Cotton field micro-domain humidity intelligent sensing method and system combined with crown temperature layer monitoring

Abstract

The application discloses a cotton field micro-domain humidity intelligent sensing method and system combined with crown temperature layer monitoring and relates to the technical field of agricultural monitoring. The method comprises the following steps: acquiring cotton field space information and dividing micro-domain plots; arranging crown temperature layer and environmental moisture monitoring devices in each micro-domain, and synchronously extracting multi-source feature sequences; constructing a cotton field heterogeneous hypergraph based on sequence data; arranging monitoring channels at each node, and combining the hyperedge set to perform micro-domain humidity zoning intelligent sensing on the heterogeneous hypergraph. The technical problem of low micro-domain humidity sensing precision and poor spatial continuity in the prior art is solved, in which cotton field micro-domain humidity monitoring relies on point type or single parameter measurement, and it is difficult to realize large-scale, high-precision and real-time dynamic sensing. The technical effect of improving the precision and stability of cotton field micro-domain humidity zoning sensing is achieved by taking crown temperature as a core physiological variable, fusing environmental moisture and other multi-source time series data, and constructing a cotton field heterogeneous hypergraph to realize micro-domain humidity spatial sensing and intelligent inference.

Description

Technical Field

[0001] This invention relates to the field of agricultural monitoring technology, specifically to a method and system for intelligent sensing of micro-area humidity in cotton fields that combines canopy temperature layer monitoring. Background Technology

[0002] As an important economic crop, cotton is highly sensitive to water conditions during its growth process, especially in arid and semi-arid regions. Water management in cotton fields directly affects cotton yield, quality, and water use efficiency. In actual production, cotton fields often exhibit significant spatial heterogeneity due to factors such as differences in soil type, uneven irrigation methods, micro-topographical variations, and fluctuations in meteorological conditions, resulting in multiple micro-regions of humidity differences at different scales.

[0003] Existing cotton field humidity monitoring technologies mostly rely on soil moisture sensors or air humidity sensors for point-based data collection, which has problems such as limited monitoring points, insufficient spatial coverage, and difficulty in reflecting the physiological state of the canopy. At the same time, these methods usually analyze environmental humidity and crop moisture status separately, lacking the characterization of the intrinsic relationship between crop transpiration, physiological regulation, and micro-area moisture distribution. This results in insufficient stability and accuracy of humidity sensing results under complex meteorological conditions, making it difficult to meet the needs of precision irrigation and intelligent decision-making.

[0004] Studies have shown that crop canopy temperature can comprehensively reflect crop transpiration intensity, water stress status, and energy balance characteristics, serving as a crucial bridge connecting crop physiological processes and environmental water conditions. However, current technologies for utilizing canopy temperature information largely remain at the level of single indicators or empirical thresholds, lacking in-depth fusion modeling with multi-source environmental water data and a systematic expression of the spatial structure and temporal evolution of cotton field micro-domains. This results in limited stability and intelligence levels in humidity zoning results. Summary of the Invention

[0005] This application provides a method and system for intelligent sensing of micro-area humidity in cotton fields that combines canopy temperature layer monitoring. It solves the technical problems in the prior art where micro-area humidity monitoring in cotton fields relies on point-based or single-parameter measurements, making it difficult to achieve large-scale, high-precision, and real-time dynamic sensing, resulting in low accuracy and poor spatial continuity of micro-area humidity sensing. It achieves the technical effect of constructing a heterogeneous hypergraph of cotton fields by using canopy temperature as the core physiological variable and integrating multi-source time-series data such as environmental moisture to realize spatial sensing and intelligent inference of micro-area humidity, thereby improving the accuracy and stability of micro-area humidity zoning sensing in cotton fields.

[0006] The first aspect of this application provides a method for intelligent sensing of micro-area humidity in cotton fields combined with canopy temperature layer monitoring, the method comprising:

[0007] The basic spatial information of the target cotton field is acquired and divided into micro-domains to obtain multiple micro-domain plots. Multiple canopy temperature monitoring devices and multiple environmental moisture monitoring devices are deployed above the cotton fields in these micro-domains to extract simultaneous multi-source features, resulting in multiple canopy temperature monitoring feature sequences and multiple environmental moisture monitoring feature sequences. Based on these multiple canopy temperature monitoring feature sequences and multiple environmental moisture monitoring feature sequences, a temporal state analysis is performed to construct a heterogeneous hypergraph of the cotton field, which includes multiple heterogeneous nodes and multiple sets of heterogeneous hyperedges. Multiple monitoring channels are deployed at the multiple heterogeneous nodes, and the cotton field micro-domain humidity zoning of the heterogeneous hypergraph is intelligently perceived using the multiple sets of heterogeneous hyperedges.

[0008] A second aspect of this application provides a smart humidity sensing system for cotton fields combining canopy temperature layer monitoring, the system comprising:

[0009] The system comprises the following modules: Micro-domain partitioning module: acquiring basic spatial information of the target cotton field and partitioning it into multiple micro-domain plots; Feature extraction module: deploying multiple canopy temperature layer monitoring devices and multiple environmental moisture monitoring devices above the cotton fields in the multiple micro-domain plots to perform simultaneous multi-source feature extraction, obtaining multiple canopy temperature layer monitoring feature sequences and multiple environmental moisture monitoring feature sequences; State analysis module: performing time-series state analysis based on the multiple canopy temperature layer monitoring feature sequences and the multiple environmental moisture monitoring feature sequences to construct a heterogeneous hypergraph of the cotton field, wherein the heterogeneous hypergraph includes multiple heterogeneous nodes and multiple sets of heterogeneous hyperedges; Intelligent sensing module: deploying multiple monitoring channels at the multiple heterogeneous nodes and combining the multiple sets of heterogeneous hyperedges to perform intelligent sensing of cotton field micro-domain humidity zoning in the heterogeneous hypergraph of the cotton field.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] First, based on spatial information such as cotton field boundaries, irrigation zones, and soil distribution, the cotton field is divided into micro-domains, forming multiple micro-domains with relatively consistent moisture characteristics. Then, canopy temperature monitoring devices and environmental moisture monitoring devices are deployed above each micro-domain to simultaneously collect multi-source time-series data on canopy temperature and environmental moisture, forming corresponding feature sequences. Next, state analysis and fusion modeling are performed on the multi-source time-series features to construct a heterogeneous hypergraph of the cotton field capable of representing the temporal correlations between different micro-domains. Finally, monitoring channels are set up at heterogeneous nodes, and combined with the micro-domain correlations reflected by the heterogeneous hyperedges, intelligent zoning perception and inference of the humidity distribution within the cotton field micro-domains are achieved. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0013] Figure 1 This is a schematic diagram of a method for intelligent sensing of humidity in cotton fields by combining canopy temperature layer monitoring, provided in an embodiment of this application.

[0014] Figure 2 This is a schematic diagram of a cotton field micro-area humidity intelligent sensing system that combines canopy temperature layer monitoring, provided as an embodiment of this application.

[0015] Figure labeling: Micro-domain partitioning module 11, feature extraction module 12, state parsing module 13, intelligent perception module 14. Detailed Implementation

[0016] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0017] Example 1, as Figure 1 As shown, this application provides a method for intelligent sensing of micro-area humidity in cotton fields combined with canopy temperature layer monitoring, wherein the method includes:

[0018] The basic spatial information of the target cotton field is obtained to divide it into micro-domains, resulting in multiple micro-domain plots.

[0019] In this embodiment, the basic spatial information of the target cotton field is first acquired, including plot boundaries, irrigation zones, and soil type distribution. After acquiring the basic spatial information, the system performs micro-domain division. The core objective of this micro-domain division is to divide the cotton field into multiple regions with similar water control conditions based on the similarity of water management and crop growth. During the division process, the system divides the regions based on factors such as soil texture consistency, irrigation control consistency, and plot boundary consistency, forming multiple partition sets. Then, the intersection of these partition sets is calculated and the regions are merged to obtain multiple micro-domain plots. These micro-domain plots can ensure that soil moisture changes within the same micro-domain are relatively consistent, reducing monitoring deviations caused by differences in soil moisture control conditions, thereby achieving more accurate micro-domain humidity perception.

[0020] Furthermore, the basic spatial information includes cotton field boundary information, irrigation zone information, and soil type distribution information.

[0021] Preferably, the acquired basic spatial information includes cotton field boundary information, irrigation zone information, and soil type distribution information, providing accurate spatial references for subsequent micro-domain delineation, humidity sensing, and intelligent decision-making. Cotton field boundary information clarifies the specific scope and regional division of the cotton field, typically obtained through remote sensing imagery, satellite maps, or field measurements. This provides physical boundaries for subsequent micro-domain delineation, clearly defining different areas within the cotton field and potential differences in water management between these areas. Irrigation zone information refers to the different irrigation zones defined based on the cotton field's irrigation system and management strategies. Each irrigation zone has similar water supply methods, frequencies, and amounts, potentially controlled by irrigation networks, drip irrigation, or sprinkler irrigation. Soil type distribution information involves the soil composition and physical properties of different areas within the cotton field. Different soil types, such as sandy soil, clay, and loam, have varying water retention and permeability, directly affecting soil moisture distribution and changes. Soil types are typically delineated based on the proportions of clay, sand, etc., determining water permeability and retention capacity. This basic spatial information not only provides comprehensive data support for the refined management of cotton fields, but also helps to achieve precise division of cotton field micro-domains, providing a scientific basis for subsequent humidity monitoring, intelligent sensing and optimization of irrigation strategies.

[0022] Furthermore, by acquiring the basic spatial information of the target cotton field, micro-domains are divided to obtain multiple micro-domain plots, including:

[0023] The target cotton field is divided into zones based on soil consistency, irrigation consistency, and plot boundary consistency to obtain a set of soil consistency zones, a set of irrigation consistency zones, and a set of plot boundary consistency zones. The intersection of the soil consistency zone set, the irrigation consistency zone set, and the plot boundary consistency zone set is obtained to obtain multiple initial micro-domains. The areas in the target cotton field other than the multiple initial micro-domains are merged according to crop row and column structure information and the principle of nearest neighbor similarity to obtain the multiple micro-domains.

[0024] Optionally, the target cotton field is first divided into zones based on soil consistency. That is, the soil texture type of each area in the cotton field is identified by the soil type distribution information in the basic spatial information. Based on the principle of similar soil texture or similar water retention capacity, the Euclidean distance of indicators such as soil type, organic matter content, salinity, and soil moisture content at each location in the soil type distribution information is calculated to quantify the similarity distance of soil characteristics at each location. Then, according to the relationship between these similarity distances and the preset similarity distance, the areas with similar soil characteristics in the target cotton field are divided into the same soil consistency zone, forming a set of soil consistency zones. The soil permeability, water retention capacity, and moisture change trend are consistent within each soil consistency zone.

[0025] Secondly, the target cotton field is divided into zones based on irrigation consistency. That is, using the same method as above, the Euclidean distance is calculated for indicators such as irrigation method, irrigation frequency, water source type, and irrigation water volume at each location in the irrigation zone information. The similarity distance of irrigation characteristics at each location is quantified, and the areas with similar irrigation characteristics in the target cotton field are divided into the same irrigation consistency zone, forming a set of irrigation consistency zones, thereby ensuring that the same water supply conditions exist within the same zone.

[0026] Secondly, the target cotton fields are divided into zones based on the consistency of plot boundaries. That is, the cotton fields are divided by boundary constraints in combination with the existing plot boundaries, roads, ditches and other physical boundaries to ensure that different plots do not overlap or merge, thereby dividing into clearly defined blocks and forming a set of plot boundary consistency zones.

[0027] After completing the three types of consistency partitioning mentioned above, the intersection of the soil consistency partition set, irrigation consistency partition set, and plot boundary consistency partition set is calculated. A partition that simultaneously satisfies all three consistency criteria is defined as an initial micro-plot, thus obtaining multiple initial micro-plots. Next, the remaining areas in the target cotton field not covered by the initial micro-plots are supplemented and allocated. Specifically, based on the crop row and column structure information and the cotton planting row and column directions, the remaining areas are regularly divided to ensure that the boundaries of the supplemented areas are consistent with the crop row and column directions, avoiding irregular areas that cross rows, are diagonally aligned, or disrupt the crop planting structure. After completing the regularized partitioning, the remaining areas are merged and allocated according to the nearest neighbor similarity allocation principle. Specifically, adjacent areas are prioritized as candidate micro-domains, and the soil and irrigation similarity between the candidate micro-domains and their adjacent initial micro-domains are calculated. The calculated similarity distances of these two characteristics are then compared with their corresponding neighborhood similarity thresholds. If both similarity distances are less than or equal to the corresponding neighborhood similarity thresholds and the micro-domains are adjacent, the candidate micro-domain is merged with its adjacent initial micro-domain. Through this multi-consistency partitioning, regularized partitioning, and nearest neighbor similarity fusion process, multiple spatially continuous micro-domains with clear boundaries, consistent with crop row structure, and uniform internal soil and irrigation conditions can be formed. This provides a stable and reliable spatial foundation for subsequent canopy temperature monitoring, multi-source feature extraction, and intelligent sensing of cotton field micro-domain humidity.

[0028] Multiple canopy temperature monitoring devices and multiple environmental moisture monitoring devices were deployed above cotton fields in multiple micro-plots to extract multi-source features in a simultaneous sequence, resulting in multiple canopy temperature monitoring feature sequences and multiple environmental moisture monitoring feature sequences.

[0029] In one embodiment, firstly, based on the spatial extent and crop row structure of each micro-plot, at least one canopy temperature monitoring device is deployed above the cotton field of each micro-plot. This canopy temperature monitoring device is preferably an infrared thermometer, and its installation height is set within a preset height range above the cotton canopy so that its monitoring field of view can cover the cotton canopy area within the corresponding micro-plot and avoid interference from bare soil or adjacent micro-plots. By continuously collecting infrared radiation information from the canopy, canopy temperature data reflecting the cotton's transpiration intensity and water stress status can be obtained in real time. Simultaneously, environmental moisture monitoring devices are deployed within micro-plots corresponding to the canopy temperature monitoring devices. These devices include air humidity sensors for monitoring air humidity within the micro-plot and / or soil moisture sensors for monitoring root zone moisture. The air humidity sensors are positioned at a predetermined height near the cotton canopy to acquire changes in air humidity within the micro-plot, while the soil moisture sensors are buried at a predetermined depth within the cotton root distribution layer to acquire soil moisture content data within the micro-plot. The deployment locations of these environmental moisture monitoring devices are spatially consistent with or adjacent to the corresponding canopy temperature monitoring devices to ensure spatial correspondence of multi-source data. During monitoring, the canopy temperature monitoring devices and environmental moisture monitoring devices synchronously collect data at a unified time sampling interval, constructing canopy temperature monitoring characteristic sequences and environmental moisture monitoring characteristic sequences in chronological order to ensure consistency between the two sequences over time. Through the deployment and synchronous acquisition process of the above-mentioned devices, corresponding canopy temperature monitoring feature sequences and environmental moisture monitoring feature sequences were obtained in multiple micro-plots, providing complete, continuous and spatially directional basic data for subsequent temporal state analysis, multi-source feature fusion and construction of heterogeneous hypermaps of cotton fields.

[0030] Based on the multiple canopy temperature layer monitoring feature sequences and the multiple environmental moisture monitoring feature sequences, a time-series state analysis is performed to construct a cotton field heterogeneous hypergraph, wherein the cotton field heterogeneous hypergraph includes multiple heterogeneous nodes and multiple heterogeneous hyperedge sets.

[0031] In one embodiment, after obtaining multiple canopy temperature layer monitoring feature sequences and multiple environmental moisture monitoring feature sequences, the canopy temperature layer monitoring feature sequences and environmental moisture monitoring feature sequences corresponding to each micro-domain are first preprocessed. Outliers are removed using anomaly detection methods based on threshold discrimination or statistical distribution. Missing values ​​are supplemented using time neighborhood interpolation or sliding window mean interpolation. Normalization is performed using min-max normalization or standard deviation normalization to eliminate the influence of different dimensions and noise on subsequent analysis. Subsequently, the processed canopy temperature layer monitoring feature sequences and environmental moisture monitoring feature sequences are analyzed for time series to determine multiple superimposed canopy temperature layer features and multiple superimposed environmental moisture features. After completing the time series analysis, the canopy temperature time series and environmental moisture time series of each micro-domain within the same time window are subjected to dual consistency iterative screening to form multiple heterogeneous nodes. Subsequently, based on these heterogeneous nodes, heterogeneous unique hyperedges were identified for multiple micro-domain plots. Heterogeneous nodes corresponding to multiple micro-domain plots with the same or similar crown temperature-environmental moisture coupling states within the same time window were connected by a heterogeneous hyperedge, thus forming multiple sets of heterogeneous hyperedges. Finally, based on the multiple sets of heterogeneous nodes and heterogeneous hyperedges, a cotton field heterogeneous hypergraph was constructed. This cotton field heterogeneous hypergraph uses heterogeneous nodes as the node units in the graph and heterogeneous hyperedges as the higher-order connection structure between nodes, providing a structured temporal-spatial representation basis for subsequent intelligent sensing and inference of cotton field micro-domain humidity zoning.

[0032] Furthermore, based on the multiple canopy temperature monitoring feature sequences and the multiple environmental moisture monitoring feature sequences, a time-series state analysis is performed to construct a cotton field heterogeneous hypergraph. This heterogeneous hypergraph includes multiple heterogeneous nodes and multiple sets of heterogeneous hyperedges, including:

[0033] The multiple canopy temperature layer monitoring feature sequences and the multiple environmental moisture monitoring feature sequences are time-series state superimposed to determine multiple canopy temperature layer superimposed features and multiple environmental moisture superimposed features; the multiple canopy temperature layer superimposed features and multiple environmental moisture superimposed features are subjected to dual consistency iterative screening to determine multiple heterogeneous nodes; based on the multiple heterogeneous nodes, the multiple micro-domain plots corresponding to the multiple canopy temperature layer superimposed features and multiple environmental moisture superimposed features are subjected to heterogeneous uniqueness hyperedge identification to determine multiple heterogeneous hyperedge sets; the cotton field heterogeneous hypergraph is constructed based on the multiple heterogeneous nodes and multiple heterogeneous hyperedge sets.

[0034] Preferably, the canopy temperature monitoring feature sequence corresponding to each micro-domain is first superimposed sequentially at consecutive sampling times, fusing the state features of the previous moment with those of the next moment to form a canopy temperature superimposed feature that reflects the trend and evolution of canopy temperature changes. Similarly, the environmental moisture monitoring feature sequence corresponding to each micro-domain is superimposed temporally, fusing the changes in air humidity and / or soil moisture over a continuous time period to form an environmental moisture superimposed feature. Through the above processing, multiple canopy temperature superimposed features and multiple environmental moisture superimposed features are obtained to characterize the comprehensive moisture state characteristics of each micro-domain within a certain time scale. Subsequently, a dual consistency iterative screening is performed on the multiple canopy temperature superimposed features and multiple environmental moisture superimposed features, that is, the canopy temperature superimposed features and environmental moisture superimposed features are jointly represented to measure the consistency of canopy temperature change trends and environmental moisture change characteristics among different micro-domains, and the center of the micro-domain features is continuously updated through iteration. During the iteration process, only feature aggregation results that satisfy the consistency condition in both the canopy temperature layer superposition feature dimension and the environmental moisture superposition feature dimension are retained. This results in multiple stable, dual-consistency feature centers, each corresponding to a heterogeneous node, representing a set of micro-plots with similar moisture state evolution characteristics. Subsequently, heterogeneous uniqueness hyperedge identification is performed based on these heterogeneous nodes. In this process, for each micro-plot, the feature distances between its corresponding canopy temperature layer superposition feature and environmental moisture superposition feature and each heterogeneous node are calculated, and the micro-plot is uniquely associated with the heterogeneous node with the smallest distance. Through this method, a heterogeneous hyperedge is established for each micro-plot and its nearest heterogeneous node, ensuring that each micro-plot in the cotton field heterogeneous hypermap is uniquely associated with only one heterogeneous node, forming a set of heterogeneous hyperedges. Finally, a heterogeneous hypergraph for cotton fields is constructed based on multiple heterogeneous nodes and multiple heterogeneous hyperedge sets. This heterogeneous hypergraph uses heterogeneous nodes as high-order feature nodes and heterogeneous unique hyperedges as the association structure connecting micro-domain plots and heterogeneous nodes. It is used to express the high-order organizational relationships of different micro-domains in cotton fields in terms of canopy temperature, moisture status and their temporal evolution characteristics. This provides a clear, stable and computable structured basis for subsequent intelligent perception and inference of cotton field micro-domain humidity zoning.

[0035] Furthermore, the multiple canopy temperature monitoring feature sequences and the multiple environmental moisture monitoring feature sequences are time-series state superimposed to determine multiple canopy temperature superimposed features and multiple environmental moisture superimposed features, including:

[0036] Extract a first crown temperature layer monitoring feature sequence from the multiple crown temperature layer monitoring feature sequences; extract the first crown temperature layer monitoring feature and the second crown temperature layer monitoring feature from the first crown temperature layer monitoring feature sequence according to time sequence; superimpose the first crown temperature layer monitoring feature and the second crown temperature layer monitoring feature according to time sequence to obtain a first-stage crown temperature layer superimposed feature; superimpose the third crown temperature layer monitoring feature from the first crown temperature layer monitoring feature sequence according to time sequence based on the first-stage crown temperature layer superimposed feature, and so on, to determine the first crown temperature layer superimposed feature, and add the first crown temperature layer superimposed feature to multiple crown temperature layer superimposed features; superimpose the multiple environmental moisture monitoring feature sequences according to time sequence to obtain multiple environmental moisture superimposed features.

[0037] Optionally, firstly, select one canopy temperature monitoring feature sequence corresponding to any micro-plot from multiple canopy temperature monitoring feature sequences as the first canopy temperature monitoring feature sequence. This first canopy temperature monitoring feature sequence consists of canopy temperature features collected in chronological order at multiple times. Subsequently, the first canopy temperature monitoring feature sequence is split in chronological order, and feature values ​​corresponding to adjacent time points are extracted sequentially to obtain the first canopy temperature monitoring feature and the immediately following second canopy temperature monitoring feature. Based on this, the first and second canopy temperature monitoring features are subjected to temporal state superposition processing. That is, by fusing the canopy temperature state at the previous time point with the canopy temperature state at the next time point, a first-stage superimposed canopy temperature feature reflecting the trend of canopy temperature change is obtained. This fusion process can be achieved through feature convolution superposition, weighted fusion, differential fusion, etc., to characterize the direction and magnitude of canopy temperature change in adjacent time periods. Subsequently, based on the obtained first-stage canopy temperature layer superposition features, the third canopy temperature layer monitoring feature in the first canopy temperature layer monitoring feature sequence is subjected to another temporal state superposition. That is, the third canopy temperature layer monitoring feature is fused with the first-stage canopy temperature layer superposition features to obtain an updated canopy temperature layer superposition feature. Then, following the chronological order, the canopy temperature layer monitoring features at each subsequent time point are sequentially superimposed with the currently obtained canopy temperature layer superposition features until the superposition processing of all time points in the first canopy temperature layer monitoring feature sequence is completed, thereby determining the first canopy temperature layer superposition feature corresponding to the micro-region. After completion, the first canopy temperature layer superposition feature is added to multiple canopy temperature layer superposition features. For the canopy temperature layer monitoring feature sequences corresponding to the remaining micro-regions, the above process is repeated to obtain multiple canopy temperature layer superposition features in sequence. Simultaneously, multiple environmental moisture monitoring feature sequences will be processed using the same temporal state overlay method as the canopy temperature layer monitoring feature sequences. That is, the environmental moisture features at adjacent time points in each environmental moisture monitoring feature sequence will be gradually merged and continuously overlaid throughout the entire time series, ultimately obtaining multiple overlaid environmental moisture features that reflect the temporal change trend of micro-domain environmental moisture status. Through this temporal state overlay processing, both the canopy temperature layer overlay features and the environmental moisture overlay features possess temporal continuity and state evolution information, providing a high-quality feature foundation for subsequent dual consistency screening, heterogeneous node determination, and cotton field heterogeneous hypergraph construction.

[0038] Furthermore, the first and second coronal temperature layer monitoring features are time-series superimposed to obtain the first-stage coronal temperature layer superimposed features, including:

[0039] Calculate the fine-grained temporal element similarity between the first and second crown temperature layer monitoring features to obtain an element similarity set; construct a temporal state superposition matrix based on the element similarity set, and use the temporal state superposition matrix to perform feature convolution superposition on the second crown temperature layer monitoring features to obtain the first stage crown temperature layer superposition features.

[0040] Optionally, the first and second canopy temperature monitoring features are first decomposed into fine-grained temporal elements. That is, each canopy temperature monitoring feature is represented as a feature vector composed of multiple temporal elements. Each temporal element corresponds to the local response of canopy temperature in different spatial sampling units or different frequency bands, reflecting the detailed changes in canopy temperature. Then, for each temporal element in the first canopy temperature monitoring feature, a similarity calculation is performed between it and the corresponding temporal element in the second canopy temperature monitoring feature. This similarity can be obtained using methods such as Euclidean distance or cosine similarity. Through the above calculations, an element similarity set consisting of multiple element-level similarity values ​​is formed, used to describe the correspondence and consistency of canopy temperature states at adjacent time points at a fine-grained level. After obtaining the element similarity set, using the temporal elements in the first and second canopy temperature monitoring features as nodes, and the element similarity as the edge weight between nodes, a temporal state superposition matrix reflecting the correlation between canopy temperature features at adjacent time points is constructed. This temporal state superposition matrix is ​​used to characterize the influence intensity and transmission relationship between different fine-grained temporal elements. Building upon this, a graph convolutional network is employed to perform feature convolutional stacking on the second canopy temperature monitoring features. Specifically, the temporal state stacking matrix is ​​viewed as a structural description of a temporal correlation graph. Each row and column of this matrix corresponds to a fine-grained temporal element. The values ​​in the matrix represent the correlation strength and mutual influence between different fine-grained temporal elements at adjacent time points. The larger the matrix value, the more similar the corresponding two temporal elements are during temporal evolution, and their information should have a higher transmission weight during stacking. Next, each fine-grained temporal element in the second canopy temperature monitoring features is treated as a node in the graph. Each node carries the canopy temperature feature information corresponding to that moment, forming the initial node feature representation of the graph convolutional network. During graph convolution calculation, the network uses a single temporal element node as the center and sequentially searches for other nodes with non-zero correlation relationships in the temporal state stacking matrix that are related to that node. These nodes are considered as adjacent nodes of the central node. During feature aggregation, the graph convolutional network weights the feature information of neighboring nodes based on the correlation strength represented in the temporal state superposition matrix. That is, the higher the correlation between neighboring nodes, the greater the proportion of their features in the aggregation process.

[0041] Subsequently, the weighted features of neighboring nodes are combined with the features of the central node itself to participate in feature fusion. This allows the central node to retain its own crown temperature information while absorbing state change information reflected by adjacent temporal elements. After the weighted fusion of neighboring node features, the graph convolutional network performs a nonlinear mapping on the fused features, forming a new representation in the feature space. This mapping process enhances the ability to characterize crown temperature change trends and temporal correlation features, thereby suppressing noise interference and highlighting representative temporal evolution information. After the above processing, each temporal element node generates an updated feature representation. This representation not only contains the original information of the second crown temperature monitoring feature but also incorporates the temporal state influence transmitted by the first crown temperature monitoring feature through the temporal state superposition matrix. As the graph convolutional network completes synchronous computation on all temporal element nodes, the second crown temperature monitoring feature is re-encoded into a new set of feature representations. These features reflect the continuity of crown temperature state at adjacent times in the time dimension and the cooperative change relationship between different temporal elements in the fine-grained element dimension. Ultimately, all the updated node features together constitute the first-stage crown temperature layer superposition features, thereby realizing the joint propagation and enhancement of crown temperature features in the time dimension and fine-grained element dimension, providing a stable and physically meaningful intermediate feature expression for subsequent multi-time-step progressive superposition processing.

[0042] Furthermore, a dual consistency iterative screening is performed on the multiple superimposed features of the canopy temperature layer and the multiple superimposed features of environmental moisture to identify multiple heterogeneous nodes, including:

[0043] The multiple canopy temperature layer superposition features and multiple environmental moisture superposition features are mapped to a two-dimensional space to obtain multiple dual-attribute particles. The horizontal axis of the two-dimensional space represents the canopy temperature layer superposition features, and the vertical axis represents the environmental moisture superposition features. Each dual-attribute particle corresponds to one canopy temperature layer superposition feature and one environmental moisture superposition feature. Multiple initial micro-domain centers in the two-dimensional space are selected from the multiple dual-attribute particles. Based on the multiple initial micro-domain centers, the two-dimensional space is subjected to dual consistency iterative screening to determine multiple heterogeneous nodes.

[0044] Optionally, each micro-plot is first used as a basic unit. Its corresponding canopy temperature layer superposition characteristics are used as the horizontal attribute value in two-dimensional space, and its corresponding environmental moisture superposition characteristics are used as the vertical attribute value, thus forming a dual-attribute particle in two-dimensional space. Each dual-attribute particle is used to comprehensively characterize the overall state of the micro-plot in both canopy temperature evolution and environmental moisture evolution. The positional differences of different micro-plots in two-dimensional space reflect the degree of difference in their water physiological response characteristics. After completing the two-dimensional spatial mapping, a dual-attribute particle is randomly selected from the set of dual-attribute particles in two-dimensional space as the first initial micro-plot center. Candidate particles are then randomly selected from the remaining dual-attribute particles, and the similarity between the candidate particle and the selected initial micro-plot center is calculated. Only when the similarity between the candidate particle and all selected initial micro-plot centers is less than a preset threshold is the candidate particle determined as a new initial micro-plot center. This combination of random selection and similarity constraints ensures that the selected initial micro-plot centers are distributed widely in two-dimensional space, avoiding excessive concentration of initial centers that could affect the subsequent iterative screening effect. After obtaining multiple initial micro-domain centers, each initial micro-domain center is used as a starting point to continuously search for dual-attribute particles with high consistency in both the canopy temperature layer superposition feature dimension and the environmental moisture superposition feature dimension in two-dimensional space. These particles are then aggregated and updated, causing the micro-domain centers to gradually move towards regions of high consistency. During the iteration process, only dual-attribute particles that simultaneously satisfy the similarity of canopy temperature layer superposition feature and environmental moisture superposition feature can be included in the influence range of the current micro-domain center, thus achieving center update under dual consistency constraints. When the iteration process meets the preset stopping condition, the initial micro-domain centers converge into stable target micro-domain centers. Each target micro-domain center corresponds to a feature aggregation result with consistency in both canopy temperature and environmental moisture features. Finally, the multiple target micro-domain centers are determined as multiple heterogeneous nodes to characterize the typical categories of different micro-domains in cotton fields in terms of moisture status and physiological response characteristics, providing a structured node foundation for subsequent heterogeneous hyperedge construction and intelligent micro-domain humidity sensing.

[0045] Furthermore, a dual consistency iterative screening is performed on the multiple superimposed features of the canopy temperature layer and the multiple superimposed features of environmental moisture to identify multiple heterogeneous nodes, including:

[0046] Randomly extract a first initial micro-domain center from the plurality of initial micro-domain centers; iterate the first initial micro-domain center according to a preset iterative filtering bandwidth to determine a first iterative micro-domain center; determine whether the adjacency density of the first iterative micro-domain center is greater than or equal to the adjacency density of the first iterative micro-domain center; if so, continue to iterate the first iterative micro-domain center according to the preset iterative filtering bandwidth until the preset filtering stop condition is met to obtain a first target micro-domain center; take the micro-domain plot corresponding to the first target micro-domain center as the first heterogeneous node, and add the first heterogeneous node into the plurality of heterogeneous nodes.

[0047] Optionally, one initial micro-domain center is randomly selected from multiple initial micro-domain centers as the first initial micro-domain center. This first initial micro-domain center corresponds to a specific location in two-dimensional space, representing a potential combination of canopy temperature layer superposition features and environmental moisture superposition features. Then, using the first initial micro-domain center as the starting point for iteration, a neighborhood search is performed in two-dimensional space according to a preset iteration filtering bandwidth. This iteration filtering bandwidth is used to limit the search range of the current micro-domain center in both the canopy temperature layer superposition feature dimension and the environmental moisture superposition feature dimension; that is, only dual-attribute particles located within this bandwidth are considered candidate neighboring particles. Based on these candidate neighboring particles, their distribution in two-dimensional space is comprehensively calculated. During the calculation, a weighting mechanism based on spatial proximity is introduced; that is, the closer a candidate neighboring particle is to the current micro-domain center in two-dimensional space, the greater its influence on the center update; conversely, the farther away the particle is, the less its influence. In this way, the update direction of the micro-domain center is more dominated by particles in local high-density regions, rather than being interfered with by edge or discrete particles. After determining the influence weights of each candidate neighboring particle, the two-dimensional coordinates of all candidate neighboring particles are aggregated according to their corresponding weights to obtain a new two-dimensional position. This new two-dimensional position reflects the weighted concentration position of dual-attribute particles in space within the current bandwidth range, representing the center of the region where candidate neighboring particles are most dense and consistent. The first initial micro-domain center is then updated based on this weighted concentration position to obtain the first iterative micro-domain center. The position of this first iterative micro-domain center is closer to the region where dual-attribute particles are densely distributed in the current neighborhood than the first initial micro-domain center.

[0048] After obtaining the first iteration micro-domain center, its adjacency density is calculated. This adjacency density characterizes the concentration of dual-attribute particles adjacent to the iteration micro-domain center within a preset iterative screening bandwidth. The adjacency density of the current iteration micro-domain center is then compared with the adjacency density of the previous iteration. When the adjacency density of the current iteration micro-domain center is greater than or equal to that of the previous iteration micro-domain center, it indicates that the micro-domain center is moving towards a region with higher consistency and higher concentration. At this point, the next iteration update continues, using the current iteration micro-domain center as a new starting point and following the same preset iterative screening bandwidth. This iterative process is repeated until a preset screening stop condition is met, such as the number of iterations reaching a preset upper limit. When the screening stop condition is met, the iterative screening process for the initial micro-domain center is terminated, ultimately obtaining a stable first target micro-domain center. After obtaining the first target micro-domain center, the set of micro-domain plots adjacent to the first target micro-domain center in two-dimensional space and satisfying the dual consistency condition is determined as a feature aggregation unit, and the entire set of micro-domain plots corresponding to this feature aggregation unit is taken as the first heterogeneous node. Subsequently, the first heterogeneous node is added to multiple heterogeneous node sets to participate in the subsequent construction of heterogeneous hyperedges and the generation of heterogeneous hypergraphs for cotton fields. For the remaining unprocessed initial micro-domain centers, the above process of random selection, iterative screening, and target center determination is repeated until all initial micro-domain centers are screened, thereby obtaining multiple stable heterogeneous nodes with dual consistency, providing more detailed and scientific support for subsequent micro-domain humidity sensing and water management.

[0049] Furthermore, the preset screening stopping condition is that the number of iterations is greater than or equal to the preset number and / or the adjacency density of the micro-domain center obtained in this iteration is less than the neighborhood density of the micro-domain center obtained in the previous iteration.

[0050] Optionally, during the dual-consistency iterative screening of the initial micro-domain centers, to ensure the controllability and terminatingability of the iteration process and the stability and reliability of the screening results, a screening stop condition is preset. Specifically, after each iteration update, the position of the micro-domain center obtained in this iteration is recorded, and its adjacency density within the preset iterative screening bandwidth is calculated. Subsequently, the adjacency density of this iteration is compared with the adjacency density of the previous iteration. If the adjacency density of the micro-domain center obtained in this iteration is less than that obtained in the previous iteration, it indicates that continuing the iteration will cause the micro-domain center to deviate from the dense region and the screening effect will begin to degrade. At this time, the iteration is stopped immediately, and the micro-domain center obtained in the previous iteration is determined as the target micro-domain center. At the same time, to avoid infinite iteration or iteration oscillation, a maximum number of iterations is also set. When the number of iterations reaches or exceeds the preset number, the iteration is stopped regardless of whether the adjacency density is still increasing, and the micro-domain center obtained in the current iteration is determined as the target micro-domain center. The above stopping condition can be triggered by either reaching a threshold number of iterations or the adjacency density no longer increasing, ensuring that the screening process can not only fully converge to the local high-density region, but also obtain stable micro-domain center results under the premise of controllable computational complexity.

[0051] Multiple monitoring channels are deployed at the multiple heterogeneous nodes, and the cotton field micro-domain humidity zoning is intelligently sensed by combining the multiple heterogeneous hyperedge sets with the multiple heterogeneous hypermaps.

[0052] In one embodiment, multiple monitoring channels are first deployed at multiple heterogeneous nodes. Each heterogeneous node corresponds to a set of micro-plots with high consistency in canopy temperature layer superposition characteristics and environmental moisture superposition characteristics. Therefore, when setting up monitoring channels at heterogeneous nodes, one or more representative monitoring points in the micro-plots associated with the heterogeneous node are used as the monitoring channel locations for that node. These monitoring channels are used to aggregate canopy temperature layer information, environmental moisture information, and their temporal variation characteristics of the micro-plots corresponding to the heterogeneous node, and serve as the state perception entry point for the heterogeneous node in the cotton field heterogeneous supermap. Subsequently, by fusing the monitoring data of multiple micro-plots under the same heterogeneous node, a node-level humidity feature that can characterize the overall humidity state of the heterogeneous node is generated. This fusion process includes a weighted integration of canopy temperature layer superposition characteristics and environmental moisture superposition characteristics, so that the node-level humidity feature simultaneously reflects the crop transpiration state and environmental moisture conditions. After obtaining the node-level humidity characteristics of each heterogeneous node, a holistic perception and inference of the cotton field's heterogeneous hypergraph is performed by combining multiple sets of heterogeneous hyperedges. These sets of heterogeneous hyperedges describe the relationships between different heterogeneous nodes in terms of spatial location, temporal evolution, or moisture response patterns. By propagating information and inferring associations along the heterogeneous hyperedges to the node-level humidity characteristics, the humidity state of a single heterogeneous node can be comprehensively influenced by the states of its associated heterogeneous nodes, thereby enhancing the continuity and stability of the overall perception results. During the information propagation and inference process, the humidity characteristics of interconnected heterogeneous nodes are jointly updated based on the association strength represented by the heterogeneous hyperedges, causing heterogeneous nodes within the same humidity evolution pattern or adjacent spatial regions to tend to form consistent or gradually changing humidity state expressions. Through this method, a high-order structured perception of the overall humidity distribution of the cotton field is achieved. Finally, based on the updated humidity status of each heterogeneous node, the associated micro-domain plots are divided into humidity levels or humidity ranges. Micro-domain plots with the same or similar humidity status are grouped into the same humidity zone, thus forming the cotton field micro-domain humidity zoning perception result. This cotton field micro-domain humidity zoning perception result can reflect the continuous spatial variation characteristics and temporal evolution trend of humidity within the cotton field, providing a reliable basis for subsequent precision irrigation decisions and cotton field water management.

[0053] In summary, the embodiments of this application have at least the following technical effects:

[0054] First, basic spatial information of the target cotton field is acquired and divided into micro-domains, resulting in multiple micro-domain plots. Next, multiple canopy temperature monitoring devices and multiple environmental moisture monitoring devices are deployed above the cotton fields in these micro-domains to perform simultaneous multi-source feature extraction, obtaining multiple canopy temperature monitoring feature sequences and multiple environmental moisture monitoring feature sequences. Then, based on these multiple canopy temperature monitoring feature sequences and multiple environmental moisture monitoring feature sequences, temporal state analysis is performed to construct a heterogeneous hypergraph of the cotton field, which includes multiple heterogeneous nodes and multiple sets of heterogeneous hyperedges. Finally, multiple monitoring channels are deployed at the multiple heterogeneous nodes, and combined with the multiple sets of heterogeneous hyperedges, intelligent sensing of micro-domain humidity zoning is performed on the heterogeneous hypergraph of the cotton field. This invention addresses the technical problem in existing cotton field micro-area humidity monitoring technologies that rely on point-based or single-parameter measurements, making it difficult to achieve large-scale, high-precision, and real-time dynamic sensing. This results in low accuracy and poor spatial continuity in micro-area humidity sensing. The invention achieves the technical effect of constructing a heterogeneous hypergraph of cotton fields by using canopy temperature as the core physiological variable and integrating multi-source time-series data such as environmental moisture to realize spatial sensing and intelligent inference of micro-area humidity, thereby improving the accuracy and stability of micro-area humidity zoning sensing in cotton fields.

[0055] Example 2, based on the same inventive concept as the cotton field micro-area humidity intelligent sensing method combined with canopy temperature layer monitoring in the aforementioned examples, such as... Figure 2 As shown, this application provides an intelligent humidity sensing system for cotton fields combining canopy temperature layer monitoring, wherein the system includes:

[0056] Micro-domain partitioning module 11: Obtains basic spatial information of the target cotton field and performs micro-domain partitioning to obtain multiple micro-domain plots; Feature extraction module 12: Deploys multiple canopy temperature layer monitoring devices and multiple environmental moisture monitoring devices above the cotton fields of multiple micro-domain plots to perform simultaneous multi-source feature extraction to obtain multiple canopy temperature layer monitoring feature sequences and multiple environmental moisture monitoring feature sequences; State analysis module 13: Performs time-series state analysis based on the multiple canopy temperature layer monitoring feature sequences and the multiple environmental moisture monitoring feature sequences to construct a heterogeneous hypergraph of the cotton field, wherein the heterogeneous hypergraph of the cotton field includes multiple heterogeneous nodes and multiple sets of heterogeneous hyperedges; Intelligent sensing module 14: Deploys multiple monitoring channels at the multiple heterogeneous nodes and performs intelligent sensing of cotton field micro-domain humidity partitioning in conjunction with the multiple sets of heterogeneous hyperedges on the heterogeneous hypergraph of the cotton field.

[0057] Furthermore, the micro-domain partitioning module 11 is used to perform the following method:

[0058] The basic spatial information includes cotton field boundary information, irrigation zone information, and soil type distribution information.

[0059] Furthermore, the micro-domain partitioning module 11 is used to perform the following method:

[0060] The target cotton field is divided into zones based on soil consistency, irrigation consistency, and plot boundary consistency to obtain a set of soil consistency zones, a set of irrigation consistency zones, and a set of plot boundary consistency zones. The intersection of the soil consistency zone set, the irrigation consistency zone set, and the plot boundary consistency zone set is obtained to obtain multiple initial micro-domains. The areas in the target cotton field other than the multiple initial micro-domains are merged according to crop row and column structure information and the principle of nearest neighbor similarity to obtain the multiple micro-domains.

[0061] Furthermore, the state resolution module 13 is used to perform the following method:

[0062] The multiple canopy temperature layer monitoring feature sequences and the multiple environmental moisture monitoring feature sequences are time-series state superimposed to determine multiple canopy temperature layer superimposed features and multiple environmental moisture superimposed features; the multiple canopy temperature layer superimposed features and multiple environmental moisture superimposed features are subjected to dual consistency iterative screening to determine multiple heterogeneous nodes; based on the multiple heterogeneous nodes, the multiple micro-domain plots corresponding to the multiple canopy temperature layer superimposed features and multiple environmental moisture superimposed features are subjected to heterogeneous uniqueness hyperedge identification to determine multiple heterogeneous hyperedge sets; the cotton field heterogeneous hypergraph is constructed based on the multiple heterogeneous nodes and multiple heterogeneous hyperedge sets.

[0063] Furthermore, the state resolution module 13 is used to perform the following method:

[0064] Extract a first crown temperature layer monitoring feature sequence from the multiple crown temperature layer monitoring feature sequences; extract the first crown temperature layer monitoring feature and the second crown temperature layer monitoring feature from the first crown temperature layer monitoring feature sequence according to time sequence; superimpose the first crown temperature layer monitoring feature and the second crown temperature layer monitoring feature according to time sequence to obtain a first-stage crown temperature layer superimposed feature; superimpose the third crown temperature layer monitoring feature from the first crown temperature layer monitoring feature sequence according to time sequence based on the first-stage crown temperature layer superimposed feature, and so on, to determine the first crown temperature layer superimposed feature, and add the first crown temperature layer superimposed feature to multiple crown temperature layer superimposed features; superimpose the multiple environmental moisture monitoring feature sequences according to time sequence to obtain multiple environmental moisture superimposed features.

[0065] Furthermore, the state resolution module 13 is used to perform the following method:

[0066] Calculate the fine-grained temporal element similarity between the first and second crown temperature layer monitoring features to obtain an element similarity set; construct a temporal state superposition matrix based on the element similarity set, and use the temporal state superposition matrix to perform feature convolution superposition on the second crown temperature layer monitoring features to obtain the first stage crown temperature layer superposition features.

[0067] Furthermore, the state resolution module 13 is used to perform the following method:

[0068] The multiple canopy temperature layer superposition features and multiple environmental moisture superposition features are mapped to a two-dimensional space to obtain multiple dual-attribute particles. The horizontal axis of the two-dimensional space represents the canopy temperature layer superposition features, and the vertical axis represents the environmental moisture superposition features. Each dual-attribute particle corresponds to one canopy temperature layer superposition feature and one environmental moisture superposition feature. Multiple initial micro-domain centers in the two-dimensional space are selected from the multiple dual-attribute particles. Based on the multiple initial micro-domain centers, the two-dimensional space is subjected to dual consistency iterative screening to determine multiple heterogeneous nodes.

[0069] Furthermore, the state resolution module 13 is used to perform the following method:

[0070] Randomly extract a first initial micro-domain center from the plurality of initial micro-domain centers; iterate the first initial micro-domain center according to a preset iterative filtering bandwidth to determine a first iterative micro-domain center; determine whether the adjacency density of the first iterative micro-domain center is greater than or equal to the adjacency density of the first iterative micro-domain center; if so, continue to iterate the first iterative micro-domain center according to the preset iterative filtering bandwidth until the preset filtering stop condition is met to obtain a first target micro-domain center; take the micro-domain plot corresponding to the first target micro-domain center as the first heterogeneous node, and add the first heterogeneous node into the plurality of heterogeneous nodes.

[0071] Furthermore, the state resolution module 13 is used to perform the following method:

[0072] The preset screening stop condition is that the number of iterations is greater than or equal to the preset number and / or the adjacency density of the micro-domain center obtained in this iteration is less than the neighborhood density of the micro-domain center obtained in the previous iteration.

[0073] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for intelligent sensing of micro-area humidity in cotton fields combined with canopy temperature layer monitoring, characterized in that, The method includes: The basic spatial information of the target cotton field is obtained to divide it into micro-domains, resulting in multiple micro-domain plots. Multiple canopy temperature monitoring devices and multiple environmental moisture monitoring devices were deployed above cotton fields in multiple micro-plots to extract multiple canopy temperature monitoring feature sequences and multiple environmental moisture monitoring feature sequences simultaneously. Based on the multiple canopy temperature layer monitoring feature sequences and the multiple environmental moisture monitoring feature sequences, time-series state analysis is performed to construct a cotton field heterogeneous hypergraph, wherein the cotton field heterogeneous hypergraph includes multiple heterogeneous nodes and multiple heterogeneous hyperedge sets; Multiple monitoring channels are deployed at the multiple heterogeneous nodes, and the cotton field heterogeneous supermap is intelligently sensed for micro-domain humidity zoning by combining the multiple heterogeneous hyperedge sets. This involves obtaining basic spatial information about the target cotton field and dividing it into micro-domains to obtain multiple micro-domain plots, including: The target cotton field is divided into zones based on soil consistency, irrigation consistency, and plot boundary consistency, respectively, to obtain a set of zones with soil consistency, a set of zones with irrigation consistency, and a set of zones with plot boundary consistency. The intersection of the soil consistency partition set, irrigation consistency partition set, and plot boundary consistency partition set is obtained to obtain multiple initial micro-domain plots; The regions in the target cotton field, excluding the multiple initial micro-plots, are merged according to crop row and column structure information and the principle of nearest neighbor similarity to obtain the multiple micro-plots; Specifically, based on the multiple canopy temperature layer monitoring feature sequences and the multiple environmental moisture monitoring feature sequences, a time-series state analysis is performed to construct a heterogeneous hypergraph of cotton fields, including: The multiple canopy temperature layer monitoring feature sequences and the multiple environmental moisture monitoring feature sequences are respectively superimposed in time sequence to determine multiple canopy temperature layer superimposed features and multiple environmental moisture superimposed features; Multiple heterogeneous nodes were identified by performing dual consistency iterative screening on the superimposed features of multiple canopy temperature layers and multiple superimposed features of environmental moisture. Based on the multiple heterogeneous nodes, heterogeneous unique hyperedges are identified for multiple micro-domain plots corresponding to the multiple superimposed features of the multiple canopy temperature layer and the multiple superimposed features of the multiple environmental moisture, and multiple sets of heterogeneous hyperedges are determined. The cotton field heterogeneous hypergraph is constructed based on the multiple heterogeneous nodes and multiple heterogeneous hyperedge sets; Specifically, a dual consistency iterative screening process is performed on the superimposed features of multiple canopy temperature layers and multiple superimposed features of environmental moisture to identify multiple heterogeneous nodes, including: The multiple canopy temperature layer superposition features and multiple environmental moisture superposition features are mapped to a two-dimensional space to obtain multiple dual-attribute particles. The horizontal axis of the two-dimensional space is the canopy temperature layer superposition feature, and the vertical axis is the environmental moisture superposition feature. Each dual-attribute particle corresponds to one canopy temperature layer superposition feature and one environmental moisture superposition feature. Multiple initial micro-domain centers in the two-dimensional space are selected from the plurality of dual-attribute particles; Based on the multiple initial micro-domain centers, a dual consistency iterative screening is performed on the two-dimensional space to determine multiple heterogeneous nodes.

2. The intelligent sensing method for micro-area humidity in cotton fields combined with canopy temperature layer monitoring as described in claim 1, characterized in that, The basic spatial information includes cotton field boundary information, irrigation zone information, and soil type distribution information.

3. The intelligent sensing method for micro-area humidity in cotton fields combined with canopy temperature layer monitoring as described in claim 1, characterized in that, The multiple canopy temperature layer monitoring feature sequences and the multiple environmental moisture monitoring feature sequences are time-series state superimposed to determine multiple canopy temperature layer superimposed features and multiple environmental moisture superimposed features, including: Extract the first coronal temperature layer monitoring feature sequence from the plurality of coronal temperature layer monitoring feature sequences; The first and second coronal temperature layer monitoring features were extracted from the first coronal temperature layer monitoring feature sequence according to time sequence. The first and second crown temperature layer monitoring features are superimposed in a time sequence to obtain the first stage crown temperature layer superimposed features. Based on the first stage crown temperature layer superposition feature, the third crown temperature layer monitoring feature of the first crown temperature layer monitoring feature sequence is superimposed in time sequence, and so on, to determine the first crown temperature layer superposition feature, and the first crown temperature layer superposition feature is added to multiple crown temperature layer superposition features. The time-series state superposition of the multiple environmental moisture monitoring feature sequences is performed to obtain multiple superimposed environmental moisture features.

4. The intelligent sensing method for micro-area humidity in cotton fields combined with canopy temperature layer monitoring as described in claim 3, characterized in that, The first and second canopy temperature layer monitoring features are superimposed in a time sequence to obtain the first-stage canopy temperature layer superimposed features, including: Calculate the fine-grained temporal element similarity between the first and second canopy temperature layer monitoring features to obtain the element similarity set; A temporal state superposition matrix is ​​constructed based on the element similarity set. The second crown temperature layer monitoring features are then superimposed using the temporal state superposition matrix to obtain the first stage crown temperature layer superposition features.

5. A smart humidity sensing system for cotton fields combining canopy temperature layer monitoring, characterized in that, For implementing the intelligent humidity sensing method for micro-domains in cotton fields combined with canopy temperature layer monitoring as described in any one of claims 1-4, the system comprises: Micro-domain partitioning module: Obtains basic spatial information of the target cotton field to partition it into micro-domains, resulting in multiple micro-domain plots; Feature extraction module: Multiple canopy temperature monitoring devices and multiple environmental moisture monitoring devices are deployed above cotton fields in multiple micro-plots to perform simultaneous multi-source feature extraction, thereby obtaining multiple canopy temperature monitoring feature sequences and multiple environmental moisture monitoring feature sequences; State analysis module: Based on the multiple canopy temperature layer monitoring feature sequences and the multiple environmental moisture monitoring feature sequences, time-series state analysis is performed to construct a cotton field heterogeneous hypergraph, wherein the cotton field heterogeneous hypergraph includes multiple heterogeneous nodes and multiple heterogeneous hyperedge sets; Intelligent sensing module: Multiple monitoring channels are deployed at the multiple heterogeneous nodes, and the cotton field heterogeneous hypermap is intelligently sensed by micro-domain humidity zoning in combination with the multiple heterogeneous hyperedge sets.

Images