Soil drought monitoring method based on dynamic optimization of multi-source data

Abstract

The application discloses a soil drought monitoring method based on dynamic optimization of multi-source data, relates to the technical field of soil monitoring, and comprises the following steps: delimiting a target region according to a preset rule, and constructing a multi-dimensional drought feature vector by using multi-source observation data. The probability of each cause type is obtained by calculating the Mahalanobis distance between the vector and the preset vector set centroid corresponding to different stress cause types. The same cause probability difference between the target region and the adjacent region is compared. If all the difference values are less than a preset mutation threshold, the soil drought operation path graph is generated by traversing and connecting. If there is a difference value greater than or equal to the threshold, mutation detection and resegmentation are performed on the adjacent region, and the new region corresponding to the minimum probability difference is taken as a new target for iterative calculation, so that the monitoring path is dynamically optimized, the monitoring range and path can be dynamically adjusted according to the regional heterogeneity of drought characteristics, and adaptive tracking and visual operation guidance of the soil humidity condition under complex drought causes are realized.

Description

Technical Field

[0001] This invention relates to the field of soil monitoring technology, and more specifically, this application relates to a soil drought monitoring method based on dynamic optimization of multi-source data. Background Technology

[0002] In the agricultural recovery following flood receding in monsoon climate zones, achieving accurate monitoring and causal diagnosis of soil drought presents significant challenges. Existing technologies have limitations in this complex scenario, and their monitoring results often fail to accurately guide differentiated agronomic decisions.

[0003] Current monitoring methods primarily rely on the analysis and threshold determination of single or limited indicators such as apparent vegetation indices or land surface temperature obtained from remote sensing. These methods are essentially based on a simple, static correlation between "symptoms" and "drought" in stable environments. However, in the initial stages of flood receding, farmland undergoes a dramatic, unsteady recovery process. The stunted growth or elevated canopy temperature exhibited by crops may be caused by a variety of distinct factors, such as soil physical water shortage, root waterlogging and oxygen deficiency, or nutrient loss. Traditional methods cannot effectively separate these conflicting signals, often attributing stresses of different causes to water deficit, leading to misjudgments. More importantly, existing methods generally neglect the rapid dynamic changes in groundwater levels—a crucial environmental driver that profoundly affects the rate of soil water receding and the effective water supply to crop roots. Ignoring this dynamic process introduces inherent biases into the assessment of drought conditions and trends. Furthermore, fixed judgment rules cannot adapt to the rapid changes in field conditions after flooding, and their rigid analytical paradigms are unable to capture the spatiotemporal heterogeneity and evolutionary dynamics of stress. They also cannot identify drought areas that may be spatially continuous or abruptly distributed due to different driving forces, thus failing to provide efficient action guidelines with spatial logic for recovery operations such as irrigation and fertilization.

[0004] To address the aforementioned issues, there is an urgent need in this field to develop a novel monitoring technology capable of deeply integrating multi-source observation information, quantitatively analyzing key environmental drivers, and providing precise diagnosis and decision support amidst dynamic changes. In summary, the core deficiency of existing technologies lies in their static and isolated analytical models, which, when faced with the complex farmland environment after flood recedes, cannot distinguish the intrinsic causes of stress, let alone provide differentiated operational pathways to guide specific agricultural practices. Summary of the Invention

[0005] To address the aforementioned technical problems, this technical solution provides a soil drought monitoring method based on dynamic optimization using multi-source data, thus resolving the issues raised in the background section.

[0006] In a first aspect, embodiments of this application provide a soil drought monitoring method based on dynamic optimization of multi-source data, comprising the following steps: S1, delineating a target area according to preset rules and acquiring multi-source observation data of the target area, and constructing a multi-dimensional drought feature vector based on the multi-source observation data; S2, acquiring a preset vector set and calculating the Mahalanobis distance between the centroids of the multi-dimensional drought feature vector and the preset vector set, and analyzing the probability of the stress cause type corresponding to the preset vector set; S3, calculating the difference in the probability of the same stress cause type between the target area and adjacent areas, and obtaining the probability difference of the same stress cause type; S4, determining whether the probability difference of all stress cause types is less than a preset mutation threshold: if the determination result is yes, then taking the target area as the center, Iterate through all probability differences, select the smallest positive value, and record the adjacent regions of the target region corresponding to the smallest positive value as the target regions to be connected; if the judgment result is negative, count the number of adjacent regions whose probability difference between the target region and any stress cause type of the adjacent region is greater than or equal to the preset mutation threshold, and re-segment the adjacent regions accordingly to obtain new adjacent regions of the target region. Record the new adjacent regions corresponding to the smallest positive value as the target regions to be connected; S5, obtain the connection vector between the geometric center of the target region and the target regions to be connected; S6, take the target regions to be connected as the new target regions, repeat steps S1 to S5, and connect all the connection vectors in sequence, which are recorded as the soil drought operation path. Generate a soil drought operation path map based on the soil drought operation path.

[0007] Secondly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned soil drought monitoring method based on dynamic optimization of multi-source data.

[0008] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0009] 1. By integrating multi-source observation data such as remote sensing spectroscopy, surface temperature, and radar-retrieved soil moisture, and incorporating the groundwater depth change rate to perform geometric transformations on the original multidimensional stress feature vector based on angle and scale parameters, the generated multidimensional drought feature vector can more sensitively and comprehensively reflect the soil moisture state under different stress causes. The probability of cause type is obtained by calculating the Mahalanobis distance between this vector and the centroid of a preset vector set, achieving quantitative differentiation of the dominant causes of drought. This solves the technical problem of insufficient identification of complex drought causes by traditional single-index monitoring, and improves the accuracy and cause identification capability of soil drought monitoring.

[0010] 2. By comparing the probability difference between the target area and adjacent areas for the same stress cause, and using a preset mutation threshold for judgment, the system can intelligently identify the spatial mutation boundaries of drought characteristics. When characteristics change continuously, it automatically generates soil drought operation paths connecting various areas; when encountering a mutation, it re-segments adjacent areas through a series of operations, including counting the number of mutation neighbors, performing equiangular pre-segmentation, and dynamically calculating the delineated area based on the standard deviation of Mahalanobis distance within the sub-region, and selects the new area corresponding to the minimum probability difference to continue. This mechanism enables the monitoring path to bypass highly heterogeneous and complex mutation areas, extending along the most continuous and significant direction of drought characteristic evolution. It solves the problems of low efficiency and inability to focus on the core drought development trend of traditional fixed grid or random path monitoring, and achieves adaptive optimization of the monitoring path and efficient resource allocation.

[0011] 3. Within a preset monitoring time window, by continuously tracking the number of newly added target areas to be connected and comparing it with a stable change threshold, the preset area of ​​the target areas is dynamically adjusted, thereby adapting to the overall rate of change in the regional drought situation without disrupting existing operational paths. By comparing the difference in new changes over consecutive time windows with an improvement judgment threshold, the effectiveness of intervention measures can be automatically determined, triggering corresponding continuous solidification operations or manual early warnings. This makes the entire monitoring not just a one-time path planning, but a dynamic optimization closed loop with feedback adjustment capabilities. It solves the technical problems of static monitoring schemes being unable to adapt to the spatiotemporal dynamic evolution of drought and lacking effectiveness evaluation and early warning capabilities, enhancing the dynamic stability and early warning function of monitoring, and designing a time-series-based update mechanism. Attached Figure Description

[0012] Figure 1 A schematic diagram illustrating the steps of the soil drought monitoring method based on dynamic optimization of multi-source data provided in this application embodiment;

[0013] Figure 2 This is a schematic diagram of the logic flow of the soil drought monitoring method based on dynamic optimization of multi-source data provided in the embodiments of this application. Detailed Implementation

[0014] This application's embodiments address the technical problem in existing technologies that, when faced with complex farmland environments after flood receding, static and isolated analysis models cannot distinguish the intrinsic causes of stress, and are even less able to provide differentiated operational paths to guide specific agricultural operations, through a soil drought monitoring method based on dynamic optimization of multi-source data.

[0015] To address the core technical challenges in soil drought monitoring, such as the difficulty in fusing multi-source data, the complexity and significant spatial heterogeneity of drought causes, and the inability of traditional static methods to dynamically track drought evolution, the approach begins with deep data fusion and feature reconstruction. Traditional single indicators cannot comprehensively characterize drought conditions; therefore, it is essential to integrate multi-source observational data, including remote sensing spectroscopy, surface temperature, and radar-retrieved soil moisture, standardizing them and constructing a multi-dimensional stress feature vector. To enhance the identification of different drought-driving mechanisms, the rate of change in groundwater depth, reflecting deep water conditions, is introduced as a regulating factor. Based on the direction and magnitude of this change, corresponding geometric transformation parameters are identified, and the original feature vector is rotated and scaled to generate a multi-dimensional drought feature vector that better separates different drought causes.

[0016] After obtaining the feature vectors that characterize the overall state and potential causes, it is necessary to quantify their matching degree with typical drought causal patterns. To this end, a predefined set of vectors, formed from expert knowledge or historical data, is defined for each type of stress causation. By calculating the Mahalanobis distance from the feature vectors to the centroids of these sets, the probability of each feature vector belonging to a particular causal type can be assessed, and the relative distance can be transformed into an intuitive probability of causal type. In this way, each region obtains a set of probability values, achieving a quantitative analysis of complex drought conditions.

[0017] Based on the above probability distribution, the focus shifts to spatial analysis and path decision-making. The spatial development of drought often exhibits continuity or abrupt changes. This spatial relationship can be detected by calculating the probability differences of the target area and all its adjacent areas for the same cause. If all differences are less than a preset abrupt change threshold, it indicates that the drought characteristics are spatially gradual. In this case, taking the current area as the center, a smooth and continuous soil drought management path is automatically generated by finding the adjacent areas with the smallest probability differences and connecting them sequentially.

[0018] When the probability difference between certain adjacent regions exceeds the mutation threshold, it indicates the encounter with a drought characteristic or dominant cause boundary. At this point, simple connections will traverse regions with excessive heterogeneity, resulting in suboptimal paths. Therefore, these mutation boundaries require precise identification and region reshaping. The number of adjacent regions satisfying the mutation conditions is counted, and this number is used as a guide to pre-segment the original adjacent regions at equal angles. Next, the dispersion of the feature vectors within each pre-segmented sub-region is calculated, and the final area of ​​each sub-region is dynamically adjusted based on the standard deviation of its Mahalanobis distance. Regions with larger standard deviations indicate more inconsistent internal features, and their defined areas are correspondingly reduced. After re-segmentation, a series of new target regions to be connected are obtained. The probability difference with these new regions is recalculated, and the region with the smallest difference is selected as the next target, thus bypassing the feature mutation boundary and achieving adaptive optimization of the monitoring path in complex spatial patterns.

[0019] To adapt the entire process to the evolving drought situation, a dynamic update mechanism was introduced. Within a preset time window, the number of newly added target areas to be connected is continuously tracked. If this number exceeds a stable change threshold, indicating an accelerated drought expansion, the initial area baseline is automatically expanded proportionally to accommodate broader dynamic changes without interrupting or resetting existing paths. In subsequent time windows, the difference in the number of newly added areas is further compared. If this difference is less than an improvement judgment threshold, indicating that the adjustment measures are effective and the situation is stabilizing, the new area baseline is fixed for a period. If the difference remains significant, a manual warning is automatically triggered, suggesting that external intervention may be necessary.

[0020] Ultimately, all the path connection vectors obtained through adaptive optimization are integrated to form a complete soil drought operation path. This path must be transformed into an intuitive and visual soil drought operation path map to have practical application value. This map translates abstract data analysis and spatial decision-making results into concrete operational guidance. It not only clearly displays the types of drought causes and their probability intensity in different regions in the form of colored grids, making the spatial differentiation and dominant factors of drought readily apparent, but also directly guides monitoring personnel, irrigation equipment, or drones to move along the optimal sequence through overlaid highlighted paths and directional arrows. Therefore, generating a soil drought operation path map is the final output and a necessary step in the entire data processing logic. It solves the problems of abstract results and poor operability in traditional monitoring methods, ultimately transforming complex multi-source data analysis into an executable and navigable dynamic operation map, achieving a key leap from data to decision-making, and from analysis to action.

[0021] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0022] like Figure 1The diagram illustrates the steps of a soil drought monitoring method based on dynamic optimization of multi-source data provided in this application embodiment, including the following steps: S1. Delineate the target area according to preset rules and acquire multi-source observation data of the target area, constructing a multi-dimensional drought feature vector based on the multi-source observation data; S2. Obtain a preset vector set and calculate the Mahalanobis distance between the centroids of the multi-dimensional drought feature vector and the preset vector set, and analyze the probability of the stress cause type corresponding to the preset vector set; S3. Calculate the difference in probability of the same stress cause type between the target area and adjacent areas to obtain the probability difference of the same stress cause type; S4. Determine whether the probability difference of all stress cause types is less than a preset mutation threshold: if the determination result is yes, then take the target area as the center... S1. Iterate through all probability differences, select the smallest positive value, and record the adjacent regions of the target region corresponding to the smallest positive value as the target regions to be connected; if the judgment result is negative, count the number of adjacent regions whose probability difference between the target region and any stress cause type of the adjacent regions is greater than or equal to the preset mutation threshold, and re-segment the adjacent regions accordingly to obtain new adjacent regions of the target region. Record the new adjacent regions corresponding to the smallest positive value as the target regions to be connected; S5. Obtain the connection vector between the geometric center of the target region and the target regions to be connected; S6. Using the target regions to be connected as the new target regions, repeat steps S1 to S5, and connect all the connection vectors in sequence, which are recorded as the soil drought operation path. Generate a soil drought operation path map based on the soil drought operation path.

[0023] By deeply fusing multi-source observation data and constructing a multidimensional drought feature vector, the limitations of traditional methods relying on single indicators can be effectively overcome, enabling refined identification and quantification of the causes of soil drought stress. Introducing Mahalanobis distance and causal type probability allows for a more accurate assessment of the impact of different stress factors. Furthermore, by dynamically detecting probability abrupt changes between regions and iteratively optimizing the approach, it can adapt to the rapid changes in the farmland environment after flood recede, identify spatially logical drought areas, and generate instructive soil drought management pathways. This provides precise and efficient action guidance for differentiated agronomic decisions in agricultural recovery in monsoon climate zones. Figure 2 This is a schematic diagram of the logic flow of the soil drought monitoring method based on dynamic optimization of multi-source data provided in the embodiments of this application.

[0024] Furthermore, the specific construction process of the multidimensional drought feature vector is as follows: extracting red band reflectance, near-infrared band reflectance, and red edge band reflectance from remote sensing spectral data; extracting surface temperature values ​​from surface temperature data; extracting topsoil volumetric water content from radar-retrieved soil moisture data; standardizing and normalizing the red band reflectance, near-infrared band reflectance, red edge band reflectance, surface temperature values, and topsoil volumetric water content respectively, and arranging them in a predetermined order to form the original multidimensional stress feature vector; and performing geometric transformations on the original multidimensional stress feature vector to obtain the multidimensional drought feature vector.

[0025] In this embodiment, red light band reflectance, near-infrared band reflectance, and red edge band reflectance are extracted from remote sensing spectral data to obtain key spectral information reflecting the health status and growth status of surface vegetation. Red light band reflectance is closely related to vegetation chlorophyll absorption, near-infrared band reflectance is mainly affected by vegetation cell structure, while red edge band reflectance shows high sensitivity to vegetation water stress and physiological changes. These spectral features are important indicators for assessing vegetation water status and drought stress levels, and can be extracted from multispectral or hyperspectral image data acquired by satellite remote sensing platforms after preprocessing steps such as radiometric calibration and atmospheric correction.

[0026] Extracting surface temperature values ​​from surface temperature data aims to obtain thermodynamic parameters reflecting the surface energy balance and water evapotranspiration. Under arid soil conditions, reduced vegetation transpiration leads to decreased surface heat dissipation, resulting in increased surface temperature. Therefore, surface temperature values ​​can serve as an effective indicator of soil moisture deficit and vegetation heat stress. Surface temperature data is typically acquired through thermal infrared remote sensing imagery, such as using thermal infrared band data, which is then processed through radiometric conversion and atmospheric correction before being retrieved.

[0027] Extracting topsoil volumetric water content from radar-retrieved soil moisture data aims to directly obtain the physical quantity of soil moisture status. Radar remote sensing technology, especially synthetic aperture radar, has the ability to penetrate the earth's surface and is sensitive to the soil's dielectric constant, thus effectively reflecting soil moisture content. It is also less affected by weather conditions such as clouds and fog, enabling all-weather observation. Topsoil volumetric water content is one of the most direct and critical indicators in soil drought monitoring, and it can be obtained from satellite radar data using specific radar retrieval models.

[0028] The extracted red-band reflectance, near-infrared-band reflectance, red-edge-band reflectance, surface temperature, and surface soil volumetric water content were standardized and normalized, then arranged in a predetermined order to form the original vector of multidimensional stress features. Because these multi-source data have different dimensions, numerical ranges, and statistical distributions, direct combination could lead to certain features with larger numerical ranges dominating in subsequent calculations, masking the influence of other important features. Standard normalization eliminates dimensional differences, transforming all features to a uniform scale, making them comparable, and ensuring that the contribution of each feature to the multidimensional vector is fair and effective. Arranging them in a predetermined order ensures the structural consistency of the vector, facilitating subsequent processing.

[0029] Geometric transformations are performed on the original multidimensional stress feature vectors to obtain multidimensional drought feature vectors. These geometric transformations aim to further optimize the expressive power of the feature vectors, making them more accurately reflect the complex state of soil drought. Such transformations can include operations such as rotation, scaling, and projection, with the goal of adjusting the direction and length of the vectors in the multidimensional feature space to highlight key information related to drought stress or eliminate redundant information, thereby enhancing the discriminative power of the features. For example, principal component analysis can be used to project the original vectors into a new coordinate system, ensuring that the new coordinate axes align with the direction of maximum data variance, thus reducing dimensionality and removing correlations; or linear discriminant analysis can be used to find the optimal projection direction, ensuring that data points with different drought levels are separated as much as possible in the new space.

[0030] Through the extraction, standardization, and geometric transformation processes of the aforementioned multi-source data, this application effectively integrates remote sensing spectral, surface temperature, and radar-retrieved soil moisture data from different physical backgrounds and dimensions into a unified and physically meaningful multidimensional drought feature vector. This refined vector construction method overcomes the feature imbalance and information redundancy problems that may result from directly combining heterogeneous data, ensuring the accuracy and reliability of subsequent Mahalanobis distance calculations. By performing geometric transformations on the original features, the feature vector's ability to characterize and distinguish soil drought conditions is further enhanced, enabling the causal type probabilities calculated in subsequent steps to more accurately reflect the actual drought stress type and its degree in the target area. This provides a more solid and accurate data foundation for subsequent abrupt change area detection and operational path planning, thereby significantly improving the accuracy and dynamic optimization capabilities of the entire soil drought monitoring method.

[0031] Furthermore, the specific process for obtaining the multidimensional drought feature vector also includes: obtaining the groundwater depth change rate and, based on the positive and negative signs and magnitude of the groundwater depth change rate, searching for the corresponding transformation parameters from a preset transformation parameter mapping table. The transformation parameters include an angle parameter for vector rotation and a scale parameter for vector scaling. Using the angle and scale parameters, a geometric transformation is performed on the original multidimensional stress feature vector to generate the multidimensional drought feature vector. The geometric transformation is as follows: ,in, Represents a multidimensional drought feature vector. Indicates the scale parameter. Indicates the angle parameter. This is a five-dimensional rotation matrix constructed using angle parameters. Representing the original vector of the multidimensional stress feature, this rotation matrix is ​​constructed by the sequential product of a set of Givens rotation matrices, which distributes the angle parameters to the rotation operations of multiple two-dimensional subplanes, thereby achieving the rotation transformation of the vector in the five-dimensional feature space.

[0032] In this embodiment, the rate of change of groundwater depth is an important indicator reflecting groundwater dynamics. Its positive or negative sign indicates the rising or falling trend of the groundwater level, while its magnitude reflects the speed and magnitude of the change. In soil drought monitoring, changes in groundwater level have a direct or indirect impact on soil moisture status, especially under drought stress, where a drop in groundwater level may exacerbate soil moisture deficit. A preset transformation parameter mapping table can be established in advance, storing the mapping relationship between the rate of change of groundwater depth in different ranges and the corresponding angle and scale parameters.

[0033] For example, when the rate of change of groundwater depth is negative and the decrease is significant, it may correspond to a larger angle parameter and a smaller scale parameter. This allows for a more significant adjustment of the drought eigenvector in the feature space, reflecting more severe drought stress. This mapping table allows for flexible selection of appropriate transformation parameters based on real-time groundwater dynamics, enabling subsequent geometric transformations to better adapt to current environmental conditions. The angle parameter adjusts the orientation of the eigenvector in multidimensional space, while the scale parameter adjusts the length or amplitude of the eigenvector.

[0034] After obtaining the original multidimensional stress feature vector, a geometric transformation is performed on it using angle and scale parameters retrieved from a pre-defined transformation parameter mapping table. This transformation is not merely a simple linear scaling or translation, but combines rotation and scaling operations to finely adjust the original feature vector, enabling it to more accurately represent the actual state of soil drought. Through this dynamic adjustment, the final generated multidimensional drought feature vector can better reflect the impact of groundwater dynamics on soil drought, thereby improving the accuracy of drought monitoring.

[0035] The Givens rotation matrix is ​​a special type of orthogonal matrix that can perform rotations on two-dimensional subplanes defined by any two coordinate axes in high-dimensional space. By cleverly distributing angular parameters across multiple two-dimensional subplanes—for example, decomposing the angular parameters into multiple sub-angles and applying them to different two-dimensional subplanes—comprehensive rotation transformations of vectors can be achieved in a five-dimensional feature space. This refined rotation operation allows feature vectors to be adjusted in direction within the feature space based on dynamic information about the rate of change of groundwater depth, thus more accurately reflecting the interactions between features of different dimensions and their contribution to drought.

[0036] By introducing the rate of change of groundwater depth as the basis for dynamically adjusting the geometric transformation, and by using a preset transformation parameter mapping table to obtain angle parameters and scale parameters, this application can achieve refined geometric transformation of the original vector of multidimensional stress features.

[0037] Specifically, by adjusting the magnitude of the vector through a scaling parameter and adjusting its orientation through a five-dimensional rotation matrix constructed based on the product of Givens rotation matrices, the resulting multidimensional drought feature vector can more accurately and dynamically reflect the actual state of soil drought. This dynamic adaptability allows the drought feature vector to incorporate not only surface observation data but also the impact of groundwater dynamics on soil moisture, thereby improving the sensitivity and accuracy of drought monitoring. Especially in areas where groundwater has a significant impact on soil moisture, it can more effectively identify and assess the degree and type of drought stress.

[0038] Furthermore, the specific process for obtaining the causal type probability is as follows: a preset vector set corresponds to a stress causal type, the preset vector set contains several preset vector combinations, and each preset vector combination represents a different quantification dimension; the Mahalanobis distance between the multidimensional drought feature vector and the centroid of the preset vector set is calculated to obtain the multidimensional reference distance; the sum of all multidimensional reference distances is recorded as the total distance; the multidimensional reference distance of a single preset vector combination corresponding to a stress causal type is recorded as the causal distance; the sum of the causal distances of different preset vector combinations is divided by the ratio of the total distance and recorded as the causal type probability of that stress causal type; thus, the causal type probabilities of different stress causal types in the target region are obtained.

[0039] In this embodiment, the present application further proposes a specific process for obtaining the probability of causal types. Specifically, a preset vector set corresponds to a stress causal type, and this preset vector set contains several preset vector combinations, where each preset vector combination aims to represent a different quantitative dimension. For example, for the stress causal type of "water deficit," its preset vector set may contain preset vector combinations that focus on different quantitative dimensions such as soil moisture, vegetation health status, and surface temperature. This multi-dimensional combination method makes the representation of stress causal types more comprehensive and detailed, and can capture the combined impact of different factors on drought stress.

[0040] Based on this, the Mahalanobis distance between the multidimensional drought feature vector and the centroid of the preset vector set is calculated to obtain the multidimensional reference distance. The multidimensional drought feature vector is a comprehensive feature representation obtained after processing the actual observation data of the current target area. The centroid of the preset vector set represents the average or typical state of this specific stress cause type in the multidimensional feature space. Calculating the Mahalanobis distance between the multidimensional drought feature vector and this centroid can effectively measure the similarity or deviation between the drought situation of the current target area and the typical state of a specific stress cause type. The Mahalanobis distance considers the correlation between the various dimensions of the data and can more accurately reflect the actual distance between multidimensional data points, thus obtaining a comprehensive "multidimensional reference distance" for subsequent probability calculations.

[0041] The sum of all multidimensional reference distances is recorded as the total distance. When calculating the causal type probability of a specific stress cause type, a benchmark is needed to measure the relative importance of that cause type relative to all possible cause types. The multidimensional reference distances calculated for each of the preset vector sets are summed to obtain a "total distance". This total distance represents the total deviation of the drought characteristics of the current target area from the typical states of all known stress cause types, providing a denominator for subsequent normalization calculations.

[0042] The multidimensional reference distance corresponding to a single preset vector combination for a stress cause type is denoted as the causal distance. To more precisely assess the contribution of a specific stress cause type to the current drought situation, it is necessary to consider the quantization dimensions represented by different "preset vector combinations" within that cause type. For a specific stress cause type, each preset vector combination within it will have a Mahalanobis distance calculated with the multidimensional drought feature vector; this distance is denoted as the "causal distance." This causal distance reflects the degree of matching between the drought characteristics of the target region and the stress cause type in a specific quantization dimension.

[0043] The sum of the causal distances for different preset vector combinations, divided by the total distance, is denoted as the causal type probability for that stress causal type. This step is the core of calculating the causal type probability. For a specific stress causal type, the causal distances corresponding to all "preset vector combinations" within it are summed to obtain the total causal distance for that stress causal type. Then, this total causal distance is divided by the previously calculated "total distance." This ratio is defined as the "causal type probability" for that stress causal type. By repeating the above calculation process, the causal type probabilities for different stress causal types in the target area can be obtained.

[0044] By employing the aforementioned technical solution, this application, when calculating the probability of stress causal types, no longer relies solely on a single Mahalanobis distance. Instead, it introduces several preset vector combinations contained in a preset vector set, each representing a different quantification dimension. By calculating the causal distance between the multidimensional drought feature vector and each preset vector combination, and summing these distances as the total causal distance for that stress causal type, and then comparing this sum with the total distance for all stress causal types, a more refined and discriminative probability of causal type is obtained. This method can more comprehensively capture the contribution of different quantification dimensions to specific stress causal types, effectively solving the problem that simple Mahalanobis distance calculation may not fully reflect the complex causal differences. Therefore, this application can more accurately identify the complex causes of soil drought in target areas, providing a more reliable quantitative basis for subsequent drought early warning, operational path planning, and precision management, significantly improving the accuracy and practicality of soil drought monitoring.

[0045] Further, the specific process of re-segmenting adjacent regions is as follows: Count the number of adjacent regions of the target region corresponding to any stress cause type whose probability difference is greater than or equal to a preset mutation threshold, denoted as the mutation neighborhood number; pre-segment the target region based on the number of mutation neighborhoods with the geometric center of all adjacent regions as the origin, obtaining pre-segmented sub-regions equal to the number of mutation neighborhoods in the adjacent region; for each pre-segmented sub-region, calculate the standard deviation of the Mahalanobis distance between the multidimensional drought feature vector of any discrete point within the sub-region and the centroid of the preset vector set; sum and average the standard deviations of all sub-regions of the adjacent region to obtain the sum of the average standard deviations of the adjacent regions; multiply the preset adjustment coefficient by the sum of the products of the standard deviations of the sub-regions, add 1, and use this as the area component; divide the preset baseline area by the area component to obtain the designated area of ​​the sub-region; use the minimum value among the designated areas of all sub-regions as the final designated area, which is used as the area of ​​the new adjacent region; segment all adjacent regions of the target region according to the final designated area to obtain the new adjacent regions of the target region, denoted as the target region to be selected.

[0046] In this embodiment, when performing the traversal connection judgment operation, the probability differences of all stress cause types need to be checked one by one. These probability differences reflect the degree of difference between the target area and its various adjacent areas in different stress cause types. The traversal operation ensures a comprehensive consideration of all potential connection possibilities. During the traversal, the smallest positive probability difference is selected. A positive value indicates that the probability of a certain stress cause type in the adjacent area is higher than that in the target area, which may indicate that the drought impact has increased or spread in that direction. Selecting the smallest positive value aims to prioritize connecting adjacent areas with the smallest difference in drought conditions from the target area but still showing some degree of increasing trend, which helps to construct a smooth and gradual operation path. After determining the smallest positive value, its corresponding adjacent area is selected as a candidate area to be connected to the target area in the current step, i.e., the target area to be connected. Once the target area to be connected is determined, the geometric center of the current target area is connected to the geometric center of the target area to be connected through a connection vector. This connection vector not only indicates the direction from the target area to the target area to be connected in space, but also represents a basic segment of the operation path. The geometric center can usually be calculated from the boundary coordinates of the region, for example, by taking the average latitude and longitude coordinates of the region.

[0047] To construct a complete operational path, this method employs a recursive or iterative process. After determining the first connection vector, the region just selected as the "target region to be connected" is promoted to a new "target region." Then, using this new target region as the center, the above-mentioned traversal, selection of the minimum positive value, determination of new target regions to be connected, and generation of connection vectors are repeated. This repetitive mechanism allows the operational path to gradually extend along the progressive direction of drought impact. Through the above iterative process, a series of continuous connection vectors are generated. Each vector connects the previous "target region" and the currently selected "target region to be connected," collectively forming the segments of the operational path. This process can continue until a preset termination condition is met, such as reaching a preset path length, covering a preset area range, or no longer having adjacent regions that meet the condition. Finally, all connection vectors generated during the iteration process are concatenated end-to-end in the order of their generation. Since each connection vector indicates the direction from one region to the next, connecting them sequentially forms a continuous path, which is the final generated soil drought operational path. This route is designed to provide spatial guidance for drought monitoring and response, such as indicating recommended routes for patrol, sampling, or irrigation operations.

[0048] By employing the aforementioned technical solution, when the probability differences between the stress cause types of the target area and adjacent areas are relatively small, indicating a relatively uniform drought condition, this application can construct a progressive soil drought operation path by traversing all probability differences and selecting the adjacent area corresponding to the smallest positive value as the next connection point. This selection strategy based on the minimum positive value difference allows the operation path to extend along the gradient direction with the gentlest and most natural drought impact or spread trend, avoiding skipping or random path planning in areas with similar drought conditions. By iteratively using the target area to be connected as the new target area and repeating the above connection judgment operation, it can be ensured that the generated path is continuous and logically progressive. Finally, by sequentially connecting all connection vectors, a clear and operable soil drought operation path is formed, providing precise spatial guidance for subsequent drought monitoring, assessment, and intervention measures, improving operational efficiency and targeting, avoiding blind operations, and thus optimizing drought response strategies.

[0049] Furthermore, the specific process of re-segmenting adjacent regions also includes: re-acquiring and calculating the difference between the probability of the same stress cause type between the target region and the target region to be selected, to obtain the probability difference of the same stress cause type; selecting the target region to be selected corresponding to the smallest absolute value of the probability difference as the target region to be connected; connecting the geometric centers of the target region and the target region to be connected through a connection vector, and taking the target region to be connected as the new target region.

[0050] In this embodiment, when a drought mutation is detected between the target region and its adjacent regions, the number of adjacent regions of the target region whose probability difference for any stress cause type is greater than or equal to a preset mutation threshold is first counted, and this number is recorded as the mutation neighborhood number. This mutation neighborhood number reflects the local density of drought mutation phenomena around the target region.

[0051] Using the geometric center of all adjacent regions of the target region as the origin, the adjacent regions are pre-segmented at equal angles based on the number of mutation neighbors. For example, if the number of mutation neighbors is N, the adjacent region can be divided into N sector-shaped sub-regions around its geometric center, with each sub-region having an equal angle. This pre-segmentation method can divide complex adjacent regions into several smaller sub-regions with a certain degree of spatial continuity, laying the foundation for subsequent refined analysis and ensuring the uniformity and accuracy of the segmentation.

[0052] Subsequently, for each pre-segmented sub-region, the standard deviation of the Mahalanobis distance between the multidimensional drought feature vector of any discrete point within that sub-region and the centroid of a predefined vector set is calculated. Mahalanobis distance measures the distance between a sample point and the distribution center, taking into account the correlation between data dimensions. The standard deviation reflects the dispersion or variability of drought features within that sub-region. A larger standard deviation indicates more uneven drought characteristics within that sub-region, potentially suggesting more complex drought conditions or more pronounced drought boundaries.

[0053] To comprehensively assess the variability of the entire adjacent region, the standard deviation of each sub-region within the adjacent region is calculated. These standard deviations are then summed and averaged to obtain a mean standard deviation representing the overall variability of the adjacent region. This mean standard deviation comprehensively reflects the heterogeneity of drought characteristics within the entire adjacent region, providing a quantitative basis for subsequent area delineation.

[0054] Next, the designated area of ​​the sub-region is dynamically adjusted based on the variability of drought characteristics. Specifically, a preset adjustment coefficient is multiplied by the standard deviation of the sub-region, and then 1 is added to obtain the area component. Then, a preset baseline area is divided by this area component to obtain the designated area of ​​the sub-region. This calculation method ensures that sub-regions with high drought characteristic variability receive smaller designated areas, thus achieving more refined segmentation; while sub-regions with low drought characteristic variability receive larger designated areas, avoiding over-segmentation.

[0055] After calculating the delineated areas of all sub-regions, the minimum value is selected as the final delineated area. Using the minimum value as the final delineated area aims to ensure that even the most complex and variable local areas with the greatest drought characteristics can be segmented sufficiently within the entire adjacent region, thereby guaranteeing the accuracy of the segmentation and sensitivity to abrupt drought changes.

[0056] Based on the final defined area, all adjacent regions of the target region are re-divided to obtain new adjacent regions of the target region, which are then recorded as the target regions to be connected. This is key to dynamically adjusting the region boundaries. Through refined segmentation, the specific range of drought mutations can be identified more accurately, providing more precise spatial units for subsequent connection and path planning.

[0057] After obtaining the new target regions to be connected, it is necessary to re-acquire the causal type probabilities of the same stress cause type between the target region and the target region to be connected, and calculate their differences. This recalculation ensures that, based on the finely segmented new regions, the drought differences between them and the target regions can be accurately assessed, providing the latest and most accurate data for selecting the best connection regions.

[0058] From the recalculated probability differences, the target region corresponding to the probability difference with the smallest absolute value is selected as the selected adjacent region for connection. Selecting the difference with the smallest absolute value means that the drought conditions of this region are most similar to those of the target region or have the gentlest transition, making it an ideal choice for connection and path extension, and helping to construct a smoother and more reasonable drought operation path.

[0059] Finally, the geometric centers of the target region and the selected adjacent region are connected by a connection vector, and the selected adjacent region is used as the new target region. This completes the local extension of the drought monitoring path and shifts the monitoring focus to the next most suitable region, thereby achieving dynamic optimization of drought monitoring and gradual path construction.

[0060] Through the above technical solution, this application can perform more refined and intelligent segmentation of adjacent regions when drought abrupt change regions are detected. By statistically counting the number of abrupt change neighborhoods, the local density of drought abrupt changes can be quantified; based on this, equiangular pre-segmentation is performed to divide adjacent regions into smaller sub-regions. By calculating the standard deviation of the Mahalanobis distance from the drought feature vector to the preset centroid within each sub-region, the heterogeneity of drought conditions within the sub-region can be accurately assessed. Based on these standard deviations, the delineated area of ​​each sub-region is dynamically adjusted, so that regions with high drought feature variability are segmented more finely, while regions with low variability are avoided from over-segmentation. By selecting the minimum value among the delineated areas of all sub-regions as the final delineated area, the sensitivity and segmentation accuracy of drought abrupt change regions are ensured. This refined segmentation method enables more accurate identification of drought boundaries and internal heterogeneity, thereby selecting the target region to be connected that is closest to the drought conditions of the target region. This effectively solves the problem of insufficient accuracy in segmenting drought abrupt change regions using traditional methods, providing more accurate and reasonable spatial units for the construction of subsequent soil drought operation paths, and significantly improving the accuracy of drought monitoring and the optimization effect of operation paths.

[0061] Furthermore, the update process for the soil drought operation path is as follows: within a preset monitoring time window, the number of target areas to be connected is continuously monitored and recorded as the new quantity; it is determined whether the new quantity exceeds a preset stable change threshold; if the new quantity exceeds the stable change threshold, the percentage difference between the new quantity and the stable change threshold is calculated and recorded as the threshold exceedance ratio; the preset delineated area in the preset rules is expanded proportionally based on the threshold exceedance ratio to obtain the expanded preset delineated area; without modifying the existing soil drought operation path, the size of the delineated target area is regenerated based on the expanded preset delineated area to obtain the modified target area; and the soil drought operation path is generated again based on the modified target area.

[0062] In this embodiment, the preset monitoring time window refers to the time interval used for periodically assessing changes in soil drought conditions, such as a day, a week, or a month. Within this time window, data is continuously collected and processed to monitor the dynamic changes in drought-stricken areas. The target region to be connected refers to the adjacent regions identified as requiring further connection and processing during the abrupt change region detection and segmentation process.

[0063] Continuous monitoring of changes in the number of these areas can directly reflect the expansion or contraction trends of arid regions. The newly added number specifically refers to the total number of newly identified target areas to be connected within the current preset monitoring time window. The preset stable change threshold is a reference value used to determine whether changes in arid regions are significant.

[0064] This threshold can be set based on historical drought data, expert experience, or statistical analysis results; for example, it can be a specific quantitative value or a percentage. When the increase exceeds this threshold, it indicates a significant change in the drought situation that warrants attention.

[0065] The exceedance ratio is an indicator that quantifies the severity of changes in arid regions. It is calculated as the percentage difference between the newly added area and the stable change threshold. The higher the ratio, the more drastic the changes in the arid region. The pre-defined delineated area serves as the basis for initially delineating the target region. When the exceedance ratio indicates a significant expansion of the arid region, the original pre-defined delineated area will be expanded proportionally based on this ratio.

[0066] For example, a functional relationship can be set such that the larger the exceedance ratio, the larger the expansion factor, thus obtaining a pre-defined delineated area that can better cover the expansion of the current drought area. This operation aims to preserve existing soil drought monitoring routes that have been generated and may be in operation during previous monitoring periods, without altering them. This means that the effectiveness of already determined monitoring areas and routes is recognized to a certain extent, avoiding unnecessary duplication of calculations and operational interruptions, and ensuring the continuity of operations.

[0067] Based on the expanded pre-defined area, the boundaries and size of the target region will be recalculated and redefined. This modified target region will encompass a wider area to accommodate the actual expansion of the drought zone. Based on the new, expanded, modified target region, steps S1 through S7 of the soil drought monitoring method will be re-executed to generate a new soil drought operational path that better reflects the current drought extent.

[0068] By introducing the aforementioned soil drought monitoring path update process, this application effectively addresses the problem of insufficient adaptability of existing methods in dealing with dynamically changing soil drought areas. Specifically, it continuously monitors the quantity changes of the target areas to be connected within a preset monitoring time window and compares them with a preset stable change threshold, enabling real-time perception of the dynamic expansion or contraction trend of drought areas. When the newly added quantity exceeds the stable change threshold, it indicates a significant change in drought conditions. At this point, the severity of this change can be quantified by calculating the threshold excess ratio. Based on this threshold excess ratio, the preset delineated area in the preset rules can be dynamically expanded, resulting in a preset delineated area that more accurately reflects the expansion of the current actual drought range. Furthermore, without altering the existing soil drought monitoring path, the continuity and efficiency of the operation are ensured. Simultaneously, a new soil drought monitoring path is generated based on the modified target area, allowing the new path to more accurately cover the actual drought range and avoiding monitoring blind spots or resource waste caused by inappropriate initial delineation areas. This dynamic adjustment mechanism significantly improves the robustness and adaptability of soil drought monitoring methods, enabling them to guide drought prevention and control operations more accurately and promptly, thereby enhancing the overall monitoring and management effectiveness.

[0069] Furthermore, the update process for the soil drought operation path also includes: if the modified target area is acquired again in the next preset monitoring time window after the modified target area, the difference between the number of new additions corresponding to the preset monitoring time window and the next preset monitoring time window is calculated and recorded as the new change difference; if the new change difference is less than the preset improvement judgment threshold, the expanded preset delineated area is fixed and continues for a preset duration; if the new change difference is greater than or equal to the improvement judgment threshold, an artificial warning is issued for the target area.

[0070] In this embodiment, the number of new additions counted within the monitoring time window that leads to the expansion of the target area is recorded, and the number of new additions is counted again in the next monitoring time window after area adjustment. By comparing the number of new additions in these two consecutive time windows, a difference can be obtained, which reflects the increase or decrease in the rate of drought area expansion. For example, if the difference is negative, it indicates that the drought expansion rate has slowed down or the area has shrunk; if the difference is positive, it indicates that the drought expansion rate is still accelerating.

[0071] To effectively assess the aforementioned changes in the number of newly added areas, this application introduces a preset improvement judgment threshold. This threshold is a pre-defined value used to define the degree of change in the number of newly added areas, thereby determining whether the drought situation has reached the standard for improvement or deterioration. This threshold can be flexibly configured based on historical drought data, expert experience, regional characteristics, or management objectives. For example, it can be set as an absolute value, such as a reduction of 5 newly added areas; or it can be set as a relative value, such as a reduction of 20% in the number of newly added areas. Its purpose is to provide an objective basis for decision-making, avoid subjective judgment, and enable automated subsequent processing.

[0072] If the aforementioned difference in the newly added area is less than the preset improvement threshold, it indicates that the trend of drought expansion has been effectively curbed, or even improved, after the target area area adjustment. In this case, the currently expanded preset area will be fixed and remain unchanged for a preset duration. The purpose of this is to stabilize the monitoring range and avoid frequent area adjustments due to slight fluctuations, thereby improving operational efficiency and stability. The preset duration can be set according to the duration of drought, monitoring frequency, or management strategy. For example, it can be set to several weeks or a complete drought monitoring cycle to ensure continuous observation of stable areas over a period of time.

[0073] Conversely, if the aforementioned increase in the difference in drought level is greater than or equal to the preset improvement threshold, it means that even after expanding the designated area of ​​the target region, the trend of drought expansion has not been effectively controlled, and may even worsen further. This indicates that the drought situation may have exceeded the scope of automatic adjustment and requires human intervention. In this case, a manual warning will be issued to the target region. This warning can be implemented in various ways, such as sending emails or SMS notifications to designated managers, or displaying a high-priority alarm on the monitoring platform interface. The warning information should include key data such as the geographical information of the target region, the current drought level, and the increase in the difference in drought level, so that managers can quickly understand the situation and take corresponding countermeasures, such as on-site inspections and resource allocation.

[0074] By introducing an evaluation mechanism for changes in the number of newly added target areas within subsequent monitoring windows, this application can dynamically determine whether the drought situation is improving or continuing to worsen. When the number of newly added target areas to be connected is detected to be stable or decreasing, the current designated area can be intelligently fixed and maintained for a period of time, thereby avoiding unnecessary frequent adjustments and improving the stability and efficiency of monitoring. Conversely, if the number of newly added areas continues to increase, it indicates that the drought situation may be beyond the scope of automatic adjustment. In this case, timely manual warnings can be issued, prompting management personnel to quickly intervene, conduct manual analysis and intervention, effectively preventing the drought disaster from further expanding, and ensuring the timeliness and effectiveness of drought monitoring and response.

[0075] Furthermore, the specific process for generating the soil drought operation path map is as follows: the target area is mapped onto the display raster layer according to the preset mapping rules; the display raster layer is color-rendered to generate a color raster image that represents the type of stress cause and the corresponding probability; on the color raster image, the soil drought operation path is displayed with a highlighted line overlay, and the operation start point is marked at the starting point of the path, and arrow symbols are added along the path extension direction to indicate the recommended operation direction.

[0076] In this embodiment, the target area is mapped onto the display raster layer according to a preset mapping rule. Specifically, this step aims to transform abstract geographic regional information into visualized image data. The target area can be a pre-divided grid cell, an administrative division, or an irregular area dynamically determined based on aridity characteristics. The preset mapping rule typically includes geographic coordinate system transformation, projection method selection, and image resolution setting, ensuring that the geometry and relative position of the target area can be accurately presented on the display raster layer. For example, Mercator projection or Gauss-Kruger projection can be used, and the latitude and longitude coordinates of the target area can be converted into pixel coordinates on the display device.

[0077] Next, the displayed raster layer is color-rendered to generate a color raster image representing the types of stress causes and their corresponding probabilities. This step visually displays detailed information about soil drought. Color rendering assigns unique colors or hues to different stress causes and adjusts the saturation or brightness of the colors according to their probabilities. For example, water stress areas may be rendered in blue tones, with a deeper blue as the probability increases; nutrient stress areas may be rendered in yellow tones, with a brighter yellow as the probability increases. In this way, users can easily identify the type, severity, and spatial distribution of drought.

[0078] On this color raster image, a soil drought operation path is displayed as a highlight line overlay. To emphasize the operation path and ensure its clear visibility against the complex background of the color raster image, this application employs a highlight line overlay. The highlight line can be a thick line, a line with a glowing effect, or a line using a color that strongly contrasts with the background. This overlay method ensures the visual salience of the operation path and prevents it from being obscured by background information.

[0079] Simultaneously, a job start point marker is placed at the beginning of the path. This marker clearly indicates the starting position of the recommended job. This marker can be a specific graphic symbol, a text label, or an icon with a specific meaning. By placing this marker at the geometric starting point of the path, users can quickly locate the starting position of the job, thereby improving job execution efficiency.

[0080] In addition, arrows are added along the path to indicate the recommended work direction. The arrows provide users with clear guidance on the work sequence and direction. These arrows can be evenly distributed along the work path or added at each turning point or key section of the path. The arrows eliminate user confusion about the work direction, ensuring that the work is carried out along the predetermined optimized path, thereby improving the accuracy and effectiveness of the work.

[0081] Abstract soil drought monitoring results and operational paths are transformed into intuitive and easily understandable visual information. Specifically, by mapping target areas onto a display raster layer and rendering them in color, users can clearly identify the spatial distribution of different stress causes and their probabilities, thus gaining a comprehensive understanding of the drought situation. Based on this, the soil drought operational path is displayed as a highlighted line overlay, supplemented by starting point markers and directional arrows, making the recommended operational plan more visually prominent and clear. This not only greatly improves the efficiency of users' understanding of drought information but also provides precise and actionable navigation guidance for actual agricultural operations, effectively avoiding operational deviations caused by ambiguous information or unclear directions, thereby significantly improving the accuracy and efficiency of soil drought operations.

[0082] This application also provides a computer-readable storage medium for storing a computer program, which, when executed by a processor, implements the steps of a soil drought monitoring method based on dynamic optimization of multi-source data.

[0083] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0084] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0085] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0086] These computer program instructions can also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0087] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0088] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A soil drought monitoring method based on dynamic optimization of multi-source data, characterized in that, Includes the following steps: S1. Delineate the target area according to preset rules, obtain multi-source observation data of the target area, and construct a multi-dimensional drought feature vector based on the multi-source observation data; S2. Obtain a preset vector set and calculate the Mahalanobis distance between the centroid of the multidimensional drought feature vector and the preset vector set, and analyze the probability of the cause type of stress corresponding to the preset vector set. S3. Calculate the difference in probability between the target area and adjacent areas for the same stress cause type to obtain the probability difference for the same stress cause type. S4. Determine whether the probability difference of all stress cause types is less than the preset mutation threshold: If the judgment result is yes, then take the target area as the center, traverse all probability differences, select the smallest positive value, and record the adjacent area of ​​the target area corresponding to the smallest positive value as the target area to be connected. If the judgment result is negative, count the number of neighboring regions whose probability difference between the target region and any stress cause type of the adjacent region is greater than or equal to the preset mutation threshold, and re-segment the adjacent regions accordingly to obtain new neighboring regions of the target region. The new neighboring region corresponding to the minimum positive value is recorded as the target region to be connected. S5. Obtain the connection vector between the geometric centers of the target region and the target regions to be connected; S6. Using the target area to be connected as the new target area, repeat steps S1 to S5, and connect all the connection vectors in sequence, which are denoted as the soil drought operation path. Generate the soil drought operation path map based on the soil drought operation path.

2. The soil drought monitoring method based on dynamic optimization of multi-source data according to claim 1, characterized in that, The specific construction process of the multidimensional drought feature vector is as follows: Extract red band reflectance, near-infrared band reflectance, and red edge band reflectance from remote sensing spectral data; Extracting surface temperature values ​​from surface temperature data; Extracting topsoil volumetric water content from radar-retrieved soil moisture data; The reflectance of the red band, the reflectance of the near-infrared band, the reflectance of the red edge band, the surface temperature value, and the volumetric water content of the surface soil were standardized and arranged in a predetermined order to form the original vector of multidimensional stress characteristics. A geometric transformation is performed on the original vector of multidimensional stress features to obtain a multidimensional drought feature vector.

3. The soil drought monitoring method based on dynamic optimization of multi-source data according to claim 2, characterized in that, The specific process for obtaining the multidimensional drought feature vector also includes: Obtain the rate of change of groundwater depth and, based on the sign and magnitude of the rate of change of groundwater depth, find the corresponding transformation parameters from the preset transformation parameter mapping table. The transformation parameters include an angle parameter for vector rotation and a scale parameter for vector scaling. By performing geometric transformations on the original vector of multidimensional stress features using angle and scale parameters, a multidimensional drought feature vector is generated. The geometric transformation is as follows: ,in, Represents a multidimensional drought feature vector. Indicates the scale parameter. Indicates the angle parameter. This is a five-dimensional rotation matrix constructed using angle parameters. Representing the original vector of the multidimensional stress feature, this rotation matrix is ​​constructed by the sequential product of a set of Givens rotation matrices, which distributes the angle parameters to the rotation operations of multiple two-dimensional subplanes, thereby achieving the rotation transformation of the vector in the five-dimensional feature space.

4. The soil drought monitoring method based on dynamic optimization of multi-source data according to claim 1, characterized in that, The specific process for obtaining the probability of the cause type is as follows: A set of preset vectors corresponds to a type of stress cause. The set of preset vectors contains several combinations of preset vectors, and each combination of preset vectors represents a different quantification dimension. Calculate the Mahalanobis distance between the multidimensional drought feature vector and the centroid of the preset vector set to obtain the multidimensional reference distance; The sum of all multidimensional reference distances is recorded as the total distance. The multidimensional reference distance corresponding to a single preset vector combination for the stress cause type is denoted as the cause distance; The sum of the causal distances of different preset vector combinations, divided by the ratio of the total distances, is recorded as the causal type probability of that stress causal type. Obtain the causal type probability of different stress causal types in the target region.

5. The soil drought monitoring method based on dynamic optimization of multi-source data according to claim 1, characterized in that, The specific process of re-segmenting adjacent regions is as follows: The number of adjacent regions of the target region corresponding to any stress cause type whose probability difference is greater than or equal to the preset mutation threshold is counted and denoted as the mutation neighborhood number. Using the geometric center of all adjacent regions of the target region as the origin, perform equal-angle pre-segmentation based on the number of mutation neighbors to obtain pre-segmented sub-regions with the number of mutation neighbors of the adjacent regions. For each pre-segmented sub-region, calculate the standard deviation of the Mahalanobis distance between the multidimensional drought feature vector of any discrete point within the sub-region and the centroid of the preset vector set; The sum of the standard deviations of all subregions of an adjacent region is calculated by averaging the sums of their standard deviations. The sum of the products of the preset adjustment coefficient and the standard deviation of the sub-region, plus 1, is used as the area component. The preset reference area is divided by the area component to obtain the designated area of ​​the sub-region. The minimum value among the defined areas of all sub-regions is taken as the final defined area, and the final defined area is used as the area of ​​the new adjacent region. Based on the final defined area, all adjacent areas of the target area are divided to obtain new adjacent areas of the target area, which are denoted as the target area to be selected.

6. The soil drought monitoring method based on dynamic optimization of multi-source data according to claim 5, characterized in that, The specific process of re-segmenting adjacent regions also includes: The probability difference of the same stress cause type is calculated by re-acquiring the target area and calculating the difference between the probability of the same stress cause type and the probability of the target area to be selected. Select the target region corresponding to the absolute value of the smallest probability difference as the target region to be connected; Connect the geometric centers of the target region and the target region to be connected by a connection vector, and then use the target region to be connected as the new target region.

7. The soil drought monitoring method based on dynamic optimization of multi-source data according to claim 1, characterized in that, The update process for the soil drought operation path is as follows: Within the preset monitoring time window, the number of target areas to be connected obtained through continuous monitoring is recorded as the number of new additions. Determine whether the number of new additions exceeds a preset stable change threshold; If the number of new additions exceeds the stable change threshold, the percentage difference between the number of new additions and the stable change threshold is calculated and recorded as the threshold exceedance ratio. The predefined area in the predefined rules is expanded by using the over-threshold ratio as a proportion to obtain the expanded predefined area. Without altering the existing soil drought operation path, the size of the target area is regenerated based on the expanded preset delineated area, resulting in a modified target area. Soil drought operation paths are then generated based on the modified target area.

8. The soil drought monitoring method based on dynamic optimization of multi-source data according to claim 7, characterized in that, The updating process for the soil drought operation path also includes: If the modified target area is acquired again in the next preset monitoring time window after the modified target area, the difference between the number of new additions corresponding to the preset monitoring time window and the next preset monitoring time window is calculated and recorded as the difference in the number of new additions. If the newly added change difference is less than the preset improvement judgment threshold, the expanded preset area will be fixed and will continue for a preset duration. If the newly added change difference is greater than or equal to the improvement judgment threshold, a manual warning is issued for the target area.

9. The soil drought monitoring method based on dynamic optimization of multi-source data according to claim 1, characterized in that, The specific process for generating the soil drought operation path map is as follows: Map the target area onto the display raster layer according to the preset mapping rules; Color rendering is performed on the displayed raster layer to generate a color raster image that represents the type of stress cause and the corresponding probability. The soil drought operation path is displayed in a highlighted linear overlay on the color raster image, with the start point of the operation marked at the beginning of the path and arrows added along the path to indicate the recommended operation direction.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-9.

Images