A rice low-temperature cold damage early warning method and system based on artificial intelligence

Abstract

The application discloses a rice low-temperature cold damage early warning method and system based on artificial intelligence, relates to the technical field of intelligent environment early warning, and comprises the following steps: collecting satellite data through a satellite and dividing a rice area into grids, synchronously deploying sensors to collect ground data, defining environment parameters, establishing a convergent gate vector field of a rice field through a three-part gate control mechanism, obtaining an initialized rice field state, using numerical integration to solve the rice field state and outputting a rice field stress index; constructing upper and lower correlation data based on the environment parameters, defining a rice field flow vector field and adding constraints, calculating the logarithmic density of the rice field state under the upper and lower correlation data, and then outputting a rice field abnormality degree, and calculating a rice field risk in combination with the rice field stress index. The application realizes real-time low-temperature risk early warning of a rice field area, guarantees early prediction of the risk of the rice field, and is helpful to optimization of a disaster prevention strategy for the rice field.

Description

Technical Field

[0001] This invention relates to the field of intelligent environmental early warning technology, and in particular to an artificial intelligence-based method and system for early warning of low-temperature chilling injury in rice. Background Technology

[0002] Over the past decade, increased climate variability has heightened the risk of low-temperature damage to rice during its critical growth stages (jointing and booting, heading and flowering). Traditional monitoring methods relying on manual field inspections and meteorological stations are insufficient to meet the demands for "high spatiotemporal resolution, minute-level updates, and field-level decision-making." With the development of high-resolution Earth observation (optical / thermal infrared) and ground-based IoT sensing, multi-source data-driven crop disaster early warning has become a hot topic. One type of method relies mainly on empirical thresholds and statistical regression, such as discrimination based on low-temperature thresholds, chilling accumulation, or indices. Another type emphasizes mechanistic modeling and data assimilation, using energy and water balance and crop physiological parameters to characterize chilling stress. A third type uses deep learning spatiotemporal models (such as spatiotemporal convolution and graph convolution) for representation learning. Although existing methods use deep models to predict the risk of chilling damage in paddy fields, they often lack stable dynamic priors and interpretable risk decomposition. Conventional probability outputs lack alignment with the actual time process, resulting in poor prediction performance. Summary of the Invention

[0003] In view of the aforementioned existing problems, the present invention is proposed.

[0004] Therefore, this invention provides an artificial intelligence-based method and system for early warning of low-temperature chilling injury in rice, which solves the problems of existing technologies lacking stable dynamic priors and interpretable risk decomposition, and conventional probability outputs lacking alignment with the real time process, resulting in poor prediction performance.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0006] In a first aspect, the present invention provides an artificial intelligence-based method for early warning of low-temperature chilling injury in rice, comprising,

[0007] By collecting satellite data and dividing the rice-growing area into a grid, simultaneously deploying sensors to collect ground data, defining environmental parameters, and establishing a convergent gated vector field for the rice field through a three-way gating mechanism, the initial state of the rice field is obtained, and the rice field stress index is output by solving the rice field state using numerical integration.

[0008] Based on environmental parameters, we construct the vertical correlation data, define the paddy field flow vector field and add constraints, calculate the log density of the paddy field state under the vertical correlation data, output the paddy field anomaly degree, and combine the paddy field stress index to calculate the paddy field risk.

[0009] A node graph is constructed based on the paddy field grid. The wind vectors of the edge nodes are measured and interpolated to obtain the wind vectors of the nodes. Connecting edges are constructed and their weights are calculated. A dynamic graph convolution algorithm is used to predict the future state of the paddy field and obtain the future risk of the paddy field. A dynamic risk map is formed based on the paddy field risk and the future risk to display.

[0010] As a preferred embodiment of the artificial intelligence-based rice low-temperature chilling injury early warning method of the present invention, the step of collecting satellite data by satellite and dividing the rice area into grids, and simultaneously deploying sensors to collect ground data refers to dividing the rice field area into grids of equal size, collecting satellite data of the rice field grid areas by satellite, including vegetation index NDVI and land surface temperature LST, and collecting rice field data by deploying sensors in the rice field area, including rice canopy temperature and air temperature difference, humidity, wind speed, net radiation and soil volumetric water content, and unifying the coordinate system and timestamp of the satellite data and rice field data.

[0011] As a preferred embodiment of the artificial intelligence-based early warning method for low-temperature chilling injury in rice described in this invention, the following steps are taken: defining environmental parameters, establishing a convergent gated vector field for the paddy field through a three-way gating mechanism, obtaining the initial paddy field state, solving for the paddy field state using numerical integration, and outputting the paddy field stress index; and integrating satellite data and paddy field data to form 7-dimensional environmental parameters. And standardize it;

[0012] Define the hour as the time unit benchmark and calculate the Fourier time characteristics. ;

[0013] The gating probability is calculated using a three-part gating mechanism based on Fourier time characteristics. ;

[0014] Define paddy field status And define the observation window to obtain standardized historical environmental parameters. The historical environmental parameters are processed and mapped to the initial paddy field state using a GRU encoder. ;

[0015] Define a fixed structure for each sub-vector field using a gating mechanism. ;

[0016] All sub-vector fields are combined into a convergent gated total vector field for the paddy field using a fixed structure.

[0017] The RK45 solver was used, and the training period, inference period and the number of maximum function evaluations were specified. The numerical integration method was used to solve the paddy field state of the convergent gated total vector field at the target time set.

[0018] The paddy field condition was mapped to a paddy field stress index using a fixed sigmoid calibration. .

[0019] As a preferred embodiment of the artificial intelligence-based early warning method for low-temperature chilling injury in rice described in this invention, the method involves: constructing vertically correlated data based on environmental parameters, defining constraints on the paddy field flow vector field, calculating the logarithmic density of the paddy field state under the vertically correlated data, outputting the paddy field anomaly degree, and calculating the paddy field risk index in conjunction with the paddy field stress index. The vertically correlated data of the paddy field state is constructed based on environmental parameters and Fourier time characteristics. By combining paddy field conditions and upstream and downstream correlation data, a two-layer MLP and residual gating structure are used to construct the paddy field flow vector field g;

[0020] Using the standard Gaussian distribution as the base distribution, the inverse flow integral is applied in the flow time. Calculate the log density of paddy field conditions under correlated data;

[0021] In the flow time Set boundary conditions in the middle;

[0022] Training is performed by minimizing the negative log-likelihood of the training set, and training is stopped after convergence occurs.

[0023] After training, the flow-time reverse flow integral was recalculated and the paddy field anomaly was defined. ;

[0024] The anomaly degree of paddy fields was conditionally standardized and combined with the paddy field stress index to calculate the paddy field risk score. .

[0025] As a preferred embodiment of the artificial intelligence-based early warning method for low-temperature chilling injury in rice as described in this invention, the following steps are taken: A node graph is constructed based on a paddy field grid; the wind vectors of edge nodes are measured and interpolated to obtain the node wind vectors; connecting edges are constructed and their weights are calculated; a dynamic graph convolution algorithm is used to predict the future state of the paddy field and obtain the future risk index of the paddy field; the center coordinates of the grid are taken as the node coordinates; the nodes of the paddy field grid are taken as the node set V; the state of the paddy field in each grid is bound to the nodes; the boundary nodes are determined according to the position of the paddy field grid and form an edge node set S; the remaining nodes are the center nodes; and the wind vectors of the edge nodes are constructed based on the wind speed and direction of the edge nodes. ;

[0026] The wind vector of the central node is obtained by using inverse distance weighted interpolation with a fixed power order based on the wind vectors of the edge nodes. ;

[0027] For each central node, the nearest neighbor method is used to select 8 neighboring nodes to form a directed edge. The weight of the connecting edge is obtained by multiplying the distance kernel and the wind direction after alignment. ;

[0028] Normalize the edge weights and construct symmetric diffusion weights. ;

[0029] Form a symmetric diffusion matrix from the symmetric diffusion weights. Meanwhile, define the degree matrix. The symmetric normalized graph Laplace operator is calculated. ;

[0030] Stack the paddy field states of each node in the node set into a state matrix. A continuous-time model was established using the graphical heat equation and a mild regression term;

[0031] Each prediction step Break it down into equal-length small steps And steadily update the state matrix ;

[0032] Get the updated prediction step size state matrix Then, the future state of the paddy field at each node is extracted and the future risk of the paddy field is calculated.

[0033] As a preferred embodiment of the artificial intelligence-based rice low-temperature cold damage early warning method of the present invention, the step of forming a dynamic risk map display based on paddy field risk and future risk refers to obtaining the current paddy field risk and the future risk of paddy field, visually marking them on the paddy field node map, and dynamically displaying them over time to form a dynamic risk map.

[0034] As a preferred embodiment of the artificial intelligence-based rice low-temperature cold damage early warning method of the present invention, the dynamic risk map of the paddy field is obtained and then synchronously stored in the database as historical data of the paddy field.

[0035] Secondly, this invention provides an artificial intelligence-based early warning system for low-temperature chilling injury in rice, comprising,

[0036] The data collection and analysis module is used to collect satellite data and divide the rice area into grids, simultaneously deploy sensors to collect ground data, define environmental parameters, establish a convergent gated vector field of the paddy field through a three-way gating mechanism, obtain the initial paddy field state, solve the paddy field state using numerical integration, and output the paddy field stress index.

[0037] The paddy field risk analysis module is used to construct vertically correlated data based on environmental parameters, define constraints for the paddy field flow vector field, calculate the logarithmic density of the paddy field state under the vertically correlated data, output the paddy field anomaly degree, and calculate the paddy field risk in combination with the paddy field stress index.

[0038] The spatiotemporal prediction module is used to construct a node graph based on the paddy field grid, measure the wind vector of the edge node and interpolate to obtain the wind vector of the node, construct connecting edges and calculate weights, use dynamic graph convolution algorithm to predict the future state of the paddy field and obtain the future risk of the paddy field, and form a dynamic risk map based on the paddy field risk and future risk for display.

[0039] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the artificial intelligence-based rice low-temperature chilling injury early warning method described in the first aspect of the present invention.

[0040] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the artificial intelligence-based rice low-temperature chilling injury early warning method as described in the first aspect of the present invention.

[0041] The beneficial effects of this invention are as follows: This invention calculates the paddy field stress index by collecting remote sensing ground collaborative data and combining it with gated vector fields, and obtains the paddy field risk by comprehensively evaluating the paddy field anomaly degree through context-conditional density assessment, thereby realizing real-time low temperature risk early warning for paddy field areas. At the same time, it predicts the state of paddy fields by using wind constraints and graph convolution algorithms, ensuring early prediction of paddy field risks and helping to optimize paddy field disaster prevention strategies. Attached Figure Description

[0042] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is a flowchart of the artificial intelligence-based early warning method for low-temperature chilling injury in rice in Example 1.

[0044] Figure 2 This is a structural diagram of the artificial intelligence-based rice low-temperature chilling injury early warning system in Example 1. Detailed Implementation

[0045] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0046] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0047] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0048] Example 1, referring to Figure 1 and Figure 2 This is the first embodiment of the present invention, which provides an artificial intelligence-based method for early warning of low-temperature chilling injury in rice, comprising the following steps:

[0049] S1. Collect satellite data and divide the rice area into grids, deploy sensors to collect ground data, define environmental parameters, and establish a convergent gated vector field of the paddy field through a three-way gating mechanism to obtain the initial state of the paddy field. Solve the state of the paddy field using numerical integration and output the paddy field stress index.

[0050] Specifically, satellite data is collected by dividing the rice paddy area into grids, and sensors are deployed simultaneously to collect ground data. This means dividing the rice paddy area into equally sized grids and collecting satellite data of the grid areas, including the vegetation index NDVI and land surface temperature LST. In addition, rice paddy data is collected by deploying sensors in the rice paddy area, including rice canopy temperature and air temperature difference, humidity, wind speed, net radiation, and soil volumetric water content. The coordinate system and timestamps of the satellite data and rice paddy data are then unified.

[0051] Furthermore, environmental parameters are defined, and a convergent gated vector field for the paddy field is established through a three-way gating mechanism to obtain the initial paddy field state. Numerical integration is used to solve for the paddy field state and output the paddy field stress index. Satellite data and paddy field data are then integrated to form a 7-dimensional environmental parameter. And standardize it:

[0052]

[0053] in and For the mean and standard deviation of the environmental parameters in each dimension, The labeled difference matrix, Standardized environmental parameters;

[0054] Define the hour as the time unit benchmark and calculate the Fourier time characteristics. :

[0055]

[0056] Where t is time;

[0057] The gating probability is calculated using a three-part gating mechanism based on Fourier time characteristics. :

[0058]

[0059] in These represent the activation intensity in the cold stress state, normal state, and recovery state, respectively. For the gated weight matrix, For gated bias;

[0060] Define paddy field status ,in and This indicates the growth and repair potential of rice, reflecting its reversible growth and repair capabilities. Indicates water potential, which is related to the water supply capacity of rice leaves / roots. It represents the energy potential, reflecting the effective energy state of the rice canopy under the influence of radiation and heat capacity. This represents cold stress memory, characterizing the memory nucleus that reflects the sustained effects of low temperatures. This indicates the degree of tissue damage, specifically the potential damage associated with the phase transition of rice cell membranes / ice nucleus loss. This represents the flux-coupled state, indicating the influence of the absorption of hydrothermal coupling on the state evolution. These are time-varying gated buffer states used to absorb the smooth transition of gating switching. They are all dimensionless latent variables, not directly observed, but solved for, and a defined observation window is used to obtain standardized historical environmental parameters. The historical environmental parameters are processed and mapped to the initial paddy field state using a GRU encoder. :

[0061]

[0062]

[0063]

[0064]

[0065]

[0066] in It is in the GRU hidden state. For gating weights, For gating parameters, For bias, To reset the door, To update the door, For the updated GRU hidden state, For linear mapping, For mapping bias, This refers to the hidden state output by the GRU encoder.

[0067] Define a fixed structure for each sub-vector field using a gating mechanism. :

[0068]

[0069] in It is a linear dissipation matrix. , For each dimension of the paddy field state, the linear dissipation parameter is... , as well as For trainable weights and biases, For the tanh function;

[0070] All sub-vector fields are combined into a convergent gated total vector field for the rice paddy using a fixed structure:

[0071]

[0072] in The total gate vector field has trainable parameters. For the nth sub-vector field, The activation intensity of the nth state includes the cold stress state, the normal state, and the recovery state;

[0073] Using the RK45 solver, with specified training, inference, and maximum function evaluation times, the rice paddy state is solved for the convergent gated total vector field at the target time set using numerical integration.

[0074]

[0075] in For the convergent gated total vector field of the rice paddy, For the target time;

[0076] The paddy field condition was mapped to a paddy field stress index using a fixed sigmoid calibration. :

[0077]

[0078] in and These are parameters for mapping the state of paddy fields.

[0079] By employing a three-part gating mechanism, this invention enables phased prediction of low-temperature chilling injury in rice based on environmental parameters. It can dynamically adjust the stress state of rice according to different environmental conditions, allowing the model to adaptively switch between cold stress, normal, and recovery states, accurately capturing the dynamic physiological changes of rice. By using a GRU encoder to process and map historical environmental data to an initial paddy field state, it can automatically extract time dependencies from historical data, capturing the impact of long-term and short-term environmental changes on rice state. This avoids the tedious process of manually extracting features required by traditional models. Simultaneously, the GRU gating mechanism effectively solves the long-term dependency problem, ensuring full utilization of the contextual information of historical data when predicting paddy field state. By establishing a convergent gating vector field for the paddy field, it can perform high-precision dynamic modeling of the response to environmental changes throughout the rice growth cycle. The convergence of the vector field ensures that the model will not produce unstable solutions during numerical calculations, while the numerical integration method can effectively solve complex nonlinear dynamic problems and calculate the real-time changes in rice state.

[0080] S2. Construct vertically correlated data based on environmental parameters, define constraints for paddy field flow vector field, calculate the logarithmic density of paddy field status under vertically correlated data, output paddy field anomaly degree, and calculate paddy field risk by combining paddy field stress index.

[0081] Specifically, based on environmental parameters, we construct vertically correlated data and define constraints for the paddy field flow vector field. After calculating the logarithmic density of the paddy field state under the vertically correlated data, we output the paddy field anomaly degree. Combined with the paddy field stress index, we calculate the paddy field risk index. We construct vertically correlated data of the paddy field state based on environmental parameters and Fourier time characteristics. By combining paddy field conditions and upstream / downstream correlation data, a two-layer MLP and residual gating structure are used to construct the paddy field flow vector field g:

[0082]

[0083] in For fixed-time frequency embedding, k is the dimension label. For time, To concatenate vectors, , , as well as For the trainable parameters of the paddy field flow vector field, at each step of training... and Execute spectral normalized projection, For a fixed dissipation matrix, ,in A fixed dissipation coefficient for each dimension;

[0084] Using the standard Gaussian distribution as the base distribution, the inverse flow integral is applied in the flow time. Calculate the log density of paddy field conditions under the correlated data:

[0085]

[0086] in As a prior distribution, To assess the condition of the paddy fields The divergence;

[0087] In the flow time Set boundary conditions in the middle, For the currently observed sample, at this time Based on the boundary conditions, we obtain:

[0088]

[0089] in As the base distribution, It is a closed-form distribution with a base distribution;

[0090] Training is performed by minimizing the negative log-likelihood of the training set, and training is stopped after convergence. Represented as:

[0091]

[0092] Where N is the total number of time points in the training set;

[0093] After training, the flow-time reverse flow integral was recalculated and the paddy field anomaly was defined. :

[0094]

[0095] The anomaly degree of paddy fields was conditionally standardized and combined with the paddy field stress index to calculate the paddy field risk score. :

[0096]

[0097] in and These are the weights for the stress index and the degree of abnormality, respectively.

[0098] By introducing a two-layer MLP and residual gating structure to construct a paddy field flow vector field, deep learning modeling of changes in paddy field state is performed. The addition of a gating mechanism allows for dynamic adjustment of state propagation within the flow vector field. The weighted flow vector field can produce different propagation effects for different paddy field states, especially during periods of chilling injury. These mechanisms effectively simulate the propagation of physical quantities such as water and energy within the paddy field. The reverse flow integral is derived in reverse time by setting boundary conditions (currently observed samples) to calculate the logarithmic density of the paddy field state at different time scales. Borrowing from the reverse calculation method in fluid dynamics, the uncertainty and anomaly degree during rice growth are obtained by integrating over the flow time. Based on a standard Gaussian distribution as the baseline distribution, the calculated logarithmic density accurately characterizes the relationship between paddy field state and environmental parameters, thus outputting the paddy field anomaly degree. The stress index and anomaly degree are weighted according to different weighting coefficients to output a paddy field risk score, forming a quantitative risk index. This risk index not only considers the current stress state of the paddy field but also incorporates its historical data, ensuring high timeliness and accuracy in risk assessment.

[0099] S3. Construct a node graph based on the paddy field grid, measure the wind vector of the edge node and interpolate to obtain the wind vector of the node, construct connecting edges and calculate weights, use dynamic graph convolution algorithm to predict the future state of the paddy field and obtain the future risk of the paddy field, and form a dynamic risk map based on the paddy field risk and future risk to display.

[0100] Specifically, a node graph is constructed based on the paddy field grid. Wind vectors at edge nodes are measured and interpolated to obtain the node wind vectors. Connecting edges are constructed and their weights calculated. A dynamic graph convolution algorithm is used to predict the future state of the paddy field and obtain its future risk index. The grid center coordinates are taken as node coordinates based on the divided paddy field grid. The nodes of the paddy field grid are used as a node set V. The state of each paddy field grid is bound to a node. Boundary nodes are determined based on the location of the paddy field grid and form an edge node set S. The remaining nodes are the center nodes. Edge node wind vectors are constructed based on the wind speed and direction of the edge nodes. :

[0101]

[0102] in Let be the wind direction at edge node s. Let be the wind speed at edge node s;

[0103] The wind vector of the central node is obtained by using inverse distance weighted interpolation with a fixed power order based on the wind vectors of the edge nodes. Define the power as 2 and the smoothing constant as 50:

[0104]

[0105] in These are the inverse distance weighting coefficients. Let be the distance between node v and edge node s. It is a smoothing constant. Power;

[0106] For each central node, the nearest neighbor method is used to select 8 neighboring nodes to form a directed edge. The weight of the connecting edge is obtained by multiplying the distance kernel and the wind direction after alignment. :

[0107]

[0108]

[0109]

[0110] in Let the distance be between the center nodes u and v. For spatial proximity, The distance kernel between the central nodes u and v This is the wind direction alignment coefficient. Let the edge direction be the unit vector. and The positions of the central nodes v and u, The wind vector is the midpoint of the edge. To prevent division by zero constant, For wind direction alignment parameters, The reference wind speed;

[0111] Normalize the edge weights and construct symmetric diffusion weights. :

[0112]

[0113] in The weights of the connected edges after row normalization;

[0114] Form a symmetric diffusion matrix from the symmetric diffusion weights. Meanwhile, define the degree matrix. The symmetric normalized graph Laplace operator is calculated. :

[0115]

[0116]

[0117] in Let the symmetric diffusion weights be those for the central nodes j and v. It is the identity matrix;

[0118] Stack the paddy field states of each node in the node set into a state matrix. Where M is the total number of nodes, a continuous-time model is established using the graph heat equation and a mild regression term:

[0119]

[0120] in The diffusion coefficient is... For regression coefficients, The long-term baseline for each node is determined by taking the historical average value for each node.

[0121] Each prediction step Break it down into equal-length small steps And steadily update the state matrix :

[0122]

[0123] in equal-length small steps Quantity;

[0124] Get the updated prediction step size state matrix Then, the future state of the paddy field at each node is extracted and the future risk of the paddy field is calculated.

[0125] By dividing the paddy field area into a grid and defining node coordinates for each grid center, a node graph is constructed. This approach provides a spatially discretized model of the paddy field state, allowing the state of each node to be dynamically updated based on environmental factors (such as wind speed and direction). Wind vector interpolation effectively overcomes the sparsity of geographical information, enabling each central node to obtain wind information from surrounding edge nodes. Through inverse distance weighted interpolation, the transition of wind vectors is smoothed, ensuring more accurate calculations of wind speed and direction, especially in large-scale paddy field areas, effectively improving computational accuracy. During edge weight calculation, the combination of wind speed direction and spatial adjacency relationships... It provides accurate local propagation effects, especially the directional influence of wind speed and direction on paddy field conditions. By combining the weights of wind direction and speed, it can more accurately simulate the propagation of paddy field conditions. In particular, under the influence of low-temperature chilling injury, the directionality of wind force has a direct impact on the propagation of cold air. By using a dynamic graph convolution algorithm, each node in the node graph (i.e., the paddy field grid) is dynamically propagated, and the future paddy field conditions and risks are predicted. This process combines the historical conditions of the paddy field, environmental factors (such as wind speed and direction), and flow patterns, so that the model can not only capture the current paddy field conditions, but also predict future risks based on environmental changes.

[0126] Furthermore, the dynamic risk map display based on paddy field risk and future risk refers to the process of obtaining the current and future risks of paddy fields, visually marking them on the paddy field node map, and dynamically displaying them over time to form a dynamic risk map.

[0127] Furthermore, the dynamic risk map of the paddy field is obtained and then synchronously stored in the database as historical data of the paddy field.

[0128] This embodiment also provides an artificial intelligence-based early warning system for low-temperature chilling injury in rice, including:

[0129] The data collection and analysis module is used to collect satellite data and divide the rice area into grids, simultaneously deploy sensors to collect ground data, define environmental parameters, establish a convergent gated vector field of the paddy field through a three-way gating mechanism, obtain the initial paddy field state, solve the paddy field state using numerical integration, and output the paddy field stress index.

[0130] The paddy field risk analysis module is used to construct vertically correlated data based on environmental parameters, define constraints for the paddy field flow vector field, calculate the logarithmic density of the paddy field state under the vertically correlated data, output the paddy field anomaly degree, and calculate the paddy field risk in combination with the paddy field stress index.

[0131] The spatiotemporal prediction module is used to construct a node graph based on the paddy field grid, measure the wind vector of the edge node and interpolate to obtain the wind vector of the node, construct connecting edges and calculate weights, use dynamic graph convolution algorithm to predict the future state of the paddy field and obtain the future risk of the paddy field, and form a dynamic risk map based on the paddy field risk and future risk for display.

[0132] This embodiment also provides a computer device applicable to the artificial intelligence-based rice low-temperature chilling injury early warning method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the artificial intelligence-based rice low-temperature chilling injury early warning method proposed in the above embodiment.

[0133] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0134] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the artificial intelligence-based early warning method for low-temperature damage to rice as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0135] In summary, this invention uses remote sensing ground collaborative data acquisition and gated vector field to calculate the paddy field stress index, and outputs the paddy field anomaly degree through context-conditional density assessment to obtain the paddy field risk, thereby realizing real-time low temperature risk early warning for paddy field areas. At the same time, it uses wind constraint and graph convolution algorithm to predict the state of paddy fields, ensuring early prediction of paddy field risks and helping to optimize paddy field disaster prevention strategies.

[0136] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An artificial intelligence-based low-temperature cold damage early warning method for rice, characterized in that: include, By collecting satellite data and dividing the rice-growing area into a grid, simultaneously deploying sensors to collect ground data, defining environmental parameters, and establishing a convergent gated vector field for the rice field through a three-way gating mechanism, the initial state of the rice field is obtained, and the rice field stress index is output by solving the rice field state using numerical integration. Based on environmental parameters, we construct the vertical correlation data, define the paddy field flow vector field and add constraints, calculate the log density of the paddy field state under the vertical correlation data, output the paddy field anomaly degree, and combine the paddy field stress index to calculate the paddy field risk. A node graph is constructed based on the paddy field grid. The wind vectors of the edge nodes are measured and interpolated to obtain the wind vectors of the nodes. Connecting edges are constructed and their weights are calculated. A dynamic graph convolution algorithm is used to predict the future state of the paddy field and obtain the future risk of the paddy field. A dynamic risk map is formed based on the paddy field risk and the future risk to display the data. The ground data includes rice paddy data; The definition environment parameter, and through three-part gate mechanism establish the convergence type gate vector field of paddy field, obtain the initialization paddy field state, use numerical integration to solve the paddy field state and output the paddy field stress index, integrate satellite data and paddy field data to form 7-dimensional environment parameter And standardization; Define the hour as the time unit benchmark and calculate the Fourier time characteristics. ; The gating probability is calculated using a three-part gating mechanism based on Fourier time characteristics. ; Define paddy field status And define the observation window to obtain standardized historical environmental parameters. The historical environmental parameters are processed and mapped to the initial paddy field state using a GRU encoder. ; Define a fixed structure for each sub-vector field using a gating mechanism. ; All sub-vector fields are combined into a convergent gated total vector field for the paddy field using a fixed structure. The RK45 solver was used, and the training period, inference period and the number of maximum function evaluations were specified. The numerical integration method was used to solve the paddy field state of the convergent gated total vector field at the target time set. The paddy field condition was mapped to a paddy field stress index using a fixed sigmoid calibration. ; The process involves constructing vertically correlated data based on environmental parameters, defining constraints on the paddy field flow vector field, calculating the logarithmic density of the paddy field state under the vertically correlated data, and outputting the paddy field anomaly degree. This is combined with the paddy field stress index to calculate the paddy field risk index. The vertically correlated data of the paddy field state is constructed based on environmental parameters and Fourier time characteristics. , To standardize environmental parameters, a paddy field flow vector field g is constructed using a two-layer MLP and residual gating structure, combining paddy field conditions and upstream / downstream correlation data. Using the standard Gaussian distribution as the base distribution, the inverse flow integral is applied in the flow time. Calculate the log density of paddy field conditions under correlated data; In the flow time Set boundary conditions in the middle; Training is performed by minimizing the negative log-likelihood of the training set, and training is stopped after convergence occurs. After training, the flow-time reverse flow integral was recalculated and the paddy field anomaly was defined. ; The anomaly degree of paddy fields was conditionally standardized and combined with the paddy field stress index to calculate the paddy field risk score. .

2. The method for early warning of low-temperature chilling injury in rice based on artificial intelligence as described in claim 1, characterized in that: The process of collecting satellite data and dividing the rice paddy area into grids, and simultaneously deploying sensors to collect ground data, refers to dividing the rice paddy area into equally sized grids and collecting satellite data of the grid areas, including the vegetation index NDVI and the land surface temperature LST. Simultaneously, it involves deploying sensors in the rice paddy area to collect rice paddy data, including rice canopy temperature and air temperature difference, humidity, wind speed, net radiation, and soil volumetric moisture content. The coordinate system and timestamps of the satellite data and the rice paddy data are then unified.

3. The artificial intelligence-based early warning method for low-temperature chilling injury in rice as described in claim 2, characterized in that: The process involves constructing a node graph based on a paddy field grid, measuring and interpolating the wind vectors of edge nodes to obtain the node wind vectors, constructing connecting edges and calculating weights, using a dynamic graph convolution algorithm to predict the future state of the paddy field and obtain the future risk index. The grid center coordinates are taken as node coordinates, and the nodes of the paddy field grid are used as a node set V. The state of the paddy field in each grid is bound to a node. Boundary nodes are determined based on the position of the paddy field grid and form an edge node set S, with the remaining nodes as center nodes. The wind vectors of the edge nodes are constructed based on the wind speed and direction of the edge nodes. ; The wind vector of the central node is obtained by using inverse distance weighted interpolation with a fixed power order based on the wind vectors of the edge nodes. ; For each central node, the nearest neighbor method is used to select 8 neighboring nodes to form a directed edge. The weight of the connecting edge is obtained by multiplying the distance kernel and the wind direction after alignment. ; Normalize the edge weights and construct symmetric diffusion weights. ; Form a symmetric diffusion matrix from the symmetric diffusion weights. Meanwhile, define the degree matrix. The symmetric normalized graph Laplace operator is calculated. ; Stack the paddy field states of each node in the node set into a state matrix. A continuous-time model was established using the graphical heat equation and a mild regression term; Each prediction step Break it down into equal-length small steps And steadily update the state matrix ; Get the updated prediction step size state matrix Then, the future state of the paddy field at each node is extracted and the future risk of the paddy field is calculated.

4. The artificial intelligence-based early warning method for low-temperature chilling injury in rice as described in claim 3, characterized in that: The dynamic risk map display based on paddy field risk and future risk refers to obtaining the current paddy field risk and future paddy field risk, visually marking them on the paddy field node map, and dynamically displaying them over time to form a dynamic risk map.

5. The artificial intelligence-based early warning method for low-temperature chilling injury in rice as described in claim 4, characterized in that: After obtaining the dynamic risk map of the paddy field, it is synchronously stored in the database as historical data of the paddy field.

6. An artificial intelligence-based early warning system for low-temperature chilling injury in rice, based on the artificial intelligence-based early warning method for low-temperature chilling injury in rice as described in any one of claims 1 to 5, characterized in that: include, The data collection and analysis module is used to collect satellite data and divide the rice area into grids, simultaneously deploy sensors to collect ground data, define environmental parameters, establish a convergent gated vector field of the paddy field through a three-way gating mechanism, obtain the initial paddy field state, solve the paddy field state using numerical integration, and output the paddy field stress index. The paddy field risk analysis module is used to construct vertically correlated data based on environmental parameters, define constraints for the paddy field flow vector field, calculate the logarithmic density of the paddy field state under the vertically correlated data, output the paddy field anomaly degree, and calculate the paddy field risk in combination with the paddy field stress index. The spatiotemporal prediction module is used to construct a node graph based on the paddy field grid, measure the wind vector of the edge node and interpolate to obtain the wind vector of the node, construct connecting edges and calculate weights, use dynamic graph convolution algorithm to predict the future state of the paddy field and obtain the future risk of the paddy field, and form a dynamic risk map based on the paddy field risk and future risk for display.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the artificial intelligence-based early warning method for rice low-temperature chilling injury as described in any one of claims 1 to 5.

8. A 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 steps of the artificial intelligence-based early warning method for low-temperature chilling injury in rice as described in any one of claims 1 to 5.

Images