Cultivated land utilization supervision system based on big data

Abstract

The invention relates to the technical field of cultivated land supervision, in particular to a big data-based cultivated land utilization supervision system, which comprises a pattern spot remote sensing interpretation module, a pixel state aggregation module, a boundary mutation recognition module, a time sequence cultivation fluctuation analysis module and a response priority pattern output module. According to the method, a time sequence label grid set is established through spatial boundary analysis and pixel index extraction of remote sensing image spots, state labels are introduced to classify pixel tillage evolution, an image spot state evolution chain and path structure is constructed on the basis of spatial aggregation, and a continuous fluctuation area is extracted through sliding window analysis of planting behavior factors in the image spots. A co-occurrence block of boundary disturbance and behavior fluctuation is positioned in combination with a space overlapping relation, a response urgency index is constructed after multiple indexes such as state density, disturbance range and duration period are quantified, multi-dimensional judgment and intervention priority ranking of the farmland utilization abnormal area are achieved, and the identification precision and response efficiency of farmland utilization changes are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of farmland monitoring technology, and in particular to a farmland use monitoring system based on big data. Background Technology

[0002] The field of farmland monitoring technology intersects with agricultural resource management and land resource monitoring, primarily involving dynamic monitoring, occupation analysis, protection assessment, and utilization status supervision of farmland resources. This field utilizes remote sensing image interpretation, geographic information systems, satellite positioning, image recognition and processing, land information modeling, data fusion, and spatiotemporal analysis to achieve precise identification and supervision of farmland area, distribution, changing trends, land use types, and illegal occupation. Application scenarios include farmland protection red line management, illegal land use early warning, farmland restoration verification, planting structure adjustment monitoring, and land use rate assessment. Service targets cover natural resource management departments, agricultural and rural affairs authorities, and local governments at the grassroots level.

[0003] The farmland use monitoring system is a monitoring and management system for the farmland use process. Specifically, it is used to collect, analyze, and provide feedback on the current status, changes, and compliance of farmland use in real time. By constructing a farmland use information database and combining remote sensing monitoring, plot identification, trend analysis, and early warning functions for violations, the system can systematically monitor situations such as farmland abandonment, illegal occupation, over-development, and restoration of abandoned farmland. This assists management departments in scientifically formulating farmland protection plans, rationally allocating farmland resources, improving farmland use efficiency, and ensuring the sustainable use of farmland resources.

[0004] Traditional monitoring systems rely on remote sensing imagery and farmland use information databases to identify behaviors such as abandonment and occupation. While they can achieve macro-level monitoring of plots and trend judgment, they lack pixel-level time-series analysis methods for farmland status. This results in insufficient accuracy in capturing fine-grained changes in farmland behavior. Furthermore, they lack structural indicators to support the identification of boundary disturbance areas, making it difficult to accurately define the scope and location of boundary abrupt changes. Traditional systems only provide management decision support based on behavioral warning results and fail to classify the urgency of responses to farmland change areas. This can easily lead to uneven distribution of monitoring resources, especially when farmland changes occur frequently in a region. Management departments may find it difficult to intervene in key areas in a timely manner, affecting the timeliness of monitoring and the rationality of resource allocation. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a big data-based farmland use monitoring system.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a big data-based farmland use monitoring system, the system comprising:

[0007] The remote sensing interpretation module for plots obtains the plot numbers and spatial boundaries of cultivated plots in the remote sensing image of a specified area. It compares the set of pixel indicators with the preset set of cultivated state reference indicators, divides the cultivated state label of each pixel at each time point, and generates a set of pixel-level time-series label grids within the cultivated plots.

[0008] The cell state aggregation module filters cell groups with consistent states in multiple consecutive time periods based on the cell-level temporal label grid set within the cultivated map patch, performs spatial merging on the group set with consistent evolution direction, and identifies the extension direction of the derivative path to obtain the nested cultivated state evolution structure information of the map patch.

[0009] The boundary mutation identification module calls the nested cultivation state evolution structure information of the map patch, extracts the direction vector of the coordinate points between continuous boundary segments, identifies land use mutation boundary fragments, compares the spatial fusion degree of the fragment boundary with the original map patch structure, marks the disturbed fragment area, and establishes spatial labeling information of the disturbed area of ​​the cultivated map patch boundary.

[0010] The temporal tillage fluctuation analysis module calculates the state fluctuation intensity value of continuous windows based on the spatial labeling information of the disturbance area of ​​the tillage map boundary. If the state fluctuation intensity value of two consecutive windows in a region exceeds the land use change warning threshold, the high fluctuation area is marked and a tillage behavior mutation block index group is generated.

[0011] The response priority pattern output module calls the high-fluctuation region in the index group of the abrupt change in the cultivation behavior and combines it with the disturbance fragment region in the spatial annotation information of the disturbance region of the cultivated land patch boundary to calculate the response urgency index value, sort them in descending order according to the response urgency index, and obtain the sequence map of priority intervention blocks for cultivated land supervision.

[0012] The present invention improves upon the following: the pixel-level temporal label grid set within the cultivated land patch includes a pixel spatial location matrix, a temporal label arrangement field, a normalized remote sensing index comparison table, and a status label matching record; the patch nested cultivated state evolution structure information includes an active-mode eclectic path unit, a secondary evolution path set, an evolution path nesting hierarchy identifier, and a spatial connection mapping map between paths; the spatial labeling information of the cultivated land patch boundary disturbance area includes a mutation boundary line segment index, a disturbance mutation degree classification layer, a statistical value of the fusion degree between the disturbance boundary and the patch, and a boundary change morphology distribution map; the cultivated behavior mutation block index group includes a behavior mutation block code, a continuous fluctuation window segment index, a status label jump identifier point set, and a mutation risk labeling level; and the cultivated land supervision priority intervention block sequence map includes a priority sorting number layer, a block urgency index value table, a co-occurrence block spatial distribution map, and an intervention level classification layer.

[0013] The present invention is improved in that the remote sensing interpretation module for the image patches includes:

[0014] The patch boundary extraction submodule obtains the patch number and spatial boundary of cultivated patches in the remote sensing image of a specified area. It numbers the patches and extracts the corresponding spatial boundary coordinate group. It locates the spatial range covered by each patch using the internal raster pixels as the index unit. After comparing and verifying the correspondence between the patch number and the spatial boundary range, it establishes an association table structure and generates a set of spatial positioning boundaries for cultivated patches.

[0015] The remote sensing index normalization submodule is based on the identified range of the cultivated map patches in the spatial positioning boundary set. It extracts three remote sensing indices—normalized vegetation index, land surface temperature, and surface roughness—from each pixel within the patch. The index values ​​are linearly normalized according to the distribution range within the patch. The three normalized index values ​​are then combined to form a three-dimensional index vector field to obtain the pixel index normalization vector information.

[0016] The tillage status determination submodule calls the three-dimensional index field of each pixel in the pixel index normalization vector information, constructs a state vector threshold range table based on the preset tillage status reference index group, compares the pixel normalization vector with the state vector threshold range in turn, filters matching intervals and marks the corresponding state labels, constructs a time-series evolution mapping table and generates a label encoding layer, and obtains the pixel-level time-series label grid set within the tillage map patch.

[0017] The present invention is improved in that the pixel state aggregation module includes:

[0018] The neighboring state extraction submodule obtains the state label sequence of pixels in the pixel-level temporal label grid within the cultivated map patch in a continuous time period, performs one-to-one label sequence comparison on spatially adjacent pixels, filters pixel groups with completely consistent state label sequences, performs spatial aggregation processing on the pixel positions in each group and constructs a pixel state aggregation unit to generate a set of groups with consistent states in a continuous period.

[0019] The evolution path determination submodule, based on the continuous periodic consistent state group set, extracts the evolution trend path sequence of each group according to the state jump direction and change order of the pixels in each group in the time series, performs comparison and judgment on the change direction vector between the path sequences, filters the path set with consistent state evolution direction and marks the evolution direction vector, and obtains the same-direction state evolution path set.

[0020] The path structure merging submodule calls the spatially continuous path units in the same-direction state evolution path set, performs overlap rate analysis on the spatial boundaries between paths, performs normalization quantization calculation on the difference between the angle between the direction vector and the position coordinate axis, and then filters the path pairs that meet the direction consistency requirements and performs spatial merging. It establishes a nested path structure index layer and extracts the path level markers to obtain the nested tillage state evolution structure information of the map patch.

[0021] The present invention is improved in that the process of performing normalization quantization calculation on the angle difference between the direction vector and the position coordinate axis is specifically as follows: the angle difference between the direction vector and the longitudinal coordinate axis of each evolution path in the same direction state evolution path set is calculated, and the angle difference is normalized by dividing the maximum absolute difference by the span of the angle change interval.

[0022] Based on the average and standard deviation of the path pairs in the set of angle differences after normalization, a directional consistency judgment threshold based on three times the standard deviation is constructed, and the path pairs are filtered using the directional consistency judgment threshold as the boundary.

[0023] A path pair is considered to meet the direction consistency requirement only if the difference in the normalized included angle between the path pairs is less than the direction consistency judgment threshold.

[0024] The present invention is improved in that the boundary mutation identification module includes:

[0025] The boundary angle extraction submodule calls the boundary coordinates of each nested path in the nested tillage state evolution structure information of the patch, calculates the direction vector for the coordinate pairs of every two consecutive boundary segments according to the path order, calculates the angle between adjacent direction vectors, constructs the angle sequence data structure according to the angle values ​​organized by the path number, and establishes a mapping table according to the timestamp and the path sequence number to generate a set of path boundary angle sequences.

[0026] The range location identification submodule constructs a boundary deformation index change curve in sequence based on each angle sequence in the path boundary angle sequence set. It compares the amplitude change of each local peak and trough node in the curve with the difference between the previous and subsequent sequences, filters out the range locations where the boundary deformation amplitude is greater than the set angle fluctuation threshold, extracts the corresponding path point number and coordinate index and records the status label type, and obtains the boundary sudden change response point information set.

[0027] The spatial disturbance labeling submodule calls the marked points in the boundary mutation response point information set, extracts the corresponding region fragments in the original patch structure according to the corresponding spatial location and label type, calculates the spatial overlap between the contour and the main structure contour of the patch, performs disturbance labeling on the region with an overlap lower than the patch fusion benchmark threshold, and superimposes the boundary of the disturbance area and classification attributes in the patch coordinates to establish the spatial labeling information of the disturbance area of ​​the cultivated patch boundary.

[0028] The present invention is improved in that the patch fusion reference threshold is a ratio threshold set based on the proportional relationship between the total boundary length of the target cultivated patch and the boundary length of the disturbed segment;

[0029] The ratio threshold is set by collecting the actual overlap ratio of unstructured fusion perturbation segments in continuous patch samples, sorting the overlap ratios in the sample group according to frequency distribution, and setting the value boundary of the overlap ratio based on the cumulative density interval of the frequency distribution.

[0030] The boundary value of the overlap ratio is set as the minimum overlap ratio value that makes the cumulative sample frequency density reach a preset confidence level, and the minimum overlap ratio value is used as the baseline threshold for patch fusion.

[0031] The present invention is improved in that the time-series tillage fluctuation analysis module includes:

[0032] The behavior factor extraction submodule collects the cultivation status label of each pixel in the continuous time series based on the spatial labeling information of the disturbance area indicated by the boundary disturbance area of ​​the cultivated map patch. Based on the label jump frequency, the state label repetition period and the duration of the empty state, it extracts three types of behavior factor values: cultivation interruption frequency, replanting interval period and planting cycle empty period, and generates a set of cultivation behavior factor indicators.

[0033] The window fluctuation calculation submodule divides the state index sequence of each patch in the set of tillage behavior factor indicators into a fixed-length sliding time window, calculates the state fluctuation intensity value of the window, and marks the continuous window tillage fluctuation screening results based on whether the state fluctuation intensity values ​​of two consecutive windows exceed the land use change warning threshold.

[0034] The high-fluctuation positioning and labeling submodule calls the marked fluctuation window index position in the continuous window tillage fluctuation screening result, performs coordinate projection according to the spatial boundary of the corresponding patch, locates the disturbance area corresponding to the position in the raster layer, and adds a labeling identification code to the spatial index table to establish a tillage behavior mutation block index group.

[0035] The present invention is improved in that the response priority pattern output module includes:

[0036] The spatial co-occurrence filtering submodule calls the high-fluctuation region in the index group of the tillage behavior mutation block, and combines it with the boundary disturbance fragment marked in the spatial annotation information of the tillage map patch boundary disturbance region to calculate the spatial overlap of the boundary range of the two types of regions. Regions with an overlap greater than the spatial fusion benchmark threshold are selected as the co-occurrence block set, and a co-occurrence block spatial identifier list is generated.

[0037] The index value normalization submodule calculates the corresponding three indicators—pixel state change density, boundary disturbance range, and continuous fluctuation duration cycle number—based on the block number in the co-occurrence block spatial identifier list, and performs intra-block set normalization on the three indicators to obtain the response urgency index value and a response urgency index value sequence.

[0038] The urgency ranking submodule calls the block number and corresponding urgency index value in the response urgency index value sequence, sorts them in descending order of urgency index value from high to low, and marks them in the spatial layer index structure after sequential numbering to establish a priority intervention block sequence map for farmland supervision.

[0039] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0040] In this invention, a time-series labeled grid is established by resolving the spatial boundaries of remote sensing patches and extracting pixel indicators. State labels are introduced to classify the cultivation evolution of pixels. Based on spatial aggregation, a state evolution chain and path structure of patches are constructed. Angle sequences of boundary direction changes are superimposed to generate boundary deformation index curves and extreme point identification is performed. Spatial labeling information of disturbance segments is established. Continuous fluctuation areas are extracted by sliding window analysis of planting behavior factors within patches. Co-occurrence blocks of boundary disturbances and behavior fluctuations are located by combining spatial overlap relationships. After quantifying multiple indicators such as state density, disturbance range, and duration period, a response urgency index is constructed to realize multi-dimensional judgment and intervention priority ranking of abnormal farmland use areas, effectively improving the identification accuracy and response efficiency of farmland use changes. Attached Figure Description

[0041] Figure 1 This is a system flowchart of the present invention;

[0042] Figure 2 This is a flowchart of the remote sensing interpretation module for image patches in this invention;

[0043] Figure 3 This is a flowchart of the pixel state aggregation module of the present invention;

[0044] Figure 4 This is a flowchart of the boundary mutation identification module of the present invention;

[0045] Figure 5 This is a flowchart of the time-series tillage fluctuation analysis module of the present invention;

[0046] Figure 6 This is a flowchart of the response priority pattern output module of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0048] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0049] Please see Figure 1 The present invention provides a technical solution: a farmland use monitoring system based on big data, the system including a map patch remote sensing interpretation module, a pixel status aggregation module, a boundary change identification module, a time-series farmland fluctuation analysis module, and a response priority pattern output module;

[0050] The remote sensing interpretation module for cultivated land parcels acquires the parcel numbers and spatial boundaries of cultivated land parcels in a remote sensing image of a specified area. It collects the normalized vegetation index, surface temperature, and surface roughness of each pixel within the parcel, classifies pixels according to the range of a single parcel, performs normalization on the three indicators, and retrieves them synchronously. It compares the pixel indicator set with the preset cultivated state reference indicator group, divides the cultivated state label of each pixel at each time point, and generates a pixel-level time-series label grid set within the cultivated land parcel.

[0051] The cell state aggregation module aggregates cell groups with consistent labels according to spatial proximity based on the cell-level temporal label grid set within the cultivated map patch. It filters cell groups with consistent states in multiple consecutive time periods, determines whether the evolution direction of each group is consistent, performs spatial merging on the group set with consistent evolution direction, constructs the dominant state chain within the patch, and identifies the extension direction of the derivative path to obtain the nested cultivated state evolution structure information of the patch.

[0052] The boundary mutation identification module calls the boundary coordinates of each nested path in the nested tillage state evolution structure information of the patch, extracts the direction vectors of the coordinate points between continuous boundary segments, calculates the included angle value of each pair of adjacent boundary segments, constructs the included angle sequence, constructs the boundary deformation index change curve of the included angle sequence, extracts the extreme value change points in the curve, cross-analyzes the spatial location of the extreme value point with the state label, determines whether it is a land use mutation boundary segment, compares the spatial fusion degree of the segment boundary with the original patch structure, marks the disturbed segment area, and establishes the spatial labeling information of the disturbed area of ​​the tillage patch boundary;

[0053] The boundary deformation index is a spatial structure index that represents the degree of geometric fluctuation of the land parcel boundary. The boundary deformation index is used to measure the degree of disturbance of the land parcel boundary. The larger the value, the stronger the boundary instability.

[0054] The temporal tillage fluctuation analysis module, based on the spatial annotation information of the disturbed area at the boundary of the tillage patch, calls the multi-period tillage status label group of the corresponding patch, extracts three behavioral factors: tillage interruption frequency, replanting interval cycle, and planting cycle gap period. According to the fixed-time sliding window segmentation method, the number of state changes within each window cycle is accumulated, and the state fluctuation intensity value is calculated for the cumulative change value of consecutive windows. If the state fluctuation intensity value of two consecutive windows in a region exceeds the land use change warning threshold, regional positioning processing is performed, high fluctuation areas are marked, and a tillage behavior mutation block index group is generated.

[0055] The response priority pattern output module calls the high-fluctuation area in the tillage behavior mutation block index group and combines it with the disturbance fragment area in the spatial annotation information of the tillage map patch boundary disturbance area. It performs a comparison processing on the spatial overlap of the two, extracts the co-occurrence blocks with strong boundary disturbance and high state fluctuation intensity value, and counts three factors for each block: pixel state change density, boundary disturbance range value, and continuous fluctuation duration cycle number. After normalizing the three factors, the response urgency index value is calculated, and the blocks are sorted in descending order according to the response urgency index to obtain the tillage supervision priority intervention block sequence map.

[0056] The set of pixel-level temporal labels within cultivated land parcels includes a pixel spatial location matrix, temporal label arrangement fields, a normalized remote sensing index comparison table, and status label matching records. The nested cultivated state evolution structure information of the parcels includes the dominant induced path unit, the set of secondary evolutionary paths, the nested evolutionary path hierarchy identifier, and the spatial connection mapping map between paths. The spatial annotation information of the disturbed areas at the boundaries of cultivated land parcels includes the mutation boundary line segment index, the disturbance mutation degree classification layer, the statistical value of the fusion degree between the disturbance boundary and the parcel, and the boundary change morphology distribution map. The cultivated behavior mutation block index group includes the behavior mutation block code, the continuous fluctuation window segment index, the status label jump marker point set, and the mutation risk labeling level. The priority intervention block sequence map for cultivated land supervision includes the priority ranking number layer, the block urgency index value table, the co-occurrence block spatial distribution map, and the intervention level classification layer.

[0057] Please see Figure 2 The remote sensing interpretation module for map features includes:

[0058] The patch boundary extraction submodule obtains the patch number and spatial boundary of cultivated patches in the remote sensing image of a specified area. It numbers the patches and extracts the corresponding spatial boundary coordinate group. It locates the spatial range covered by each patch using the internal raster pixels as the index unit. After comparing and verifying the correspondence between the patch number and the spatial boundary range, it establishes an association table structure and generates a set of spatial positioning boundaries for cultivated patches.

[0059] The system acquires the patch numbers and spatial boundaries of cultivated land parcels in a remote sensing image of a specified area. It receives a geographic coordinate range representing the target area, for example, a rectangular area defined by longitude 116.30°E to 116.32°E and latitude 39.90°N to 39.92°N. Then, based on the spectral characteristics and texture information of ground features, cultivated land parcels are identified on the 0.5-meter resolution remote sensing image of this area. For each identified cultivated land parcel, the system assigns a unique identifier, denoted as patch number TB001. Next, along the outline of patch TB001, the geographic coordinates of its boundaries are extracted point by point in a counter-clockwise direction, with an extraction interval of 1 meter. This yields a sequence of continuous coordinate pairs. For example, the starting point of the boundary coordinate group of TB001 is (116.3105°E, 39.9112°N), and the second point is (116.3... The system traces the boundary from 105°E to 39.9113°N until it closes back to the starting point, forming a complete spatial boundary. Then, using the raster pixels within the patch as the basic unit, the system determines the entire set of pixels covered by the patch. For example, a 10×10 meter patch contains 100 pixels on a 1-meter resolution image. The system records the row and column indices of these 100 pixels. Finally, the patch number TB001 is associated with its corresponding spatial boundary coordinate set and internal pixel index set, and a data consistency check is performed. The check involves recalculating the bounded pixel set based on the boundary coordinate set and comparing it with the directly extracted pixel index set. When the difference in the number of pixels in the two sets is less than 1% and the spatial overlap is higher than 99%, the check passes. Finally, the association is stored in the database, generating a spatial positioning boundary set for cultivated land patches.

[0060] The remote sensing index normalization submodule is based on the spatial positioning boundary of cultivated land patches and the range of identified patches. It extracts three remote sensing indices—normalized vegetation index, land surface temperature, and surface roughness—from each pixel within the patch. The index values ​​are linearly normalized according to the distribution range within the patch, and the three normalized index values ​​are combined to form a three-dimensional index vector field to obtain pixel index normalization vector information.

[0061] The system selects patch TB001 from the boundary set and locks the spatial location of all pixels within it. For the first pixel within patch TB001, with index (i=1, j=1), the system extracts its Normalized Difference Vegetation Index (NDVI) from multispectral image data (value 0.68), its Land Surface Temperature (LST) from thermal infrared image data (value 25.2℃), and its Surface Roughness (SR) based on radar image data (value 0.035 meters). Subsequently, the system statistically analyzes these three index values ​​for all pixels within patch TB001 to determine the distribution range of each index. For example, the statistical analysis shows that the NDVI value range for all pixels within patch TB001 is [0.20, 0.80], the LST value range is [22.0℃, 28.0℃], and the SR value range is [0.010 meters, 0.050 meters]. Next, linear normalization is performed on the index values ​​of pixels (i=1, j=1). The calculation process for the NDVI normalized value is (0.68-0.20) / (0.80-0.20), resulting in 0.80. The calculation process for the LST normalized value is (25.2-22.0) / (28.0-22.0), resulting in 0.53. The calculation process for the SR normalized value is (0.035-0.010) / (0.050-0.010), resulting in 0.625. The system combines these three normalized index values ​​of 0.80, 0.53, and 0.625 into a three-dimensional index vector field, denoted as [0.80, 0.53, 0.625]. The system then iterates through all other pixels within patch TB001, repeating the above extraction, statistical, and normalization calculation process to obtain the pixel index normalized vector information.

[0062] The tillage status determination submodule calls the three-dimensional index field of each pixel in the pixel index normalization vector information, constructs a state vector threshold range table based on the preset tillage status reference index group, compares the pixel normalization vector with the state vector threshold range in turn, filters the matching interval and marks the corresponding state label, constructs a time-series evolution mapping table and generates a label encoding layer, and obtains the pixel-level time-series label grid set within the tillage map patch.

[0063] A threshold range table for state vectors is constructed based on a pre-defined set of reference indicators for tillage states. This set of indicators is established using expert knowledge and historical sample data, defining the value ranges of the three-dimensional indicator vectors corresponding to different tillage states. For example, the normalized NDVI value range for the "sowing period" state is [0.1, 0.3], the normalized LST value range is [0.6, 0.9], and the normalized SR value range is [0.7, 1.0]. The range for the "vigorous growth period" state is NDVI [0.7, 1.0], LST […]. [0.3, 0.6], SR[0.2, 0.5], the system retrieves the three-dimensional index vector [0.80, 0.53, 0.625] of the pixel (i=1, j=1) at time point t1 in the previous step, and compares this vector with each range in the state vector threshold range table in turn. The comparison process is to determine whether the three components of the vector fall within the three intervals defined by a certain state. For the vector [0.80, 0.53, 0.625], its NDVI component 0.80 falls within the [0.7] interval of the "vigorous growth period". Within the range of [1.0], its LST component 0.53 falls within the range of [0.3, 0.6] of the "vigorous growth period," but its SR component 0.625 does not fall within the range of [0.2, 0.5] of the "vigorous growth period." Therefore, this pixel does not match the "vigorous growth period" state. The system continues to compare with other state ranges until a completely matching range is found. It is then set to finally match the "mid-growth" state. The system then marks this pixel with the "mid-growth" state label at time point t1 and assigns it the corresponding label code, such as "2." Subsequently, the system repeats this determination process for the next time point t2, t3, ..., tn in the time series to construct the temporal evolution mapping table of the pixel. For example, the state label encoding sequence of the pixel (i=1, j=1) in five consecutive time periods (t1 to t5) is [2,2,3,3,1], where "3" represents "maturity" and "1" represents "fallow period". Finally, the system integrates the state label encoding of all pixels at each time point to generate a label encoding layer and obtain the pixel-level temporal label grid set within the cultivated map patch.

[0064] Please see Figure 3 The cell status aggregation module includes:

[0065] The neighboring state extraction submodule obtains the state label sequence of pixels in the pixel-level temporal label grid within the cultivated map patch within a continuous time period. It performs one-to-one label sequence comparison on spatially adjacent pixels, filters pixel groups with completely consistent state label sequences, performs spatial aggregation processing on the pixel positions within each group, constructs pixel state aggregation units, and generates a set of groups with consistent states in a continuous period.

[0066] The system extracts two spatially adjacent cells from the grid set, such as cell A (index i=1, j=1) and cell B (index i=1, j=2), and obtains the state label sequence over five consecutive time periods (t1 to t5). Let the label sequence of cell A be [2,2,3,3,1] and the label sequence of cell B be [2,2,3,3,1]. The system performs a one-to-one comparison of these two sequences, starting from the first position of each sequence and comparing the label codes at each corresponding position to ensure they are identical. In this example, the labels of the two sequences are completely identical across all five time positions. Therefore, cells A and B are determined to be a cell combination with completely identical state label sequences. Subsequently, the system continues to check other adjacent cells, such as cell C (index i=1, j=3), using cell B as the center. The label sequence of cell A is [2,2,3,3,2], which is different from the sequence of cell B at the fifth position. Therefore, B and C are not combined. The system traverses all cells in the patch, continuously performing this pairwise comparison and expansion, and filters out all cells with completely consistent state label sequences and spatial adjacency, forming several cell combination groups. For example, a group G1 is finally formed, which contains cells A, B and four other adjacent cells, all with the label sequence [2,2,3,3,1]. Next, the system performs spatial aggregation processing on the position coordinates of all cells in group G1, calculates the polygon boundary formed by these cells, and constructs a cell state aggregation unit. This unit represents a region where the evolution of cultivation state behavior is completely synchronized within a specific time period, and finally generates a set of continuous periodic state consistent groups.

[0067] The evolution path judgment submodule is based on a set of groups with consistent states in a continuous period. According to the state jump direction and change order of the pixels in each group in the time series, it extracts the evolution trend path sequence of each group, performs a comparison judgment on the change direction vector between the path sequences, filters the path set with consistent state evolution direction and marks the evolution direction vector to obtain the set of state evolution paths in the same direction.

[0068] Selecting one group, for example, group G1 containing six pixels, its state label sequence over five consecutive time periods (t1 to t5) is [2,2,3,3,1], where label "2" represents "mid-growth stage", "3" represents "mature stage", and "1" represents "fallow stage". Based on the direction and order of state transitions of the pixels within this group over time, the system extracts its evolutionary trend path sequence. The state transitions occur from t2 to t3 (from "2" to "3") and from t4 to t5 (from "3" to "1"). Therefore, the evolutionary trend path sequence of this group can be represented as mid-growth stage -> mature stage -> fallow stage. The system performs the same processing on another spatially non-adjacent group G2, setting its state label sequence to [2,3,3,1]. [1], its evolution trend path sequence is also mid-growth stage -> mature stage -> fallow stage. Subsequently, the system performs a comparative judgment on the change direction vector between these two path sequences. Here, the direction vector is an abstract representation, and its dimension is determined by the state type. For example, the vector from "mid-growth stage" to "mature stage" in the state space is denoted as V(2,3), and the vector from "mature stage" to "fallow stage" is denoted as V(3,1). The evolution paths of groups G1 and G2 are both composed of V(2,3) and V(3,1) in sequence. Therefore, the evolution direction is determined to be consistent. The system selects G1 and G2 and classifies them into a path set with consistent state evolution direction, and labels this set with a unified evolution direction vector [V(2,3), V(3,1)], thus obtaining a set of state evolution paths in the same direction.

[0069] The path structure merging submodule calls the path units with spatial continuity in the same-direction state evolution path set, performs overlap rate analysis on the spatial boundaries between paths, and performs normalized quantization calculation on the difference between the angle between the direction vector and the position coordinate axis before filtering. Path pairs that meet the direction consistency requirements are filtered and spatial merging is performed. Nested path structure index layer is established and path level markers are extracted to obtain the nested tillage state evolution structure information of the map patch.

[0070] The process of performing normalization quantization calculation on the angle difference between the direction vector and the position coordinate axis is as follows: calculate the angle difference between the direction vector and the vertical coordinate axis of each evolution path in the same direction state evolution path set, and normalize the angle difference by dividing the maximum absolute difference by the span of the angle change interval.

[0071] Based on the average and standard deviation of the path pairs in the set of angle differences after normalization, a directional consistency judgment threshold based on three times the standard deviation is constructed, and the path pairs are filtered using the directional consistency judgment threshold as the boundary.

[0072] A path pair is considered to meet the direction consistency requirement only if the difference in the normalized included angle between the path pairs is less than the direction consistency judgment threshold.

[0073] Two spatially adjacent or partially overlapping path units, denoted as path P1 and path P2, are selected. These belong to the same set of co-evolutionary paths. First, an overlap rate analysis is performed on the spatial boundaries between paths P1 and P2. This is calculated by dividing the intersection area of ​​the regions occupied by the two path units by the area of ​​their union. For example, if the area of ​​P1 is 50m²... 2 The area of ​​P2 is 60m² 2 The intersection area is 15m 2 The combined area is 95m² 2 The overlap rate is approximately 15 / 95 ≈ 0.158. Subsequently, the angle difference between the direction vector and the longitudinal coordinate axis of each evolution path in the set of unidirectional state evolution paths is calculated. Here, the direction vector refers to the geometric orientation of the path unit in space, obtained by calculating the displacement direction of the centroid of the path unit in the time series. Assuming the longitudinal coordinate axis is due north (0°), the angle between the direction vector of path P1 and the longitudinal coordinate axis is 15°, the angle of path P2 is 25°, and the angle of path P3 is 18°. The angle range of all paths in this set is...

[0074] [10°, 30°], meaning the range of angle variation is 20°. The angle difference is normalized by dividing the maximum absolute difference by the range of angle variation. For P1 and P2, the absolute value of their angle difference is...

[0075] |15-25| = 10°, with a normalized value of 10 / 20 = 0.5. For P1 and P3, the absolute value of the angle difference is |15-18| = 3°, with a normalized value of 3 / 20 = 0.15. Next, based on the average and standard deviation of the path pairs in the set of normalized angle differences, a directional consistency judgment threshold based on three times the standard deviation is constructed. The average of the normalized angle difference samples calculated for all path pairs in the set is set to be 0.25, and the standard deviation to be 0.08. Therefore, the directional consistency judgment threshold is calculated as 0.25 + 3 × 0.08 = 0.49. This threshold is then used as the boundary for judging the path pairs. The normalized angle difference between P1 and P2 is greater than the threshold of 0.49 (0.5), and the normalized angle difference between P1 and P3 is less than the threshold of 0.49 (0.15). The path pair is considered to meet the direction consistency requirement only when the normalized angle difference between the path pairs is less than the direction consistency judgment threshold. Therefore, P1 and P3 meet the direction consistency requirement. If the spatial overlap rate also meets the preset value (e.g., greater than 0.1), then spatial merging is performed on P1 and P3 to form a larger nested path structure. An index is built on this structure and it is marked as a higher-level path. Finally, the nested cultivation state evolution structure information of the map patch is obtained.

[0076] Please see Figure 4The boundary mutation identification module includes:

[0077] The boundary angle extraction submodule calls the boundary coordinates of each nested path in the nested tillage state evolution structure information of the patch, calculates the direction vector for the coordinate pairs of every two consecutive boundary segments according to the path order, calculates the angle between adjacent direction vectors, constructs the angle sequence data structure according to the angle values ​​organized by the path number, and establishes a mapping table based on the timestamp and the path sequence number to generate a set of path boundary angle sequences.

[0078] A nested path numbered NP01 is selected, and the ordered sequence of coordinate points constituting its spatial boundary is extracted from its attributes. For example, the first five points of the sequence are P1(x1,y1), P2(x2,y2), P3(x3,y3), P4(x4,y4), and P5(x5,y5). Following the path order, the system calculates the direction vector for each pair of coordinates of consecutive boundary segments. The first boundary segment is P1P2, and its direction vector V1 is calculated as (x2-x1,y2-y1). The second boundary segment is P2P3, and its direction vector V2 is calculated as (x3-x2,y3-y2). The third boundary segment is P3P4, and its direction vector V3 is calculated as (x4-x3,y4-y2), and so on. Subsequently, the system calculates the angle between adjacent direction vectors; for example, it calculates the angle θ between V1 and V2. 12 The calculation formula is: The calculation yields θ 12 =175.2°, then calculate the angle θ between V2 and V3. 23 , to obtain θ 23 =168.5°. Calculate the included angle of all adjacent boundary segments along the entire boundary in sequence to form a sequence of included angle values ​​organized by path number. For path NP01, this sequence may be [175.2, 168.5, 179.1, 92.3, 178.8, ...]. Finally, the system establishes a mapping table based on the timestamp associated with the path (e.g., the start time of the evolution cycle) and the path number NP01, stores the included angle sequence in a data structure, and generates a set of path boundary included angle sequences.

[0079] The range location identification submodule constructs a boundary deformation index change curve in sequence based on each angle sequence in the path boundary angle sequence set. It compares the amplitude change of each local peak and trough node in the curve with the difference between the previous and subsequent sequences, filters out the range locations where the boundary deformation amplitude is greater than the set angle fluctuation threshold, extracts the corresponding path point number and coordinate index and records the status label type, and obtains the boundary sudden change response point information set.

[0080] The system selects the included angle sequence [175.2, 168.5, 179.1, 92.3, 178.8, ...] corresponding to path NP01 and constructs a boundary deformation index variation curve in sequence. The horizontal axis represents the index of the boundary point, and the vertical axis represents the corresponding included angle value. The system analyzes the curve to identify local peaks (values ​​higher than the adjacent points on both sides) and troughs (values ​​lower than the adjacent points on both sides). For example, in the sequence, 179.1 is a local peak, and 92.3 is a local trough. Subsequently, the amplitude change of each local peak and trough node in the curve is compared with the difference between the preceding and following sequences. For the trough point 92.3, its preceding value is 179.1 and its following value is 178.8. Its amplitude change is calculated as the average of the forward difference and the backward difference, i.e., (|92.3-179.1). 1|+|92.3-178.8|) / 2=(86.8+86.5) / 2=86.65°. Next, the system compares this change amplitude with the set angle fluctuation threshold. The angle fluctuation threshold is set based on the statistical analysis of a large number of regularly shaped cultivated land patch boundary angle sequences, taking four times the standard deviation of the angle change as the threshold. Assuming that the threshold calculated by this method is 45°, since the calculated amplitude change of 86.65° is greater than the set angle fluctuation threshold of 45°, the system filters this position as a range position and extracts the path point number corresponding to the trough point, i.e., P4, and its coordinate index (x4, y4). At the same time, it records the status label type of the point at a specific time point, such as "fallow period". Finally, the information is integrated to obtain the boundary change response point information set.

[0081] The spatial disturbance labeling submodule calls the marked points in the boundary mutation response point information set, extracts the corresponding region fragments in the original patch structure according to the corresponding spatial location and label type, calculates the spatial overlap between the contour and the main structure contour of the patch, performs disturbance labeling on the region with an overlap lower than the patch fusion benchmark threshold, and superimposes the disturbance area boundary and classification attributes into the patch coordinates to establish the spatial labeling information of the cultivated patch boundary disturbance region.

[0082] The baseline threshold for patch fusion is a ratio threshold set based on the proportional relationship between the total boundary length of the target cultivated patch and the boundary length of the disturbed segment.

[0083] The ratio threshold is set by collecting the actual overlap ratio of unstructured fusion perturbation segments in continuous patch samples, sorting the overlap ratios in the sample group according to their frequency distribution, and setting the boundary of the overlap ratio value based on the cumulative density interval of the frequency distribution.

[0084] The boundary value of the overlap ratio is set to the minimum overlap ratio value that makes the cumulative sample frequency density reach the preset confidence level, and the minimum overlap ratio value is used as the baseline threshold for patch fusion.

[0085] A marker point P4 is selected, with coordinates (x4, y4), and its state is "fallow". Based on its spatial location, a neighborhood region segment centered on P4 is extracted from the structure of the original patch TB001. For example, a circular region with a radius of 10 meters is denoted as R1, and the contour of this region is extracted. Then, the spatial overlap between the contour of R1 and the main structural contour of patch TB001 is calculated. The calculation method is to divide the overlap length of the two contour lines by the total contour length of the R1 region. The total contour length of R1 is set to 62.8 meters, of which 5.0 meters overlap with the main structural contour of TB001. Therefore, the overlap is 5.0 / 62.8≈0.08. Next, the system compares this overlap with the patch fusion benchmark threshold. The patch fusion benchmark threshold is a ratio threshold set based on the proportional relationship between the total boundary length of the target cultivated patch and the boundary length of the disturbed segment. The method for setting this ratio threshold is as follows: The formula is as follows: collect the actual overlap ratio of unstructured fusion perturbation segments in continuous map patch samples. For example, collect 100 samples to obtain 100 overlap ratio values, sort these values ​​according to frequency distribution, and calculate the cumulative density. The preset confidence level is 90%, and the value boundary is set as the minimum overlap ratio value that makes the cumulative sample frequency density reach 90%. After setting the statistical analysis, when the overlap ratio value is 0.15, the cumulative sample frequency density reaches 90%, so 0.15 is used as the map patch fusion benchmark threshold. Since the calculated overlap ratio of 0.08 is lower than the map patch fusion benchmark threshold of 0.15, the system performs perturbation labeling on region R1, and overlays the boundary coordinates of the perturbation area R1 and its classification attributes (e.g., type: "field ridge collapse", source status: "fallow period") in the coordinate data of map patch TB001 to establish spatial labeling information of the perturbation area of ​​cultivated map patch boundary.

[0086] Please see Figure 5 The time-series tillage fluctuation analysis module includes:

[0087] The behavior factor extraction submodule collects the cultivation status label of each pixel in the continuous time series based on the spatial labeling information of the disturbance area indicated by the boundary disturbance area of ​​the cultivated map patch. Based on the label jump frequency, the state label repetition period and the duration of the empty state, it extracts three types of behavior factor values: cultivation interruption frequency, replanting interval period and planting cycle empty period, and generates a set of cultivation behavior factor indicators.

[0088] The disturbance area R1 is located and labeled, and the cultivation status labels of each pixel in R1 are collected in 24 consecutive time series (e.g., observation once every half month for one year). For example, the label sequence of one pixel is [1, 1, 1, 2, 2, 3, 3, 1, 1, 1, 1, 1, 2, 2, 3, 1, 1, 1, 1, 1, 1, 1, 1], where "1" is a fallow window, "2" is sowing / growing, and "3" is maturity / harvest. Based on this label sequence, the system extracts three types of behavioral factor values. First, the cultivation interruption frequency is extracted. Cultivation interruption is defined as the period after transitioning from a non-fallow state (2 or 3) to a fallow state (1), with the fallow window lasting for more than one cycle before entering the next cultivation. In the sequence, the first cultivation ends in the 8th cycle, and the second ends in the 16th cycle. There are a total of 2 interruptions within one year. The frequency of crop interruption is 2 times / year. Next, the replanting interval cycle is extracted, referring to the time interval between two cultivation behaviors (the occurrence of state 2). The first sowing occurs in cycle 4, and the second in cycle 13, with an interval of 13-4=9 observation cycles, or 4.5 months. Finally, the planting cycle gap period is extracted, referring to the duration of the gap before and after a complete planting activity. Before and after the first cultivation (cycles 4 to 8), the gap period is in cycles 1 to 3 and 9 to 12, with a total length of 3+4=7 cycles. Before and after the second cultivation (cycles 13 to 16), the gap period is in cycles 9 to 12 and 17 to 24, with a total length of 4+8=12 cycles. The average gap period is (7+12) / 2=9.5 cycles. Finally, these three types of behavioral factor values ​​(2 times / year, 4.5 months, 9.5 cycles) are integrated to generate a set of cultivation behavior factor indicators.

[0089] The window fluctuation calculation submodule divides the sequence of state indicators for each patch in the tillage behavior factor index set into segments using a fixed-length sliding time window, employing the formula:

[0090]

[0091] The state fluctuation intensity value of the window is obtained by calculation. The state fluctuation intensity values ​​of two consecutive windows are marked according to whether they exceed the land use change warning threshold, and the continuous window cultivation fluctuation screening results are obtained.

[0092] Among them, R wThe values ​​represent the state fluctuation intensity of the window, F represents the number of times the tillage state label jumps within the current time window, obtained by accumulating jump events in the state label sequence within the sliding window, F′ represents the number of times the state label jumps within the previous time window, obtained by the previous window in the sliding time series, G represents the normalized comprehensive index of tillage behavior within the current window, obtained by integrating behavioral factors such as tillage interruption frequency, multiple cropping interval cycle, and planting cycle gap period in a linear normalization manner, M represents the total number of time windows corresponding to the current patch, calculated based on the time series length and window step size, S represents the number of patch behavioral factors, defined according to the fixed number of extracted indicators, and the land use change warning threshold is the value corresponding to the 85th percentile state fluctuation intensity value in the time series change statistical distribution extracted based on the training block, used to mark the unstable boundary of cultivated land use, and used as the comparison benchmark for fluctuation intensity in subsequent judgments;

[0093] The sequence is divided into segments using a sliding time window of fixed length of 6 time points, and the following formula is used:

[0094]

[0095] The calculation obtains the state fluctuation intensity value of the window, where R... wThe value representing the intensity of state fluctuations within a window is a scalar; a larger value indicates more drastic changes in farming behavior within the time window. The innovation of the formula lies in quantifying the acceleration of the rate of change by introducing the absolute value of the difference between the number of state jumps in the current window and the previous window, |FF′|. Simultaneously, a nonlinear transformation of the comprehensive index G of farming behavior is performed using the logarithmic function ln(1+G), amplifying the sensitivity in the low index range. The fluctuation intensity is scaled and normalized using the square root of the product of the total number of time windows M and the number of behavioral factors S as the denominator. The specific calculation process is as follows: For a time series, the total observation period is 24, and the window step size is 1. Therefore, the total number of time windows M is 24-6+1=19. The number of behavioral factors S is defined as 3 (farming interruption frequency, replanting interval cycle, and planting cycle gap period). For the current... The 10th time window analyzed previously (covering time points 10 to 15) has a state label sequence of [1, 1, 2, 2, 3, 1]. State transitions occur at time points 11 to 12 (1->2) and 14 to 15 (3->1). Therefore, the number of state label transitions F in the current window is 2. The sequence of the previous time window (covering time points 9 to 14) is [1, 1, 1, 2, 2, 3], and the number of transitions F′ is 1. The normalized tillage behavior composite index G in the current window is obtained by linearly normalizing the three behavioral factors calculated within this window: tillage interruption frequency (e.g., 0.3), multiple cropping interval cycle (e.g., 0.6), and planting cycle gap period (e.g., 0.4), and then weighting and summing them (with each factor set to 1 / 3). Its value is (0.3+0.6+0.4) / 3≈0.433. Substituting this value into the formula:

[0096]

[0097] The calculated R w The values ​​are compared with a land use change warning threshold, which is based on the time-series change statistical distribution extracted from 200 training blocks with stable use, and their respective R values ​​are calculated. w The values ​​are sorted, and the 85th percentile is taken. The 85th percentile of these 200 values ​​is set to 0.060. Therefore, the land use change warning threshold is set to 0.060. (Note: The last sentence appears to be incomplete and possibly refers to a different context.) w If the value is 0.0477, which does not exceed the threshold, the system will not flag it. However, for the next window, if its R... w If the value is calculated to be 0.071, it is marked because it exceeds the threshold. By traversing and comparing all windows, the continuous window tillage fluctuation screening results are obtained.

[0098] The high-fluctuation positioning and labeling submodule calls the marked fluctuation window index position in the continuous window tillage fluctuation screening results, performs coordinate projection according to the spatial boundary of the corresponding patch, locates the disturbance area corresponding to the position in the raster layer, and adds the label identification code to the spatial index table to establish a tillage behavior mutation block index group.

[0099] The 11th and 12th time windows were identified as high-fluctuation areas, covering the time period from time 11 to 17. Based on the spatial boundary coordinates of the corresponding patch TB001, the system projects the vector boundary of the patch onto the rasterized land use layer and precisely locates the set of pixels within the patch. Since the fluctuation is calculated from the behavior factor of the entire patch, this labeling is first associated with the entire patch TB001. If the system design supports sub-patch level analysis, it can further locate the disturbed areas consisting of specific pixels whose state changed during time points 11 to 17. For example, if a region in the northeast corner of a map patch changes from "fallow land" to "construction land" during this time period, the system will identify the pixels in that region as a disturbed area. Subsequently, in the spatial index table (such as R-Tree or Quadtree) associated with the raster layer, the system will query the entry containing the pixels in the disturbed area and append a label identifier code to the attribute field of the entry, such as "HFS-2024-001", where HFS represents "high fluctuation state", 2024 is the year, and 001 is the sequence number. Finally, an index group of cultivated behavior mutation blocks will be established.

[0100] Please see Figure 6 The response priority pattern output module includes:

[0101] The spatial co-occurrence filtering submodule calls the high-fluctuation region in the tillage behavior mutation block index group and combines it with the boundary disturbance fragment marked in the spatial annotation information of the tillage map patch boundary disturbance region to calculate the spatial overlap of the boundary range of the two types of regions. Regions with an overlap greater than the spatial fusion benchmark threshold are selected as the co-occurrence block set and a co-occurrence block spatial identifier list is generated.

[0102] The system retrieves the high-fluctuation area from the tillage behavior mutation block index group, specifically the disturbed area in the northeast corner of patch TB001, denoted as Area_H. It then combines this area with the boundary disturbance fragments marked in the spatial annotation information of the tillage patch boundary disturbance area, such as area R1 identified in the previous step due to abrupt boundary angle changes, denoted as Area_B. The system obtains the geographic boundary coordinates of Area_H and Area_B and calculates the spatial overlap of these two types of area boundaries. The calculation method is to divide the intersection area of ​​the two areas by the area of ​​the smaller of the two areas. The area of ​​Area_H is set to 120. m 2 The area of ​​Area_B is 100m².2 The area of ​​the intersecting region was calculated to be 95m² using GIS spatial analysis. 2 The spatial overlap is 95 / min(120,100) = 95 / 100 = 0.95. The system compares this overlap with the spatial fusion benchmark threshold. The threshold is set with reference to typical land use change cases. Generally, when the area caused by internal behavior changes overlaps with the area characterized by boundary deformation by more than 80%, the two are considered to have a strong correlation. Therefore, the spatial fusion benchmark threshold is set to 0.80. Since the calculated overlap of 0.95 is greater than the threshold of 0.80, the system filters the overlapping area as a co-occurrence block and assigns it a unique number, such as CXBK-001. Finally, all the block numbers that meet the conditions are summarized to generate a list of co-occurrence block spatial identifiers.

[0103] The index value normalization submodule calculates three indicators—pixel state change density, boundary disturbance range, and continuous fluctuation duration—based on the block number in the co-occurring block spatial identifier list. It then performs intra-block set normalization on these three indicators using the following formula:

[0104]

[0105] The response urgency index value is obtained through calculation, resulting in a sequence of response urgency index values.

[0106] Among them, I r The response urgency index is represented by D, which represents the normalized density of cell state changes in the current block. It is calculated by dividing the number of state transitions per unit area by the total number of cells in the block. μ D The value of D represents the mean of the normalized density values ​​of pixel state changes in all co-occurring blocks. It is calculated as the arithmetic mean of the D values ​​of all co-occurring blocks. The value of E represents the normalized range of the current block boundary perturbation. It is calculated as the range of the current boundary angle sequence divided by the maximum range of all ranges. E The mean of the normalized range of boundary perturbations in all co-occurring blocks is represented by μ, where T represents the normalized number of consecutive fluctuation duration periods, calculated as the number of consecutive jump time windows in the current block divided by the total number of observation time windows. T V represents the normalized mean of the number of consecutive fluctuation durations of all co-occurring blocks. D V E V T These represent the sample standard deviations of the three indicators D, E, and T in the co-occurrence block, respectively. The calculation method is to take the square root of the sample variance of each indicator, and the sum of the squares of the three is used to represent the overall fluctuation intensity.

[0107] Based on the block number in the co-occurrence block spatial identifier list, such as CXBK-001, three indicators are calculated for this block: pixel state change density, boundary perturbation range, and number of continuous fluctuation duration periods. These three indicators are then normalized within the block set using the following formula:

[0108]

[0109] The urgency index value is obtained by calculation, where I r The response urgency index measures the degree of anomaly by calculating the standardized Euclidean distance from the indicator vector of a block to the mean vector of indicators of all blocks. The innovation of the formula is that the denominator uses the square root of the sum of the squares of the standard deviations of each indicator. This is equivalent to projecting the variability in the multidimensional indicator space onto a comprehensive dimension, so that the distance measurement is not dominated by the drastic fluctuations of a single indicator, but comprehensively considers the overall deviation trend. To calculate this index, the statistical parameters of the entire co-occurring block set (set to 10 blocks) need to be obtained first, as shown in Table 1.

[0110] Table 1. Statistics of Co-occurrence Block Indicators

[0111]

[0112] As shown in Table 1, the system calculates the mean and standard deviation of the three indicators for all 10 blocks, and obtains the mean μ of the normalized value of the pixel state change density. D =0.65, standard deviation V D =0.12, the mean μ of the normalized value of the boundary disturbance range. E =60.50, standard deviation V E =15.30, the mean μ of the normalized value of the duration of continuous fluctuations. T =1.8, standard deviation V T =0.40, for block CXBK-001, its current block pixel state change density D is 0.8, the current block boundary perturbation range E is 86.65, and the continuous fluctuation duration T is 2. Substituting these values ​​into the formula:

[0113]

[0114] The results indicate that the overall anomaly level of block CXBK-001 is approximately 1.709 times the average variation level of all blocks. The system repeats this calculation for all co-occurring blocks to obtain a sequence of response urgency index values.

[0115] The urgency ranking submodule calls the block number and corresponding urgency index value in the urgency index value sequence, sorts them in descending order from high to low urgency index value, and marks them in the spatial layer index structure after sequential numbering to establish a priority intervention block sequence map for farmland supervision.

[0116] The system retrieves a list containing all co-occurring blocks and their corresponding urgency indices by calling the block numbers and corresponding urgency index values ​​in the response urgency index value sequence. The entries in the list are then sorted in descending order of urgency index value, and each block is assigned a number based on this sorted order. For example, CXBK-001 is numbered 1, CXBK-003 is numbered 2, and so on. This priority number is then marked in the spatial layer index structure associated with each block. For instance, a new field "Intervention Priority" is added to the attribute data of block CXBK-001 and assigned the value "1". Finally, the blocks with priority markings are visualized and rendered in a Geographic Information System (GIS), with different priorities identified by different colors or symbols, thus establishing a sequence map of priority intervention blocks for farmland supervision.

[0117] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A farmland use monitoring system based on big data, characterized in that, The system includes: The remote sensing interpretation module for plots obtains the plot numbers and spatial boundaries of cultivated plots in the remote sensing image of a specified area. It compares the set of pixel indicators with the preset set of cultivated state reference indicators, divides the cultivated state label of each pixel at each time point, and generates a set of pixel-level time-series label grids within the cultivated plots. The cell state aggregation module filters cell groups with consistent states in multiple consecutive time periods based on the cell-level temporal label grid set within the cultivated map patch, performs spatial merging on the group set with consistent evolution direction, and identifies the extension direction of the derivative path to obtain the nested cultivated state evolution structure information of the map patch. The boundary mutation identification module calls the nested cultivation state evolution structure information of the map patch, extracts the direction vector of the coordinate points between continuous boundary segments, identifies land use mutation boundary fragments, compares the spatial fusion degree of the fragment boundary with the original map patch structure, marks the disturbed fragment area, and establishes spatial labeling information of the disturbed area of ​​the cultivated map patch boundary. The temporal tillage fluctuation analysis module calculates the state fluctuation intensity value of continuous windows based on the spatial labeling information of the disturbance area of ​​the tillage map boundary. If the state fluctuation intensity value of two consecutive windows in a region exceeds the land use change warning threshold, the high fluctuation area is marked and a tillage behavior mutation block index group is generated. The response priority pattern output module calls the high-fluctuation region in the index group of the abrupt change in the cultivation behavior and combines it with the disturbance fragment region in the spatial annotation information of the disturbance region of the cultivated land patch boundary to calculate the response urgency index value, sort them in descending order according to the response urgency index, and obtain the sequence map of priority intervention blocks for cultivated land supervision.

2. The farmland use monitoring system based on big data according to claim 1, characterized in that, The pixel-level temporal label grid set within the cultivated land parcel includes a pixel spatial location matrix, a temporal label arrangement field, a normalized remote sensing index comparison table, and a status label matching record. The nested cultivated state evolution structure information of the parcel includes an active-directed evolutionary path unit, a set of secondary evolutionary paths, nested evolutionary path hierarchy identifiers, and a spatial connection mapping map between paths. The spatial labeling information of the boundary disturbance area of ​​the cultivated land parcel includes a mutation boundary line segment index, a disturbance mutation degree classification layer, statistical values ​​of the fusion degree between the disturbance boundary and the parcel, and a boundary change morphology distribution map. The cultivated behavior mutation block index group includes behavior mutation block codes, continuous fluctuation window segment indexes, a set of status label jump marker points, and mutation risk labeling levels. The priority intervention block sequence map for cultivated land supervision includes a priority sorting number layer, a block urgency index value table, a co-occurrence block spatial distribution map, and an intervention level classification layer.

3. The farmland use monitoring system based on big data according to claim 2, characterized in that, The remote sensing interpretation module for the image features includes: The patch boundary extraction submodule obtains the patch number and spatial boundary of cultivated patches in the remote sensing image of a specified area. It numbers the patches and extracts the corresponding spatial boundary coordinate group. It locates the spatial range covered by each patch using the internal raster pixels as the index unit. After comparing and verifying the correspondence between the patch number and the spatial boundary range, it establishes an association table structure and generates a set of spatial positioning boundaries for cultivated patches. The remote sensing index normalization submodule is based on the identified range of the cultivated map patches in the spatial positioning boundary set. It extracts three remote sensing indices—normalized vegetation index, land surface temperature, and surface roughness—from each pixel within the patch. The index values ​​are linearly normalized according to the distribution range within the patch. The three normalized index values ​​are then combined to form a three-dimensional index vector field to obtain the pixel index normalization vector information. The tillage status determination submodule calls the three-dimensional index field of each pixel in the pixel index normalization vector information, constructs a state vector threshold range table based on the preset tillage status reference index group, compares the pixel normalization vector with the state vector threshold range in turn, filters matching intervals and marks the corresponding state labels, constructs a time-series evolution mapping table and generates a label encoding layer, and obtains the pixel-level time-series label grid set within the tillage map patch.

4. The farmland use monitoring system based on big data according to claim 3, characterized in that, The pixel state aggregation module includes: The neighboring state extraction submodule obtains the state label sequence of pixels in the pixel-level temporal label grid within the cultivated map patch in a continuous time period, performs one-to-one label sequence comparison on spatially adjacent pixels, filters pixel groups with completely consistent state label sequences, performs spatial aggregation processing on the pixel positions in each group and constructs a pixel state aggregation unit to generate a set of groups with consistent states in a continuous period. The evolution path determination submodule, based on the continuous periodic consistent state group set, extracts the evolution trend path sequence of each group according to the state jump direction and change order of the pixels in each group in the time series, performs comparison and judgment on the change direction vector between the path sequences, filters the path set with consistent state evolution direction and marks the evolution direction vector, and obtains the same-direction state evolution path set. The path structure merging submodule calls the spatially continuous path units in the same-direction state evolution path set, performs overlap rate analysis on the spatial boundaries between paths, performs normalization quantization calculation on the difference between the angle between the direction vector and the position coordinate axis, and then filters the path pairs that meet the direction consistency requirements and performs spatial merging. It establishes a nested path structure index layer and extracts the path level markers to obtain the nested tillage state evolution structure information of the map patch.

5. The farmland use monitoring system based on big data according to claim 4, characterized in that, The process of performing normalization quantization calculation on the angle difference between the direction vector and the position coordinate axis is as follows: calculate the angle difference between the direction vector and the longitudinal coordinate axis of each evolution path in the same-direction state evolution path set, and normalize the angle difference by dividing the maximum absolute difference by the span of the angle change interval. Based on the average and standard deviation of the path pairs in the set of angle differences after normalization, a directional consistency judgment threshold based on three times the standard deviation is constructed, and the path pairs are filtered using the directional consistency judgment threshold as the boundary. A path pair is considered to meet the direction consistency requirement only if the difference in the normalized included angle between the path pairs is less than the direction consistency judgment threshold.

6. The farmland use monitoring system based on big data according to claim 5, characterized in that, The boundary mutation identification module includes: The boundary angle extraction submodule calls the boundary coordinates of each nested path in the nested tillage state evolution structure information of the patch, calculates the direction vector for the coordinate pairs of every two consecutive boundary segments according to the path order, calculates the angle between adjacent direction vectors, constructs the angle sequence data structure according to the angle values ​​organized by the path number, and establishes a mapping table according to the timestamp and the path sequence number to generate a set of path boundary angle sequences. The range location identification submodule constructs a boundary deformation index change curve in sequence based on each angle sequence in the path boundary angle sequence set. It compares the amplitude change of each local peak and trough node in the curve with the difference between the previous and subsequent sequences, filters out the range locations where the boundary deformation amplitude is greater than the set angle fluctuation threshold, extracts the corresponding path point number and coordinate index and records the status label type, and obtains the boundary sudden change response point information set. The spatial disturbance labeling submodule calls the marked points in the boundary mutation response point information set, extracts the corresponding region fragments in the original patch structure according to the corresponding spatial location and label type, calculates the spatial overlap between the contour and the main structure contour of the patch, performs disturbance labeling on the region with an overlap lower than the patch fusion benchmark threshold, and superimposes the boundary of the disturbance area and classification attributes in the patch coordinates to establish the spatial labeling information of the disturbance area of ​​the cultivated patch boundary.

7. The farmland use monitoring system based on big data according to claim 6, characterized in that, The patch fusion benchmark threshold is a ratio threshold set based on the proportional relationship between the total boundary length of the target cultivated patch and the boundary length of the disturbed segment. The ratio threshold is set by collecting the actual overlap ratio of unstructured fusion perturbation segments in continuous patch samples, sorting the overlap ratios in the sample group according to frequency distribution, and setting the value boundary of the overlap ratio based on the cumulative density interval of the frequency distribution. The boundary value of the overlap ratio is set as the minimum overlap ratio value that makes the cumulative sample frequency density reach a preset confidence level, and the minimum overlap ratio value is used as the baseline threshold for patch fusion.

8. The farmland use monitoring system based on big data according to claim 7, characterized in that, The time-series tillage fluctuation analysis module includes: The behavior factor extraction submodule collects the cultivation status label of each pixel in the continuous time series based on the spatial labeling information of the disturbance area indicated by the boundary disturbance area of ​​the cultivated map patch. Based on the label jump frequency, the state label repetition period and the duration of the empty state, it extracts three types of behavior factor values: cultivation interruption frequency, replanting interval period and planting cycle empty period, and generates a set of cultivation behavior factor indicators. The window fluctuation calculation submodule divides the state index sequence of each patch in the set of tillage behavior factor indicators into a fixed-length sliding time window, calculates the state fluctuation intensity value of the window, and marks the continuous window tillage fluctuation screening results based on whether the state fluctuation intensity values ​​of two consecutive windows exceed the land use change warning threshold. The high-fluctuation positioning and labeling submodule calls the marked fluctuation window index position in the continuous window tillage fluctuation screening result, performs coordinate projection according to the spatial boundary of the corresponding patch, locates the disturbance area corresponding to the position in the raster layer, and adds a labeling identification code to the spatial index table to establish a tillage behavior mutation block index group.

9. The farmland use monitoring system based on big data according to claim 8, characterized in that, The response priority pattern output module includes: The spatial co-occurrence filtering submodule calls the high-fluctuation region in the index group of the tillage behavior mutation block, and combines it with the boundary disturbance fragment marked in the spatial annotation information of the tillage map patch boundary disturbance region to calculate the spatial overlap of the boundary range of the two types of regions. Regions with an overlap greater than the spatial fusion benchmark threshold are selected as the co-occurrence block set, and a co-occurrence block spatial identifier list is generated. The index value normalization submodule calculates the corresponding three indicators—pixel state change density, boundary disturbance range, and continuous fluctuation duration cycle number—based on the block number in the co-occurrence block spatial identifier list, and performs intra-block set normalization on the three indicators to obtain the response urgency index value and a response urgency index value sequence. The urgency ranking submodule calls the block number and corresponding urgency index value in the response urgency index value sequence, sorts them in descending order of urgency index value from high to low, and marks them in the spatial layer index structure after sequential numbering to establish a priority intervention block sequence map for farmland supervision.

Images