Arable land loss evaluation method and system based on arable land spatio-temporal variation characteristic analysis

Abstract

The application provides a cultivated land loss evaluation method and system based on cultivated land space-time change characteristic analysis, first acquires cultivated land space-time change tracking data set containing large-scale cultivated land spatial distribution, utilization state and surrounding environment data of different observation periods, then executes cultivated land spatial feature modeling processing, obtains large-scale cultivated land spatial coupling feature set in batches, executes cultivated land time sequence feature modeling processing, combines the spatial coupling feature set to obtain large-scale cultivated land space-time interaction feature set in batches, then inputs the two into a cultivated land loss risk conduction model, efficiently identifies large-scale cultivated land loss risk conduction paths, generates cultivated land loss risk evaluation results in batches, divides large-scale cultivated land loss risk areas, designs batch cultivated land protection intervention schemes and generates large-scale cultivated land protection strategy set, can efficiently process large-scale data, comprehensively and accurately evaluate large-scale cultivated land loss risk and scientifically and reasonably formulate cultivated land protection strategies.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and more specifically, to a method and system for assessing farmland loss based on the analysis of the spatiotemporal variation characteristics of farmland. Background Technology

[0002] In large-scale practical scenarios of agricultural resource management and sustainable development, arable land, as a core basic resource for agricultural production, plays a decisive role in food security, ecological environment stability, and the overall sustainable development of the socio-economic system due to its dynamic changes in quantity and quality. With the rapid advancement of urbanization, the profound adjustment of agricultural industrial structure, and the complex impact of natural environmental changes, arable land loss is occurring on a large scale and at a high frequency. How to efficiently and accurately assess the risk of large-scale arable land loss and formulate arable land protection strategies with batch applicability and systematicity has become a major challenge that urgently needs to be addressed in the field of agricultural resource management.

[0003] Currently, existing methods for assessing farmland loss have significant limitations when dealing with large-scale data. Most methods focus on single-factor analysis of a single region or small-scale sample, such as considering only the reduction in local farmland area or simple land use type conversions, lacking a comprehensive and integrated analysis of the spatiotemporal characteristics of large-scale farmland changes. When faced with massive amounts of data, these methods struggle to effectively integrate the adjacency relationships of farmland spatial distribution, the temporal changes in use status, and the dynamic impact of surrounding environmental factors on farmland loss, resulting in inaccurate and incomplete assessment results. Furthermore, in terms of farmland protection strategy formulation, due to a lack of in-depth understanding of the risk transmission mechanisms of large-scale farmland loss, existing protection measures often lack specificity and systematicity, making it difficult to effectively curb the trend of farmland loss on a large scale. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, the present invention provides a method for assessing farmland loss based on the analysis of spatiotemporal variation characteristics of farmland, the method comprising:

[0005] Acquire a set of farmland spatiotemporal change tracking data containing farmland spatial distribution data, utilization status data and surrounding environment data for different observation periods. The farmland spatial distribution data includes farmland plot boundary coordinates and farmland plot adjacency relationships. The utilization status data includes planting type and cultivation frequency. The surrounding environment data includes soil moisture and terrain slope.

[0006] The spatial feature modeling process of cultivated land is performed on the data set of cultivated land spatiotemporal change tracking. The adjacency relationship corresponding to the boundary coordinates of cultivated land plots is extracted. The fluctuation of soil moisture and topographic slope between adjacent cultivated land plots is associated with the adjacency relationship of cultivated land plots to obtain the set of cultivated land spatial coupling features.

[0007] The temporal and spatial changes of cultivated land are tracked and processed by modeling the temporal and spatial characteristics of cultivated land. Combined with the spatial coupling feature set of cultivated land, the conversion pattern of planting type with the observation period and the fluctuation pattern of cultivation frequency with the observation period are analyzed to obtain the temporal and spatial interaction feature set of cultivated land.

[0008] The set of spatial coupling features of cultivated land and the set of spatiotemporal interaction features of cultivated land are input into the cultivated land loss risk transmission model. The transmission path of cultivated land loss risk between adjacent cultivated land plots is identified through the dynamic correlation of spatial and temporal features, and cultivated land loss risk assessment results containing the loss tendency and risk transmission factors of each cultivated land plot are generated.

[0009] Based on the risk assessment results of farmland loss, continuous farmland loss risk areas are divided, and farmland protection intervention schemes are designed in combination with the transmission path of each farmland loss risk area to generate a set of farmland protection strategies.

[0010] Furthermore, this invention also provides a farmland loss assessment system based on the analysis of spatiotemporal changes in farmland, characterized by comprising:

[0011] A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the above-described method for assessing farmland loss based on the analysis of spatiotemporal changes in farmland by executing the machine-executable instructions.

[0012] In another aspect, the present invention also provides a computer program product, the computer program product including machine-executable instructions, the machine-executable instructions being stored in a computer-readable storage medium, a processor of a computer device reading the machine-executable instructions from the computer-readable storage medium, the processor executing the machine-executable instructions, causing the computer device to execute the above-mentioned method for assessing farmland loss based on the analysis of the spatiotemporal change characteristics of farmland.

[0013] Based on the above, this invention possesses powerful batch data processing capabilities. By acquiring a dataset of farmland spatiotemporal change tracking data, including large-scale farmland spatial distribution data, utilization status data, and surrounding environmental data from different observation periods, it can efficiently extract the adjacency relationships corresponding to the boundary coordinates of farmland plots during farmland spatial feature modeling. Furthermore, by combining the adjacency relationships of farmland plots with the fluctuations in soil moisture and terrain slope between adjacent farmland plots, it quickly generates a comprehensive and accurate large-scale farmland spatial coupling feature set. This precisely reveals the intrinsic connection between the large-scale farmland spatial distribution and the surrounding environment, significantly improving data processing efficiency and analytical accuracy. Further, by performing farmland temporal feature modeling and combining the farmland spatial coupling feature set, it can batch analyze the conversion patterns of planting types and the fluctuation patterns of cultivation frequency during observation periods, obtaining a large-scale farmland spatiotemporal interaction feature set with high dynamism and correlation, avoiding analytical biases caused by excessive data volume. By inputting a large-scale set of spatially coupled features and a set of spatiotemporally interactive features of cultivated land into a model for the transmission of cultivated land loss risk, and through the dynamic correlation of spatial and temporal features, the transmission paths of large-scale cultivated land loss risk between adjacent cultivated land plots can be efficiently identified. This allows for the batch generation of cultivated land loss risk assessment results containing the loss tendency and risk transmission factors of each cultivated land plot, achieving a comprehensive and detailed assessment of large-scale cultivated land loss risk and effectively improving the timeliness and accuracy of the assessment. Finally, based on the large-scale cultivated land loss risk assessment results, continuous cultivated land loss risk areas are divided, and batch-based cultivated land protection intervention schemes are designed in conjunction with the transmission paths of each cultivated land loss risk area. This generates a targeted and systematic set of large-scale cultivated land protection strategies, effectively improving the efficiency and effectiveness of large-scale cultivated land protection. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the execution flow of the method for assessing farmland loss based on the analysis of the spatiotemporal change characteristics of farmland provided in this embodiment of the invention.

[0015] Figure 2 This is a schematic diagram of exemplary hardware and software components of a farmland loss assessment system based on spatiotemporal change characteristics analysis of farmland provided in an embodiment of the present invention. Detailed Implementation

[0016] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a method for assessing farmland loss based on the analysis of spatiotemporal changes in farmland, provided in one embodiment of the present invention. The following is a detailed description of this method for assessing farmland loss based on the analysis of spatiotemporal changes in farmland.

[0017] Step S110: Obtain a set of farmland spatiotemporal change tracking data containing farmland spatial distribution data, utilization status data and surrounding environment data for different observation periods. The farmland spatial distribution data includes farmland plot boundary coordinates and farmland plot adjacency relationships. The utilization status data includes planting type and cultivation frequency. The surrounding environment data includes soil moisture and terrain slope.

[0018] In this embodiment, the farmland spatiotemporal change tracking dataset covers at least three consecutive observation periods, with each period maintaining a consistent time span, such as an annual observation cycle, and using uniform data collection standards across all periods. Farmland spatial distribution data is obtained through a combination of high-resolution remote sensing image interpretation and field mapping. The boundary coordinates of farmland plots are represented using the nationally unified geodetic coordinate system, expressed as a sequence of polygon vertex coordinates, with each vertex containing both longitude and latitude parameters. Adjacency relationships between farmland plots are determined through spatial topology analysis; that is, when two farmland plots share a common line segment or point on their boundaries, they are considered adjacent, and the direction and relative position of the adjacent boundaries are recorded.

[0019] Status data is collected through a combination of farmland management ledgers and field surveys. Planting types are categorized into several standard categories, including but not limited to food crops, cash crops, and horticultural crops. Each category contains several specific crop types, and each category uses a unified coding system. Tillage frequency is recorded based on the actual number of times tillage is carried out within a year, distinguishing between normal tillage, fallow, and crop rotation. Fallow refers to no tillage throughout the year, while crop rotation refers to changing the planting type seasonally within the same year.

[0020] Surrounding environmental data was collected through a combination of fixed monitoring stations and mobile monitoring. Soil moisture was monitored at fixed points using soil moisture sensors, with monitoring depths divided into surface, middle, and deep layers. Data from each layer was averaged daily. Topographic slope was calculated using digital elevation model data and expressed as a percentage slope, with the calculation unit consistent with the boundaries of cultivated land plots. All collected data were accompanied by metadata such as timestamps, acquisition device numbers, and data quality identifiers. The data quality identifiers were categorized into four levels: excellent, good, average, and poor, for subsequent data screening.

[0021] During data collection, data anonymization techniques were employed for sensitive data such as farmland ownership information and farmer contact information. Specifically, this included: anonymizing the names of farmland owners by replacing them with randomly generated letter combinations; partially masking farmer contact information, retaining only the first three and last four digits; and encrypting all private data using a symmetric encryption algorithm compliant with national cryptographic management standards, with the key changed regularly and access strictly controlled. Encrypted transmission protocols were used during data transmission to ensure that data was not leaked or tampered with during transmission.

[0022] Step S120: Perform farmland spatial feature modeling processing on the farmland spatiotemporal change tracking data set, extract the adjacency relationship corresponding to the boundary coordinates of farmland plots, and combine the adjacency relationship of farmland plots to associate the fluctuation of soil moisture and topographic slope between adjacent farmland plots to obtain the farmland spatial coupling feature set.

[0023] Step S121: Extract the spatial distribution data of cultivated land for the same observation period from the cultivated land spatiotemporal change tracking data set, separate the boundary coordinates of each cultivated land plot according to the cultivated land plot identifier, extract the adjacent relationships of all cultivated land plots within the observation period, and determine the identifiers of all adjacent cultivated land plots corresponding to each cultivated land plot.

[0024] Step S1211: Select the spatial distribution data of cultivated land marked as the same observation period from the cultivated land spatiotemporal change tracking data set. The cultivated land spatial distribution data contains information on all cultivated land plots in the observation period, and each cultivated land plot has a unique cultivated land plot identifier.

[0025] In this embodiment, the data filtering module is first invoked from the farmland spatiotemporal change tracking dataset to perform precise matching using the observation period field. Each observation period has a unique time identifier, formatted as year plus quarter; for example, the identifier for a certain observation period is a combination of year and quarter information. During the filtering process, the integrity of the data needs to be verified, i.e., checking whether the farmland spatial distribution data for that observation period contains three parts: a data header file, a data body, and data endnotes. The data header file contains information such as the observation period, data collection range, and coordinate system parameters, while the data endnotes contain the data checksum and data generation time.

[0026] Farmland plot information is stored in a structured data format. Each record corresponds to an independent farmland plot and includes attributes such as plot identifier, number of boundary coordinates, coordinate sequence, area, and perimeter. The farmland plot identifier uses a 10-character code: the first two characters represent the administrative division code, the middle four characters represent the township code, and the last four characters are a sequential code, ensuring uniqueness and consistency across all observation periods. Preliminary verification is performed on the selected data to check for anomalies such as duplicate plot identifiers, zero boundary coordinates, or empty area fields. If abnormal data is found, it is marked as pending verification, and the reason for the anomaly is recorded.

[0027] Step S1212: Extract the boundary coordinates of each cultivated land plot according to the cultivated land plot identifier. The coordinate information includes the specific location records of each vertex around the cultivated land plot. For cultivated land plot boundary coordinates with missing vertex positions, derive and complete them by the boundary coordinates of adjacent cultivated land plots. Organize the cultivated land plot boundary coordinates of each cultivated land plot according to the cultivated land plot identifier to form a list of cultivated land plot boundary coordinates.

[0028] For each farmland parcel identifier, a boundary coordinate sequence is read from the structured data record. This sequence is arranged clockwise, with the starting and ending vertices sharing the same coordinate point, forming a closed polygon. The number of coordinate points is determined by the complexity of the farmland parcel's shape; regularly shaped parcels require fewer coordinate points, while irregularly shaped parcels require more. When a missing value is detected in the coordinate sequence—that is, when the longitude or latitude field of a vertex is empty—a coordinate completion mechanism is activated.

[0029] The coordinate completion mechanism first identifies all adjacent cultivated land parcels and extracts their boundary coordinate sequences. Then, by comparing the overlap of boundary segments, it determines the boundary segment containing the missing coordinate point. If this boundary segment completely overlaps with a boundary segment of an adjacent cultivated land parcel, the corresponding coordinate point from the adjacent cultivated land parcel is used to complete the missing value. If there is partial overlap, linear interpolation is used to calculate the missing coordinate point, based on the rate of change of the overlapping portion. The completed coordinate sequence needs to be recalculated for closure to ensure that the starting and ending coordinates are consistent. If they are inconsistent, the last coordinate point is adjusted to coincide with the starting point.

[0030] The list of farmland plot boundary coordinates is organized in tabular form and includes five fields: farmland plot identifier, coordinate point number, longitude, latitude, and coordinate source. The coordinate source field indicates whether the coordinates were originally collected or derived and supplemented. The derived and supplemented coordinate points must also record the identifiers of the adjacent farmland plots they referenced.

[0031] Step S1213: Extract an information module that specifically records the adjacent relationships of cultivated land plots from the cultivated land spatial distribution data during the same observation period. This information module contains the corresponding records of each cultivated land plot identifier and its adjacent cultivated land plot identifiers.

[0032] In the spatial distribution data of cultivated land during the same observation period, the information module on the adjacency relationship of cultivated land plots is stored in the form of an association table, which includes four fields: main cultivated land plot identifier, adjacent cultivated land plot identifier, adjacent boundary length, and adjacent boundary type. The adjacent boundary type is divided into three types: completely adjacent, partially adjacent, and point adjacent. Completely adjacent means that two cultivated land plots share a complete boundary line segment, partially adjacent means that they share a part of the boundary line segment, and point adjacent means that they share only one boundary vertex.

[0033] During the extraction process, the integrity of the association table needs to be verified, meaning that each farmland plot identifier must appear at least once in the main farmland plot identifier field, and the adjacent farmland plot identifier fields must not be empty. Simultaneously, the symmetry of adjacency relationships needs to be checked; that is, if farmland plot A's adjacent farmland plot includes farmland plot B, then farmland plot B's adjacent farmland plot should also include farmland plot A. Any asymmetry must be marked as an anomaly and manually verified.

[0034] Step S1214: Match the adjacent relationship records of cultivated land plots one by one according to the cultivated land plot identifier. For each cultivated land plot identifier, extract all the adjacent cultivated land plot identifiers from the corresponding record to form a list of adjacent cultivated land plot identifiers for each cultivated land plot.

[0035] For each farmland plot identifier, an exact query is performed in the adjacency relationship table to retrieve all records where the main farmland plot identifier is equal to that identifier. The values ​​of the adjacent farmland plot identifier field are then extracted to form a preliminary list of adjacent farmland plot identifiers. This list is then deduplicated to remove duplicate adjacent farmland plot identifiers, and sorted by the length of the adjacent boundary in descending order. For those with the same length, they are arranged in alphabetical order of the adjacent farmland plot identifiers.

[0036] The list should record the adjacent boundary length and type for each adjacent farmland plot identifier. For example, in the list of adjacent farmland plot identifiers for farmland plot A, the adjacent boundary length for farmland plot B is a certain value, and the adjacent boundary type is "partially adjacent". Additionally, a total count of adjacent farmland plots should be added to the end of the list for subsequent data validation.

[0037] Step S1215: Check the list of adjacent cultivated land parcel identifiers for each cultivated land parcel. If any adjacent cultivated land parcel identifier is missing, compare the boundary coordinates of the cultivated land parcel with the boundary coordinates of other cultivated land parcels in the surrounding area to determine whether there are any unrecorded adjacent cultivated land parcels. If so, supplement the corresponding adjacent cultivated land parcel identifiers.

[0038] Step S1215-1: Perform a preliminary check on the list of adjacent cultivated land parcels for each cultivated land parcel. If the list is empty or contains only a few markers, and the boundary coordinates of the cultivated land parcel show that there are other cultivated land parcels around it, then it is determined that the markers of the adjacent cultivated land parcels of the cultivated land parcel may be missing.

[0039] The preliminary inspection includes quantity checks and spatial rationality checks. In the quantity check, if the area of ​​a cultivated land plot exceeds a certain threshold and the number of adjacent cultivated land plot identifiers in the identifier list is zero, it is considered an anomaly. If a cultivated land plot is a regular polygon with more than four sides, and the number of adjacent identifiers is less than half the number of sides, it is considered potentially missing. The spatial rationality check involves plotting the boundary coordinates of the cultivated land plot on a spatial canvas and visually inspecting whether there are any unidentified adjacent cultivated land plots around it; if so, they are marked as potentially missing.

[0040] Step S1215-2: Extract the complete boundary coordinates of the cultivated land plot, determine the position of all vertices of the cultivated land plot boundary, and extract the boundary coordinates of all other cultivated land plots within the same observation period to determine the position of the vertices of the other cultivated land plots.

[0041] The boundary coordinate sequences of the farmland plots to be inspected are converted into spatial geometric objects, represented using a polygon data structure, where each vertex serves as a node of the polygon. Simultaneously, the boundary coordinate sequences of all other farmland plots within the same observation period are also converted into corresponding polygon geometric objects and stored in a spatial index data structure to improve spatial query efficiency. The spatial index uses a quadtree index structure, dividing the spatial region into different levels of grids based on coordinate ranges. Each grid corresponds to storing the geometric objects of the farmland plots contained within that grid.

[0042] Step S1215-3: Compare the boundary coordinates of this cultivated land plot with those of each other cultivated land plot to check if there are any overlapping boundary segments. If there are overlapping boundary segments and the length of the overlapping segments exceeds the preset judgment length, then the two cultivated land plots are determined to be adjacent cultivated land plots.

[0043] For each farmland parcel geometry to be compared, an intersection operation in spatial overlay analysis is performed to calculate the boundary intersection between the farmland parcel to be checked and the comparison farmland parcels. If the intersection result is a line segment, the length of the line segment is calculated. If the length is greater than a preset judgment length, the parcels are considered adjacent. The judgment length is set as a percentage of the average boundary length of farmland parcels, for example, one-tenth of the average boundary length. If the intersection result is a point, it is further checked whether the point is a vertex of the two farmland parcels. If so, the points are considered adjacent; otherwise, they are ignored.

[0044] Step S1215-4: For cultivated land plots that are identified as adjacent cultivated land plots but do not appear in the adjacent cultivated land plot identifier list of the cultivated land plot, extract their cultivated land plot identifiers and add them to the adjacent cultivated land plot identifier list of the cultivated land plot.

[0045] Add the identifiers of farmland parcels identified as adjacent by spatial analysis but not in the original list to the end of the list of adjacent farmland parcel identifiers, and mark them as spatial derivation supplements. Simultaneously, record the adjacent boundary lengths (i.e., the calculated lengths of overlapping line segments) and adjacent boundary types (determined as completely adjacent, partially adjacent, or point adjacent based on the overlap). The supplemented list needs to be reordered, maintaining the order from largest to smallest adjacent boundary length.

[0046] Step S1215-5: Repeat the above comparison process (i.e., step S1215-4) until the boundary coordinates of the cultivated land plot are compared with the boundary coordinates of all other cultivated land plots in the same observation period.

[0047] The comparison process is performed using a loop iterative approach. In each iteration, an uncompared farmland plot geometric object is selected from the spatial index, and intersection and adjacency checks are performed until all farmland plot geometric objects have been compared. To avoid duplicate comparisons, a list of already compared flags is set up, and farmland plots that have already been compared are not compared again.

[0048] Step S1215-6: Re-examine the supplemented list of adjacent cultivated land plot identifiers to confirm that each identifier in the list corresponds to a cultivated land plot with an overlapping boundary line segment. Remove misjudged non-adjacent cultivated land plot identifiers to form a list of adjacent cultivated land plot identifiers.

[0049] The second check employs a two-way verification mechanism. For each identifier in the supplemented list of adjacent farmland parcel identifiers, the boundary coordinates of the corresponding farmland parcel are extracted and re-intersected with the boundary coordinates of the farmland parcel to be checked. If the intersection length is less than the judgment length, it is considered a false positive, and the identifier is removed from the list. Simultaneously, the symmetry of adjacency relationships is checked. If the supplemented list contains farmland parcel B, but the adjacent farmland parcel identifier lists of farmland parcel B do not contain the identifier of the farmland parcel to be checked, then the identifier of the farmland parcel to be checked is added to the list of farmland parcel B.

[0050] Step S1216: Associate the boundary coordinates of each cultivated land plot with the corresponding list of adjacent cultivated land plot identifiers. For any information on the association deviation, rematch the cultivated land plot identifiers to form a spatial basic information unit for cultivated land plots.

[0051] During the association process, the list of farmland plot boundary coordinates is joined with the list of adjacent farmland plot identifiers using the farmland plot identifier as the key. If the joined data shows a single boundary coordinate corresponding to multiple adjacent lists or a single adjacent list corresponding to multiple boundary coordinates, it is considered an association deviation. In this case, a re-matching mechanism is initiated, comparing metadata such as data collection timestamps and data quality identifiers, selecting the data with the highest quality level as the benchmark, and correcting the deviation data.

[0052] The spatial basic information units of cultivated land parcels are represented using an object-oriented data structure. Each unit contains two parts: attribute information and geometric information. The attribute information includes the cultivated land parcel identifier, the number of adjacent cultivated land parcels, and the main adjacent directions; the geometric information is a polygon object composed of a sequence of boundary coordinates. All units are arranged in ascending order by cultivated land parcel identifier and stored in a dedicated data table in the spatial database, with a spatial index established to accelerate subsequent query operations.

[0053] Step S122: Based on the boundary coordinates of each cultivated land plot and the boundary coordinates of the corresponding adjacent cultivated land plots, determine the overlapping part of their boundaries, record the specific range of the boundary overlap, and count the total range of boundary overlap between each cultivated land plot and all adjacent cultivated land plots to establish a boundary association file between cultivated land plots and adjacent cultivated land plots.

[0054] For each farmland parcel, iterate through all adjacent farmland parcel identifiers in its list of adjacent farmland parcel identifiers. For each adjacent farmland parcel identifier, extract the corresponding boundary coordinate sequence from the farmland parcel spatial basic information unit, and convert the boundary coordinate sequences of the two farmland parcels into polygonal geometric objects. Then, perform an intersection operation on the geometric objects, calculate the intersection of the two polygons, and the intersection result is the boundary overlap portion.

[0055] If the overlapping boundary portion is a line segment, record the starting and ending coordinates of the line segment, as well as its length and direction angle; if it consists of multiple discontinuous line segments, record the parameters of each line segment separately. The specific range of the boundary overlap is described relative to the boundary of the main cultivated land plot, that is, record the starting and ending vertex numbers of the overlapping line segments in the coordinate sequence of the main cultivated land plot boundary.

[0056] When calculating the total overlap range, the lengths of the overlapping line segments between the same cultivated land parcel and all its adjacent cultivated land parcels are summed to obtain the total overlap length. Simultaneously, the ratio of the total overlap length to the perimeter of the main cultivated land parcel is calculated; this is called the boundary overlap rate, used to measure the degree of adjacency between cultivated land parcels. Boundary association files are stored in XML format, with each cultivated land parcel corresponding to a separate XML file. These files contain information such as the main cultivated land parcel identifier, the number of adjacent cultivated land parcels, the identifier of each adjacent cultivated land parcel, overlapping boundary parameters, total overlap length, and boundary overlap rate.

[0057] Step S123: Extract surrounding environmental data for the same observation period, match soil moisture data and topographic slope data to the corresponding cultivated land plots according to the cultivated land plot identifiers, form an environmental data file for each cultivated land plot, and associate the environmental data file of each cultivated land plot with the environmental data files of the corresponding adjacent cultivated land plots according to the adjacent cultivated land plot identifiers.

[0058] Data matching the current observation period is selected from surrounding environmental data, based on data timestamps. Soil moisture data is matched according to farmland plot identifiers. The matching method is as follows: when a monitoring station is located inside a farmland plot, the monitoring data of that station is directly matched to the corresponding farmland plot; when a monitoring station is located outside a farmland plot but less than a certain threshold from the farmland plot boundary, a distance-weighted average method is used to interpolate the monitoring data of multiple stations to the farmland plot, with the interpolation weight inversely proportional to the square of the distance.

[0059] Topographic slope data is matched to farmland plots through spatial overlay analysis. This involves overlaying topographic slope calculation units with farmland plot boundaries and using an area-weighted average method to calculate the average slope of each farmland plot. The weight is the area proportion of each calculation unit within the farmland plot. For farmland plots containing multiple slope types, the maximum slope, minimum slope, and slope standard deviation are recorded simultaneously.

[0060] The environmental data archive includes fields such as farmland plot identifier, soil moisture (divided into three levels), topographic slope (average, maximum, minimum, and standard deviation), data collection time, and data quality identifier. When linking environmental data archives of adjacent farmland plots, an adjacent environmental data field is added to the environmental data archive of each farmland plot to store the soil moisture and topographic slope data of adjacent farmland plots, forming an environmental data association table. Each row in the table corresponds to a comparison record of environmental data between a farmland plot and an adjacent farmland plot.

[0061] Step S124: Compare the soil moisture data of each cultivated land plot with its adjacent cultivated land plots, record the difference in soil moisture between the two, and compare the topographic slope data of the two, record the difference in topographic slope between the two, and organize them into an environmental difference record of adjacent cultivated land plots according to the cultivated land plot identification.

[0062] Soil moisture differences were compared at three levels. The absolute and relative differences in soil moisture between the main cultivated land plot and adjacent cultivated land plots at the same level were calculated. The absolute difference was the soil moisture of the main cultivated land plot minus the soil moisture of the adjacent cultivated land plot, and the relative difference was the ratio of the absolute difference to the soil moisture of the main cultivated land plot. The direction of the difference was also recorded, i.e., whether the soil moisture of the main cultivated land plot was higher or lower than that of the adjacent cultivated land plot.

[0063] The absolute and relative differences in average slope were calculated by comparing topographic slope differences, using the same method as for soil moisture. In addition, the consistency of slope types was compared, i.e., whether the slopes of the main cultivated land plot and adjacent cultivated land plots belonged to the same slope grade. Slope grades are classified into five levels according to regulations: flat slope, gentle slope, sloping slope, steep slope, and sharp slope.

[0064] The environmental differences between adjacent cultivated land plots are recorded in tabular form, including fields such as the main cultivated land plot identifier, the adjacent cultivated land plot identifier, the absolute difference in surface soil moisture, the relative difference in surface soil moisture, the direction of the difference in surface soil moisture, the absolute difference in middle soil moisture, etc. (listed in order of level), the average absolute difference in topographic slope, the average relative difference in topographic slope, and the consistency indicator of topographic slope level, where the consistency indicator of topographic slope level is "consistent" or "inconsistent".

[0065] Step S125: Integrate the total overlap range of the boundaries of each cultivated land plot with the corresponding environmental difference records of adjacent cultivated land plots, remove redundant information that only repeatedly records the cultivated land plot identifiers, retain the core association logic of boundary association and environmental difference, and form a spatial association unit of cultivated land plots.

[0066] The integration process employs association rule mining to analyze the correlation between the total overlap range of boundaries and environmental difference records. First, the total overlap range is divided into several intervals, such as low overlap rate, medium overlap rate, and high overlap rate. Similarly, the difference values ​​in each environmental difference record are divided into several intervals, such as no difference, small difference, medium difference, and large difference. Then, the co-occurrence frequency between different boundary overlap intervals and environmental difference intervals is analyzed using a contingency table, and association rules with co-occurrence frequencies exceeding a threshold are extracted.

[0067] When removing redundant information, the boundary overlap line segment parameters, specific environmental difference values, and association strength parameters obtained through association rule mining are retained for each adjacent cultivated land plot. The association strength parameters are represented by support and confidence. The spatial association unit of cultivated land plots includes four parts: the main cultivated land plot identifier, the set of adjacent cultivated land plot identifiers, boundary overlap features (total length, overlap rate, association rule support), and environmental difference features (soil moisture differences at each level, topographic slope differences, association rule confidence). These are stored in JSON format for subsequent processing.

[0068] Step S126: Classify the spatial association units of all cultivated land plots according to the observation period. For the missing spatial association units of cultivated land plots after classification, complete them by deriving the boundary coordinates of cultivated land plots in the same observation period. Mark the observation period and boundary coordinate range of each spatial association unit of cultivated land plots, and finally form a set of spatial coupling features of cultivated land.

[0069] The classification process groups the spatial association units of cultivated land parcels according to the observation period, with one group corresponding to each observation period. The completeness of the spatial association units of cultivated land parcels in each group is checked, i.e., whether it contains spatial association units of all cultivated land parcels within that observation period. If any are missing, a completion mechanism is initiated. The completion mechanism uses the boundary coordinates of cultivated land parcels within the same observation period to re-execute the processing flow of steps S121 to S125, generating the missing spatial association units of cultivated land parcels.

[0070] Each spatially associated unit of cultivated land parcels is assigned an observation period identifier and a boundary coordinate range field. The boundary coordinate range is represented by the smallest bounding rectangle of the cultivated land parcel boundary, including four parameters: minimum longitude, maximum longitude, minimum latitude, and maximum latitude. The cultivated land spatial coupling feature set is a collection of spatially associated units of cultivated land parcels for all observation periods. A secondary index is established based on the observation period and the cultivated land parcel identifier, supporting queries for all cultivated land parcels by observation period or queries for spatially associated units of all observation periods by cultivated land parcel identifier.

[0071] Step S130: Perform temporal feature modeling processing on the farmland spatiotemporal change tracking data set, and combine the farmland spatial coupling feature set to analyze the conversion pattern of planting type with the observation period and the fluctuation pattern of cultivation frequency with the observation period, so as to obtain the farmland spatiotemporal interaction feature set.

[0072] Step S131: Extract utilization status data for all observation periods from the farmland spatiotemporal change tracking data set. Assign the planting type data and cultivation frequency data for each observation period to the same farmland plot according to the farmland plot identifier. For utilization status data for farmland plots without a corresponding plot, complete the data by matching the farmland plot boundary coordinates to form a time-series record of utilization status for each farmland plot.

[0073] Utilization status data for all observation periods were extracted from the farmland spatiotemporal change tracking dataset. This data includes planting type and tillage frequency. Data was integrated according to farmland plot identifiers, mapping planting type and tillage frequency data from different observation periods to the same farmland plot. For utilization status data without a clearly defined corresponding farmland plot, the farmland plot containing those coordinates was identified by comparing the data collection location coordinates with the farmland plot boundary coordinates, and the data was matched to that plot. If a data collection location coordinate falls within the boundaries of multiple farmland plots, the farmland plot with the closest timestamp was selected for matching based on the data collection timestamp and the update timestamp of the farmland plot boundary coordinates.

[0074] When generating the time-series record of land use status, the planting type and cultivation frequency data for each farmland plot are arranged in chronological order of the observation periods. The time-series record includes farmland plot identification, observation period sequence, planting type code for each observation period, and cultivation frequency code for each observation period. Missing observation period data in the time-series record are marked as "unrecorded" and will be processed later.

[0075] Step S132: Extract the spatial association units of each cultivated land plot from the set of cultivated land spatial coupling features at different observation periods. Match the spatial association units of cultivated land plots with the time series records of the utilization status of the corresponding cultivated land plots in the order of observation periods. For the observation period data with matching deviation, realign them by the cultivated land plot identifier so that the spatial association information and utilization status information of the same cultivated land plot correspond to each other in the same observation period.

[0076] From the set of spatial coupling features of cultivated land, spatial association units of each cultivated land plot are extracted at different observation periods based on the cultivated land plot identifier and observation time period. The extracted spatial association units of cultivated land plots are arranged in chronological order of observation time periods and compared with the observation time period sequence of the corresponding use status time series records. If the two observation time period sequences are not completely consistent, and there are cases of early or late observation, the cultivated land plot identifier is used as a key to find the same observation time period in the two sequences, and the spatial association units and use status data are rearranged according to the same observation time period.

[0077] For observation period data with matching bias, i.e., a certain observation period exists in the spatial association unit of cultivated land plots, but the data for that observation period is missing in the utilization state time series record, or vice versa, the following methods are used: If the spatial association unit exists but the utilization state data is missing, add the observation period to the utilization state time series record and mark the planting type and cultivation frequency as "unknown"; if the utilization state data exists but the spatial association unit is missing, re-query the spatial association unit of the cultivated land plot in that observation period from the cultivated land spatial coupling feature set. If it still does not exist, mark the spatial association unit as "not generated".

[0078] Step S133: Traverse the time-series records of utilization status of each cultivated land plot in chronological order of observation time periods, compare the planting type data of adjacent observation time periods, if the planting type is different in adjacent time periods, record the observation time period in which the conversion occurs, and record the planting type before and after the conversion, and organize them into a planting type conversion record according to the cultivated land plot identifier.

[0079] Step S1331: Extract all planting type data for a single cultivated land plot from the utilization status time series record according to the cultivated land plot identifier, and arrange them in chronological order of the observation time. For the observation time data with arrangement deviation, reorder them by timestamp to form the planting type time series chain of the cultivated land plot.

[0080] Planting type data for individual cultivated land plots is extracted from the time-series records of cultivated land use status based on the plot identifier. This planting type data includes the observation period identifier and the corresponding planting type code. This data is then arranged chronologically according to the observation period, based on the start timestamp of each observation period. If the data is found to be disorganized, for example, if the start timestamp of one observation period is earlier than the previous one, the data is reordered using the timestamps to ensure that the planting type time-series chain increases sequentially.

[0081] The planting type time series is stored in a linear list structure, with each node containing three elements: observation period identifier, planting type code, and data quality identifier. For data from repeated observation periods in the time series, the record with the highest data quality identifier is retained; if the data quality identifiers are the same, the most recently collected record is retained.

[0082] Step S1332: Starting from the initial observation period of the planting type time series chain, compare the planting type data of the current observation period with the next observation period one by one. If the planting types are different in adjacent periods, record the observation period in which the conversion occurs, and record the planting types before and after the conversion. Organize the data according to the farmland plot identifier to form a planting type conversion record.

[0083] Starting from the first observation period node in the planting type time sequence chain, the planting type code of the current node is compared with the planting type code of the next node. If the two codes are different, it is determined that a planting type conversion has occurred, the observation period in which the conversion occurs is recorded as the next observation period, the planting type before the conversion is the planting type code of the current node, and the planting type after the conversion is the planting type code of the next node.

[0084] All planting type conversion events are compiled into planting type conversion records based on the farmland plot identifiers. Each record contains five fields: farmland plot identifier, conversion observation period, planting type code before conversion, planting type code after conversion, and conversion confirmation identifier. The conversion confirmation identifier indicates whether the conversion event has been verified on-site, and is categorized as confirmed or unconfirmed. Unconfirmed conversion events require verification using other data in subsequent steps.

[0085] Step S134: Collect planting type conversion records for all cultivated land plots, classify conversion categories according to the combination of planting types before and after conversion, record the occurrence of each conversion category in different observation periods, analyze the distribution pattern of the same conversion category in continuous observation periods, and form a planting type conversion pattern.

[0086] The planting type conversion records of all cultivated land plots are statistically analyzed. Conversion categories are divided according to the combination of the planting type code before and after conversion, and each conversion category corresponds to a unique combination code. For example, if the land was classified as a food crop before conversion and became a cash crop after conversion, it corresponds to a specific combination code.

[0087] Record the number of occurrences and frequency of each transformation category in different observation periods. The frequency is the ratio of the number of occurrences of that transformation category in a certain observation period to the total number of transformations in that observation period. Analyze the distribution pattern of the same transformation category in consecutive observation periods, including: in which observation periods the transformation category appears, the trend of the number of occurrences (increasing, decreasing, fluctuating), and the correlation of the number of occurrences in adjacent observation periods.

[0088] Planting type conversion patterns are represented by pattern descriptors. Each pattern descriptor contains information such as conversion category code, main observation period sequence, frequency variation trend, and typical associated farmland characteristics (e.g., topographic slope range, soil moisture range). Conversion categories that appear in multiple consecutive observation periods and whose frequency exceeds a certain threshold are marked as stable conversion patterns; conversion categories that appear only in individual observation periods or whose frequency fluctuates greatly are marked as temporary conversion patterns.

[0089] Step S135: Traverse the time-series records of utilization status of each cultivated land plot in chronological order of observation time periods, extract the cultivation frequency data of adjacent observation time periods, analyze the changes in cultivation frequency data, record the observation time periods in which the cultivation frequency changes, and record the cultivation frequency data before and after the changes, and organize them into a cultivation frequency change record according to the cultivated land plot identifier.

[0090] The utilization status time-series records of each cultivated land plot are traversed sequentially according to the observation period, and the tillage frequency data of adjacent observation periods are extracted. The tillage frequency data is expressed as the actual number of tillages within the year, and it is necessary to distinguish between normal tillage, fallow, crop rotation, etc. The changes in tillage frequency data are analyzed, and the absolute and relative changes in tillage frequency between adjacent observation periods are calculated. The absolute change is the tillage frequency of the later observation period minus the tillage frequency of the earlier observation period, and the relative change is the ratio of the absolute change to the tillage frequency of the earlier observation period (if the tillage frequency of the earlier observation period is zero, the relative change is handled according to special rules).

[0091] When the absolute value of the absolute change is greater than zero, it is determined that the tillage frequency has changed. The observation period in which the change occurred is recorded as the next observation period. The tillage frequency before the change is the tillage frequency of the previous observation period, and the tillage frequency after the change is the tillage frequency of the next observation period. Tillage frequency change records are compiled according to the farmland plot identification and include fields such as farmland plot identification, observation period of change, tillage frequency before change, tillage frequency after change, and change magnitude (absolute change and relative change).

[0092] Step S136: Collect records of tillage frequency changes for all cultivated land plots, analyze the tillage frequency change trend of the same cultivated land plot during continuous observation periods, compare the tillage frequency changes of different cultivated land plots during the same observation period, summarize the overall change pattern of tillage frequency with the observation period, and form a tillage frequency fluctuation pattern.

[0093] Records of tillage frequency changes for all cultivated land plots were compiled and grouped according to plot identification. The trend of tillage frequency changes for the same cultivated land plot over continuous observation periods was analyzed. The trend was determined using linear regression analysis, with the observation period number as the independent variable and tillage frequency as the dependent variable. The sign of the regression coefficient indicated whether the trend was increasing, decreasing, or stationary. The significance of the trend was also calculated using a significance test. A trend was considered significant if the test result was below a certain significance level.

[0094] By comparing the changes in tillage frequency of different farmland plots during the same observation period, the average, standard deviation, maximum, and minimum values ​​of tillage frequency changes for all farmland plots within the same observation period are calculated. The degree of concentration and dispersion of the changes are analyzed. The overall variation pattern of tillage frequency over the observation period is summarized, including: the average magnitude of tillage frequency changes in each observation period, the main direction of change (increase or decrease), and the spatial distribution characteristics of the changes (which areas have larger changes and which areas have smaller changes).

[0095] The fluctuation pattern of tillage frequency is represented by a pattern descriptor. Each pattern descriptor contains information such as the observation period sequence, average variation amplitude, distribution of variation direction, spatial distribution characteristics, and typical influencing factors (such as soil moisture changes and differences in topographic slope). Regions that show the same variation trend in multiple observation periods are marked as stable fluctuation regions; regions with unstable variation trends are marked as sensitive fluctuation regions.

[0096] Step S137: Integrate the planting type conversion pattern and the fluctuation pattern of cultivation frequency according to the observation period, label the cultivated land plot identifier and spatial correlation information corresponding to each pattern and pattern, and fill in the missing observation period data after integration by inferring the pattern of adjacent time periods, and finally form a set of cultivated land spatiotemporal interaction features.

[0097] Planting type conversion patterns and tillage frequency fluctuation patterns are integrated according to observation periods. The integration uses the observation period as the keyword to match planting type conversion patterns and tillage frequency fluctuation patterns within the same observation period. A farmland plot identifier field is added to each pattern and pattern to clearly identify the farmland plot corresponding to that pattern or pattern. A spatial association information field is added, referencing the spatial association unit of that farmland plot in the farmland spatial coupling feature set, including adjacent farmland plot identifiers, boundary overlap features, and environmental difference features.

[0098] For data from missing observation periods after integration—that is, when a planting type conversion pattern exists but the fluctuation pattern of cultivation frequency is missing in a certain observation period, or vice versa—the missing data is filled in by deducing the patterns of adjacent observation periods. The derivation method is as follows: if the patterns of the preceding and following observation periods are consistent, then the missing period is filled in using that pattern; if the patterns of the preceding and following observation periods are different, then a weighted average method is used, with the weight being the reciprocal of the time distance from the missing period.

[0099] The spatiotemporal interaction feature set of cultivated land is an integrated collection of planting type conversion patterns and cultivation frequency fluctuation patterns. Indexed by observation period and cultivated land plot identifier, it supports multi-dimensional querying and analysis. Each element in the set comprises five parts: observation period identifier, cultivated land plot identifier, planting type conversion pattern descriptor, cultivation frequency fluctuation pattern descriptor, and spatial association information reference. It is stored using a relational database table structure for easy integration with subsequent cultivated land loss risk transmission models.

[0100] Step S140: Input the set of spatial coupling features of cultivated land and the set of spatiotemporal interaction features of cultivated land into the cultivated land loss risk transmission model. Through the dynamic correlation of spatial and temporal features, identify the transmission path of cultivated land loss risk between adjacent cultivated land plots, and generate cultivated land loss risk assessment results containing the loss tendency and risk transmission factors of each cultivated land plot.

[0101] Step S141: Classify the spatial association units of cultivated land plots in the cultivated land spatial coupling feature set according to the observation period, so that the spatial association information of the same observation period is concentrated. Also classify the planting type conversion pattern and the cultivation frequency fluctuation pattern in the cultivated land spatiotemporal interaction feature set according to the observation period, and establish the correspondence between the observation period and the feature information.

[0102] The spatial correlation units of cultivated land plots in the cultivated land spatial coupling feature set are classified according to the observation period. Each observation period corresponds to a classification group, which contains all spatial correlation units of cultivated land plots in that observation period (including boundary overlap features, environmental difference features, etc.). At the same time, the planting type conversion patterns and tillage frequency fluctuation patterns in the cultivated land spatiotemporal interaction feature set are also classified according to the observation period. Each observation period corresponds to a set of planting type conversion patterns and a set of tillage frequency fluctuation patterns.

[0103] A correspondence table between observation periods and characteristic information is established. Each row in the table corresponds to one observation period and includes four fields: observation period identifier, spatial association information group reference, planting type conversion mode group reference, and tillage frequency fluctuation pattern group reference. This correspondence table allows for the rapid acquisition of all spatial and temporal characteristic information for any given observation period.

[0104] Step S142: Input the spatial correlation unit of cultivated land plots, planting type conversion pattern and cultivation frequency fluctuation pattern of cultivated land plots in the same observation period into the cultivated land loss risk transmission model. First, correlate the boundary overlap information and planting type conversion pattern in the spatial correlation unit of cultivated land plots, and analyze the correlation logic between the boundary overlap range and planting type conversion between adjacent cultivated land plots.

[0105] The spatial correlation units of cultivated land plots, planting type conversion patterns, and cultivation frequency fluctuations during the same observation period are used as input data and fed into the cultivated land loss risk transmission model. The cultivated land loss risk transmission model consists of four layers: data preprocessing layer, feature association layer, path identification layer, and risk assessment layer.

[0106] In the feature association layer, the association analysis between boundary overlap information and planting type conversion patterns is first performed. Boundary overlap range data, including the total overlap length and overlap rate, is extracted from the spatial association units of cultivated land plots. The cultivated land plot identifier pairs (source cultivated land plot identifier and target cultivated land plot identifier) ​​are extracted from the planting type conversion patterns. By associating the two through the cultivated land plot identifier pairs, the correspondence between the boundary overlap range of adjacent cultivated land plot pairs and the planting type conversion is obtained.

[0107] When analyzing the correlation logic, the probability of planting type conversion within different boundary overlap ranges is calculated. This involves counting the number of adjacent farmland plots that undergo planting type conversion within a certain boundary overlap range, and comparing this to the total number of all adjacent farmland plot pairs within that boundary overlap range. Simultaneously, the relationship between the boundary overlap direction and the conversion direction is analyzed. For example, when the boundary overlap portion is located in a specific direction of a farmland plot (such as the eastern boundary), is the probability of planting type conversion higher than in other directions?

[0108] Step S143: Connect the environmental differences and the fluctuation patterns of cultivation frequency in the spatial correlation units of cultivated land plots, analyze the correlation logic between soil moisture and topographic slope differences and cultivation frequency changes between adjacent cultivated land plots, and form a dual correlation relationship of spatial and temporal characteristics.

[0109] Environmental difference information is extracted from the spatial association units of cultivated land plots, including soil moisture differences (divided into three levels) and topographic slope differences (average, maximum, and minimum). Cultivated land plot identifier pairs (source cultivated land plot identifier and target cultivated land plot identifier) ​​are extracted from the fluctuation patterns of cultivation frequency. By linking the two cultivated land plot identifier pairs, the correspondence between environmental difference information and cultivation frequency changes of adjacent cultivated land plot pairs is obtained.

[0110] When analyzing correlation logic, intervals are divided according to the degree of environmental difference, and the probability of changes in tillage frequency within each interval is calculated. For example, the relative difference in soil moisture is divided into several intervals, and the proportion of adjacent farmland plots experiencing changes in tillage frequency within each interval is statistically analyzed. Simultaneously, the relationship between the direction of environmental difference and the direction of tillage frequency change is analyzed; for example, the probability of an increase or decrease in tillage frequency when the soil moisture of the main farmland plot is higher than that of adjacent farmland plots.

[0111] By integrating the correlation logic between boundary overlap and planting type conversion, and the correlation logic between environmental differences and changes in tillage frequency, a dual correlation relationship of spatial and temporal characteristics is formed. The dual correlation relationship is represented in matrix form, with rows representing adjacent farmland plots, columns representing correlation characteristics (boundary overlap range, soil moisture difference, terrain slope difference, probability of planting type conversion, probability of tillage frequency change, etc.), and matrix elements being the corresponding feature values.

[0112] Step S144: Based on the dual association, identify the potential transmission path of risk from cultivated land plots that have undergone planting type conversion or cultivation frequency change to their adjacent cultivated land plots, record the adjacent cultivated land plots through which the risk starts from the source cultivated land plot, and form a preliminary risk transmission path.

[0113] Step S1441: Select farmland plots that have undergone planting type conversion from the dual association relationship, mark the farmland plots as source farmland plots of planting type conversion, and select farmland plots that have undergone changes in cultivation frequency, mark the farmland plots as source farmland plots of cultivation frequency change.

[0114] Pairs of farmland parcels with a planting type conversion probability greater than a certain threshold are extracted from the dual association relationships. The farmland parcel in the pair that undergoes conversion first is marked as the source farmland parcel for the planting type conversion. Similarly, pairs of farmland parcels with a tillage frequency change probability greater than a certain threshold are extracted, and the farmland parcel that changes first is marked as the source farmland parcel for the tillage frequency change. The threshold is determined based on statistical analysis of historical data; a threshold is set when the conversion or change probability is higher than the average probability of all farmland parcel pairs plus twice the standard deviation.

[0115] Step S1442: For each source cultivated land plot undergoing planting type conversion, extract the identifiers of all its adjacent cultivated land plots from the cultivated land plot spatial association unit, and check whether the adjacent cultivated land plots also undergo planting type conversion in subsequent observation periods. If any adjacent cultivated land plot undergoes conversion within the observation period after the source cultivated land plot conversion, then mark the source cultivated land plot and the adjacent cultivated land plot as potential risk transmission nodes.

[0116] For each source farmland plot undergoing a planting type conversion, all adjacent farmland plot identifiers are extracted from the list of adjacent farmland plot identifiers in the spatial association unit of farmland plots. Then, the planting type data of these adjacent farmland plots during the observation period after the planting type conversion of the source farmland plot is queried. If the planting type conversion occurs in the first or second observation period after the conversion of the source farmland plot, it is determined that there is a possibility of risk transmission, and the source farmland plot and the adjacent farmland plot are marked as potential risk transmission nodes.

[0117] Step S1443: For each source farmland plot with a change in tillage frequency, extract the identifiers of all its adjacent farmland plots from the spatial association unit of the farmland plot, and check whether the adjacent farmland plots also experienced changes in tillage frequency in subsequent observation periods. If any adjacent farmland plot changes during the observation period after the source farmland plot changes, mark the source farmland plot and the adjacent farmland plot as potential risk transmission nodes.

[0118] For each source farmland plot where the tillage frequency changes, perform operations similar to step S1442, extract the identifiers of adjacent farmland plots, query the tillage frequency data for subsequent observation periods, and mark the tillage frequency as a potential risk transmission node if a tillage frequency change occurs during the observation period after the source farmland plot changes.

[0119] Step S1444: Connect potential risk transmission nodes under the same risk transmission logic according to the order of observation time periods. If the adjacent cultivated land plots of the source cultivated land plot of planting type conversion are converted during the observation time period after the source cultivated land plot is converted, a continuous preliminary risk transmission path is formed.

[0120] Based on the chronological order of the observation periods, potential risk transmission nodes belonging to the same risk transmission logic (planting type conversion or change in cultivation frequency) are connected together. For the risk of planting type conversion, starting from the source cultivated land plot, if its adjacent cultivated land plot A converts in a subsequent observation period, and the adjacent cultivated land plot B of cultivated land plot A also converts in the observation period after cultivated land plot A converts, then a preliminary risk transmission path is formed: source cultivated land plot → cultivated land plot A → cultivated land plot B.

[0121] Step S1445: If the adjacent cultivated land plots of the source cultivated land plot change during the observation period after the change of the source cultivated land plot, a continuous preliminary risk transmission path is formed.

[0122] For the risk of changes in tillage frequency, a method similar to step S1444 is used to connect potential risk transmission nodes to form a preliminary risk transmission path. The preliminary risk transmission path is represented in the form of a directed graph, where nodes are identified as farmland plots, directed edges represent the direction of risk transmission, and the observation period and transmission probability (based on the conversion or change probability in the dual association relationship) are marked on the edges.

[0123] Step S145: Verify the preliminary risk transmission path. By comparing the risk transmission situation in different observation periods, confirm the continuity and stability of the preliminary risk transmission path, remove temporary transmission paths that only appear in a single observation period, and retain stable risk transmission paths that exist in multiple observation periods.

[0124] All preliminary risk transmission paths were collected and grouped by path identifier. Each group contained path records for all observation periods in the same transmission direction. Continuity verification was performed on each group of path records, i.e., checking whether the path existed in consecutive observation periods. For example, if a path existed in observation periods 1 and 3, but not in observation period 2, it was considered discontinuous.

[0125] Stability verification is performed by calculating the frequency of a path's occurrence across multiple observation periods. The frequency is the ratio of the number of observation periods in which the path exists to the total number of observation periods. When the frequency exceeds a certain stability threshold, the path is considered a stable risk transmission path; otherwise, it is a temporary transmission path. The stability threshold is determined based on the total number of observation periods; the more observation periods there are, the lower the threshold can be.

[0126] After removing temporary transmission paths, stable risk transmission paths are retained. For stable risk transmission paths, their transmission strength is further analyzed. The transmission strength comprehensively considers factors such as transmission probability, boundary overlap range, and degree of environmental difference, and is calculated using a weighted summation method. The weights are determined by the analytic hierarchy process (AHP), with transmission probability having the highest weight, boundary overlap range second, and degree of environmental difference the lowest.

[0127] Step S146: Based on the stable risk transmission path, analyze the role of each cultivated land plot in the stable risk transmission path. If the cultivated land plot is the starting point of the risk, calculate the possibility of it causing the loss of surrounding cultivated land plots. If the cultivated land plot is the receiving point of the risk, calculate the possibility of it being lost due to the influence of the source cultivated land plot. The loss tendency of each cultivated land plot is obtained by combining the results.

[0128] Based on the directed graph structure of the stable risk transmission path, the role of each farmland plot is analyzed. If a farmland plot is the starting node of a certain path and not the ending node of other paths, it is determined to be a pure source farmland plot; if it is both the starting node and the ending node, it is determined to be a mixed role farmland plot; if it is only the ending node, it is determined to be a pure receiving farmland plot.

[0129] For pure source farmland plots, the probability of causing loss to surrounding farmland plots is calculated based on the sum of the transmission strength of all directed edges of its output; the greater the transmission strength, the higher the probability of loss. For pure receiving farmland plots, the probability of loss due to the influence of source farmland plots is calculated based on the sum of the transmission strength of all directed edges of its input. For mixed-role farmland plots, the probability of causing loss and the probability of being affected are calculated separately, and the larger of the two values ​​is taken as the overall loss tendency.

[0130] The tendency to lose land use is assessed using a five-level scoring system, ranging from very low to very high: low, medium, high, and very high. The scoring is based on the division of probability values ​​into intervals, with the interval boundaries determined by percentiles. For example, the probability values ​​of all cultivated land plots are ranked, and the 20%, 40%, 60%, and 80% percentiles are used as the interval boundaries.

[0131] Step S147: Extract the feature information that has the most direct impact on the loss tendency of each arable land plot. If the loss tendency is mainly affected by the change of planting type of adjacent arable land plots, then the relevant information of the change of planting type is the risk transmission factor. If it is mainly affected by the change of farming frequency caused by environmental differences, then the relevant information of the environmental differences is the risk transmission factor.

[0132] Sensitivity analysis was used to extract the most direct features affecting the tendency to leach soil. The sensitivity analysis method involved sequentially changing the value of each feature (within a reasonable range) and observing the magnitude of the change in the tendency to leach soil. The greater the magnitude of the change, the more direct the impact of that feature. Feature information included the probability of planting type conversion, boundary overlap rate, relative difference in soil moisture, and absolute difference in terrain slope.

[0133] If the change in loss tendency is mainly caused by changes in the probability of planting type conversion, then the relevant information on planting type conversion is identified as the risk transmission factor, including the planting type code before conversion, the planting type code after conversion, and the observation interval of the conversion. If it is mainly caused by changes in the relative difference in soil moisture or the absolute difference in topographic slope, then the relevant information on environmental differences is identified as the risk transmission factor, including the level of soil moisture difference and the type of topographic slope difference (average, maximum, or minimum).

[0134] Risk transmission factors are represented by factor descriptors, which include information such as factor type (transformation of planting type or environmental differences), specific factor content, and impact weight (the magnitude of change obtained from sensitivity analysis).

[0135] Step S148: Organize the loss tendency and corresponding risk transmission factors of each cultivated land plot according to the cultivated land plot identification. For the missing cultivated land plot information after sorting, derive and supplement it through the characteristics of surrounding cultivated land plots, and mark the corresponding stable risk transmission path to finally form the cultivated land loss risk assessment result.

[0136] The loss tendency and risk transmission factors of cultivated land are compiled into an assessment record table based on the cultivated land plot identification. Each row in the table corresponds to a cultivated land plot and includes fields such as cultivated land plot identification, loss tendency level, risk transmission factor descriptor, and reference to stable risk transmission path. For missing cultivated land plot information, i.e., a cultivated land plot does not appear in the assessment record table, it is supplemented by deducing and filling in the missing information by looking up the characteristic information of surrounding cultivated land plots.

[0137] The derivation and completion method is as follows: Select all adjacent cultivated land parcels of the cultivated land parcel, calculate the weighted average of the loss tendency levels of these adjacent cultivated land parcels, with the weight being the overlap rate of adjacent boundaries; the risk transmission factor adopts the factor type that appears most frequently among the adjacent cultivated land parcels. When labeling the corresponding stable risk transmission path, the directed graph identifier of the stable risk transmission path is referenced to indicate the position of the cultivated land parcel in the path (source node, intermediate node, or receiving node).

[0138] The results of the farmland loss risk assessment are a set of assessment record tables and directed graphs of stable risk transmission paths. They support querying farmland plots by loss tendency level, or querying their risk transmission factors and the stable risk transmission paths they belong to by farmland plot identifier.

[0139] Step S150: Based on the results of the farmland loss risk assessment, divide the continuous farmland loss risk areas, design farmland protection intervention schemes in combination with the transmission paths of each farmland loss risk area, and generate a set of farmland protection strategies.

[0140] Step S151: Analyze the loss tendency of each arable land plot in the arable land loss risk assessment results, and classify the arable land plots according to the manifestation of the loss tendency. If the loss tendency of the arable land plot is mainly reflected in the easy occurrence of planting type conversion, it is classified as the first type of arable land plot; if it is mainly reflected in the easy occurrence of changes in cultivation frequency, it is classified as the second type of arable land plot.

[0141] The results of the farmland loss risk assessment were analyzed to extract the loss tendency level and risk transmission factor descriptors for each farmland plot. Based on the factor type of the risk transmission factor descriptors, farmland plots were classified: if the factor type is a change in planting type, they are classified as Class I farmland plots; if it is a change in farming frequency due to environmental differences, they are classified as Class II farmland plots. For farmland plots with mixed factor types, the dominant factor type was determined based on the magnitude of its influence weight, and the plots were classified into the corresponding categories.

[0142] After classification, statistics were collected on the two types of cultivated land plots separately, and the number, area proportion, and spatial distribution characteristics of each type of cultivated land plot were calculated.

[0143] Step S152: Based on the boundary coordinates and classification results of cultivated land plots, adjacent cultivated land plots belonging to the same category are merged at the spatial level. Referring to the transmission path, adjacent cultivated land plots with stable risk transmission paths are preferentially included in the same merging scope to form continuous cultivated land loss risk areas.

[0144] Based on the boundary coordinates and classification results of cultivated land parcels, spatial merging operations are performed in the geographic information system. The merging rule is as follows: cultivated land parcels belonging to the same category that are adjacent and have a boundary overlap rate greater than a certain merging threshold are merged into a preliminary region. The merging threshold is determined based on the average area of ​​cultivated land parcels; the merging threshold for smaller cultivated land parcels can be appropriately lowered.

[0145] Referring to stable risk transmission paths, adjacent farmland plots with direct directed edge connections are preferentially included in the same merging scope, regardless of whether the boundary overlap rate reaches the merging threshold. For example, if farmland plot A is a Class I farmland plot, and its adjacent farmland plot B is also Class I, and there is a stable risk transmission path from A to B, then A and B will be forcibly merged into the same area.

[0146] The merging process employs a regional growth method, starting with a seed farmland plot and gradually incorporating eligible adjacent farmland plots into the region until expansion is impossible. Seed farmland plots are selected based on their extremely high loss tendency, ensuring that the highest-risk areas are merged first. The resulting contiguous regions must have an area greater than a certain minimum area threshold; areas smaller than this threshold are merged with the adjacent largest region.

[0147] Step S153: For each farmland loss risk area, extract the transmission path of all farmland plots in the farmland loss risk area, determine the main diffusion direction of the risk in the farmland loss risk area, and extract the spatial association unit and spatiotemporal interaction characteristics of the farmland plots corresponding to the farmland loss risk area to determine the core spatial and temporal factors that cause the risk in the farmland loss risk area.

[0148] For example, step S1531: Based on the boundary range of the farmland loss risk area, determine the identifiers of all farmland plots included in the farmland loss risk area, and extract the transmission path corresponding to the farmland plot identifiers from the farmland loss risk assessment results.

[0149] Based on the boundary range of the farmland loss risk area, spatial overlay analysis is used to identify all farmland plots within the area. The boundary range is represented by the minimum bounding rectangle, which includes four parameters: minimum longitude, maximum longitude, minimum latitude, and maximum latitude. From the directed graph of stable risk transmission paths in the farmland loss risk assessment results, the transmission paths of all farmland plots at each node belonging to that area are extracted to form a subgraph of transmission paths within the area.

[0150] Step S1532: Classify the extracted transmission paths according to risk type, such as planting type conversion risk or cultivation frequency change risk. Count the number of transmission paths of each risk type in the farmland loss risk area. The transmission path corresponding to the risk type with the highest number is the dominant risk transmission path in the farmland loss risk area.

[0151] The transmission path submaps within the region are categorized by risk type, which is determined based on the risk transmission factor type of the path. The number of paths for each risk type is counted, and the percentage (the ratio of the number of paths of a certain type to the total number of paths in the region) is calculated. The transmission path corresponding to the risk type with the highest percentage is the dominant risk transmission path in the region. If the difference between the percentages of two risk types is less than a certain threshold, it is determined to be a mixed dominant risk transmission path.

[0152] Step S1533: Analyze the distribution of farmland plots in the dominant risk transmission path of the farmland loss risk area, determine the concentrated area of ​​the source farmland plots at the starting node of the dominant risk transmission path, and determine the extension direction of the affected farmland plots at subsequent nodes of the dominant risk transmission path. This extension direction is the main diffusion direction of the risk in the farmland loss risk area.

[0153] The spatial distribution of farmland plots at key nodes in the dominant risk transmission path was analyzed. Kernel density estimation was used to calculate the distribution density of the source farmland plots at the initial node; the area with the highest density was identified as the concentration area. The directional angles of all affected farmland plots at subsequent nodes relative to the initial node were calculated, and the frequency of these directional angles was statistically analyzed. The range of directional angles with the highest frequency was identified as the main diffusion direction. The directional angles were divided into eight directions (east, south, west, north, southeast, northeast, southwest, and northwest), with each direction corresponding to a specific angle range.

[0154] Step S1534: Extract the spatial association units of all cultivated land plots in the cultivated land loss risk area from the cultivated land spatial coupling feature set, analyze the boundary overlap information and environmental difference information in the cultivated land plot spatial association units, and if the boundary overlap range between the source cultivated land plot and the affected cultivated land plot in the dominant risk transmission path of the cultivated land loss risk area meets the preset range, then the boundary overlap related information is the core spatial factor causing the risk of the cultivated land loss risk area.

[0155] From the set of spatial coupling features of cultivated land, spatial association units of all cultivated land parcels within the region are extracted, and boundary overlap information (total length, overlap rate) and environmental difference information (soil moisture difference, topographic slope difference) are analyzed. For the source cultivated land parcel and affected cultivated land parcel pair in the dominant risk transmission path, it is checked whether their boundary overlap range is within the preset range. The preset range is determined based on the boundary overlap range of risk transmission in historical data, and is usually the average overlap rate plus or minus the standard deviation.

[0156] If more than 70% of the farmland plots have boundary overlap within a preset range, then boundary overlap-related information is determined to be a core spatial factor, including the average boundary overlap rate and distribution characteristics.

[0157] Step S1535: If the difference in soil moisture or topographic slope between the source farmland plot and the affected farmland plot in the dominant risk transmission path of the farmland loss risk area meets the preset difference range, then the environmental difference-related information is the core spatial factor causing the risk of the farmland loss risk area.

[0158] Similar to step S1534, check whether the relative difference in soil moisture or the absolute difference in topographic slope between the source farmland plot and the affected farmland plot is within the preset difference range. The preset difference range is also determined based on historical data statistics. If more than 70% of the farmland plots have environmental differences within the preset range, then environmental difference-related information is determined as the core spatial factor, including the hierarchical distribution of soil moisture differences and the type distribution of topographic slope differences.

[0159] Step S1536: Extract the planting type conversion pattern and the fluctuation pattern of cultivation frequency corresponding to the arable land loss risk area from the arable land spatiotemporal interaction feature set. If the risk type of the arable land loss risk area is planting type conversion risk, then analyze the conversion category with the highest proportion in the planting type conversion pattern. The relevant information of this conversion category is the core temporal factor that leads to the risk of the arable land loss risk area.

[0160] Extract planting type conversion patterns and cultivation frequency fluctuations corresponding to areas at risk of arable land loss from the spatiotemporal interaction feature set of arable land. If the risk type of the region is planting type conversion risk, count the proportion of all planting type conversion categories in the region. The conversion category with the highest proportion (such as the conversion of food crops to cash crops) is the core temporal factor. Relevant information includes the main observation period of the conversion and the spatial clustering characteristics of the conversion.

[0161] Step S1537: If the risk type of the arable land loss risk area is the risk of changes in tillage frequency, then analyze the most frequent trend in the tillage frequency fluctuation pattern. The relevant information of this trend is the core temporal factor that leads to the risk of arable land loss in this area.

[0162] If the regional risk type is the risk of changes in tillage frequency, count the number of occurrences of all tillage frequency change trends in the region. The trend with the most occurrences (such as a continuous decreasing trend) is the core time series factor. Relevant information includes the starting observation period of the trend and the spatial distribution of the magnitude of change.

[0163] Step S154: If the main arable land in the arable land loss risk area is Class I arable land, then design protection intervention content for planting type conversion, including planning the planting types that can be retained in the arable land loss risk area, determining the range of planting type conversion restrictions in the arable land loss risk area, and setting up planting type monitoring points in arable land plots at key nodes of the transmission path according to the transmission path.

[0164] If Class I arable land accounts for more than 60% of the area, it is considered primarily Class I arable land. When planning the permitted planting types, priority should be given to planting types that are highly compatible with the region's soil and climate conditions and have high ecological value, taking into account local agricultural development plans and ecological protection requirements. The list of permitted planting types must include specific crop types and their percentage requirements; for example, food crops must account for no less than a certain percentage.

[0165] The scope of conversion restrictions includes prohibited conversion types and restricted conversion types. Prohibited conversion types refer to those that are not allowed to be converted from a permitted planting type, such as those prohibited from being converted to a planting type for non-arable land use; restricted conversion types refer to those that require approval and whose conversion area cannot exceed a certain proportion of the total area of ​​the region.

[0166] Based on the stable risk transmission path, planting type monitoring points are set up in key node farmland plots. Key nodes include central farmland plots in areas with concentrated source farmland plots, intermediate farmland plots in the main diffusion direction, and receiving farmland plots at the regional boundary. Monitoring points adopt the form of fixed sample plots with a standard area. Planting type data are collected monthly, and data recording is done using a combination of photographs and text descriptions.

[0167] Step S155: If the main arable land in the arable land loss risk area is Class II arable land, then design protection intervention content based on changes in tillage frequency, including regulating the soil moisture and topographic slope related environmental conditions in the arable land loss risk area, setting the regulation range of soil moisture and topographic slope in the arable land loss risk area, and setting environmental monitoring points at key nodes of arable land in the transmission path according to the transmission path.

[0168] If the proportion of Class II arable land in a region exceeds 60%, it is considered to be primarily Class II arable land. When regulating soil moisture environmental conditions, irrigation or drainage plans are formulated based on soil moisture difference information in the core spatial factors. For areas where soil moisture is below the suitable range, the frequency and amount of irrigation are increased; for areas above the suitable range, drainage facilities are constructed or agronomic measures such as deep tillage and loosening of the soil are adopted. The suitable range is determined based on the water requirement characteristics of the main crop types in the region.

[0169] When regulating environmental conditions related to topographic slope, for cultivated land with a slope exceeding a certain critical value, soil and water conservation measures such as terracing and contour farming are adopted; for areas with large slope differences, local land leveling is used to reduce slope differences. The regulation range includes the target range of soil moisture (e.g., surface soil moisture within a certain range) and the target range of topographic slope (e.g., average slope less than a certain value).

[0170] Based on the stable risk transmission path, environmental monitoring points are set up in key farmland plots. Automatic monitoring equipment is deployed at the monitoring points to monitor soil moisture (three levels) and topographic slope changes in real time. Data is uploaded to the management platform every hour, and abnormal data will trigger an automatic alarm.

[0171] Step S156: Integrate the protection and intervention content for areas at risk of farmland loss with the monitoring point setting plan, determine the specific farmland plots to be implemented for each intervention operation, the pattern of the observation period corresponding to the implementation timing, and mark the boundary range and core risk factors of the farmland loss risk area corresponding to the farmland protection and intervention plan.

[0172] The protection intervention content and monitoring point setting plan are integrated into a farmland protection intervention plan, which includes two parts: an intervention measure list and a monitoring plan. In the intervention measure list, each intervention operation is clearly identified as a specific farmland plot, and the timing of implementation is determined based on core time-series factors. For example, the planting type conversion intervention is implemented one month before the observation period when the conversion is more likely to occur, and the tillage frequency change intervention is implemented during the season when soil moisture differences are greater.

[0173] The boundary of areas at risk of farmland loss is marked using the coordinates of the least circumscribed rectangle, and the core risk factors reference descriptors for core spatial factors and core temporal factors. The monitoring plan includes monitoring point numbers, deployment locations (farmland plot identifiers and specific coordinates), monitoring indicators, monitoring frequency, data transmission methods, and data processing procedures.

[0174] Step S157: Collect all farmland protection intervention plans for all farmland loss risk areas, perform matching analysis on the intervention logic of each farmland protection intervention plan and the risk characteristics of the corresponding farmland loss risk areas, remove intervention content that does not match the risk characteristics of farmland loss risk areas, and finally form a set of farmland protection strategies.

[0175] We collected all farmland protection intervention plans for areas at risk of farmland loss and established an evaluation index system for these plans. The indicators included the matching degree between intervention measures and risk types, the feasibility of intervention measures (technical and economic feasibility), and the comprehensiveness of the monitoring plan. We used an expert scoring method to rate the indicators of each plan on a scale of 1 to 5.

[0176] During the matching analysis, if the risk type of several preventive measures is inconsistent with the regional risk type, or if the feasibility score is less than 3, the intervention will be removed. For example, for a region with a risk of crop type conversion, if an intervention to regulate soil moisture is included but its impact on crop type conversion is not significant, the intervention will be removed.

[0177] The final set of farmland protection strategies includes all farmland protection intervention programs that have passed matching analysis. These programs are categorized and stored according to risk area type and loss tendency level, supporting queries for programs by region or applicable areas by intervention type. The farmland protection strategy set is regularly evaluated and updated, with the evaluation cycle coinciding with the observation period, adjusting intervention programs and strategies based on new farmland spatiotemporal change tracking data.

[0178] Based on the same inventive concept, please refer to Figure 2 The diagram shows a schematic block diagram of a farmland loss assessment system 100 based on the spatiotemporal change characteristics analysis of farmland, which is used to perform the above-described inspection video stream processing method according to an embodiment of this application. The farmland loss assessment system 100 based on the spatiotemporal change characteristics analysis of farmland may include a communication unit 110, a machine-readable storage medium 120, and a processor 130.

[0179] In this embodiment, both the machine-readable storage medium 120 and the processor 130 are located in the farmland loss assessment system 100 based on the analysis of farmland spatiotemporal change characteristics and are separately configured. However, it should be understood that the machine-readable storage medium 120 may also be independent of the farmland loss assessment system 100 based on the analysis of farmland spatiotemporal change characteristics and may be accessed by the processor 130 through a bus interface. Alternatively, the machine-readable storage medium 120 may also be integrated into the processor 130 and may communicate and interact with external systems through the communication unit 110.

[0180] The processor 130 is the control center of the farmland loss assessment system 100 based on the analysis of farmland spatiotemporal change characteristics. It connects various parts of the system via various interfaces and lines, and performs various functions and processes data by running or executing software programs and / or modules stored in the machine-readable storage medium 120, and by calling data stored in the machine-readable storage medium 120, thereby providing overall monitoring of the farmland loss assessment system 100. Optionally, the processor 130 may include one or more processing cores; for example, the processor 130 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may also not be integrated into the processor. The machine-readable storage medium 120 is used to store machine-executable instructions for executing the scheme of this application, and the processor 130 is used to execute the machine-executable instructions stored in the machine-readable storage medium 120 to implement the inspection video stream processing method provided in the aforementioned method embodiments.

[0181] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A cultivated land loss evaluation method based on cultivated land space-time variation characteristic analysis, characterized in that, The method comprises: acquiring a cultivated land spatio-temporal change tracking data set containing different observation period cultivated land spatial distribution data, utilization state data and surrounding environment data, the cultivated land spatial distribution data containing cultivated land plot boundary coordinates and cultivated land plot adjacent relationship, the utilization state data containing planting type and tillage frequency, and the surrounding environment data containing soil humidity and terrain slope; performing cultivated land spatial feature modeling processing on the cultivated land spatio-temporal change tracking data set, extracting the adjacent relationship corresponding to the cultivated land plot boundary coordinates, combining the fluctuation of the soil humidity and the terrain slope between adjacent cultivated land plots according to the cultivated land plot adjacent relationship, and obtaining a cultivated land spatial coupling feature set; performing cultivated land time series feature modeling processing on the cultivated land spatio-temporal change tracking data set, combining the cultivated land spatial coupling feature set, analyzing the conversion mode of the planting type with the observation period and the fluctuation law of the tillage frequency with the observation period, and obtaining a cultivated land spatio-temporal interaction feature set; inputting the cultivated land spatial coupling feature set and the cultivated land spatio-temporal interaction feature set into a cultivated land loss risk conduction model, identifying the conduction path of the cultivated land loss risk between adjacent cultivated land plots through the dynamic association of spatial and time series features, and generating a cultivated land loss risk assessment result containing the loss tendency of each cultivated land plot and a risk conduction factor; dividing continuous cultivated land loss risk regions according to the cultivated land loss risk assessment result, combining the conduction path of each cultivated land loss risk region to design a cultivated land protection intervention scheme, and generating a cultivated land protection strategy set. 2.The cultivated land loss evaluation method based on cultivated land space-time variation feature analysis according to claim 1, characterized in that, The cultivated land spatial feature modeling processing performed on the cultivated land spatio-temporal change tracking data set, the extraction of the adjacent relationship corresponding to the cultivated land plot boundary coordinates, the combination of the fluctuation of the soil humidity and the terrain slope between adjacent cultivated land plots according to the cultivated land plot adjacent relationship, and the obtaining of the cultivated land spatial coupling feature set comprise: extracting the cultivated land spatial distribution data of the same observation period from the cultivated land spatio-temporal change tracking data set, separating the cultivated land plot boundary coordinates of each cultivated land plot according to the cultivated land plot identifier, and extracting the cultivated land plot adjacent relationship of all cultivated land plots in the observation period to determine all adjacent cultivated land plot identifiers corresponding to each cultivated land plot; determining the boundary overlap part of each cultivated land plot and the corresponding adjacent cultivated land plot according to the cultivated land plot boundary coordinates of each cultivated land plot and the cultivated land plot boundary coordinates of the corresponding adjacent cultivated land plot, recording the specific range of the boundary overlap, and counting the total range of the boundary overlap between each cultivated land plot and all adjacent cultivated land plots to establish a boundary association file of the cultivated land plot and the adjacent cultivated land plot; extracting the surrounding environment data of the same observation period, matching the soil humidity data and the terrain slope data to the corresponding cultivated land plot according to the cultivated land plot identifier to form an environment data file of each cultivated land plot, and associating the environment data file of each cultivated land plot with the environment data file of the corresponding adjacent cultivated land plot according to the adjacent cultivated land plot identifier; comparing the soil humidity data of each cultivated land plot and its adjacent cultivated land plot, recording the difference of the soil humidity of the two, and comparing the terrain slope data of the two, recording the difference of the terrain slope of the two, and arranging to form an adjacent cultivated land plot environment difference record according to the cultivated land plot identifier. Integrate the total range of the boundary of each farmland plot with the corresponding adjacent farmland plot environment difference record, remove redundant information that only repeats the farmland plot identification, retain the core association logic of the boundary association and the environment difference, and form a farmland plot spatial association unit; Classify the farmland plot spatial association units of all farmland plots according to observation periods, complete the missing farmland plot spatial association units after classification by using the farmland plot boundary coordinates of the same observation period, and mark the corresponding observation period and farmland plot boundary coordinate range of each farmland plot spatial association unit, to finally form a farmland spatial coupling feature set. 3.The cultivated land loss evaluation method based on cultivated land spatio-temporal variation feature analysis according to claim 2, characterized in that, The farmland spatial distribution data of the same observation period is extracted from the farmland space-time change tracking data set, the farmland plot boundary coordinates of each farmland plot are separated according to the farmland plot identification, and the adjacent relationship of all farmland plots in the observation period is extracted, to determine the corresponding adjacent farmland plot identification of each farmland plot, including: The farmland spatial distribution data of the same observation period is extracted from the farmland space-time change tracking data set, the farmland plot boundary coordinates of each farmland plot are separated according to the farmland plot identification, and the adjacent relationship of all farmland plots in the observation period is extracted, to determine the corresponding adjacent farmland plot identification of each farmland plot, including: The farmland spatial distribution data of the same observation period is extracted from the farmland space-time change tracking data set, the farmland plot boundary coordinates of each farmland plot are separated according to the farmland plot identification, and the adjacent relationship of all farmland plots in the observation period is extracted, to determine the corresponding adjacent farmland plot identification of each farmland plot, including: The farmland plot boundary coordinates of each farmland plot are extracted one by one according to the farmland plot identification, the coordinate information includes the specific position record of each vertex on the periphery of the farmland plot, the missing vertex position of the farmland plot boundary coordinates is completed by using the adjacent farmland plot boundary coordinates, and the farmland plot boundary coordinates of each farmland plot are arranged according to the farmland plot identification to form a farmland plot boundary coordinate list; In the farmland spatial distribution data of the same observation period, the information module recording the adjacent relationship of the farmland plots is extracted, and the information module includes the corresponding record of each farmland plot identification and its adjacent farmland plot identification; The adjacent relationship record of the farmland plots is matched one by one according to the farmland plot identification, and the adjacent farmland plot identification of each farmland plot is extracted from the corresponding record to form a list of adjacent farmland plot identifications of each farmland plot; The adjacent farmland plot identification list of each farmland plot is checked, if there is a missing adjacent farmland plot identification of any farmland plot, whether there is an unrecorded adjacent farmland plot is judged by comparing the farmland plot boundary coordinates of the farmland plot with the farmland plot boundary coordinates of other surrounding farmland plots, and if there is, the corresponding adjacent farmland plot identification is supplemented; 4. The cultivated land loss evaluation method based on cultivated land space-time variation feature analysis according to claim 3, characterized in that, The farmland plot boundary coordinates of each farmland plot are associated with the corresponding adjacent farmland plot identification list, and the information of the association deviation is matched again by using the farmland plot identification to form a farmland plot spatial basic information unit. The adjacent farmland plot identification list of each farmland plot is checked, if there is a missing adjacent farmland plot identification of any farmland plot, whether there is an unrecorded adjacent farmland plot is judged by comparing the farmland plot boundary coordinates of the farmland plot with the farmland plot boundary coordinates of other surrounding farmland plots, and if there is, the corresponding adjacent farmland plot identification is supplemented, including: performing preliminary inspection on the list of adjacent cultivated land plot identifiers of each cultivated land plot, if the list is empty or contains only a small number of identifiers, and the cultivated land plot boundary coordinates of the cultivated land plot show that there are other cultivated land plots around the cultivated land plot, it is determined that the adjacent cultivated land plot identifiers of the cultivated land plot are missing; extracting the complete cultivated land plot boundary coordinates of the cultivated land plot, determining the positions of all vertices of the cultivated land plot boundary, and extracting the cultivated land plot boundary coordinates of all other cultivated land plots in the same observation period to determine the vertex positions of the other cultivated land plots; comparing the cultivated land plot boundary coordinates of the cultivated land plot with those of each of the other cultivated land plots to check whether there are overlapping boundary line segments, if there are overlapping boundary line segments, and the length of the overlapping line segment exceeds a preset determination length, it is determined that the two cultivated land plots are adjacent cultivated land plots; extracting the cultivated land plot identifiers of the cultivated land plots determined to be adjacent cultivated land plots but not present in the list of adjacent cultivated land plot identifiers of the cultivated land plot, and supplementing them to the list of adjacent cultivated land plot identifiers of the cultivated land plot; repeating the comparison until the cultivated land plot boundary coordinates of the cultivated land plot and the cultivated land plot boundary coordinates of all other cultivated land plots in the same observation period are compared; performing re-inspection on the supplemented list of adjacent cultivated land plot identifiers to confirm that each identifier in the list of adjacent cultivated land plot identifiers corresponds to a cultivated land plot that has overlapping boundary line segments with the cultivated land plot, removing misjudged non-adjacent cultivated land plot identifiers, and forming the list of adjacent cultivated land plot identifiers.

5. The cultivated land loss evaluation method based on cultivated land space-time variation feature analysis according to claim 1, characterized in that, The cultivated land time-space change tracking data set is subjected to cultivated land time sequence feature modeling processing, combined with the cultivated land space coupling feature set, the conversion mode of planting type with observation period and the fluctuation law of cultivation frequency with observation period are analyzed, and a cultivated land space-time interaction feature set is obtained, including: extracting the utilization state data of all observation periods from the cultivated land space-time change tracking data set, corresponding the planting type data and the cultivation frequency data of each observation period to the same cultivated land plot according to the cultivated land plot identifier, and matching and completing the utilization state data that is not determined to correspond to the cultivated land plot through the cultivated land plot boundary coordinates to form the utilization state time sequence record of each cultivated land plot; extracting the cultivated land plot space correlation unit of each cultivated land plot in different observation periods from the cultivated land space coupling feature set, and matching the cultivated land plot space correlation unit with the utilization state time sequence record of the corresponding cultivated land plot in order of observation period, and re-aligning the observation period data with matching deviation through the cultivated land plot identifier, so that the spatial correlation information and the utilization state information of the same cultivated land plot in the same observation period correspond to each other; traversing the utilization state time sequence record of each cultivated land plot in order of observation period, comparing the planting type data of adjacent observation periods, if the planting types of adjacent periods are different, recording the observation period when the conversion occurs, and recording the planting types before and after the conversion, and arranging the planting type conversion record according to the cultivated land plot identifier; The planting type conversion records of all the farmland plots are counted, the conversion categories are divided according to the combination of the planting types before and after the conversion, the occurrence of each conversion category in different observation periods is recorded, the distribution rule of the same conversion category in continuous observation periods is analyzed, and a planting type conversion mode is formed; The utilization state time sequence records of each farmland plot are traversed in the order of observation periods, the tillage frequency data of adjacent observation periods is extracted, the change of the tillage frequency data is analyzed, the observation period in which the tillage frequency changes is recorded, and the tillage frequency data before and after the change is recorded, and the tillage frequency change record is formed by arranging according to the farmland plot identifier; The tillage frequency change records of all the farmland plots are counted, the tillage frequency change trend of the same farmland plot in continuous observation periods is analyzed, and the tillage frequency change of different farmland plots in the same observation period is compared, the overall change rule of the tillage frequency with the observation period is summarized, and the tillage frequency fluctuation rule is formed; The planting type conversion mode and the tillage frequency fluctuation rule are integrated according to the observation period, the farmland plot identifier and the spatial correlation information corresponding to each mode and rule are marked, the missing observation period data after integration is completed by the adjacent period rule, and finally the farmland space-time interaction feature set is formed.

6. The cultivated land loss evaluation method based on cultivated land space-time variation feature analysis according to claim 5, characterized in that, The utilization state time sequence records of each farmland plot are traversed in the order of observation periods, the planting type data of adjacent observation periods is compared, if the planting types of adjacent periods are different, the observation period in which the conversion occurs is recorded, and the planting types before and after the conversion are recorded, and the planting type conversion record is arranged according to the farmland plot identifier, including: The planting type data of all observation periods of a single farmland plot is extracted from the utilization state time sequence record according to the farmland plot identifier, and arranged in the order of observation periods, the observation period data deviating from the arrangement is reordered by time stamp, and the planting type time sequence chain of the farmland plot is formed; The planting type data of the current observation period and the next observation period is compared one by one from the starting observation period of the planting type time sequence chain, and whether they are consistent is judged; If the planting type data of the current observation period and the next observation period is inconsistent, it is determined that the planting type conversion occurs in the next observation period, the specific identifier of the next observation period is recorded, and the planting type before the conversion of the current observation period and the planting type after the conversion of the next observation period are recorded; The subsequent observation periods are continuously traversed, and the planting type data of adjacent periods is repeatedly compared, until all the planting type data of the farmland plot in all observation periods is traversed, and all the records of the farmland plot in which the planting type conversion occurs are collected; The utilization state time sequence records of all the farmland plots are traversed, and the planting type conversion records of each farmland plot are collected; The planting type conversion records of all cultivated land plots are classified according to the cultivated land plot identifiers, and the missing cultivated land plot conversion records after classification are completed by deriving adjacent cultivated land plot conversion rules. Each cultivated land plot identifier corresponds to an independent planting type conversion record, which contains all observation periods, the type before conversion and the type after conversion of the cultivated land plot conversion, forming a complete set of planting type conversion records. 7.The cultivated land loss assessment method based on cultivated land spatio-temporal variation feature analysis according to claim 1, characterized in that, The cultivated land space coupling feature set and the cultivated land space-time interaction feature set are input into the cultivated land loss risk transmission model to identify the transmission path of the cultivated land loss risk between adjacent cultivated land plots through the dynamic association of spatial and temporal features, and generate the cultivated land loss risk assessment results including the loss tendency of each cultivated land plot and the risk transmission factor, including: The cultivated land plot spatial association units in the cultivated land space coupling feature set are classified according to the observation period, so that the spatial association information of the same observation period is concentrated, and the planting type conversion mode and the tillage frequency fluctuation rule in the cultivated land space-time interaction feature set are also classified according to the observation period, establishing the correspondence between the observation period and the feature information; The cultivated land plot spatial association units, the planting type conversion mode and the tillage frequency fluctuation rule of the same observation period are input into the cultivated land loss risk transmission model. First, the boundary overlap information in the cultivated land plot spatial association unit is associated with the planting type conversion mode to analyze the association logic of the boundary overlap range between adjacent cultivated land plots and the planting type conversion; The environmental difference information in the cultivated land plot spatial association unit is associated with the tillage frequency fluctuation rule to analyze the association logic of the soil moisture and terrain slope difference between adjacent cultivated land plots and the change of tillage frequency, forming a double association relationship between spatial and temporal features; Based on the double association relationship, the potential transmission path of the risk from the cultivated land plot that has undergone planting type conversion or tillage frequency change to its adjacent cultivated land plot is identified, and the adjacent cultivated land plots passed by the risk from the source cultivated land plot are recorded to form a preliminary risk transmission path; The preliminary risk transmission path is verified by comparing the risk transmission at different observation periods to confirm the continuity and stability of the preliminary risk transmission path, remove the temporary transmission path that only appears at a single observation period, and retain the stable risk transmission path that exists at multiple observation periods; According to the stable risk transmission path, the role of each cultivated land plot in the stable risk transmission path is analyzed. If the cultivated land plot is the starting point of the risk, the possibility of causing the loss of the surrounding cultivated land plots is calculated, and if the cultivated land plot is the receiving point of the risk, the possibility of causing the loss of the source cultivated land plot is calculated. The loss tendency of each cultivated land plot is obtained by comprehensively considering the loss tendency of each cultivated land plot; The feature information that most directly affects the loss tendency of each cultivated land plot is extracted. If the loss tendency is affected by the planting type conversion of the adjacent cultivated land plot, the related information of the planting type conversion is the risk transmission factor, and if it is affected by the tillage frequency change caused by the environmental difference, the related information of the environmental difference is the risk transmission factor. The loss tendency and the corresponding risk transmission factor of each cultivated land plot are sorted according to the cultivated land plot identifier, the missing cultivated land plot information after sorting is completed by deducing from the characteristics of the surrounding cultivated land plots, and the corresponding stable risk transmission path is marked, and finally the cultivated land loss risk assessment result is formed. 8.The method of claim 7, wherein, Based on the double correlation relationship, the potential transmission path of the risk spreading from the cultivated land plot which has occurred planting type conversion or change of tillage frequency to its adjacent cultivated land plot is identified, the risk starting from the source cultivated land plot and passing through the adjacent cultivated land plot is recorded, and a preliminary risk transmission path is formed, including: From the double correlation relationship, the cultivated land plot which has occurred planting type conversion is screened out, the cultivated land plot is marked as a planting type conversion source cultivated land plot, and the cultivated land plot which has occurred change of tillage frequency is screened out, and the cultivated land plot is marked as a tillage frequency change source cultivated land plot; For each planting type conversion source cultivated land plot, all adjacent cultivated land plot identifiers thereof are extracted from the cultivated land plot spatial correlation unit, and whether the adjacent cultivated land plot has also occurred planting type conversion in the subsequent observation period is checked; if any one of the adjacent cultivated land plots has occurred conversion in the observation period after the source cultivated land plot conversion, the source cultivated land plot and the adjacent cultivated land plot are marked as potential risk transmission nodes; For each tillage frequency change source cultivated land plot, all adjacent cultivated land plot identifiers thereof are also extracted from the cultivated land plot spatial correlation unit, and whether the adjacent cultivated land plot has also occurred change of tillage frequency in the subsequent observation period is checked; if any one of the adjacent cultivated land plots has occurred change in the observation period after the source cultivated land plot change, the source cultivated land plot and the adjacent cultivated land plot are marked as potential risk transmission nodes; According to the order of observation periods, the potential risk transmission nodes under the same risk transmission logic are connected; if the adjacent cultivated land plot of the planting type conversion source cultivated land plot has occurred conversion after the source cultivated land plot conversion, and another adjacent cultivated land plot of the adjacent cultivated land plot has also occurred conversion after the adjacent cultivated land plot conversion, a continuous preliminary risk transmission path is formed; If the adjacent cultivated land plot of the tillage frequency change source cultivated land plot has occurred change after the source cultivated land plot change, and another adjacent cultivated land plot of the adjacent cultivated land plot has also occurred change after the adjacent cultivated land plot change, a continuous preliminary risk transmission path is formed; All marked potential risk transmission nodes are combed to form different preliminary risk transmission paths, each preliminary risk transmission path is marked with the corresponding risk type planting type conversion risk or tillage frequency change risk, and the identifier of each node cultivated land plot in the preliminary risk transmission path and the corresponding change observation period are recorded; the missing node information after combing is completed by deducing from the preliminary risk transmission path rule. 9.The method for cultivated land loss assessment based on cultivated land space-time variation feature analysis according to claim 1, characterized in that, According to the cultivated land loss risk assessment result, continuous cultivated land loss risk regions are divided, cultivated land protection intervention schemes are designed in combination with the transmission paths of the cultivated land loss risk regions, a cultivated land protection strategy set is generated, including: analyzing each arable land plot loss tendency in the arable land loss risk assessment result, classifying the arable land plots according to the manifestation of the loss tendency, if the arable land plot loss tendency is manifested as easy conversion of planting type, the arable land plot is classified into the first type; if it is manifested as easy change of tillage frequency, the arable land plot is classified into the second type; According to the arable land plot boundary coordinates and the classification results, adjacent arable land plots belonging to the same category are merged at the spatial level, and the adjacent arable land plots with stable risk transmission paths are preferentially included in the same merging range to form continuous arable land loss risk areas; For each arable land loss risk area, extract the transmission paths of all arable land plots in the arable land loss risk area, determine the diffusion direction of the risk in the arable land loss risk area, and extract the spatial correlation unit and the time-space interaction characteristics corresponding to the arable land plot in the arable land loss risk area. Determine the core space and time sequence factors that cause the risk of the arable land loss risk area; If the arable land loss risk area is the first type of arable land plot, design the protection intervention content for the conversion of planting type, including planning the planting type allowed to be retained in the arable land loss risk area, determining the conversion restriction range of the planting type in the arable land loss risk area, and setting planting type monitoring points at the key node arable land plots of the transmission path according to the transmission path; If the arable land loss risk area is the second type of arable land plot, design the protection intervention content for the change of tillage frequency, including regulating the soil humidity and topographic slope related environmental conditions in the arable land loss risk area, setting the regulation range of soil humidity and topographic slope in the arable land loss risk area, and setting environmental monitoring points at the key node arable land plots of the transmission path according to the transmission path; Integrate the protection intervention content and monitoring point setting scheme for the arable land loss risk area, determine the implementation object specific arable land plot and the implementation time corresponding to the observation period of each intervention operation, and label the arable land loss risk area boundary range and core risk factors corresponding to the arable land protection intervention scheme; Collect the arable land protection intervention schemes of all arable land loss risk areas, match and analyze the intervention logic of each arable land protection intervention scheme and the risk characteristics of the corresponding arable land loss risk area, remove the intervention content that does not match the risk characteristics of the arable land loss risk area, and finally form a set of arable land protection strategies.

10. An arable land loss evaluation system based on analysis of spatiotemporal variation characteristics of arable land, characterized in that, It includes: a processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the machine-executable instructions to perform the arable land loss assessment method based on the arable land time-space variation characteristic analysis of any one of claims 1 to 9.

Images