is mainly manifested as an abnormally high canopy temperature data while the soil water potential data is normal, the abnormal stress source is determined to be atmospheric stress (such as high temperature and dryness); when the deviation item is manifested as an abnormally low soil water potential data, the abnormal stress source is determined to be soil water stress.

[0087] After identifying the anomalous stress source, the correlation between this source and various data types in the multi-source heterogeneous data is determined based on a pre-defined knowledge base. The correlation level quantifies the sensitivity or reliability of different data types in indicating the specific stress source. For example, for atmospheric stress (high temperature and dryness), canopy temperature data, stomatal conductance data, and stem flow rate data show a high correlation; while for soil moisture stress, soil water potential data, soil moisture data, and stem flow rate data show a high correlation. The correlation level between each data type and the specific anomalous stress source is quantified into a numerical value (e.g., between 0 and 1), and a pre-defined correlation threshold is used for subsequent judgment.

[0088] In response to the identification of a specific anomalous stress source, a Bayesian probability update model is initiated to dynamically reallocate weights. This model uses the current weights of each data type as prior probabilities and "identifying a specific anomalous stress source" as new evidence to calculate the posterior probabilities of the weights of each data type.

[0089] For example, reduce the weight of data types whose correlation with identified abnormal stress sources is less than a preset correlation threshold; at the same time, increase the weight of data types whose correlation with identified abnormal stress sources is greater than a preset correlation threshold.

[0090] Specifically, for a given data type, the update of its posterior weights follows the basic idea of ​​Bayes' theorem: the posterior probability is proportional to the product of the prior probability and the likelihood. In this context, the likelihood reflects the degree of matching between the observations of this data type and the currently identified anomalous stress source patterns, which is determined by the aforementioned degree of association. The higher the degree of association, the larger the likelihood value, and the greater the boost this data type receives in the weight update.

[0091] This probability update process redistributes the weights of all data types, resulting in a redistributed weight set. This set reflects the reordering of the importance of each data type in assessing crop demand under the current specific stress scenario.

[0092] The redistributed weight set is normalized (ensuring the sum of all weights is 1), and its output is used as the adaptive weight set. This adaptive weight set is a weight vector optimized for the currently detected anomalies. It will be used in subsequent weighted fusion calculations, making the generated "demand dynamic index" more sensitive to the actual crop stress state and significantly improving the robustness and situational adaptability of the evaluation model.

[0093] According to the present invention, a method for generating a dynamic crop demand index based on multi-sensor data fusion generates a time-weighted sequence based on the current growth stage information of the target crop and a spatial weight matrix based on the spatial environment information of the field where the target crop is located. The method then adjusts the adaptive weight set using the time-weighted sequence and the spatial weight matrix to obtain a spatiotemporal weight set, including: Based on the current growth stage information of the target crop, and based on the pre-defined sensitivity differences of each growth stage of the crop to various data types in multi-source heterogeneous data, a time-weighted sequence is generated by fitting the logistic growth function. Based on the spatial environment information of the target crop field, a spatial weight matrix is ​​generated using a spatial interpolation algorithm. The spatial environment information includes: crop planting density, vegetation coverage, and spatial distribution of vegetation index. The spatiotemporal weight set is obtained by dynamically adjusting the adaptive weight set based on the time weight sequence and the spatial weight matrix.

[0094] In this embodiment of the invention, the dynamics and spatial heterogeneity of crop growth are fully considered. In the temporal dimension, based on the physiological characteristics of different crop growth stages (e.g., seedling stage sensitivity to nitrogen, heading stage critical sensitivity to water), the response sensitivity of each factor is fitted using a logistic growth function to dynamically generate a time-weighted sequence. In the spatial dimension, combining crop planting density maps, vegetation cover, and NDVI spatial distribution maps, algorithms such as Kriging interpolation are used to generate a weighted spatial distribution matrix. By constructing a spatiotemporal dynamic weight adjustment model, the evaluation system can achieve refined management "on demand, on time, and on location," significantly improving the accuracy of variable operations.

[0095] In this embodiment of the invention, a time-weighted sequence is generated based on the current growth period information of the target crop.

[0096] For each data type in the multi-source heterogeneous data, a baseline sensitivity weight was pre-defined based on agronomic knowledge for each growth stage of the crop. For example, during the seedling stage, the crop is sensitive to nitrogen demand, so the baseline weight of stomatal conductance data (related to nitrogen metabolism) is relatively high; during critical water periods (such as the jointing stage), the crop has an urgent need for water, so the baseline weights of stem flow rate data and canopy temperature data are relatively high.

[0097] To smoothly handle the transition between different growth stages, a logistic growth function is used to dynamically fit these baseline sensitivity weights. This function can simulate the gradual changes in crop sensitivity to various factors during growth. Using time as the independent variable and the baseline sensitivity weights of each data type as the dependent variable, an independent logistic growth function model is constructed for each data type. By inputting the current growth stage information (such as the growth stage number or number of growing days), the function outputs the weight value of that data type at the current moment. Arranging the weight values ​​of all data types in sequence generates a time-weight sequence. This time-weight sequence ensures that weight adjustments conform to the actual physiological development patterns of the crop.

[0098] In this embodiment of the invention, a spatial weight matrix is ​​generated based on the spatial environment information of the field where the target crop is located.

[0099] Obtain multi-source spatial environmental information in the field, including at least: crop planting density distribution maps, vegetation cover distribution maps obtained through remote sensing or field measurements, and spatial distribution maps of vegetation indices (such as NDVI) calculated from multispectral images.

[0100] Spatial interpolation algorithms (preferably Kriging interpolation) are used to process this spatial information. This algorithm predicts values ​​at unsampled locations based on the spatial correlation of known points and generates a continuous spatial distribution surface. First, spatial interpolation is performed on each type of spatial environmental information (e.g., crop planting density, vegetation cover, NDVI) to obtain their respective whole-field rasterized distribution maps. Then, according to preset rules (e.g., areas with high crop planting density, large vegetation cover, or low NDVI values ​​typically require more refined management and should be assigned higher spatial weights), the values ​​of each raster layer are converted into spatial weight coefficients. Finally, through layer overlay calculations, a spatial weight matrix that comprehensively reflects the spatial heterogeneity of the field is generated. Each element in this spatial weight matrix corresponds to the overall spatial weight adjustment coefficient for a specific location in the field (e.g., a raster cell).

[0101] In this embodiment of the invention, the obtained adaptive weight set is dynamically adjusted in time and space using the generated time weight sequence and spatial weight matrix, and finally outputs the spatiotemporal weight set.

[0102] The adaptive weight set is element-wise multiplied with the time weight sequence (Hadamard product) to obtain a time-corrected weight vector. This time-corrected weight vector is then integrated with the spatial weight matrix. For each spatial location in the field, the final spatiotemporal weight set equals the time-corrected weight vector multiplied by the corresponding weight coefficient in the spatial weight matrix for that location.

[0103] Through the above adjustments, the spatiotemporal weight set not only considers the adaptive weight adjustment due to real-time data anomalies, but also integrates the temporal patterns of crop growth and the spatial differences in field management, thus achieving a refined dynamic allocation of weights in both time and space dimensions.

[0104] This invention mainly consists of multiple sensors deployed in the field (including soil moisture sensors, soil water potential sensors, stem flow sensors, canopy infrared thermal imaging sensors, stomatal conductance monitors, and atmospheric CO2 concentration and photosynthetically active radiation (PAR) sensors), a data acquisition and transmission unit, a data processor, and a cloud analysis platform. The system continuously collects data from each sensor. First, through precise sensing at the soil-plant interface, it simultaneously acquires soil available water information (volume water content + matrix potential) and the plant's actual transpiration demand (stem flow rate). Second, through extended sensing, it captures early physiological signals of crop water stress (canopy temperature, stomatal conductance) and correlates them with carbon metabolism processes (CO2, PAR). Finally, the data processor runs an adaptive weighted nonlinear dynamic fusion algorithm, combined with a spatiotemporal dynamic weight adjustment model, to fuse multidimensional data into a unified "crop demand dynamic index." The higher the index value, the more urgent the crop's current demand for water and fertilizer.

[0105] For example, when this system is applied to a wheat field, it continuously collects various data. During the spring greening stage, the system focuses on changes in soil moisture availability and stomatal conductance; at this time, the "demand dynamic index" primarily reflects water demand. Entering the jointing and booting stage (the critical water period), the system automatically increases the weight of stem flow rate and canopy temperature. Although soil moisture sensors show good moisture levels, the system detects a continuous decline in stem flow rate, while canopy infrared thermal imaging shows abnormally high temperatures in some areas. The fusion algorithm identifies this abnormal pattern, automatically increases the weight of meteorological and thermal imaging data, calculates a higher "demand dynamic index," and issues a "potential water stress" warning to the manager. This aligns with the assessment that the weather was sunny, windy, and dry, leading to increased transpiration. Managers can then use this information for precise watering to avoid yield loss.

[0106] This invention proposes a method for generating dynamic crop demand indices based on multi-sensor data fusion. By constructing an integrated "soil-plant-atmosphere" sensing network and intelligent fusion algorithm, it achieves early, accurate, and quantitative assessment of crop water and fertilizer requirements. Compared with existing technologies, this invention can not only detect stress risks in advance but also provide personalized management suggestions based on crop growth stages and differences in field microenvironments. The method features an advanced model, clear logic, and strong practicality, possessing significant application value and broad prospects for promoting smart agriculture and achieving green and sustainable agricultural development.

[0107] The crop demand dynamic index generation device based on multi-sensor data fusion provided by the present invention will be described below. The crop demand dynamic index generation device based on multi-sensor data fusion described below can be referred to in correspondence with the crop demand dynamic index generation method based on multi-sensor data fusion described above.

[0108] refer to Figure 5 , Figure 5 This is a schematic diagram of the module of the crop demand dynamic index generation device based on multi-sensor data fusion provided by the present invention.

[0109] The acquisition module 501 is used to acquire multi-source heterogeneous data of the target crop obtained from soil-plant-atmosphere continuum detection; The comparison module 502 is used to compare multi-source heterogeneous data with a normal trend relationship preset based on historical data, and to determine the deviation item of at least one type of data in the multi-source heterogeneous data. The adaptive weight module 503 is used to adjust the weights of each data type in multi-source heterogeneous data based on the bias term through a Bayesian probability update model to obtain an adaptive weight set. The spatiotemporal weight module 504 is used to generate a time weight sequence based on the current growth stage information of the target crop and to generate a spatial weight matrix based on the spatial environment information of the field where the target crop is located; the time weight sequence and the spatial weight matrix are used to adjust the adaptive weight set to obtain the spatiotemporal weight set. The weighted fusion module 505 is used to perform weighted fusion calculations on multi-source heterogeneous data based on spatiotemporal weight sets to generate a dynamic demand index for the target crop.

[0110] Specifically, the crop demand dynamic index generation device based on multi-sensor data fusion provided by the present invention can realize all the method steps implemented in the above-mentioned crop demand dynamic index generation method embodiment based on multi-sensor data fusion, and can achieve the same technical effect. Here, the parts that are the same as those in the method embodiment and the beneficial effects will not be described in detail.

[0111] Figure 6 This is a schematic diagram of the physical structure of the electronic device provided by the present invention, such as... Figure 6 As shown, the electronic device may include: a processor 610, a communications interface 620, a memory 630, and a communications bus 640, wherein the processor 610, the communications interface 620, and the memory 630 communicate with each other through the communications bus 640. The processor 610 can call logic instructions in the memory 630 to execute a method for generating a dynamic crop demand index based on multi-sensor data fusion. This method includes: acquiring multi-source heterogeneous data of the target crop from soil-plant-atmosphere continuum detection; comparing the multi-source heterogeneous data with a pre-defined normal trend relationship based on historical data to determine a deviation term for at least one type of data in the multi-source heterogeneous data; adjusting the weights of each data type in the multi-source heterogeneous data based on the deviation term using a Bayesian probability update model to obtain an adaptive weight set; generating a time weight sequence based on the current growth stage information of the target crop and generating a spatial weight matrix based on the spatial environment information of the field where the target crop is located; adjusting the adaptive weight set using the time weight sequence and spatial weight matrix to obtain a spatiotemporal weight set; and performing weighted fusion calculation on the multi-source heterogeneous data based on the spatiotemporal weight set to generate a dynamic demand index for the target crop.

[0112] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0113] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the crop demand dynamic index generation method based on multi-sensor data fusion provided by the above methods. The method includes: acquiring multi-source heterogeneous data of the target crop from soil-plant-atmosphere continuum detection; comparing the multi-source heterogeneous data with a pre-set normal trend relationship based on historical data to determine at least one type of deviation term in the multi-source heterogeneous data; adjusting the weights of each data type in the multi-source heterogeneous data based on the deviation term using a Bayesian probability update model to obtain an adaptive weight set; generating a time weight sequence based on the current growth stage information of the target crop and generating a spatial weight matrix based on the spatial environment information of the field where the target crop is located; adjusting the adaptive weight set using the time weight sequence and the spatial weight matrix to obtain a spatiotemporal weight set; and performing weighted fusion calculation on the multi-source heterogeneous data based on the spatiotemporal weight set to generate a demand dynamic index of the target crop.

[0114] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the crop demand dynamic index generation method based on multi-sensor data fusion provided by the above methods. The method includes: acquiring multi-source heterogeneous data of a target crop obtained from soil-plant-atmosphere continuum detection; comparing the multi-source heterogeneous data with a pre-defined normal trend relationship based on historical data to determine a deviation term for at least one type of data in the multi-source heterogeneous data; adjusting the weights of each data type in the multi-source heterogeneous data based on the deviation term using a Bayesian probability update model to obtain an adaptive weight set; generating a time weight sequence based on the current growth stage information of the target crop and generating a spatial weight matrix based on the spatial environment information of the field where the target crop is located; adjusting the adaptive weight set using the time weight sequence and the spatial weight matrix to obtain a spatiotemporal weight set; and performing weighted fusion calculation on the multi-source heterogeneous data based on the spatiotemporal weight set to generate a demand dynamic index for the target crop.

[0115] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0116] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for generating a dynamic crop demand index based on multi-sensor data fusion, characterized in that, include: Acquire multi-source heterogeneous data of the target crop from soil-plant-atmosphere continuum detection; The multi-source heterogeneous data is compared with the normal trend relationship preset based on historical data to determine the deviation item of at least one type of data in the multi-source heterogeneous data. Based on the aforementioned bias term, the weights of each data type in the multi-source heterogeneous data are adjusted using a Bayesian probability update model to obtain an adaptive weight set; A time-weighted sequence is generated based on the current growth stage information of the target crop, and a spatial weight matrix is ​​generated based on the spatial environment information of the field where the target crop is located; the adaptive weight set is adjusted using the time-weighted sequence and the spatial weight matrix to obtain a spatiotemporal weight set; The multi-source heterogeneous data is weighted and fused based on the spatiotemporal weight set to generate the demand dynamic index of the target crop.

2. The method for generating a dynamic crop demand index based on multi-sensor data fusion according to claim 1, characterized in that, The acquisition of multi-source heterogeneous data of the target crop obtained from soil-plant-atmosphere continuum detection includes: Acquire soil monitoring data from the soil layer, wherein the soil monitoring data includes at least one of the following: soil moisture data measured by a soil moisture sensor and soil water potential data measured by a soil water potential sensor; Acquire plant monitoring data from the plant level, which includes at least one of the following: stem flow rate data measured by a stem flow sensor, canopy temperature data measured by a canopy infrared thermal imaging sensor, and stomatal conductance data measured by a stomatal conductance monitor. Acquire atmospheric monitoring data from the atmospheric level, which includes at least one of the following: carbon dioxide concentration data measured by a carbon dioxide concentration sensor and photosynthetically active radiation data measured by a photosynthetically active radiation sensor.

3. The method for generating a dynamic crop demand index based on multi-sensor data fusion according to claim 2, characterized in that, The step of comparing the multi-source heterogeneous data with a pre-defined normal trend relationship based on historical data to determine at least one type of deviation term in the multi-source heterogeneous data includes: Based on historical data, a normal variation trend relationship model is established among multiple target parameters, and a deviation threshold is set for each target parameter. The multiple target parameters include at least: the soil water potential data, the stem flow rate data, and the canopy temperature data. Based on the comparison of multiple target parameters in the real-time collected multi-source heterogeneous data with the normal change trend relationship model, the target parameters with a deviation degree greater than the deviation threshold are determined as deviation items.

4. The method for generating a dynamic crop demand index based on multi-sensor data fusion according to claim 2, characterized in that, The method further includes: Based on the soil moisture data and the soil water potential data, the effective water range of the soil is determined; Based on the available soil water range and the stem flow rate data, the actual transpiration requirements of the target crop are determined.

5. The method for generating a dynamic crop demand index based on multi-sensor data fusion according to claim 1, characterized in that, The step of adjusting the weights of each data type in the multi-source heterogeneous data based on the bias term using a Bayesian probability update model to obtain an adaptive weight set includes: Based on the deviation term, identify the abnormal stress sources that cause data deviation and determine the degree of correlation between the abnormal stress sources and each data type in the multi-source heterogeneous data; In response to the identification of the abnormal stress source, based on the Bayesian probability update model, the weight of data types whose correlation with the abnormal stress source is less than the correlation threshold is reduced, and the weight of data types whose correlation with the abnormal stress source is greater than the correlation threshold is increased, to obtain a redistributed weight set; The redistributed weight set is used as the adaptive weight set.

6. The method for generating a dynamic crop demand index based on multi-sensor data fusion according to claim 1, characterized in that, The process involves generating a time-weighted sequence based on the current growth stage information of the target crop and generating a spatial weight matrix based on the spatial environment information of the field where the target crop is located. The adaptive weight set is then adjusted using the time-weighted sequence and the spatial weight matrix to obtain a spatiotemporal weight set, including: Based on the current growth stage information of the target crop, and based on the preset sensitivity differences of each growth stage of the crop to various data types in the multi-source heterogeneous data, a time-weighted sequence is generated by fitting the logistic growth function. Based on the spatial environment information of the field where the target crop is located, the spatial weight matrix is ​​generated using a spatial interpolation algorithm. The spatial environment information includes: crop planting density, vegetation coverage, and spatial distribution of vegetation index. Based on the time weight sequence and the spatial weight matrix, the adaptive weight set is dynamically adjusted in time and space to obtain the spatiotemporal weight set.

7. A crop demand dynamic index generation device based on multi-sensor data fusion, characterized in that, include: The acquisition module is used to acquire multi-source heterogeneous data of the target crop obtained from soil-plant-atmosphere continuum detection; The comparison module is used to compare the multi-source heterogeneous data with a preset normal change trend relationship based on historical data, and to determine the deviation item of at least one type of data in the multi-source heterogeneous data. An adaptive weighting module is used to adjust the weights of each data type in the multi-source heterogeneous data based on the bias term using a Bayesian probability update model, thereby obtaining an adaptive weighting set. The spatiotemporal weighting module is used to generate a time weighting sequence based on the current growth stage information of the target crop, and to generate a spatial weighting matrix based on the spatial environment information of the field where the target crop is located; the adaptive weighting set is adjusted using the time weighting sequence and the spatial weighting matrix to obtain the spatiotemporal weighting set. The weighted fusion module is used to perform weighted fusion calculations on the multi-source heterogeneous data based on the spatiotemporal weight set to generate the demand dynamic index of the target crop.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the crop demand dynamic index generation method based on multi-sensor data fusion as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the crop demand dynamic index generation method based on multi-sensor data fusion as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the crop demand dynamic index generation method based on multi-sensor data fusion as described in any one of claims 1 to 6.

Images