Prediction model training method and system based on land survey data

Abstract

The embodiments of the present application provide a prediction model training method and system based on land survey data. First, a land survey dataset including historical land attribute distribution data and real-time land status record data is obtained. Then, the historical land attribute distribution data is spatially gridded, and the real-time land status record data is timestamp aligned. Then, cross-modal feature fusion is performed to generate a land spatial feature map. Abnormal fluctuation nodes in the land spatial feature map are extracted to determine a target training sample set. Finally, based on the target training sample set, an initial prediction model is iteratively trained in multiple stages to obtain a land status prediction model. This can effectively improve the accuracy of land status prediction and provide a scientific basis for land management.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more specifically, to a prediction model training method and system based on land surveying data. Background Art

[0002] Accurately predicting changes in land conditions is crucial for land management and planning. With increasing environmental change and the impact of human activities, soil properties, vegetation cover, and other aspects of land are facing significant uncertainty. For example, irrational agricultural practices and accelerated urbanization can lead to soil degradation, impacting vegetation growth and ultimately altering the balance of the entire ecosystem.

[0003] Traditional land assessment methods often rely on limited sample data and simple statistical analysis, making it difficult to accurately and dynamically predict land conditions. Modern surveying technologies, however, can acquire vast amounts of land measurement data, including historical land attribute distribution data and real-time land condition records. This data enables the construction of more accurate land condition prediction models. However, effectively utilizing this multi-source, multi-modal data remains a key challenge. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a prediction model training method and system based on land surveying data.

[0005] In conjunction with the first aspect of the present application, a prediction model training method based on land measurement data is provided, which is applied to a prediction model training system based on land measurement data. The method comprises:

[0006] Obtain a land survey dataset for the target area, comprising historical land attribute distribution data and real-time land status record data. The historical land attribute distribution data includes soil composition information and terrain characteristic parameters for multiple geographic blocks, and the real-time land status record data includes surface temperature distribution and humidity change trajectories collected by current environmental monitoring equipment.

[0007] Performing spatial grid division on the historical land attribute distribution data to generate attribute coding units with hierarchical association relationships, and performing timestamp alignment processing on the real-time land status record data to obtain a dynamic state sequence corresponding to the attribute coding units;

[0008] Performing cross-modal feature fusion on the attribute coding unit and the dynamic state sequence to generate a land spatial feature map, wherein each node in the land spatial feature map represents the attribute evolution trend of the corresponding geographical block within a preset time range;

[0009] Extracting abnormal fluctuation nodes from the land spatial characteristic map, and determining a target training sample set based on soil composition information and surface temperature distribution corresponding to the abnormal fluctuation nodes;

[0010] The initial prediction model is subjected to multi-stage iterative training based on the target training sample set to obtain a land status prediction model, wherein the land status prediction model is used to output soil degradation risk assessment results and vegetation cover change probability in a future time period based on the input geographic block identifier.

[0011] In a possible implementation of the first aspect, performing spatial grid division on the historical land attribute distribution data to generate attribute coding units with hierarchical association relationships includes:

[0012] Dividing the target area into a plurality of geographic grid cells of equal area according to the latitude and longitude coordinate ranges in the historical land attribute distribution data;

[0013] For each of the geographic grid cells, extract the soil pH distribution, organic matter content gradient, and mineral concentration change curve contained therein to generate a soil attribute vector corresponding to the geographic grid cell;

[0014] Based on the differences in soil property vectors between adjacent geographic grid cells, a multi-level spatial correlation network is constructed, in which high-level network nodes represent the mean of large-scale soil properties, and low-level network nodes represent the fluctuation characteristics of local soil properties.

[0015] The network nodes at each level in the spatial association network are mapped and associated with corresponding geographic grid units to generate the attribute coding units.

[0016] In a possible implementation of the first aspect, performing cross-modal feature fusion on the attribute encoding unit and the dynamic state sequence to generate a land spatial feature map includes:

[0017] Performing frequency domain transformation on the surface temperature distribution data in the dynamic state sequence to extract temperature fluctuation patterns at different time scales;

[0018] Aligning the temperature fluctuation pattern with the soil property vector of the corresponding geographic grid cell in the time domain to generate a hybrid feature vector integrating the soil-temperature correlation feature;

[0019] Determining feature diffusion weights between nodes based on a transmission path of the hybrid feature vector in a multi-level spatial association network;

[0020] A spatial interpolation calculation is performed on the mixed feature vector based on the feature diffusion weight to generate a continuous feature distribution map covering the entire target area, and the continuous feature distribution map is converted into graph structure data to form the land spatial feature map.

[0021] In a possible implementation of the first aspect, extracting abnormal fluctuation nodes in the land spatial characteristic map includes:

[0022] Calculating the characteristic change rate of each node in the land spatial characteristic map, and constructing an anomaly detection threshold based on the difference in characteristic change rates between adjacent nodes;

[0023] Identifying nodes whose characteristic change rate exceeds the anomaly detection threshold, and tracing back their soil composition mutation records in historical land attribute distribution data;

[0024] The nodes that meet both the characteristic rate anomaly and soil composition mutation are marked as the abnormal fluctuation nodes, and their corresponding latitude and longitude coordinates and timestamp information are recorded.

[0025] In a possible implementation of the first aspect, the performing multi-stage iterative training on the initial prediction model based on the target training sample set includes:

[0026] Dividing the target training sample set into a training subset and a validation subset, wherein the training subset contains geographic block data with clear soil degradation labels, and the validation subset contains geographic block data without soil state change labels;

[0027] In a first training phase, the parameters of the initial prediction model are initialized using the training subset, and initial prediction weights are generated based on the correlation between soil composition information and surface temperature distribution;

[0028] In the second training phase, the validation subset is input into the initialized prediction model, and error back propagation is performed based on the output vegetation cover change probability and real-time satellite image data to adjust the dynamic parameters in the initial prediction weights;

[0029] In the third training phase, the adjusted prediction model is deployed to the edge computing device, the prediction results are calibrated online through the real-time received environmental sensor data, and the final parameters of the land state prediction model are updated.

[0030] In a possible implementation of the first aspect, the online calibration of the prediction result using the environmental sensor data received in real time includes:

[0031] Get the soil moisture sensor readings and light intensity monitoring values ​​of the target geographic block in the current time period;

[0032] Comparing the soil moisture sensor reading with the vegetation cover change probability output by the land state prediction model to determine a moisture deviation coefficient;

[0033] Performing weighted correction on the humidity deviation coefficient according to the light intensity monitoring value to generate an environmental calibration factor;

[0034] The environmental calibration factor is superimposed on the output layer of the land state prediction model to dynamically adjust the soil degradation risk assessment results in the future time period.

[0035] In a possible implementation of the first aspect, the method further includes:

[0036] When the soil degradation risk assessment results of the target geographical area are detected to exceed the preset threshold, a list of recommended land management options is generated;

[0037] Extract the resource consumption data and expected effect data required for each plan in the recommended list of land management plans, and build a multi-objective optimization decision-making model;

[0038] According to the currently available resource inventory data and environmental carrying capacity constraints, the optimal land management plan is selected from the multi-objective optimization decision model and output to the user terminal.

[0039] In a possible implementation of the first aspect, generating a recommended list of land management solutions includes:

[0040] Query the historical land management case database for successful case data that matches the current soil degradation type;

[0041] Analyze the treatment technology paths, material ratio parameters and implementation cycle information used in the successful case data;

[0042] Combined with the terrain slope data and water source distribution data of the target geographical block, the above-mentioned treatment technology path is adaptively modified to generate multiple candidate treatment plans with priority ranking;

[0043] The candidate management plans are evaluated for matching with real-time weather forecast data, and plans that do not meet climate conditions are eliminated to form a recommended list of land management plans.

[0044] In combination with the second aspect of the present application, a prediction model training system based on land measurement data is provided. The prediction model training system based on land measurement data includes a machine-readable storage medium and a processor. The machine-readable storage medium stores machine-executable instructions. When the processor executes the machine-executable instructions, the prediction model training system based on land measurement data implements the aforementioned prediction model training method based on land measurement data.

[0045] In conjunction with the third aspect of the present application, a computer-readable storage medium is provided, in which computer-executable instructions are stored. When the computer-executable instructions are executed, the aforementioned prediction model training method based on land surveying data is implemented.

[0046] Combining any of the above aspects, first, a land measurement dataset containing historical land attribute distribution data and real-time land status record data is obtained, then the historical land attribute distribution data is spatially gridded, and the real-time land status record data is timestamp aligned. Subsequently, cross-modal feature fusion is performed to generate a land spatial feature map, and abnormal fluctuation nodes in the land spatial feature map are extracted to determine the target training sample set. Finally, based on the target training sample set, the initial prediction model is iteratively trained in multiple stages to obtain a land status prediction model, thereby effectively improving the accuracy of land status prediction and providing a scientific basis for land governance. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained by combining these drawings without paying any creative work.

[0048] Figure 1 A flowchart of a prediction model training method based on land survey data provided in an embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0050] The terms "first," "second," and so on, in the specification and claims of the present invention and the accompanying drawings are used to distinguish between different items, not to describe a specific order. Furthermore, the terms "including," "having," and any variations thereof, are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product, or end comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed therein, or may optionally include other steps or elements inherent to such process, method, product, or end.

[0051] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present invention. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute a separate or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0052] Figure 1 The flowchart of the prediction model training method based on land survey data provided by an embodiment of the present application is shown. It should be understood that in other embodiments, the order of some steps in the prediction model training method based on land survey data of this embodiment can be shared based on actual needs, or some steps can be omitted or maintained. The details of the prediction model training method based on land survey data include:

[0053] Step S110: Acquire a land survey dataset for the target area. The land survey dataset includes historical land attribute distribution data and real-time land status record data. The historical land attribute distribution data includes soil composition information and terrain characteristic parameters for multiple geographic blocks. The real-time land status record data includes surface temperature distribution and humidity change trajectory collected by current environmental monitoring equipment.

[0054] In a real-world scenario, assume the target area is a large agricultural land covering multiple geographical areas. Historical land attribute distribution data includes soil composition information, including the proportions of various minerals, pH, and organic matter content. For example, in certain geographical areas, the nitrogen, phosphorus, and potassium content are 15 mg / kg, 10 mg / kg, and 20 mg / kg, respectively, with a pH of 6.5 and an organic matter content of 3%. Topographical parameters include altitude, slope, and aspect. For example, some areas have an altitude of 200 meters, a slope of 5 degrees, and a southeasterly orientation. Regarding real-time land status data, environmental monitoring equipment is widely distributed throughout this agricultural land. Surface temperature distribution data can be collected by multiple temperature sensors, which record temperatures at regular intervals (e.g., hourly), generating temperature distribution data at different geographic locations. For example, at 10:00 AM, the temperature near the irrigation water source is 22°C, while the temperature at the edge of the farmland near the woods is 20°C. The humidity change trajectory is recorded by the humidity sensor, which can reflect the changes in humidity over a day or a period of time. For example, the humidity is higher at 80% in the morning, gradually decreasing to 40% in the afternoon as the sun rises, and then rising again as night falls.

[0055] Step S120: performing spatial grid division on the historical land attribute distribution data to generate attribute coding units with hierarchical association relationships, and performing timestamp alignment processing on the real-time land status record data to obtain a dynamic state sequence corresponding to the attribute coding units.

[0056] First, the target area is divided into multiple geographic grid cells of equal area based on the longitude and latitude coordinate ranges in the historical land attribute distribution data. Taking the agricultural land mentioned above as an example, the longitude and latitude range of this agricultural land is fixed. Assume that the longitude range is 110°-115° East longitude and the latitude range is 30°-35° North latitude. According to certain division rules, it is divided into square geographic grid cells with a side length of 1 km. In this way, the entire agricultural land is divided into many small grid cells of equal area.

[0057] For each of the geographic grid cells, the soil pH distribution, organic matter content gradient, and mineral concentration change curve contained therein are extracted to generate a soil attribute vector corresponding to the geographic grid cell. In a specific geographic grid cell, after detailed soil testing, it was found that the soil pH showed a distribution that gradually decreased from the northeast to the southwest within the grid, with the pH value gradually decreasing from 7.0 to 6.0; the organic matter content gradually decreased from 5% near the farmland entrance to the grid center to 3%; the mineral concentration change curve showed that the concentration of potassium gradually decreased from 25 mg / kg in the northwest corner of the grid to 15 mg / kg in the southeast corner. Based on these data, the soil attribute vector corresponding to the geographic grid cell can be generated.

[0058] Based on the differences in soil property vectors between adjacent geographic grid cells, a multi-level spatial association network is constructed, in which high-level network nodes represent the large-scale soil property mean, and low-level network nodes represent the local soil property fluctuation characteristics. For example, there are certain differences in soil property vectors between several adjacent geographic grid cells. If 10 adjacent geographic grid cells are viewed as a group, their overall soil pH mean, organic matter content mean, and mineral concentration mean can be used as the large-scale soil property mean represented by the high-level network nodes. The fluctuations such as the deviation of each individual geographic grid cell from this group of means are represented by the low-level network nodes.

[0059] The network nodes at each level of the spatial association network are mapped and associated with corresponding geographic grid cells to generate the attribute coding units. In this way, each geographic grid cell has a corresponding attribute coding unit. These attribute coding units not only contain their own soil attribute information, but also establish connections with surrounding geographic grid cells through the spatial association network.

[0060] At the same time, real-time land status data is timestamped. In environmental monitoring of this agricultural land, the timing of data collection by different sensors may vary. For example, some temperature sensors may have inconsistent data collection times with other sensors due to equipment failure or transmission delays. Through timestamp alignment, all data corresponding to geographic grid cells, such as surface temperature distribution and humidity change trajectories, are aligned in time, resulting in a dynamic state sequence corresponding to the attribute encoding unit.

[0061] Step S130: performing cross-modal feature fusion on the attribute coding unit and the dynamic state sequence to generate a land spatial feature map, wherein each node in the land spatial feature map represents the attribute evolution trend of the corresponding geographical block within a preset time range.

[0062] The surface temperature distribution data in the dynamic state sequence is transformed into a frequency domain to extract temperature fluctuation patterns at different time scales. In this agricultural land, the surface temperature distribution data will have different fluctuation patterns within a day. Through frequency domain transformation, the temperature data in the time domain can be converted to the frequency domain, so that the temperature fluctuations under different frequency components can be more clearly seen. For example, during the day when solar radiation is strong, temperature fluctuations may be mainly characterized by shorter cycles (such as hourly fluctuations), while at night, the temperature fluctuation cycles may become longer (such as fluctuations every 3-4 hours).

[0063] The temperature fluctuation pattern is aligned in the time domain with the soil property vector of the corresponding geographic grid cell to generate a hybrid feature vector that integrates the soil-temperature correlation characteristics. In each geographic grid cell, the previously obtained soil property vector and the extracted temperature fluctuation pattern are aligned in chronological order. For example, when the soil pH in the soil property vector remains stable over a certain period of time, the corresponding temperature fluctuation pattern is also what it looks like during that period. In this way, the two are fused to generate a hybrid feature vector.

[0064] The feature diffusion weights between nodes are determined based on the transmission paths of the hybrid feature vectors in the multi-level spatial association network. Since the transmission paths of the hybrid feature vectors in the spatial association network are different, the diffusion effects of the features between different nodes are also different. For example, in a high-level network node, if a node is connected to more low-level network nodes and the hybrid feature vectors of these nodes are similar, then its feature diffusion weight to other nodes will be larger; while in a low-level network node, if the hybrid feature vector of a node is significantly different from that of the surrounding nodes, then its feature diffusion weight to other nodes will be smaller.

[0065] Based on the feature diffusion weights, the mixed feature vectors are spatially interpolated to generate a continuous feature distribution map covering the entire target area, and the continuous feature distribution map is converted into graph structure data to form the land spatial feature map. Through spatial interpolation calculations, the mixed feature vectors of each geographic grid unit and the feature diffusion weights between them are used to extend the features of the discrete geographic grid units to the entire target area, forming a continuous feature distribution map. The continuous feature distribution map is then converted into graph structure data, where the nodes of the graph correspond to the geographic grid units and the edges represent the relationships between the nodes. In this way, a land spatial feature map is obtained, in which each node in the map can characterize the evolution trend of properties such as soil and temperature in the corresponding geographic block within a preset time range.

[0066] Step S140: extracting abnormal fluctuation nodes in the land spatial characteristic map, and determining a target training sample set based on soil composition information and surface temperature distribution corresponding to the abnormal fluctuation nodes.

[0067] Calculate the characteristic change rate of each node in the land spatial characteristic map, and construct an anomaly detection threshold based on the difference in characteristic change rates between adjacent nodes. In the land spatial characteristic map, each node has its corresponding characteristic change rate. Taking the soil pH characteristic as an example, over a period of time, the soil pH of the geographical block represented by a node changes from 6.5 to 6.0, and the rate of change is calculated. Then observe the soil pH change rate of adjacent nodes. If the change rate of most adjacent nodes is within a smaller range, such as between -0.01 and 0.01, then an anomaly detection threshold can be constructed based on this range, such as setting the anomaly detection threshold to a change rate greater than 0.05 or less than -0.05.

[0068] Nodes whose characteristic change rates exceed the anomaly detection threshold are identified, and their soil composition mutation records in the historical land attribute distribution data are traced. When a node's characteristic change rate exceeds the anomaly detection threshold, for example, a node's soil pH change rate is 0.1, the corresponding geographic block is searched in the historical land attribute distribution data. It may be discovered that this geographic block has experienced large-scale fertilizer application in the past, resulting in a mutation in certain soil components, such as a sudden increase in soil nitrogen content by 50%.

[0069] Nodes that meet both the characteristic rate anomaly and soil composition mutation criteria are marked as abnormal fluctuation nodes, and their corresponding longitude and latitude coordinates and timestamp information are recorded. If a node meets both the abnormal soil pH change rate and soil composition mutation records, then the node is marked as an abnormal fluctuation node. The longitude and latitude coordinates of the geographic block corresponding to the node are also recorded, such as 112° East longitude and 32° North latitude, as well as the timestamp of the abnormal fluctuation, such as 10:00 AM on May 10, 2023. Based on the soil composition information and surface temperature distribution data corresponding to these abnormal fluctuation nodes, the target training sample set can be determined, which will be used for subsequent prediction model training.

[0070] Step S150: Perform multi-stage iterative training on the initial prediction model based on the target training sample set to obtain a land status prediction model, wherein the land status prediction model is used to output the soil degradation risk assessment results and the probability of vegetation cover change in a future time period according to the input geographic block identifier.

[0071] The target training sample set is divided into a training subset and a validation subset, wherein the training subset contains data from geographic blocks with clear soil degradation labels, and the validation subset contains data from geographic blocks without soil state changes labeled. In the previously determined target training sample set, it is assumed that a portion of the geographic blocks have undergone detailed soil testing and long-term observation, and it is clearly known that these geographic blocks have soil degradation. The degree of soil degradation has been quantified and labeled, for example, according to the degree of soil fertility decline, with labels such as mild, moderate, and severe degradation. The data from these geographic blocks constitute the training subset. While the data from another portion of the geographic blocks do not have clear soil state changes labeled, they still contain relevant data such as soil composition information and surface temperature distribution. The data from these geographic blocks constitute the validation subset.

[0072] In the first training stage, the training subset is used to initialize the parameters of the initial prediction model, and the initial prediction weights are generated based on the correlation between the soil composition information and the surface temperature distribution. The initial prediction model can be a neural network model, such as a multi-layer perceptron model. The data of the training subset is input into the initial prediction model, and the parameters of the initial prediction model are initialized based on the relationship between the soil composition information (such as various mineral content, organic matter content, etc.) and the surface temperature distribution. For example, when the organic matter content in the soil is high, the surface temperature is relatively stable. This relationship can be reflected in the setting of weights when the model is initialized. If the correlation between the organic matter content and the stability of the surface temperature is very strong, then a larger weight value will be given in the weight setting.

[0073] In the second training phase, the validation subset is input into the initialized prediction model. Error backpropagation is performed based on the output vegetation cover change probability and real-time satellite image data to adjust the dynamic parameters in the initial prediction weights. The validation subset data is input into the initialized prediction model, which outputs the vegetation cover change probability for each geographic block. Simultaneously, the actual vegetation cover situation is obtained through satellite image data. For example, satellite imagery may show that the vegetation cover area of ​​a certain geographic block has dropped from 80% to 70% within a month, while the vegetation cover area corresponding to the model's predicted vegetation cover change probability has dropped to 75%, resulting in an error. Based on this error, the dynamic parameters in the model's initial prediction weights are adjusted through an error backpropagation algorithm to make the model's prediction results closer to the actual situation.

[0074] In the third training phase, the adjusted prediction model is deployed to an edge computing device. The prediction results are calibrated online using real-time environmental sensor data, and the final parameters of the land state prediction model are updated. The prediction model, trained and adjusted in the first two phases, is deployed to an edge computing device located near the agricultural land. This edge computing device can receive real-time data from environmental sensors, such as soil moisture sensors and light intensity monitoring devices. For example, the current soil moisture sensor reading is 30%, and the model's predicted probability of vegetation cover change is based on previous training data and model parameters. However, this moisture value may affect vegetation cover. The soil moisture sensor reading is compared with the model's output of the probability of vegetation cover change to determine the moisture deviation coefficient. If the soil moisture corresponding to the model's predicted probability of vegetation cover change is assumed to be 40%, the moisture deviation coefficient can be calculated based on the difference between the two. This moisture deviation coefficient is then weighted and corrected based on the light intensity monitoring value to generate an environmental calibration factor. For example, when light intensity is high, the effect of moisture on vegetation cover may be magnified or reduced. Based on this relationship, the moisture deviation coefficient is weighted and corrected. Finally, the environmental calibration factor is superimposed on the output layer of the land status prediction model to dynamically adjust the soil degradation risk assessment results in the future time period.

[0075] When the soil degradation risk assessment results for a target geographic area exceed a preset threshold, a list of recommended land management options is generated. For example, if the preset soil degradation risk threshold is 50%, a list of recommended land management options will be generated when the soil degradation risk assessment results for a target geographic area reach 60%.

[0076] First, search the historical land management case database for successful cases that match the current soil degradation type. If the current soil degradation type is due to declining soil fertility, search the historical case database for all successful cases of soil fertility reduction due to declining soil fertility. These cases include information on the treatment technology used, material ratio parameters, and implementation period. For example, one successful case involved improving soil fertility through the application of organic fertilizers. The organic fertilizer ratio was a mixture of nitrogen, phosphate, and potassium in a ratio of 3:2:1, and the implementation period was six months.

[0077] Analyze the treatment technology paths, material ratio parameters, and implementation cycle information used in the successful case data. For the organic fertilizer case mentioned above, analyze the technical path in detail, namely how the organic fertilizer was evenly applied to the soil, whether it was spread mechanically or manually; how the material ratio parameter of 3:2:1 was determined, and whether it was related to local soil type and crop needs; and how soil fertility changed at different stages during the six-month implementation period.

[0078] Combined with the terrain slope data and water source distribution data of the target geographic area, the aforementioned governance technology path is adaptively modified to generate multiple candidate governance solutions with priority ranking. If the terrain slope of the target geographic area is large, then when applying organic fertilizer, it may be necessary to consider preventing fertilizer loss. The original technical path can be modified by adopting layered fertilization or fertilization along contour lines. At the same time, if the water source is distributed far away, then irrigation issues need to be considered after fertilizer application, and some water-saving irrigation measures may need to be added. Based on these adaptive modifications, multiple candidate governance solutions are generated and prioritized according to factors such as feasibility and cost.

[0079] The candidate treatment options are evaluated for compatibility with real-time weather forecast data, and options that do not meet climatic conditions are eliminated to form a recommended list of land treatment options. If the weather forecast predicts heavy rainfall in the coming period, some fertilization options that require a dry environment will be unsuitable. After eliminating these options, the remaining options form the recommended list of land treatment options.

[0080] Extract the resource consumption data and expected effect data required for each plan in the recommended list of land management plans and construct a multi-objective optimization decision-making model. For each candidate management plan, determine the resources required, such as manpower, material resources (fertilizer, irrigation equipment, etc.), and financial resources (such as the cost of purchasing fertilizer and equipment), as well as the expected effects, such as the degree of improvement in soil fertility and the proportion of increased vegetation cover. Based on this data, construct a multi-objective optimization decision-making model, for example, with minimizing resource consumption and maximizing expected effects as the objective functions.

[0081] Based on the currently available resource inventory data and environmental carrying capacity constraints, the optimal land management solution is selected from the multi-objective optimization decision model and output to the user terminal. Assume that the currently available organic fertilizer inventory is limited, and the local environmental carrying capacity is limited, which cannot withstand excessive fertilizer application to avoid environmental pollution. Based on these constraints, the optimal land management solution is selected from the multi-objective optimization decision model and then output to the user terminal responsible for managing the agricultural land, so that they can carry out land management operations.

[0082] In the above embodiment, the land survey data-based prediction model training system for executing the above method embodiment has at least one processor, a control module (chip set) coupled to at least one of the (at least one) processors, a memory coupled to the control module, a non-volatile memory (NVM) / storage device coupled to the control module, at least one load / output device coupled to the control module, and a network interface coupled to the control module.

[0083] The processor may include at least one single-core or multi-core processor, and the processor may include any combination of general-purpose processors or dedicated processors (e.g., graphics processors, application processors, baseband processors, etc.). For some alternative implementations, the prediction model training system based on land surveying data can be based on electronic devices such as the gateway described in the embodiments of this application.

[0084] For some alternative embodiments, a predictive model training system based on land survey data may include at least one computer-readable medium (e.g., a memory or NVM / storage device) having instructions and at least one processor integrated with the at least one computer-readable medium and configured to execute the instructions to implement a module to perform the actions described in the present disclosure.

[0085] For one embodiment, the control module may include any suitable interface controller to provide any suitable interface to at least one of the processor(s) and / or any suitable device or component in communication with the control module.

[0086] The control module may include a memory controller module to provide an interface to the memory. The memory controller module may be a hardware module, a software module, and / or a firmware module.

[0087] The memory may be used, for example, to load and store data and / or instructions for a land survey data based prediction model training system. For one embodiment, the memory may include any suitable volatile memory, such as a suitable DRAM.

[0088] For one embodiment, the control module may include at least one load / output controller to provide an interface to the NVM / storage device and the (at least one) load / output device.

[0089] For example, NVM / storage devices may be used to store data and / or instructions. The NVM / storage devices may include any suitable non-volatile memory (e.g., flash memory) and / or may include any suitable (at least one) non-volatile storage device (e.g., at least one hard disk drive (HDD), at least one compact disc (CD) drive, and / or at least one digital versatile disc (DVD) drive).

[0090] The NVM / storage device may include storage resources that are physically part of the device on which the land survey data-based predictive model training system is installed, or may be accessible to the device without being part of the device. For example, the NVM / storage device may be accessed via (at least one) load / output device over a network.

[0091] The (at least one) load-in / output device may provide an interface for the land measurement data-based prediction model training system to communicate with any other appropriate device. The load-in / output device may include a communication component, a phonetic component, a sensor component, etc. The network interface may provide an interface for the land measurement data-based prediction model training system to communicate based on at least one network. The land measurement data-based prediction model training system may wirelessly communicate with at least one component of a wireless network based on any of at least one wireless network priors and / or protocols, for example, access a wireless network based on a communication prior.

[0092] For one embodiment, at least one of the (at least one) processors may be loaded together with the logic of at least one controller of a control module (e.g., a memory controller module). For one embodiment, at least one of the (at least one) processors may be loaded together with the logic of at least one controller of a control module to form a system-level load. For one embodiment, at least one of the (at least one) processors may be fused on the same die with the logic of at least one controller of a control module. For one embodiment, at least one of the (at least one) processors may be fused on the same die with the logic of at least one controller of a control module to form a system-on-chip (SoC).

[0093] The above is a detailed introduction to the embodiments of the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for those skilled in the art, based on the ideas of the present application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

[0094] An embodiment of the present invention discloses a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program enables a computer to execute the steps of the land survey data-based prediction model training method described in the aforementioned embodiment.

[0095] An embodiment of the present invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute the steps in the land survey data-based prediction model training method described in the aforementioned embodiment.

[0096] The device embodiments described above are merely illustrative, wherein the modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, i.e., they may be located in one place or distributed across multiple modules. Some or all of the modules may be selected based on actual needs to achieve the purpose of the present embodiment. Those skilled in the art can understand and implement the present invention without inventive effort.

[0097] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus the necessary general hardware platform, or of course, by hardware. Based on this understanding, the above technical solution, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, the storage medium including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electronically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium capable of storing or storing data.

[0098] Finally, it should be noted that what is disclosed above is only a preferred embodiment of the present invention, which is only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A prediction model training method based on land survey data, characterized in that: The method comprises: Obtain a land survey dataset for the target area, comprising historical land attribute distribution data and real-time land status record data. The historical land attribute distribution data includes soil composition information and terrain characteristic parameters for multiple geographic blocks, and the real-time land status record data includes surface temperature distribution and humidity change trajectories collected by current environmental monitoring equipment. Performing spatial grid division on the historical land attribute distribution data to generate attribute coding units with hierarchical association relationships, and performing timestamp alignment processing on the real-time land status record data to obtain a dynamic state sequence corresponding to the attribute coding units; Performing cross-modal feature fusion on the attribute coding unit and the dynamic state sequence to generate a land spatial feature map, wherein each node in the land spatial feature map represents the attribute evolution trend of the corresponding geographical block within a preset time range; Extracting abnormal fluctuation nodes from the land spatial characteristic map, and determining a target training sample set based on soil composition information and surface temperature distribution corresponding to the abnormal fluctuation nodes; The initial prediction model is subjected to multi-stage iterative training based on the target training sample set to obtain a land status prediction model, wherein the land status prediction model is used to output soil degradation risk assessment results and vegetation cover change probability in a future time period based on the input geographic block identifier.

2. The prediction model training method based on land surveying data according to claim 1, characterized in that: The spatial grid division of the historical land attribute distribution data to generate attribute coding units with hierarchical association relationships includes: Dividing the target area into a plurality of geographic grid cells of equal area according to the latitude and longitude coordinate ranges in the historical land attribute distribution data; For each of the geographic grid cells, extract the soil pH distribution, organic matter content gradient, and mineral concentration change curve contained therein to generate a soil attribute vector corresponding to the geographic grid cell; Based on the differences in soil property vectors between adjacent geographic grid cells, a multi-level spatial correlation network is constructed, in which high-level network nodes represent the mean of large-scale soil properties, and low-level network nodes represent the fluctuation characteristics of local soil properties. The network nodes at each level in the spatial association network are mapped and associated with corresponding geographic grid units to generate the attribute coding units.

3. The prediction model training method based on land surveying data according to claim 2, characterized in that: The cross-modal feature fusion of the attribute coding unit and the dynamic state sequence to generate a land spatial feature map includes: Performing frequency domain transformation on the surface temperature distribution data in the dynamic state sequence to extract temperature fluctuation patterns at different time scales; Aligning the temperature fluctuation pattern with the soil property vector of the corresponding geographic grid cell in the time domain to generate a hybrid feature vector integrating the soil-temperature correlation feature; Determining feature diffusion weights between nodes based on a transmission path of the hybrid feature vector in a multi-level spatial association network; A spatial interpolation calculation is performed on the mixed feature vector based on the feature diffusion weight to generate a continuous feature distribution map covering the entire target area, and the continuous feature distribution map is converted into graph structure data to form the land spatial feature map.

4. The prediction model training method based on land survey data according to claim 3, characterized in that: The extracting of abnormal fluctuation nodes in the land spatial characteristic map includes: Calculating the characteristic change rate of each node in the land spatial characteristic map, and constructing an anomaly detection threshold based on the difference in characteristic change rates between adjacent nodes; Identifying nodes whose characteristic change rate exceeds the anomaly detection threshold, and tracing back their soil composition mutation records in historical land attribute distribution data; The nodes that meet both the characteristic rate anomaly and soil composition mutation are marked as the abnormal fluctuation nodes, and their corresponding latitude and longitude coordinates and timestamp information are recorded.

5. The prediction model training method based on land surveying data according to claim 1, characterized in that: The multi-stage iterative training of the initial prediction model based on the target training sample set includes: Dividing the target training sample set into a training subset and a validation subset, wherein the training subset contains geographic block data with clear soil degradation labels, and the validation subset contains geographic block data without soil state change labels; In a first training phase, the parameters of the initial prediction model are initialized using the training subset, and initial prediction weights are generated based on the correlation between soil composition information and surface temperature distribution; In the second training phase, the validation subset is input into the initialized prediction model, and error back propagation is performed based on the output vegetation cover change probability and real-time satellite image data to adjust the dynamic parameters in the initial prediction weights; In the third training phase, the adjusted prediction model is deployed to the edge computing device, the prediction results are calibrated online through the real-time received environmental sensor data, and the final parameters of the land state prediction model are updated.

6. The prediction model training method based on land surveying data according to claim 5, characterized in that: The online calibration of the prediction result using the environmental sensor data received in real time includes: Get the soil moisture sensor readings and light intensity monitoring values ​​of the target geographic block in the current time period; Comparing the soil moisture sensor reading with the vegetation cover change probability output by the land state prediction model to determine a moisture deviation coefficient; Performing weighted correction on the humidity deviation coefficient according to the light intensity monitoring value to generate an environmental calibration factor; The environmental calibration factor is superimposed on the output layer of the land state prediction model to dynamically adjust the soil degradation risk assessment results in the future time period.

7. The prediction model training method based on land surveying data according to claim 1, characterized in that: The method further comprises: When the soil degradation risk assessment results of the target geographical area are detected to exceed the preset threshold, a list of recommended land management options is generated; Extract the resource consumption data and expected effect data required for each plan in the recommended list of land management plans, and build a multi-objective optimization decision-making model; According to the currently available resource inventory data and environmental carrying capacity constraints, the optimal land management plan is selected from the multi-objective optimization decision model and output to the user terminal.

8. The prediction model training method based on land surveying data according to claim 7, characterized in that: The generating of the recommended list of land management solutions includes: Query the historical land management case database for successful case data that matches the current soil degradation type; Analyze the treatment technology paths, material ratio parameters and implementation cycle information used in the successful case data; Combined with the terrain slope data and water source distribution data of the target geographical block, the above-mentioned treatment technology path is adaptively modified to generate multiple candidate treatment plans with priority ranking; The candidate management plans are evaluated for matching with real-time weather forecast data, and plans that do not meet climate conditions are eliminated to form a recommended list of land management plans.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores machine-executable instructions, which, when executed by a computer, implement the prediction model training method based on land surveying data according to any one of claims 1 to 8.

10. A prediction model training system based on land survey data, characterized in that: The method comprises a processor and a computer-readable storage medium, wherein the computer-readable storage medium stores machine-executable instructions, and when the machine-executable instructions are executed by a computer, the method for training a prediction model based on land surveying data according to any one of claims 1 to 8 is implemented.

Images