A scab environmental factor intelligent monitoring and early warning device and method

Abstract

The present application relates to the technical field of agricultural environment monitoring, and particularly relates to a scab environmental factor intelligent monitoring and early warning device and method. The real-time scab distribution overview is divided into multiple sub-distribution profiles, each real-time scab sample between the various sub-distribution profiles is semi-variogram space correlation hot spot interpolated, and the average dynamic layout flow of the sub-distribution profile is variational deduced to obtain a dynamic hot spot change map of the real-time scab distribution in a specified monitoring area. The past environmental factors that potentially caused past scab infection are tested to obtain existing occurrence rules of the scab affected by the environmental factors, and the dynamic hot spot change map is updated based on the hazard rating thermal dynamic balance of the existing occurrence rules to determine a control area and make the intelligent monitoring and early warning device make a warning response. The present application can monitor the scab environmental factors in the wheat planting process in real time and make a control and early warning response to unreasonable planting environment, effectively improving the planting environment and quality of wheat.

Description

Technical Field

[0001] This invention relates to the field of agricultural environmental monitoring technology, and in particular to an intelligent monitoring and early warning device and method for environmental factors of Fusarium head blight. Background Technology

[0002] Fusarium head blight (FHB) is a major disease affecting grain crops such as wheat. Caused by fungi of the genus *Fusarium*, it not only severely impacts crop yields but also produces toxins that contaminate grains, threatening human and animal health. The occurrence of FHB is closely related to various environmental factors, such as temperature, humidity, rainfall, soil moisture, air humidity, and crop growth stage. Under specific climatic conditions, the disease can easily break out and spread. Therefore, real-time monitoring and intelligent early warning of FHB's environmental conditions are crucial for early disease control. With the development of IoT and AI technologies, agricultural disease monitoring and early warning are gradually shifting from traditional experience-based judgment to data-driven and model-based approaches. FHB monitoring and early warning systems based on environmental factors are becoming an important component of smart agriculture.

[0003] However, many existing environmental factor early warning devices for Fusarium head blight on the market rely primarily on temperature and humidity as core parameters, lacking comprehensive collection and analysis of multi-source environmental factors. This makes it difficult to accurately determine the multi-dimensional conditions for disease occurrence. Some devices depend on static threshold judgments or empirical formulas, lacking the ability to dynamically learn and adapt to changing trends of environmental factors, resulting in low early warning accuracy. Furthermore, traditional Fusarium head blight environmental factor early warning devices struggle to accurately capture the temporal dynamic changes of the tiny Fusarium head blight fungus, significantly reducing the accuracy of suitability assessments for real-time environmental factors leading to Fusarium head blight occurrence within a designated area. This makes it difficult to accurately determine the rationality of environmental factors, easily leading to false alarms or missed alarms. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides an intelligent monitoring and early warning device and method for environmental factors of Fusarium head blight.

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

[0006] The first aspect of this invention provides an early warning method for an intelligent monitoring and early warning device for environmental factors of Fusarium head blight, comprising the following steps:

[0007] S102: Using the tensor dimension of the established monitoring requirements, perform temporal singular decomposition on the past environmental factor data and past Fusarium head blight infection data of the specified monitoring area within a preset historical time period to obtain the core variable tensor structure of the past environmental factor data and past Fusarium head blight infection data with the preset time period as the alignment benchmark.

[0008] S104: Based on the past spatiotemporal node map of the core variable tensor structure environmental factor - Fusarium head blight infection, the temporal convolution block and spatial convolution block contained in the past spatiotemporal node map are extracted by convolution and superimposed alternately to obtain the real-time Fusarium head blight distribution overview when real-time environmental factor data is generated in the specified monitoring area within the advanced monitoring time segment.

[0009] S106: Divide the real-time Fusarium head blight distribution overview into multiple sub-distribution overviews, use semi-variable spatial correlation hotspot interpolation for each real-time Fusarium head blight sample between each sub-distribution overview, and use variational deduction to calculate the average dynamic layout flow of the sub-distribution overviews to obtain a dynamic hotspot transition map of the real-time Fusarium head blight distribution within the specified monitoring area.

[0010] S108: Test the potential directional relationship between past environmental factors and past Fusarium head blight infections, obtain the existing occurrence patterns of Fusarium head blight under the influence of environmental factors, update the dynamic hotspot change map based on the hazard rating thermal dynamic balance of the existing occurrence patterns, thereby determining the control area and enabling the intelligent monitoring and early warning device to issue an early warning response.

[0011] More specifically, step S102 includes the following steps:

[0012] Obtain the designated monitoring areas and monitoring records for wheat cultivation, and extract past environmental factor data and past Fusarium head blight infection data for the designated monitoring areas from the intelligent monitoring and early warning devices within a preset historical time period through the monitoring records;

[0013] The established monitoring requirements are obtained when the intelligent monitoring and early warning device senses past environmental factor data and past Fusarium head blight infection data within a preset historical time period. Based on the established monitoring requirements, the tensor dimensions and the target monitoring rank of each tensor dimension are preset.

[0014] A singular decomposition algorithm is introduced, in which the past environmental factor data and past Fusarium head blight infection data are matrix-expanded in each tensor dimension to obtain the tensor flattened matrix of environmental factor monitoring. Based on the target monitoring rank, the tensor flattened matrix is ​​subjected to temporal singular decomposition in the singular decomposition algorithm.

[0015] By using temporal singular decomposition, the temporal tensor singular values ​​of each tensor dimension are obtained. Based on the temporal tensor singular values ​​of each tensor dimension, a core variable tensor structure is established using the preset time period as the alignment benchmark for past environmental factor data and past Fusarium head blight infection data.

[0016] More specifically, step S104 includes the following steps:

[0017] Based on the core variable tensor structure topology, a spatiotemporal node graph of environmental factors-Fusarium head blight infection is constructed. A one-dimensional dilated causal convolutional layer is introduced to extract the temporal changes of each past spatiotemporal node by convolution in the time dimension, generating several temporal convolutional blocks.

[0018] An adaptive graph convolutional layer is introduced to calculate the spatial relationships of each past spatiotemporal node, resulting in an adaptive adjacency matrix. Based on the adaptive adjacency matrix, convolution captures the dependency relationships between each past spatiotemporal node in space, generating several spatial convolutional blocks.

[0019] Obtain wheat planting requirements, and based on these requirements, obtain the advanced monitoring time segments of Fusarium head blight infection in the designated monitoring area through the intelligent monitoring and early warning device. Then, obtain real-time environmental factor data of the designated monitoring area at the advanced monitoring time segments through the intelligent monitoring and early warning device.

[0020] Based on the advanced monitoring time segment, a convolutional stacking order of real-time environmental factor data is constructed. Based on the convolutional stacking order, several temporal convolutional blocks and several spatial convolutional blocks are alternately stacked to form a deep spatiotemporal prediction structure. Based on the deep spatiotemporal prediction structure, a real-time Fusarium head blight distribution overview is determined when real-time environmental factor data is generated in the specified monitoring area within the advanced monitoring time segment.

[0021] More specifically, step S106 includes the following steps:

[0022] The preset monitoring strategy of the intelligent monitoring and early warning device for environmental factors of Fusarium head blight is obtained, and the instantaneous monitoring time step of the intelligent monitoring and early warning device is extracted according to the preset monitoring strategy.

[0023] Based on the instantaneous monitoring time step, the advanced monitoring time segment is divided into several uniform auxiliary time segments. The real-time Fusarium head blight distribution corresponding to each auxiliary time segment is extracted from the real-time Fusarium head blight distribution overview and marked as a sub-distribution overview.

[0024] A Gaussian model is introduced to calculate the Mahalanobis distance between each real-time Fusarium head blight sample in each sub-distribution profile. Based on the Mahalanobis distance, the spatial correlation between each real-time Fusarium head blight sample is fitted in the Gaussian model to obtain the semi-variable distribution function of the real-time Fusarium head blight sample.

[0025] A semi-variant interpolation model is constructed using the Kriging interpolation algorithm. In the semi-variant interpolation model, interpolation calculations are performed on each half-variant distribution function to obtain the Kriging hotspot interpolation equation for real-time Fusarium head blight samples. The Kriging hotspot interpolation equation is solved to obtain the relative static fitting error matrix of each real-time Fusarium head blight sample at the corresponding time segment among the various sub-distribution profiles.

[0026] Based on the dynamic potential index constrained by the relative static fitting error matrix, for each real-time Fusarium head blight sample, the dynamic energy index is minimized to perform variational deduction of the average dynamic layout flow of the neighborhood samples in each sub-distribution profile. If the dynamic potential index is 0, the variational deduction operation is stopped and the layout direction clue field is output.

[0027] Obtain a geographic schematic diagram of the designated monitoring area, construct a multi-dimensional spatial domain of the designated monitoring area based on the geographic schematic diagram, and stitch together all sub-distribution overviews in the multi-dimensional spatial domain according to the attached time segments based on the layout trend clue field to generate a dynamic hotspot change map of real-time Fusarium head blight distribution within the designated monitoring area.

[0028] More specifically, step S108 includes the following steps:

[0029] By using the timeline of past Fusarium head blight infections within a preset historical period and the directed edges of the historical factor parameter intervals generated by the corresponding historical environmental factor data, the potential occurrence relationship between the past environmental factor data and the past Fusarium head blight infection data is tested, and the existing occurrence pattern of Fusarium head blight under the influence of environmental factors is obtained.

[0030] The random sampling strategy of the intelligent monitoring and early warning device for environmental factors of Fusarium head blight is obtained. Based on the random sampling strategy, the multidimensional spatial domain of the specified monitoring area is divided into N sub-spatial domains. At the same time, a thermal color gamut rating system that fits the risk assessment of Fusarium head blight hazards is obtained based on big data.

[0031] Based on the thermal color gamut rating system, a binary color gamut search tree with existing occurrence patterns is constructed. The root base and current root value of the binary color gamut search tree are obtained. Starting from the root base, the hotspot chromaticity coefficient of each dynamic hotspot in the dynamic hotspot transition graph is traversed.

[0032] If the hotspot chromaticity coefficient is greater than the current root value, then the dynamic hotspot is iteratively inserted into the left trunk of the binary color gamut search tree to generate the left subtree; if the hotspot chromaticity coefficient is less than the current root value, then the dynamic hotspot is iteratively inserted into the right trunk of the binary color gamut search tree to generate the right subtree.

[0033] Obtain the heights of the left and right subtrees respectively, and get the height values ​​of the left and right trees. Calculate the balance deviation of each original trunk node in the binary search tree structure based on the height values ​​of the left and right trees to obtain several balance factors.

[0034] If the balance factor exceeds the dynamic balance factor threshold of the dynamic hotspot, the original trunk node where the balance factor is located is extracted and marked as the deleted trunk node. The leaf growth path of the deleted trunk node is obtained. The height of the binary color gamut search tree is continuously updated from the leaf growth path upwards until the balance factor does not exceed the dynamic balance factor threshold. The appropriate thermal index for the proliferation of Fusarium head blight in each subspace is output.

[0035] A prevention and control heat map query table is established based on planting needs. If a suitable heat index can be found in the prevention and control heat map query table, the subspace area is marked as a prevention and control area and uploaded to the intelligent monitoring and early warning device to issue an early warning signal response.

[0036] More specifically, the method of using the timeline of past Fusarium head blight infections within a preset historical period and the directed edges of the corresponding historical factor parameter intervals generated from past environmental factor data to test the potential correlation between past environmental factor data and past Fusarium head blight infection data, and obtaining the existing occurrence patterns of Fusarium head blight under the influence of environmental factors, specifically includes the following steps:

[0037] By extracting the timeline of past Fusarium head blight infection data in the specified monitoring area within a preset historical time period from the monitoring archives, the corresponding historical factor parameter ranges generated by each past environmental factor data within the preset historical time period are obtained simultaneously.

[0038] Historical environmental factor data is defined as Class I milestones, and historical Fusarium head blight infection data is defined as Class II milestones. Based on the occurrence timeline, anti-disturbance independent bins are preset, and independent recursive condition sets for each Class I milestone are constructed according to the historical factor parameter intervals.

[0039] Each of the first-class mileage points is discretely combined one by one to obtain several sets of first-class mileage point discrete pairs. The perturbation degree of each set of first-class mileage point discrete pairs for second-class mileage points under the triggering independent recursive condition set is tested and calculated to obtain the perturbation frequency of each set of first-class mileage point discrete pairs.

[0040] If the disturbance frequency is within the independent bin for disturbance resistance, then delete the discrete pairs of Class I mileage points in that group, and mark the independent recursive condition set triggered by the deleted discrete pairs of Class I mileage points as a separation set; if the disturbance frequency is not within the independent bin for disturbance resistance, then skip the discrete pairs of Class I mileage points in that group and execute the next set of decision analysis, and finally obtain the potential undirected edges that cause each Class I mileage point to appear in each Class II mileage point.

[0041] The historical factor parameter intervals of each occurrence node in the occurrence timeline are obtained and defined as the hazard factor parameter intervals. Sovereign hazard conditions for each Class II mileage point are constructed based on the hazard factor parameter intervals.

[0042] If the separation set has at least one or more sovereignty hazard conditions, then the interval capacity based on the sovereignty hazard conditions is the direction and intensity applied to the potential undirected edges corresponding to the two types of mileage points to obtain several potential directed edges.

[0043] By constructing a potential occurrence directed acyclic graph using several potential directed edges, the existing occurrence patterns of Fusarium head blight under the influence of environmental factors can be determined based on the topological structure of the potential occurrence directed acyclic graph.

[0044] The second aspect of this invention provides an intelligent monitoring and early warning device for Fusarium head blight environmental factors, applicable to an early warning method for any of the intelligent monitoring and early warning devices for Fusarium head blight environmental factors described in any one of the claims, specifically including:

[0045] A Fusarium wilt environmental factor monitoring module, which is responsible for real-time monitoring of environmental factors around wheat planting in the field;

[0046] The monitoring and early warning module is used to monitor and issue early warning signals for areas with unreasonable and unsuitable thermal indices.

[0047] The spatial modeling module is responsible for acquiring a geographic schematic diagram of the specified monitoring area and constructing a multi-dimensional spatial domain of the specified monitoring area based on the geographic schematic diagram.

[0048] The data storage and extraction module is used to store and extract past environmental factor data and past Fusarium head blight infection data of the intelligent monitoring and early warning device for a specified monitoring area within a preset historical time period.

[0049] The data query module and the data calculation module are used to query whether a suitable heat index exists in the heat index query table for epidemic prevention and control.

[0050] This invention addresses the technical deficiencies in the prior art, and its beneficial technical effects are as follows:

[0051] By performing temporal singular decomposition on past environmental factor data and past Fusarium head blight infection data within a preset historical time period in the designated monitoring area according to the tensor dimensions of the predetermined monitoring requirements, the core variable tensor structure of the past environmental factor data and past Fusarium head blight infection data with the preset time period as the alignment benchmark is obtained. Based on the past spatiotemporal node graph of environmental factor-Fusarium head blight infection in the core variable tensor structure, the temporal convolutional blocks and spatial convolutional blocks contained in the past spatiotemporal node graph are extracted by convolution and alternately superimposed to obtain the real-time Fusarium head blight distribution overview in the designated monitoring area when real-time environmental factor data is generated within the advanced monitoring time segment. The invention divides the real-time Fusarium head blight distribution overview into multiple sub-distribution profiles. Semi-variable spatial correlation hotspot interpolation is used to interpolate the real-time Fusarium head blight samples between each sub-distribution profile, and variational deduction is performed to extrapolate the average dynamic layout flow of the sub-distribution profiles, resulting in a dynamic hotspot transition map of the real-time Fusarium head blight distribution within a specified monitoring area. The invention also tests the directed relationships between past environmental factors and potential past Fusarium head blight infections, obtaining existing occurrence patterns of Fusarium head blight under the influence of environmental factors. Based on the existing occurrence patterns, the dynamic hotspot transition map is updated using the hazard rating thermodynamic dynamic balance, thereby determining the control area and enabling the intelligent monitoring and early warning device to issue an early warning response. This invention can perform real-time monitoring of environmental factors related to Fusarium head blight during wheat cultivation and calculate the suitability and rationality of environmental factors for Fusarium head blight growth. Based on this suitability and rationality, the intelligent monitoring and early warning device can promptly issue a control and early warning response, achieving scientific prevention and control of wheat cultivation and avoiding Fusarium head blight infection that reduces wheat cultivation quality. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0053] Figure 1 A flowchart of the first method for an early warning method using an intelligent monitoring and early warning device for environmental factors of Fusarium head blight is shown.

[0054] Figure 2 A second method flowchart of an early warning method for an intelligent monitoring and early warning device for environmental factors of Fusarium head blight is shown;

[0055] Figure 3 A schematic diagram of the overall structure of an intelligent monitoring and early warning device for environmental factors of Fusarium head blight is shown. Detailed Implementation

[0056] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0057] 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 therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0058] The first aspect of this invention provides an early warning method for an intelligent monitoring and early warning device for environmental factors of Fusarium head blight, such as... Figure 1 As shown, it includes the following steps:

[0059] S102: Using the tensor dimension of the established monitoring requirements, perform temporal singular decomposition on the past environmental factor data and past Fusarium head blight infection data of the specified monitoring area within a preset historical time period to obtain the core variable tensor structure of the past environmental factor data and past Fusarium head blight infection data with the preset time period as the alignment benchmark.

[0060] S104: Based on the past spatiotemporal node map of the core variable tensor structure environmental factor - Fusarium head blight infection, the temporal convolution block and spatial convolution block contained in the past spatiotemporal node map are extracted by convolution and superimposed alternately to obtain the real-time Fusarium head blight distribution overview when real-time environmental factor data is generated in the specified monitoring area within the advanced monitoring time segment.

[0061] S106: Divide the real-time Fusarium head blight distribution overview into multiple sub-distribution overviews, use semi-variable spatial correlation hotspot interpolation for each real-time Fusarium head blight sample between each sub-distribution overview, and use variational deduction to calculate the average dynamic layout flow of the sub-distribution overviews to obtain a dynamic hotspot transition map of the real-time Fusarium head blight distribution within the specified monitoring area.

[0062] S108: Test the potential directional relationship between past environmental factors and past Fusarium head blight infections, obtain the existing occurrence patterns of Fusarium head blight under the influence of environmental factors, update the dynamic hotspot change map based on the hazard rating thermal dynamic balance of the existing occurrence patterns, thereby determining the control area and enabling the intelligent monitoring and early warning device to issue an early warning response.

[0063] More specifically, step S102 includes the following steps:

[0064] Obtain the designated monitoring areas and monitoring records for wheat cultivation, and extract past environmental factor data and past Fusarium head blight infection data for the designated monitoring areas from the intelligent monitoring and early warning devices within a preset historical time period through the monitoring records;

[0065] The established monitoring requirements are obtained when the intelligent monitoring and early warning device senses past environmental factor data and past Fusarium head blight infection data within a preset historical time period. Based on the established monitoring requirements, the tensor dimensions and the target monitoring rank of each tensor dimension are preset.

[0066] A singular decomposition algorithm is introduced, in which the past environmental factor data and past Fusarium head blight infection data are matrix-expanded in each tensor dimension to obtain the tensor flattened matrix of environmental factor monitoring. Based on the target monitoring rank, the tensor flattened matrix is ​​subjected to temporal singular decomposition in the singular decomposition algorithm.

[0067] By using temporal singular decomposition, the temporal tensor singular values ​​of each tensor dimension are obtained. Based on the temporal tensor singular values ​​of each tensor dimension, a core variable tensor structure is established using the preset time period as the alignment benchmark for past environmental factor data and past Fusarium head blight infection data.

[0068] It should be noted that certain environmental conditions generated during wheat cultivation can make Fusarium head blight more comfortable for its spread and growth. Therefore, the prevention and early warning of Fusarium head blight environmental factors usually requires predictive methods to analyze whether the gains of real-time environmental factors in promoting Fusarium head blight spread are reasonable. However, the prediction of Fusarium head blight growth under real-time environmental factors mainly relies on information from historical data. However, there is a situation where the tensor dimensions of the extracted historical environmental factor data and historical Fusarium head blight infection data cannot be aligned in time. For example, during real-time prediction, a certain historical Fusarium head blight infection data at a certain historical time point may be based on a spatial tensor, while a certain historical environmental factor data that creates that spatial tensor is not at that historical time point. This creates a temporal matching error in the data information prediction, resulting in a large information asymmetry in real-time prediction. This leads to misaligned pattern, vector, and matrix calculation errors in the distribution of Fusarium head blight under real-time changes in historical environmental factors, reducing the reliability of real-time distribution prediction of Fusarium head blight. To address this, this method pre-defines tensor dimensions and corresponding target monitoring ranks based on established monitoring requirements. The tensor dimension can be arbitrary, depending on the monitoring indicators for environmental factors and Fusarium head blight infection. For example, if the intelligent monitoring and early warning device monitors environmental factor data from a spatial and environmental variable perspective over a historical period, then the tensor dimension would be a three-dimensional tensor of longitude × latitude × environmental variable. The target monitoring rank specifies the main feature dimensions to be retained on each tensor dimension, controlling the composition of the tensor decomposition. Subsequently, past environmental factor data and past Fusarium head blight infection data are matrix-expanded along the tensor dimensions, thus unfolding the tensor into a matrix along each tensor dimension. This matrix format clearly demonstrates the tensor symmetry between past environmental factor data and past Fusarium head blight infection data, thereby improving the accuracy of tensor unification across different data variables.

[0069] It should be noted that the past environmental factor data includes temperature, humidity, light intensity, soil texture, and soil pH. Then, using the tensor flattening matrix obtained from the temporal singular decomposition algorithm, the most important principal components (orthogonal directions) under each tensor dimension are extracted. This effectively removes redundant or low-energy dimensions and avoids interference from unnecessary tensor component noise. This ensures more accurate alignment of tensor components generated from past Fusarium head blight infection data based on a unified temporal series of past environmental factor data. It guarantees the core tensor alignment of past environmental factor data and past Fusarium head blight infection data under the constraint of a preset time period as the alignment benchmark. This provides a reliable temporal data structure foundation for subsequent real-time prediction of Fusarium head blight distribution, avoids prediction disorder, and improves the accuracy of intelligent monitoring and early warning devices in predicting Fusarium head blight distribution.

[0070] More specifically, step S104, as follows: Figure 2 As shown, the specific steps include:

[0071] S202: Based on the core variable tensor structure topology, construct the past spatiotemporal node graph of environmental factors - Fusarium head blight infection, introduce a one-dimensional dilated causal convolutional layer to extract the temporal changes of each past spatiotemporal node in the time dimension, and generate several temporal convolutional blocks.

[0072] S204: Simultaneously introduce an adaptive graph convolutional layer to calculate the spatial relationships of each past spatiotemporal node, obtain an adaptive adjacency matrix, and capture the dependency relationships between each past spatiotemporal node in space based on the adaptive adjacency matrix convolution to generate several spatial convolutional blocks.

[0073] S206: Obtain wheat planting requirements, and based on these requirements, obtain the advanced monitoring time segments of Fusarium head blight infection in the designated monitoring area through the intelligent monitoring and early warning device, and obtain real-time environmental factor data of the designated monitoring area at the advanced monitoring time segments through the intelligent monitoring and early warning device;

[0074] S208: Construct the convolutional stacking order of real-time environmental factor data according to the advanced monitoring time segment, and alternately stack several temporal convolutional blocks and several spatial convolutional blocks based on the convolutional stacking order to form a deep spatiotemporal prediction structure. Determine the real-time Fusarium head blight distribution overview in the specified monitoring area when real-time environmental factor data is generated within the advanced monitoring time segment based on the deep spatiotemporal prediction structure.

[0075] It should be noted that, due to the temporal and spatial characteristics of both environmental factors and Fusarium head blight, traditional methods for predicting the trend of environmental factors and Fusarium head blight infection typically require extensive training and validation with historical data. This reliance on data satisfying certain linear spatiotemporal relationships and spatiotemporal normal distributions significantly degrades the performance of traditional models in predicting the spatiotemporal distribution of Fusarium head blight. Furthermore, these models struggle to capture the nonlinear relationships between spatiotemporal data and are weak in recognizing spatiotemporal patterns of complex environmental factors affecting Fusarium head blight, making it difficult to handle dynamic changes in spatiotemporal data, such as structural abrupt changes and seasonal variations. Consequently, they cannot ensure high accuracy in predicting the real-time distribution of Fusarium head blight. To address this, this method first constructs a historical spatiotemporal node graph of environmental factors and Fusarium head blight infection based on the tensor structure topology of the core variables. This graph depicts the temporal characteristics and dependencies of past Fusarium head blight transmission and infection under the influence of past environmental factors, serving as a crucial basis for understanding the historical data trajectory. Next, a one-dimensional dilated causal convolutional layer is used to extract the temporal features of each past spatiotemporal node through convolution in the time dimension. This is achieved by stacking multiple layers to model and capture the long-term dependence of past environmental factors on Fusarium head blight infection, from short-term to long-term temporal changes, maintaining predictive causality by only considering past events. Simultaneously, an adaptive graph convolutional layer is used to calculate the spatial relationships between each past spatiotemporal node to obtain an adaptive adjacency matrix. This allows convolution to capture the dependencies between each past spatiotemporal node in space, making it more flexible than traditional prediction models. On the one hand, it can automatically learn hidden spatial dependencies, enhancing generalization ability and combining prior knowledge and learning structures to improve spatial prediction accuracy; on the other hand, it reduces unnecessary computational steps, improving prediction efficiency. The adaptive adjacency matrix can be an adaptive learning graph, allowing the prediction of Fusarium head blight distribution to be independent of a fixed graph structure and possessing the ability to automatically discover potential node relationships.

[0076] It should be noted that temporal and spatial convolutions generate several temporal convolutional blocks and several spatial convolutional blocks. These convolutional blocks reveal the historical development trend of Fusarium head blight transmission and infection under the influence of past environmental factors, enhancing predictive expressiveness and preventing gradient vanishing. Therefore, by alternately stacking these convolutional blocks based on real-time environmental factor data and the temporal sequence of monitored real-time data, a real-time distribution of Fusarium head blight with the spatiotemporal trend of real-time environmental factors as the main context can be generated. This method can analyze the historical development trend of historical environmental factor data and historical Fusarium head blight infection data based on the spatiotemporal convolution of the core variable tensor structure, thereby making the prediction of real-time environmental factor monitoring data more in line with the known spatiotemporal characteristics of Fusarium head blight distribution and improving the accuracy of real-time Fusarium head blight distribution prediction.

[0077] More specifically, step S106 includes the following steps:

[0078] The preset monitoring strategy of the intelligent monitoring and early warning device for environmental factors of Fusarium head blight is obtained, and the instantaneous monitoring time step of the intelligent monitoring and early warning device is extracted according to the preset monitoring strategy.

[0079] Based on the instantaneous monitoring time step, the advanced monitoring time segment is divided into several uniform auxiliary time segments. The real-time Fusarium head blight distribution corresponding to each auxiliary time segment is extracted from the real-time Fusarium head blight distribution overview and marked as a sub-distribution overview.

[0080] A Gaussian model is introduced to calculate the Mahalanobis distance between each real-time Fusarium head blight sample in each sub-distribution profile. Based on the Mahalanobis distance, the spatial correlation between each real-time Fusarium head blight sample is fitted in the Gaussian model to obtain the semi-variable distribution function of the real-time Fusarium head blight sample.

[0081] A semi-variant interpolation model is constructed using the Kriging interpolation algorithm. In the semi-variant interpolation model, interpolation calculations are performed on each half-variant distribution function to obtain the Kriging hotspot interpolation equation for real-time Fusarium head blight samples. The Kriging hotspot interpolation equation is solved to obtain the relative static fitting error matrix of each real-time Fusarium head blight sample at the corresponding time segment among the various sub-distribution profiles.

[0082] Based on the dynamic potential index constrained by the relative static fitting error matrix, for each real-time Fusarium head blight sample, the dynamic energy index is minimized to perform variational deduction of the average dynamic layout flow of the neighborhood samples in each sub-distribution profile. If the dynamic potential index is 0, the variational deduction operation is stopped and the layout direction clue field is output.

[0083] Obtain a geographic schematic diagram of the designated monitoring area, construct a multi-dimensional spatial domain of the designated monitoring area based on the geographic schematic diagram, and stitch together all sub-distribution overviews in the multi-dimensional spatial domain according to the attached time segments based on the layout trend clue field to generate a dynamic hotspot change map of real-time Fusarium head blight distribution within the designated monitoring area.

[0084] It should be noted that the main component of Fusarium head blight is the conidia of wheat pathogens. While these conidia exhibit a certain point-like distribution and infection trajectory during real-time spread, their tiny size makes them difficult to observe with the naked eye. Therefore, to more accurately capture and clarify the infection trend and distribution pattern of Fusarium head blight in real time, this method fragments the obtained real-time Fusarium head blight distribution overview into uniform time segments according to the instantaneous monitoring time step of the intelligent monitoring and early warning device. This yields sub-distribution profiles for different time periods. Furthermore, semi-variant distribution analysis and hotspot interpolation are performed on each real-time Fusarium head blight sample within each sub-distribution profile. This generates the hotspot distribution of real-time Fusarium head blight at specific time nodes within each sub-distribution profile, thereby visualizing the temporal static activity of Fusarium head blight at different times within the advanced monitoring time segment. Next, the relative static fitting error matrix of each real-time Fusarium head blight sample in the corresponding time segment is calculated among the sub-distribution profiles. Based on this relative static fitting error matrix, the average dynamic layout flow of each sub-distribution profile in the advance monitoring time segment is derived variationally. This is a temporal dynamic clue connecting the real-time Fusarium head blight activity in all sub-distribution profiles, thereby establishing a dynamic relationship between the real-time Fusarium head blight activity distribution at different times, and finally forming a hotspot activity transition map expressing the dynamic distribution of real-time Fusarium head blight in the specified monitoring area according to the advance monitoring time segment. The dynamic hotspot transition map constructed by this method can clearly describe the real-time infection trend and distribution scale of Fusarium head blight in a specific time period, making it possible to visualize the distribution of real-time Fusarium head blight under the influence of real-time environmental factors. This can effectively improve the accuracy of the suitability analysis of environmental factors for Fusarium head blight growth and avoid false alarms and missed alarms in the device's early warning.

[0085] It should be noted that the semivariogram distribution function can quantitatively describe the spatial autocorrelation of all real-time Fusarium head blight samples in a certain sub-distribution profile, that is, how the sample attribute values ​​change with increasing distance. This function is the core parameter of hotspot interpolation, and its assignment quantifies the hotspot location of real-time Fusarium head blight samples. The Kriging hotspot interpolation equation established based on the semivariogram distribution function and the spatial location of the samples can ensure that the hotspot interpolation result is a linear unbiased estimate, improving the accuracy of hotspot location while obtaining a static confidence assessment between each hotspot sample, that is, a relative static fitting error matrix. This relative static fitting error matrix expresses the dynamic smoothness of real-time Fusarium head blight samples between each sub-distribution profile. When the smoothness is high, this method uses variational inference to minimize this dynamic potential exponent to track the dynamic layout trajectory of hotspot samples with the temporal sequence of the advance monitoring time segment, so that the sub-distribution profiles at different times have a reasonable and smooth temporal connection basis, making the dynamic hotspot changes in the advance monitoring time segment more accurate and reliable.

[0086] More specifically, step S108 includes the following steps:

[0087] By using the timeline of past Fusarium head blight infections within a preset historical period and the directed edges of the historical factor parameter intervals generated by the corresponding historical environmental factor data, the potential occurrence relationship between the past environmental factor data and the past Fusarium head blight infection data is tested, and the existing occurrence pattern of Fusarium head blight under the influence of environmental factors is obtained.

[0088] The random sampling strategy of the intelligent monitoring and early warning device for environmental factors of Fusarium head blight is obtained. Based on the random sampling strategy, the multidimensional spatial domain of the specified monitoring area is divided into N sub-spatial domains. At the same time, a thermal color gamut rating system that fits the risk assessment of Fusarium head blight hazards is obtained based on big data.

[0089] Based on the thermal color gamut rating system, a binary color gamut search tree with existing occurrence patterns is constructed. The root base and current root value of the binary color gamut search tree are obtained. Starting from the root base, the hotspot chromaticity coefficient of each dynamic hotspot in the dynamic hotspot transition graph is traversed.

[0090] If the hotspot chromaticity coefficient is greater than the current root value, then the dynamic hotspot is iteratively inserted into the left trunk of the binary color gamut search tree to generate the left subtree; if the hotspot chromaticity coefficient is less than the current root value, then the dynamic hotspot is iteratively inserted into the right trunk of the binary color gamut search tree to generate the right subtree.

[0091] Obtain the heights of the left and right subtrees respectively, and get the height values ​​of the left and right trees. Calculate the balance deviation of each original trunk node in the binary search tree structure based on the height values ​​of the left and right trees to obtain several balance factors.

[0092] If the balance factor exceeds the dynamic balance factor threshold of the dynamic hotspot, the original trunk node where the balance factor is located is extracted and marked as the deleted trunk node. The leaf growth path of the deleted trunk node is obtained. The height of the binary color gamut search tree is continuously updated from the leaf growth path upwards until the balance factor does not exceed the dynamic balance factor threshold. The appropriate thermal index for the proliferation of Fusarium head blight in each subspace is output.

[0093] A prevention and control heat map query table is established based on planting needs. If a suitable heat index can be found in the prevention and control heat map query table, the subspace area is marked as a prevention and control area and uploaded to the intelligent monitoring and early warning device to issue an early warning signal response.

[0094] It should be noted that when the dynamic hotspot change map shows a large frequency, quantity, and range of hotspot activity in a certain area, it indicates that the environmental factors present there are more suitable for the growth and spread of Fusarium head blight. However, traditional early warning methods usually directly assess the hazard level based on known real-time Fusarium head blight distribution. This method struggles to consider the suitability and rationality of environmental factors for Fusarium head blight infection, thus failing to make accurate early warning judgments based on real-time monitored environmental factor data. To address this, this method first utilizes the timeline of past Fusarium head blight infections within a preset historical period and the corresponding historical factor parameter intervals generated from past environmental factor data to conduct directed edge testing on the potential occurrence relationships between past environmental factor data and past Fusarium head blight infection data. This further refines the existing occurrence patterns of Fusarium head blight under the influence of environmental factors. Next, a thermal color gamut rating system suitable for Fusarium head blight (FHB) risk assessment is obtained. This system is a logical rule reference that accurately delineates the risk level of FHB through thermal colors; for example, thermal red represents a high risk level. A binary color gamut search tree based on this system, which constructs a tree showing existing occurrence patterns, can assign corresponding thermal colors to the trends of environmental factors leading to different FHB infections. This allows the real-time FHB distribution expressed using dynamic hotspot transition maps to accurately index the corresponding risk level's thermal color rating within the binary color gamut search tree. Specifically, if the hotspot's chromaticity coefficient is greater than the current rootstock value, it indicates that the dynamic hotspot's RPG color value within the color gamut is low, meaning it's in a light-colored area. Therefore, this dynamic hotspot is iteratively inserted into the left branch of the binary color gamut search tree, which is primarily responsible for recording light-colored areas, thus generating a left subtree belonging to the lighter-colored hotspot. Conversely, if the RPG color value of the dynamic hotspot is too high within the color gamut, it indicates that the hotspot is in a dark area. Therefore, the dynamic hotspot is iteratively inserted into the right trunk, which is mainly responsible for recording dark colors, generating a right subtree belonging to the darker hotspot. The tree structure formed by the left and right subtrees then constitutes the global color gamut description of the dynamic hotspot. This method enables accurate color assignment of dynamic hotspot samples for real-time Fusarium head blight distribution, resulting in a more precise rating of the hazard risk of the corresponding environmental factors.

[0095] It should be noted that the balance factor is used to measure the balance deviation of the color gamut change relative to the next time sequence when each dynamic hotspot changes over time, reflecting the reliability of the color gamut update of the intelligent monitoring and early warning device for real-time dynamic monitoring of Fusarium head blight. If the balance factor exceeds the dynamic balance factor threshold of the dynamic hotspot, it indicates that the color gamut update of the dynamic hotspot is incorrect when the time sequence changes. In this case, the RPG color value of the dynamic hotspot should be updated again. This method deletes the original trunk node where the balance factor is located and continuously updates the height of the binary color gamut search tree upward according to the original leaf growth path, thereby reassigning the RPG color value of the next time sequence node, thus achieving the dynamic hazard rating update effect of environmental factors under the real-time distribution of Fusarium head blight. Among them, if a suitable heat index can be found in the prevention and control heat query table, it means that the environmental factors in this area are more conducive to the growth, reproduction, and spread of Fusarium head blight. Therefore, this area is regarded as a high pathogen hazard area, and the intelligent monitoring and early warning device should be controlled to issue corresponding early warnings for this area. This method enables the real-time environmental factors to be assigned color values ​​using a thermal color gamut-based real-time Fusarium head blight distribution hazard rating system. This allows for an accurate assessment of whether environmental factors within a region are suitable or reasonable for Fusarium head blight growth, resulting in more precise prevention and control warnings, avoiding false alarms or missed alarms, and improving warning performance.

[0096] More specifically, the method of using the timeline of past Fusarium head blight infections within a preset historical period and the directed edges of the corresponding historical factor parameter intervals generated from past environmental factor data to test the potential correlation between past environmental factor data and past Fusarium head blight infection data, and obtaining the existing occurrence patterns of Fusarium head blight under the influence of environmental factors, specifically includes the following steps:

[0097] By extracting the timeline of past Fusarium head blight infection data in the specified monitoring area within a preset historical time period from the monitoring archives, the corresponding historical factor parameter ranges generated by each past environmental factor data within the preset historical time period are obtained simultaneously.

[0098] Historical environmental factor data is defined as Class I milestones, and historical Fusarium head blight infection data is defined as Class II milestones. Based on the occurrence timeline, anti-disturbance independent bins are preset, and independent recursive condition sets for each Class I milestone are constructed according to the historical factor parameter intervals.

[0099] Each of the first-class mileage points is discretely combined one by one to obtain several sets of first-class mileage point discrete pairs. The perturbation degree of each set of first-class mileage point discrete pairs for second-class mileage points under the triggering independent recursive condition set is tested and calculated to obtain the perturbation frequency of each set of first-class mileage point discrete pairs.

[0100] If the disturbance frequency is within the independent bin for disturbance resistance, then delete the discrete pairs of Class I mileage points in that group, and mark the independent recursive condition set triggered by the deleted discrete pairs of Class I mileage points as a separation set; if the disturbance frequency is not within the independent bin for disturbance resistance, then skip the discrete pairs of Class I mileage points in that group and execute the next set of decision analysis, and finally obtain the potential undirected edges that cause each Class I mileage point to appear in each Class II mileage point.

[0101] The historical factor parameter intervals of each occurrence node in the occurrence timeline are obtained and defined as the hazard factor parameter intervals. Sovereign hazard conditions for each Class II mileage point are constructed based on the hazard factor parameter intervals.

[0102] If the separation set has at least one or more sovereignty hazard conditions, then the interval capacity based on the sovereignty hazard conditions is the direction and intensity applied to the potential undirected edges corresponding to the two types of mileage points to obtain several potential directed edges.

[0103] By constructing a potential occurrence directed acyclic graph using several potential directed edges, the existing occurrence patterns of Fusarium head blight under the influence of environmental factors can be determined based on the topological structure of the potential occurrence directed acyclic graph.

[0104] It should be noted that the method for testing the potential occurrence relationship of directed edges for past environmental factor data and past Fusarium head blight infection data involves obtaining the occurrence timeline of each past Fusarium head blight infection data within a preset historical time period in the specified monitoring area, as well as the corresponding historical factor parameter intervals generated by each past environmental factor data within the preset historical time period. Since the occurrence of historical factor parameter intervals generated by certain past environmental factors does not necessarily lead to the occurrence of past Fusarium head blight infection, this method first tests and calculates the degree of perturbation of the second-class mileage points under the triggering independent recursive condition set for each set of discrete pairs of first-class mileage points. If the perturbation frequency is within the disturbance-resistant independent bin, it indicates that the historical parameters applied by the set of past environmental factors are located on the timeline of past Fusarium head blight infection, that is, the set of past environmental factors caused the occurrence of past Fusarium head blight infection. Therefore, the set of past environmental factors does not change the independence of Fusarium head blight infection, so the set of discrete pairs of first-class mileage points is deleted, and the applied historical factor parameter interval is regarded as a separation set. If it is not located, it means that when the past environmental factors produce specific historical factor parameters, they do not affect the change of Fusarium head blight infection and are independent. Therefore, the combination of past environmental factor data of this group of first-class milestone points is ignored. In this way, the potential irregular relationship between past environmental factors and the occurrence of Fusarium head blight infection can be preliminarily identified. The potential undirected edge of occurrence lays the foundation for subsequent exploration of the pattern.

[0105] It should be noted that after clarifying the fuzzy potential patterns, it is necessary to further explore and reveal the specific direction and intensity of the potential undirected connections using the historical factor parameter intervals of past Fusarium head blight occurrences. To this end, this method constructs sovereign hazard conditions for each type II milestone based on the hazard factor parameter intervals. If the separator set contains at least one or more sovereign hazard conditions, it indicates a high degree of consistency between the historical factor parameters that led to past Fusarium head blight occurrences and the historical factor parameters applied to the historical environmental factor data with potential undirected connections. Therefore, it can be basically determined that the past environmental factors at that time of occurrence caused the past Fusarium head blight infection. Therefore, the interval capacity of the sovereign hazard conditions is further used as the direction and intensity of the potential undirected edges applied to the type II milestones, thereby clarifying the existing occurrence patterns of Fusarium head blight under the influence of environmental factors. This method can explore and analyze the potential occurrence patterns between historical environmental factor data and historical Fusarium head blight infection data, providing reliable and highly credible hazard rating logic clues for the suitability of real-time environmental factors for the growth of real-time Fusarium head blight, making the hazard rating more accurate.

[0106] The second aspect of this invention provides an intelligent monitoring and early warning device for Fusarium head blight environmental factors, applicable to the early warning method of any of the intelligent monitoring and early warning devices for Fusarium head blight environmental factors described in any one of the claims, such as... Figure 3 As shown, it specifically includes:

[0107] The Fusarium wilt environmental factor monitoring module 1011 is responsible for real-time monitoring of environmental factors around wheat planting in the field.

[0108] The monitoring and early warning module 1012 is used to monitor and issue early warning signals for areas with unreasonable and unsuitable thermal indices.

[0109] The spatial modeling module 1013 is responsible for obtaining a geographic schematic diagram of the specified monitoring area and constructing a multi-dimensional spatial domain of the specified monitoring area based on the geographic schematic diagram.

[0110] Data storage and extraction module 1014, the data storage and extraction module is used to store and extract past environmental factor data and past Fusarium head blight infection data of the intelligent monitoring and early warning device for the specified monitoring area within a preset historical time period;

[0111] The data query module and the data calculation module are used to query whether a suitable heat index exists in the heat index query table for epidemic prevention and control.

[0112] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A warning method of a scab environmental factor intelligent monitoring and early warning device, characterized in that, Includes the following steps: S102: Using the tensor dimension of the established monitoring requirements, perform temporal singular decomposition on the past environmental factor data and past Fusarium head blight infection data of the specified monitoring area within a preset historical time period to obtain the core variable tensor structure of the past environmental factor data and past Fusarium head blight infection data with the preset time period as the alignment benchmark. S104: Based on the past spatiotemporal node map of the core variable tensor structure environmental factor - Fusarium head blight infection, the temporal convolution block and spatial convolution block contained in the past spatiotemporal node map are extracted by convolution and superimposed alternately to obtain the real-time Fusarium head blight distribution overview when real-time environmental factor data is generated in the specified monitoring area within the advanced monitoring time segment. S106: Divide the real-time Fusarium head blight distribution overview into multiple sub-distribution overviews, use semi-variable spatial correlation hotspot interpolation for each real-time Fusarium head blight sample between each sub-distribution overview, and use variational deduction to calculate the average dynamic layout flow of the sub-distribution overviews to obtain a dynamic hotspot transition map of the real-time Fusarium head blight distribution within the specified monitoring area. S108: Test the potential directional relationship between past environmental factors and past Fusarium head blight infections, obtain the existing occurrence patterns of Fusarium head blight under the influence of environmental factors, update the dynamic hotspot change map based on the hazard rating thermal dynamic balance of the existing occurrence patterns, thereby determining the control area and enabling the intelligent monitoring and early warning device to issue an early warning response. Specifically, step S108 includes the following steps: By using the timeline of past Fusarium head blight infections within a preset historical period and the directed edges of the historical factor parameter intervals generated by the corresponding historical environmental factor data, the potential occurrence relationship between the past environmental factor data and the past Fusarium head blight infection data is tested, and the existing occurrence pattern of Fusarium head blight under the influence of environmental factors is obtained. The random sampling strategy of the intelligent monitoring and early warning device for environmental factors of Fusarium head blight is obtained. Based on the random sampling strategy, the multidimensional spatial domain of the specified monitoring area is divided into N sub-spatial domains. At the same time, a thermal color gamut rating system that fits the risk assessment of Fusarium head blight hazards is obtained based on big data. Based on the thermal color gamut rating system, a binary color gamut search tree with existing occurrence patterns is constructed. The root base and current root value of the binary color gamut search tree are obtained. Starting from the root base, the hotspot chromaticity coefficient of each dynamic hotspot in the dynamic hotspot transition graph is traversed. If the hotspot chromaticity coefficient is greater than the current root value, then the dynamic hotspot is iteratively inserted into the left trunk of the binary color gamut search tree to generate the left subtree; if the hotspot chromaticity coefficient is less than the current root value, then the dynamic hotspot is iteratively inserted into the right trunk of the binary color gamut search tree to generate the right subtree. Obtain the heights of the left and right subtrees respectively, and get the height values ​​of the left and right trees. Calculate the balance deviation of each original trunk node in the binary search tree structure based on the height values ​​of the left and right trees to obtain several balance factors. If the balance factor exceeds the dynamic balance factor threshold of the dynamic hotspot, the original trunk node where the balance factor is located is extracted and marked as the deleted trunk node. The leaf growth path of the deleted trunk node is obtained. The height of the binary color gamut search tree is continuously updated from the leaf growth path upwards until the balance factor does not exceed the dynamic balance factor threshold. The appropriate thermal index for the proliferation of Fusarium head blight in each subspace is output. A prevention and control heat map query table is established based on planting needs. If a suitable heat index can be found in the prevention and control heat map query table, the subspace area is marked as a prevention and control area and uploaded to the intelligent monitoring and early warning device to issue an early warning signal response.

2. The early warning method of the scab environmental factor intelligent monitoring and early warning device according to claim 1, characterized in that, Step S102 specifically includes the following steps: Obtain the designated monitoring areas and monitoring records for wheat cultivation, and extract past environmental factor data and past Fusarium head blight infection data for the designated monitoring areas from the intelligent monitoring and early warning devices within a preset historical time period through the monitoring records; The established monitoring requirements are obtained when the intelligent monitoring and early warning device senses past environmental factor data and past Fusarium head blight infection data within a preset historical time period. Based on the established monitoring requirements, the tensor dimensions and the target monitoring rank of each tensor dimension are preset. A singular decomposition algorithm is introduced, in which the past environmental factor data and past Fusarium head blight infection data are matrix-expanded in each tensor dimension to obtain the tensor flattened matrix of environmental factor monitoring. Based on the target monitoring rank, the tensor flattened matrix is ​​subjected to temporal singular decomposition in the singular decomposition algorithm. By using temporal singular decomposition, the temporal tensor singular values ​​of each tensor dimension are obtained. Based on the temporal tensor singular values ​​of each tensor dimension, a core variable tensor structure is established using the preset time period as the alignment benchmark for past environmental factor data and past Fusarium head blight infection data.

3. The early warning method of the scab environmental factor intelligent monitoring and early warning device according to claim 1, characterized in that, Step S104 specifically includes the following steps: Based on the core variable tensor structure topology, a spatiotemporal node graph of environmental factors-Fusarium head blight infection is constructed. A one-dimensional dilated causal convolutional layer is introduced to extract the temporal changes of each past spatiotemporal node by convolution in the time dimension, generating several temporal convolutional blocks. An adaptive graph convolutional layer is introduced to calculate the spatial relationships of each past spatiotemporal node, resulting in an adaptive adjacency matrix. Based on the adaptive adjacency matrix, convolution captures the dependency relationships between each past spatiotemporal node in space, generating several spatial convolutional blocks. Obtain wheat planting requirements, and based on these requirements, obtain the advanced monitoring time segments of Fusarium head blight infection in the designated monitoring area through the intelligent monitoring and early warning device. Then, obtain real-time environmental factor data of the designated monitoring area at the advanced monitoring time segments through the intelligent monitoring and early warning device. Based on the advanced monitoring time segment, a convolutional stacking order of real-time environmental factor data is constructed. Based on the convolutional stacking order, several temporal convolutional blocks and several spatial convolutional blocks are alternately stacked to form a deep spatiotemporal prediction structure. Based on the deep spatiotemporal prediction structure, a real-time Fusarium head blight distribution overview is determined when real-time environmental factor data is generated in the specified monitoring area within the advanced monitoring time segment.

4. The early warning method of the scab environmental factor intelligent monitoring and early warning device according to claim 1, characterized in that, Step S106 specifically includes the following steps: The preset monitoring strategy of the intelligent monitoring and early warning device for environmental factors of Fusarium head blight is obtained, and the instantaneous monitoring time step of the intelligent monitoring and early warning device is extracted according to the preset monitoring strategy. Based on the instantaneous monitoring time step, the advanced monitoring time segment is divided into several uniform auxiliary time segments. The real-time Fusarium head blight distribution corresponding to each auxiliary time segment is extracted from the real-time Fusarium head blight distribution overview and marked as a sub-distribution overview. A Gaussian model is introduced to calculate the Mahalanobis distance between each real-time Fusarium head blight sample in each sub-distribution profile. Based on the Mahalanobis distance, the spatial correlation between each real-time Fusarium head blight sample is fitted in the Gaussian model to obtain the semi-variable distribution function of the real-time Fusarium head blight sample. A semi-variant interpolation model is constructed using the Kriging interpolation algorithm. In the semi-variant interpolation model, interpolation calculations are performed on each half-variant distribution function to obtain the Kriging hotspot interpolation equation for real-time Fusarium head blight samples. The Kriging hotspot interpolation equation is solved to obtain the relative static fitting error matrix of each real-time Fusarium head blight sample at the corresponding time segment among the various sub-distribution profiles. Based on the dynamic potential index constrained by the relative static fitting error matrix, for each real-time Fusarium head blight sample, the dynamic energy index is minimized to perform variational deduction of the average dynamic layout flow of the neighborhood samples in each sub-distribution profile. If the dynamic potential index is 0, the variational deduction operation is stopped and the layout direction clue field is output. Obtain a geographic schematic diagram of the designated monitoring area, construct a multi-dimensional spatial domain of the designated monitoring area based on the geographic schematic diagram, and stitch together all sub-distribution overviews in the multi-dimensional spatial domain according to the attached time segments based on the layout trend clue field to generate a dynamic hotspot change map of real-time Fusarium head blight distribution within the designated monitoring area.

5. The early warning method of the scab environmental factor intelligent monitoring and early warning device according to claim 1, characterized in that, The method involves using the timeline of past Fusarium head blight infections within a preset historical period and the directed edges of historical factor parameter intervals generated from past environmental factor data to test the potential correlation between past environmental factor data and past Fusarium head blight infection data, thereby obtaining the existing occurrence patterns of Fusarium head blight under the influence of environmental factors. This specifically includes the following steps: By extracting the timeline of past Fusarium head blight infection data in the specified monitoring area within a preset historical time period from the monitoring archives, the corresponding historical factor parameter ranges generated by each past environmental factor data within the preset historical time period are obtained simultaneously. Historical environmental factor data is defined as Class I milestones, and historical Fusarium head blight infection data is defined as Class II milestones. Based on the occurrence timeline, anti-disturbance independent bins are preset, and independent recursive condition sets for each Class I milestone are constructed according to the historical factor parameter intervals. Each of the first-class mileage points is discretely combined one by one to obtain several sets of first-class mileage point discrete pairs. The perturbation degree of each set of first-class mileage point discrete pairs for second-class mileage points under the triggering independent recursive condition set is tested and calculated to obtain the perturbation frequency of each set of first-class mileage point discrete pairs. If the disturbance frequency is within the independent bin for disturbance resistance, then delete the discrete pairs of Class I mileage points in that group, and mark the independent recursive condition set triggered by the deleted discrete pairs of Class I mileage points as a separation set; if the disturbance frequency is not within the independent bin for disturbance resistance, then skip the discrete pairs of Class I mileage points in that group and execute the next set of decision analysis, and finally obtain the potential undirected edges that cause each Class I mileage point to appear in each Class II mileage point. The historical factor parameter intervals of each occurrence node in the occurrence timeline are obtained and defined as the hazard factor parameter intervals. Sovereign hazard conditions for each Class II mileage point are constructed based on the hazard factor parameter intervals. If the separation set has at least one or more sovereignty hazard conditions, then the interval capacity based on the sovereignty hazard conditions is the direction and intensity applied to the potential undirected edges corresponding to the two types of mileage points to obtain several potential directed edges. By constructing a potential occurrence directed acyclic graph using several potential directed edges, the existing occurrence patterns of Fusarium head blight under the influence of environmental factors can be determined based on the topological structure of the potential occurrence directed acyclic graph.

6. The intelligent monitoring and early warning device for Gibberella environmental factors, applied to the early warning method of the intelligent monitoring and early warning device for Gibberella environmental factors according to any one of claims 1-5, characterized in that, Specifically, it includes: A Fusarium wilt environmental factor monitoring module, which is responsible for real-time monitoring of environmental factors around wheat planting in the field; The monitoring and early warning module is used to monitor and issue early warning signals for areas with unreasonable and unsuitable thermal indices. The spatial modeling module is responsible for acquiring a geographic schematic diagram of the specified monitoring area and constructing a multi-dimensional spatial domain of the specified monitoring area based on the geographic schematic diagram. The data storage and extraction module is used to store and extract past environmental factor data and past Fusarium head blight infection data of the intelligent monitoring and early warning device for a specified monitoring area within a preset historical time period. The data query module is used to query whether a suitable heat index exists in the heat index query table for epidemic prevention and control.

Images