Method and system for analyzing functional fragmentation of cultivated land based on multi-source remote sensing time series data

Abstract

This application relates to the field of farmland functional fragmentation analysis technology, and discloses a method and system for farmland functional fragmentation analysis based on multi-source remote sensing time-series data. The method includes: acquiring multi-source remote sensing time-series data and digital elevation models of the target area and its corresponding adjacent areas for farmland functional fragmentation analysis; based on the multi-source remote sensing time-series data, analyzing the physical geometric fragmentation of the target area, the phenological heterogeneity fragmentation determined based on time-series trajectories, and the micro-topography and operational resistance fragmentation used to characterize the continuity of farmland operations, to obtain a multi-dimensional feature fragmentation index set for the target area; and based on the multi-dimensional feature fragmentation index set, obtaining the functional fragmentation degree used to characterize the quantified functional fragmentation of the target area for functional fragmentation analysis. This system corresponds to this method. This application solves the technical problems of existing technologies that can only measure static two-dimensional physical boundaries, ignore multi-source time-series characteristics, and address implicit functional degradation.

Description

Technical Field

[0001] This application relates to the field of farmland functional fragmentation analysis technology, specifically a method and system for farmland functional fragmentation analysis based on multi-source remote sensing time series data. Background Technology

[0002] With the development of high-resolution remote sensing technology and cloud computing platforms such as Google Earth Engine, the spatial measurement of farmland fragmentation has shifted from traditional field sampling surveys to full-domain spatial analysis based on multi-source remote sensing images, including optical and radar data.

[0003] Currently, the calculation of farmland fragmentation, both domestically and internationally, mainly relies on landscape ecology methods such as those based on Fragstats software. Existing techniques typically first classify land use in single-period remote sensing images, extract farmland boundaries, and then use geometric formulas to calculate geometric indices such as patch density (PD), edge density (ED), or landscape shape index (LSI). However, relying solely on this approach has technical limitations, including only measuring static two-dimensional physical boundaries and neglecting multi-source temporal characteristics and latent functional degradation. Summary of the Invention

[0004] The purpose of this application is to provide a method and system for analyzing the functional fragmentation of cultivated land based on multi-source remote sensing time-series data, so as to solve the technical problems of existing technologies that can only measure static two-dimensional physical boundaries, ignore multi-source time-series characteristics and implicit functional degradation.

[0005] To achieve the above objectives, this application provides a method for functional fragmentation analysis of cultivated land based on multi-source remote sensing time-series data, including: Acquire multi-source remote sensing time-series data and digital elevation models of the target area and its adjacent areas to be analyzed for functional fragmentation of arable land. The multi-source remote sensing time-series data includes optical data with time series and radar data with time series from the same period as the optical data. Based on multi-source remote sensing time-series data, the physical geometric fragmentation of the target area, the phenological heterogeneity fragmentation determined based on time-series trajectories, and the micro-topography and operational resistance fragmentation used to characterize the continuity of farmland operations are analyzed to obtain a multi-dimensional feature fragmentation index set for the target area. Based on a multidimensional feature fragmentation index set and combined with nonlinear machine learning, the functional fragmentation degree of the target region is obtained for functional fragmentation analysis. This functional fragmentation degree is used to characterize the functional fragmentation status of the quantified target region.

[0006] Preferably, the analysis of the physical geometric fracture includes: A sliding window is constructed centered on the target area and extends to adjacent areas; The physical geometric fragmentation is calculated based on the length of the cultivated land boundary and the proportion of centrally connected cultivated land patches within the sliding window. This physical geometric fragmentation is used to quantify the physical geometric fragmentation situation.

[0007] Preferably, the analysis of the phenological heterogeneity fragmentation specifically involves analyzing the hidden functional fragmentation that reflects the physical connection but operational heterogeneity between the target area and the adjacent areas.

[0008] Preferably, the analysis of the micro-topography and operational resistance fragmentation specifically involves analyzing the micro-resistance of agricultural machinery operating between the target area and adjacent areas in three-dimensional space.

[0009] Preferably, the fragmentation of the hidden function is quantified by phenological heterogeneity fragmentation, which is determined based on a dynamic time warping algorithm.

[0010] Preferably, the micro-resistance is quantified by micro-topography and operational resistance fragmentation, which is determined by the surface roughness variation coefficient based on radar data and the slope based on a digital elevation model.

[0011] Preferably, the phenological heterogeneity fragmentation is determined based on a dynamic time warping algorithm, specifically: the similarity of vegetation index trajectories between the target area and adjacent areas in the time series is determined by the dynamic time warping algorithm, and the phenological heterogeneity fragmentation is determined based on the similarity of vegetation index trajectories.

[0012] Preferably, the radar data includes co-polarized data and cross-polarized data; the determination of the surface roughness variation coefficient includes determination based on co-polarized data; the determination of the surface roughness variation coefficient also includes determination based on co-polarized data and cross-polarized data.

[0013] Preferably, based on the multidimensional feature fragmentation index set and the functional fragmentation degree, and combined with the SHAP framework, local attribution analysis is performed on the functional fragmentation analysis to output a driving force distribution map for the target region.

[0014] To achieve the above objectives, this application also provides a farmland functional fragmentation analysis system based on multi-source remote sensing time-series data, which applies the farmland functional fragmentation analysis method based on multi-source remote sensing time-series data as described above, including: The data cube construction module is used to acquire multi-source remote sensing time-series data and digital elevation models of the target area and adjacent areas of the target area to be analyzed for functional fragmentation of arable land. The multi-source remote sensing time-series data includes optical data with time series and radar data with time series from the same period as the optical data. The multidimensional feature fragmentation index set construction module is used to analyze the physical geometric fragmentation of the target area based on multi-source remote sensing time series data, the phenological heterogeneity fragmentation determined based on time series trajectory, and the micro-topography and operation resistance fragmentation used to characterize the continuity of farmland operations, so as to obtain a multidimensional feature fragmentation index set for the target area. The functional fragmentation analysis module is used to obtain the functional fragmentation degree of the target area based on a multidimensional feature fragmentation index set and combined with nonlinear machine learning for functional fragmentation analysis. This functional fragmentation degree is used to characterize the functional fragmentation status of the quantified target area.

[0015] Beneficial Effects: The method and system for functional fragmentation analysis of cultivated land based on multi-source remote sensing time-series data proposed in this application solves the technical problems of existing technologies that can only measure static two-dimensional physical boundaries, ignore multi-source time-series characteristics, and address implicit functional degradation. Specifically, it proposes a functional fragmentation measurement framework that integrates time-series DTW phenological distance and radar topographic resistance, breaking through the limitations of two-dimensional geometry and incorporating physical inaccessibility and operational heterogeneity into the fragmentation consideration, thus overcoming multi-dimensional physical and temporal constraints. Through high-frequency multi-source time-series data, it effectively identifies physically connected but fragmented cultivated land units in actual planting structures, improving the accuracy of change detection in complex hilly areas and enabling the capture of implicit precursors to abandonment. Through the SHAP framework, it not only outputs quantitative functional fragmentation but also accurately indicates to agricultural managers whether the fragmentation of a plot is due to steep terrain or complex ownership and management, providing direct evidence for comprehensive land consolidation and high-standard farmland construction, and achieving white-box decision support. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A flowchart of a method for analyzing the functional fragmentation of cultivated land based on multi-source remote sensing time-series data, provided as an example; Figure 2 This is a structural block diagram of the farmland functional fragmentation analysis system based on multi-source remote sensing time-series data provided in this embodiment.

[0018] The implementation, functional features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] The technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0020] In this document, the term "comprising" is intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0021] This embodiment discloses a method and system for functional fragmentation analysis of cultivated land based on multi-source remote sensing time-series data. It relates to the fields of cultivated land functional fragmentation analysis, intelligent interpretation of remote sensing images, agricultural resource monitoring, and land consolidation decision support. Specifically, it involves a technical solution that comprehensively evaluates the spatial morphological fragmentation of cultivated land, the heterogeneity of phenological management of adjacent cultivated land, and the operational resistance of micro-topography using optical remote sensing time-series data, radar remote sensing time-series data, digital elevation model data, and cultivated land mask data. The aim is to solve the technical problems in existing cultivated land fragmentation assessments, which can only measure two-dimensional map boundaries, have difficulty identifying the heterogeneity of phenological management of adjacent cultivated land, have difficulty using SAR time-series roughness to characterize operational resistance, and have difficulty unifying multiple fragmentation mechanisms into functional fragmentation.

[0022] In the application field of this embodiment, farmland fragmentation is an important indicator for measuring the degree of contiguousness of farmland spatial distribution, the potential for large-scale agricultural operations, and assessing the risk of farmland abandonment. Analysis of existing technologies reveals that they only measure static two-dimensional physical boundaries, neglecting multi-source temporal characteristics and implicit functional degradation. In response to this problem, this embodiment discloses a method and system for analyzing farmland functional fragmentation based on multi-source remote sensing time-series data. In summary, this embodiment aims to provide a technical solution for measuring farmland functional fragmentation that can deeply integrate coarseness features obtained from radar data with phenological features obtained from optical data, truly reflecting the three-dimensional spatial resistance and dynamic utilization status of farmland.

[0023] The key terms used in this embodiment will now be explained in detail.

[0024] Functional Fragmentation Index The functional fragmentation index (FFI) is a 0-1 normalized index output by an existing machine learning model, taking into account the physical and geometric contiguity of the target area and adjacent areas, temporal phenological heterogeneity, and three-dimensional terrain-roughness operational resistance. Combined with a preset functional fragmentation threshold, it determines whether the target area has a high risk of land abandonment. It should be noted that this functional fragmentation threshold can be adjusted based on the classification results of the validation samples. The curve or management requirements are determined.

[0025] Phenological Heterogeneity Index In this embodiment, The Dynamic Time Warping (DTW) algorithm quantifies the differences in vegetation index time-series curves between adjacent cultivated land pixels in the target area and adjacent areas, thereby reflecting the implicit functional fragmentation of the target area and adjacent areas that are physically connected but have heterogeneous management. Specifically, this implicit functional fragmentation refers to cultivated land units that are physically adjacent but have highly heterogeneous planting structures and phenological characteristics due to differences in management entities.

[0026] Micro-topography and Terrain Resistance Index In this embodiment, The surface roughness variation coefficient with temporal characteristics, obtained by fusing the slope obtained from the Digital Elevation Model (DEM) and the backscattering coefficient from the Synthetic Aperture Radar (SAR) data, is used to quantify the micro-resistance of agricultural machinery operations in three-dimensional space. This micro-resistance is affected by the spatial heterogeneity of soil texture and stubble distribution, exhibiting irregular fluctuations and abrupt changes. A sudden increase in local resistance can easily cause agricultural machinery to exceed its load limit and obstruct its operation, which can lead to substandard tillage, poor land preparation quality, increased farming costs for farmers, decreased willingness to invest in production, and thus increased risk of land abandonment.

[0027] The functional fragmentation analysis method and system for cultivated land based on multi-source remote sensing time-series data in this embodiment will now be described.

[0028] Reference Figure 1 , Figure 1 This is a flowchart illustrating the method for analyzing the functional fragmentation of cultivated land based on multi-source remote sensing time-series data provided in this embodiment.

[0029] like Figure 1 As shown, in a first aspect, this embodiment discloses a method for functional fragmentation analysis of cultivated land based on multi-source remote sensing time-series data, including: S10: Acquire multi-source remote sensing time-series data and digital elevation models of the target area and its adjacent areas to be analyzed for functional fragmentation of cultivated land. The multi-source remote sensing time-series data includes optical data with time series and radar data with time series from the same period as the optical data.

[0030] In summary, S10 of this embodiment constructs a data cube for the target region and its adjacent regions. In a specific application of this embodiment, S10 can be implemented based on the following steps: The steps for acquiring optical data specifically involve obtaining time-series multispectral optical images, such as Sentinel-2, covering the target area and its adjacent areas, and calculating and acquiring a set of time-series vegetation indices. ,in, This represents the total number of time steps. From this, the optical data is obtained.

[0031] The steps for acquiring radar data, specifically, involve acquiring synthetic aperture radar (SAR) images from the same period as the optical data, such as Sentinel-1, and extracting a time series set of polarization backscattering coefficients, which includes data with the same polarization. and cross-polarization data .

[0032] For the target area and its digital elevation model (DEM), existing DEMs can be used directly.

[0033] In practical applications, the Sentinel-2 in this embodiment uses a 10m resolution, and the DEM is an SRTM 30m resampled to 10m, with a timing length of... Preferred periods are 12-36 and synthesized monthly.

[0034] Thus, this embodiment has fully constructed a data cube for the target area and its adjacent areas, providing a data foundation for analyzing the physical geometric fragmentation, phenological heterogeneous fragmentation, and micro-topography and operational resistance fragmentation of the target area.

[0035] The motivation for improving the analysis of the phenological heterogeneity fragmentation and the fragmentation of micro-topography and operational resistance in the target area in this embodiment will now be explained.

[0036] In summary, existing technologies for processing time-series data exhibit static, slice-like characteristics, lacking dynamic considerations of time-series phenology. Traditional calculations, based on single-period classification data, cannot reflect the functional inconsistencies of cultivated land throughout long-term crop growth cycles. Specifically, if two physically adjacent cultivated lands exhibit highly heterogeneous planting structures and phenological characteristics due to different management entities—i.e., implicit abandonment or intercropping—then these two physically adjacent cultivated lands have actually formed implicit functional fragmentation, which existing geometric techniques cannot identify.

[0037] In summary, current technologies address the fragmentation caused by micro-topography and operational resistance by focusing on a single dimension and neglecting the impact of terrain and micro-physical resistance. Existing technologies rely solely on optical classification results to calculate the geometric fragmentation of a two-dimensional plane, failing to consider the three-dimensional spatial connectivity resistance caused by topographic undulations, as well as micro-physical characteristics such as surface roughness reflected by radar data. It's important to clarify that the three-dimensional spatial connectivity resistance here refers to the obstacles to cultivation formed by the spatial differences in terrain undulations, field ridges and depressions, and soil texture, resulting from the proximity of adjacent plots, vertical terrain differences, and the movement of agricultural machinery through the field. Specifically, due to the existence of three-dimensional spatial connectivity resistance, large and medium-sized agricultural machinery is limited from operating across large and medium-sized areas in actual farmland management. This increases labor and fuel inputs in cultivation, hinders large-scale production, and results in low economic benefits for small-scale farming. Therefore, the three-dimensional connectivity resistance caused by terrain exacerbates the disadvantages of farmland utilization, effectively increasing the aforementioned risk of land abandonment. Therefore, it is necessary to analyze the micro-topography and operational resistance fragmentation, integrate topographic data and SAR roughness data, establish a farmland fragmentation evaluation system based on micro-topography and operational resistance, and accurately assess the potential risks of farmland abandonment.

[0038] The physical geometric fragmentation, phenological heterogeneous fragmentation, and micro-topography and operational resistance fragmentation of the target area analyzed in this embodiment will now be described.

[0039] S20: Based on multi-source remote sensing time-series data, analyze the physical geometric fragmentation of the target area, the phenological heterogeneity fragmentation determined based on time-series trajectories, and the micro-topography and operational resistance fragmentation used to characterize the continuity of farmland operations, to obtain a multi-dimensional feature fragmentation index set for the target area.

[0040] S20 of this embodiment aims to construct a multi-dimensional feature fragmentation index set for use in the synthesis and mapping of functional fragmentation.

[0041] The analysis of the physical geometric fracture condition specifically includes: A sliding window is constructed centered on the target area and extends to adjacent areas; The physical geometric fragmentation is calculated based on the length of the cultivated land boundary and the proportion of centrally connected cultivated land patches within the sliding window. This physical geometric fragmentation is used to quantify the physical geometric fragmentation situation.

[0042] This embodiment uses cultivated land patches corresponding to the target area to be solved. Define a sliding window in space around the center. The physical geometric fragmentation of cultivated land corresponding to a window is calculated by fusing two geometric dimensions: patch edge fragmentation characteristics and cultivated land connectivity and clustering characteristics. , A higher value indicates a greater degree of geometric fragmentation of the cultivated land parcels within the window. The specific calculation formula and step-by-step calculation process are as follows: Calculate the original edge density of the window : in: Represents a sliding window Within the range, the total length of the boundary between cultivated land pixels and non-cultivated land pixels; Represents a sliding window The total area of ​​the entire region. In practice, It represents the total development of farmland plot boundaries per unit area; the more fragmented the plots, the more fragmented the boundaries. The larger the value, the better.

[0043] Original edge density of the window Normalization is performed to obtain the standardized edge density. : in: This represents the maximum original edge density corresponding to all sliding windows in the global domain to be solved; This represents the minimum original edge density corresponding to all sliding windows in the global domain to be solved; This is an infinitesimal constant used to avoid calculation errors caused by a denominator of 0. Thus far, The value is constrained to The interval eliminates the difference in edge density dimensions, satisfying the premise of multi-index weighted calculation.

[0044] Calculate the proportion of connected cultivated land in the calculation window : in: Represents a sliding window Within, and with the central cultivated land patch A contiguous area of ​​farmland that is spatially adjacent and connected; This indicates the area of ​​contiguous cultivated land connected to the central patch within the window; Represents a sliding window The total area of ​​all arable land contained within it; It is an infinitesimal constant to avoid calculation errors caused by a denominator of zero; value range The higher the proportion of connected cultivated land in the total cultivated land area of ​​the window, the better the contiguous clustering and the lower the degree of fragmentation of the cultivated land. This embodiment adopts... As a reverse indicator of connectivity breakdown, the worse the connectivity, the larger the value of this indicator.

[0045] Therefore, the mathematical expression for the physical geometric fragmentation degree in this embodiment can be: in: and These are preset geometric weight coefficients, and their sum is unique and both are greater than or equal to 0. Therefore, in this embodiment... Farmland fragmentation is quantified by coupling two types of geometric features. Specifically, the first relies on normalized edge density to characterize the degree of fragmentation at plot boundaries, and the second relies on the inverse term of connectivity ratio to characterize the degree of contiguous farmland damage. After weighted fusion of the two indicators, the higher the edge density and the lower the connectivity ratio, the more accurate the final result. The higher the calculation result, the more severe the geometric fragmentation of the corresponding arable land.

[0046] The analysis of phenological heterogeneity fragmentation in this embodiment will now be explained. Specifically, the analysis of phenological heterogeneity fragmentation refers to the analysis of latent functional fragmentation that reflects the physical connection but heterogeneous management between the target area and adjacent areas. In a simple example, two geographically contiguous plots of farmland without field ridges and with completely connected spatial physical boundaries are owned by two different farmers. One farmer cultivates rice year-round, resulting in a high-growth vegetation timeline curve in summer, while the other farmer leaves the land fallow year-round, resulting in low-growing and disordered vegetation timelines. Although the two plots are physically connected, their interannual vegetation phenological timeline changes differ significantly. This type of latent fragmentation is quantified by multi-temporal optical vegetation index sequences and DTW timeline distance, which is the functional farmland fragmentation in the phenological dimension.

[0047] In a preferred embodiment of this invention, the fragmentation of the hidden function is quantified by phenological heterogeneity fragmentation, which is determined based on a dynamic time warping algorithm. Specifically, the similarity of vegetation index trajectories between the target area and adjacent areas in the time series is determined by the dynamic time warping algorithm, and the phenological heterogeneity fragmentation is determined based on the similarity of vegetation index trajectories.

[0048] The calculation of phenological heterogeneity fragmentation degree in this embodiment will now be explained.

[0049] This embodiment uses a multi-temporal vegetation index time series combined with a dynamic time warping (DTW) algorithm to quantify the hidden functional fragmentation caused by contiguous cultivated land areas but inconsistent farming patterns and phenological rhythms. Before performing DTW distance calculations, the vegetation index time series corresponding to each pixel undergoes preprocessing including missing data interpolation, temporal smoothing filtering, and 0-1 value range standardization to avoid index fluctuations caused by cloud and fog interference, missing time series data, and phenological phase misalignment, thus ensuring the degree of phenological heterogeneity fragmentation. The calculation results are stable and reliable. Phenological heterogeneity fragmentation. The calculation results are as follows: Filter center pixel Effective farmland neighborhood set : in: Center pixel The set of all spatially adjacent pixels; For a single neighboring pixel surrounding the central pixel; Represents a pixel Belongs to the central pixel The spatial neighborhood range; Preset land category identification parameters, Used to characterize the corresponding pixel Pixels representing cultivated land are excluded; non-cultivated land pixels are discarded and not included in subsequent phenological difference calculations. The selected pixels will be... Defined as Perform subsequent calculations.

[0050] Calculating the time difference distance based on DTW : in: and These are two sets of vegetation index time series to be compared, corresponding to the vegetation time series of the central pixel and the vegetation time series of the neighboring pixels, respectively. It is the set of all feasible regular paths in two time series. This indicates selecting the optimal matching path with the minimum cumulative cost among all regularized paths; This represents the total number of matching nodes on the optimal regularized path. The sequence number of the matched node within the path; For time sequence In the The vegetation index values ​​corresponding to each matching node; For time sequence In the The vegetation index values ​​corresponding to each matching node; The cost function for single-point matching can be selected as the absolute distance between node values. Or Euclidean distance; The minimum cumulative distance between two vegetation time series after optimal path matching is DTW. The greater the difference in the phenological change trajectory of the time series, the larger the DTW calculation value.

[0051] Solving for coordinates The central farmland pixel Corresponding phenological heterogeneity fragmentation : in: For set Total number of arable land pixels included; This indicates traversing all valid adjacent farmland cells of the center cell. .

[0052] Therefore, this embodiment The arithmetic mean of the temporal DTW distances between the center pixel and all neighboring cultivated land pixels; The larger the value, the more significant the differences between adjacent arable lands that are physically connected in space in terms of crop planting type, field management mode, and interannual phenological development pattern. The higher the degree of fragmentation of the hidden function of arable land, the stronger the resistance to large-scale agricultural mechanization operations.

[0053] This document describes the analysis of micro-topography and operational resistance in this embodiment.

[0054] Specifically, the analysis of the micro-topography and operational resistance fragmentation involves analyzing the micro-resistance of agricultural machinery operating between the target area and adjacent areas in three-dimensional space. It should be noted that this micro-resistance is the aforementioned three-dimensional spatial connectivity resistance.

[0055] In a preferred embodiment of this invention, the micro-resistance is quantified by micro-topography and operational resistance fragmentation, which is determined by the surface roughness variation coefficient based on radar data and the slope based on a digital elevation model. Considering the aforementioned radar data includes both co-polarized and cross-polarized data, the determination of the surface roughness variation coefficient in this embodiment includes determination based on co-polarized data, and further includes determination based on both co-polarized and cross-polarized data.

[0056] The micro-topography and operational resistance fragmentation of this embodiment will now be described in detail.

[0057] This embodiment uses coordinate-based methods. Micro-topography and operational resistance fragmentation at the location This indicates the quantitative relationship between micro-topography and operational resistance. Topographic slope parameters extracted from DEM, SAR polarization Temporal roughness parameters, SAR cross-polarization The temporal roughness parameter is obtained by weighted coupling of three indices; among them, the surface roughness variation coefficient is characterized by the sample standard deviation of the SAR temporal backscattering coefficient, and one of them can be used alone. Same polarization data calculation can also be combined , Dual-polarization data collaborative calculation. When the raw SAR backscattering coefficients are stored in dB logarithmic units, the dB values ​​can be pre-converted to a linear scale before statistical calculations are performed. Alternatively, standard deviation statistics can be directly completed based on the dB time series, ensuring that the roughness calculation results closely match the actual undulations of the field surface and the tillage resistance characteristics caused by stubble distribution.

[0058] When the determination of the surface roughness variation coefficient in this embodiment includes determination based on homopolarization data, the micro-topography and operational resistance fragmentation in this embodiment... The mathematical expression can be: in: and These are preset weighting coefficients, and their sum is one, with both being greater than or equal to 0. At this point, for... The mathematical expression: coordinates The micro-topography and operational resistance fragmentation corresponding to each pixel. The higher the fragmentation value, the greater the micro-connectivity resistance of agricultural machinery operating across plots in three-dimensional space, and the stronger the degree of farmland resistance fragmentation. Extracted from the Digital Elevation Model (DEM) Location, surface slope The slope is a positive switching term used to quantify the resistance of agricultural machinery to climbing and passing through slopes caused by slope undulations. The greater the slope, the higher the resistance to slope cultivation. The total number of time steps mentioned above. Number the time steps; For the first Co-polarized backscattering coefficients acquired by time-step SAR satellites. For all The time-step average value of the time-step polarization backscattering coefficient; The standard deviation of the same polarization time series is the surface roughness variation index obtained from the same polarization data in this embodiment, which characterizes the tillage friction resistance caused by the unevenness of the surface soil and the changes in the surface soil quality.

[0059] When the determination of the surface roughness variation coefficient in this embodiment also includes determination based on homopolarization data and crosspolarization data, the micro-topography and operational resistance fragmentation in this embodiment... The mathematical expression can be: in: , and The preset weighting coefficients are such that the sum of the three is one, and all three are greater than or equal to 0. At this point, for... In the mathematical expression, For the first Cross-polarization backscattering coefficients acquired by time-step SAR satellites. For all The time-step average value of the cross-polarization backscattering coefficient; The standard deviation of the same polarization time series was used to further obtain a more comprehensive surface roughness variation index based on cross-polarization data. Specifically, cross-polarization is more sensitive to crop residues, shallow roots, and fine surface obstacles, and is used to supplement the characterization of additional tillage resistance on the shallow surface. It should be noted that the determination of the surface roughness variation coefficient in this embodiment includes two roughness selection implementation methods: one is to use only the same polarization data, that is, to use only the same polarization data to solve for surface roughness, making... Discarding the cross-polarization calculation terms, only retaining the terrain and co-polarization roughness terms to complete the calculation. The solution involves two approaches: first, combining dual-polarization data and simultaneously utilizing two sets of time-series data—one homopolarized and the other cross-polarized—to integrate multi-dimensional features of topography and dual-polarization roughness and comprehensively quantify micro-level tillage resistance.

[0060] Therefore, this embodiment The resistance to connectivity in three-dimensional space was quantified; The higher the value, the greater the micro-level resistance to cross-regional and contiguous agricultural machinery operations between adjacent farmlands, the worse the economic efficiency of large-scale farming, and the higher the potential risk of farmland being left idle or abandoned.

[0061] Thus, this embodiment has constructed a multidimensional feature fragmentation index set. Existing methods for quantifying farmland fragmentation generally rely on single-period optical image classification results, calculating the apparent fragmentation of plots solely from two-dimensional planar geometric parameters—a conventional approach to fragmentation assessment in this field. Guided by this conventional approach, those skilled in the art typically focus only on geometric indicators such as patch length and width, and boundary density, lacking a research and development direction that integrates multi-source heterogeneous data from time-series remote sensing, microwave radar, and digital elevation data to construct a fragmentation index. This embodiment departs from the traditional single-geometric quantification paradigm, combining three independent evaluation indicators to form a combined system: The approach of using geometry and connectivity is used to quantify the explicit spatial fragmentation of land parcels; The long-term vegetation index and DTW time-series matching algorithm were optimized and introduced to establish a quantitative formula for the implicit management fragmentation caused by the interconnected land surfaces but different planting systems of farmers. The construction idea of ​​this index cannot be simply derived from the traditional geometric fragmentation algorithm. By integrating DEM slope parameters with temporal roughness statistical features of dual-polarization SAR, a quantitative characterization of agricultural machinery operation resistance and fragmentation is achieved in three-dimensional space, overcoming the inherent limitation of existing technologies that can only perform planar measurements. Three categories of indicators comprehensively cover different causes of farmland fragmentation from three dimensions: spatial morphology, cultivation management, and field traffic resistance. Compared with traditional single-dimensional evaluation schemes, the quantitative results are more consistent with the actual utilization status of farmland and the patterns of abandonment.

[0062] The functional fragmentation of this embodiment will now be explained.

[0063] S30: Based on a multidimensional feature fragmentation index set and combined with nonlinear machine learning, the functional fragmentation degree of the target region is obtained for functional fragmentation analysis. This functional fragmentation degree is used to characterize the functional fragmentation status of the quantified target region.

[0064] In the specific application of this embodiment, S30 completes the functional fragmentation based on nonlinear machine learning. Composition and mapping. In a specific implementation, functional fragmentation. Synthesis and mapping, including: obtaining a multidimensional feature fragmentation index set Historical farmland use efficiency data were used as the input feature and as the target variable. For example, 0 indicates good contiguous planting, while 1 indicates marginal abandonment or inefficient utilization. Existing XGBoost models can be used, where the objective function is to iterate by minimizing the objective loss function. Furthermore, other existing nonlinear supervised learning models, such as random forests, gradient boosting trees, LightGBM, or CatBoost, can also be used. The obtained multidimensional feature fragmentation index set is input into the trained machine learning model, and the predicted probability value output by the model is the comprehensive functional fragmentation index. The mapping range is .

[0065] Therefore, the farmland functional fragmentation analysis method based on multi-source remote sensing time-series data in this embodiment is based on coupled modeling of three heterogeneous remote sensing data sources: optical time-series vegetation remote sensing, DEM topographic data, and dual-polarization microwave SAR, and is supported by... , and The three differentiated fragmentation quantification indicators enable multi-dimensional synergistic quantification of farmland fragmentation from two-dimensional geometric explicit fragmentation, phenological management implicit fragmentation, and three-dimensional topographic resistance fragmentation. This fully restores the true causes of farmland fragmentation and accurately correlates with the risk of farmland abandonment. It overcomes the technical bias of existing technologies that rely solely on single-period optical images, are limited to two-dimensional geometric calculations, unilaterally ignore the heterogeneity of field management and the implicit fragmentation caused by topographic surface resistance, and unilaterally sever the inherent connection between fragmentation mechanism and abandonment.

[0066] Further investigation revealed a lack of nonlinear interpretation mechanisms for end-point outcomes such as abandonment in the analysis of functional fragmentation of arable land. Specifically, existing geometric indices exhibit linear superposition, making it difficult to directly quantify the actual driving force of multidimensional fragmentation on arable land degradation, such as marginal abandonment, and failing to analyze complex nonlinear spatial mechanisms. Therefore, this embodiment designs a method using the SHAP framework to achieve spatial pixel-level stripping of driving factors.

[0067] Specifically, based on the multidimensional feature fragmentation index set and the functional fragmentation degree, and combined with the SHAP framework, local attribution analysis is performed on the functional fragmentation analysis to output a driving force distribution map for the target region.

[0068] In this specific application, the existing SHAP (SHapley Additive exPlanations) framework is introduced to process each spatial cell of the aforementioned output. Local attribution analysis of predicted values: in: Indicates the first The Shapley marginal contribution value corresponding to the item fragmentation feature; in this embodiment Corresponding in sequence , and ,Right now , and The larger the absolute value of the value, the higher the probability that this indicator is the dominant driving factor for current grid fragmentation. The complete set consisting of all fragmented features. Represents the total number of features within the entire set; Indicates from the complete collection Remove the first The remaining feature set after 1 feature, This indicates traversing all data that do not contain features. Feature subset ; Representative subset The number of features contained within; These are the combined weight coefficients for each subset, used to evenly distribute features across different subsets. The resulting marginal increase; Represents a subset of input features At that time, the model output value of the comprehensive evaluation model for breakage; This indicates adding new features based on subset S. Then, the corresponding output value of the model; Indicates the addition of features The resulting model output increment, representing features The marginal contribution in this subset of scenarios. Therefore, Features The weighted average of marginal increments under all feasible subset combinations eliminates coupling interference between features and objectively quantifies the independent contribution of individual indicators to the overall fragmentation result. In this embodiment, the solutions are obtained separately. , and Three Shapley contribution coefficients; comparing the absolute values ​​of the three coefficients within the same spatial grid, the index with the largest absolute value is the core dominant driving factor for high farmland fragmentation in that grid; if When the absolute value is the largest, geometric fragmentation is the primary cause; if The largest absolute value indicates that the fragmentation of heterogeneous phenological management is the primary contributing factor; if The largest absolute value indicates that micro-topography and operational resistance are the primary causes; the three types of driving factors are assigned values ​​to the RGB single color channels respectively, generating a three-channel fused spatial distribution map of the driving force of farmland breakage, which intuitively shows the spatial differentiation law of the dominant mechanism of breakage in the whole area.

[0069] Thus, the farmland functional fragmentation analysis method based on multi-source remote sensing time-series data in this embodiment solves the shortcomings of existing technologies that can only measure static two-dimensional physical boundaries, ignore multi-source time-series characteristics and implicit functional degradation, and constructs a technical solution for measuring farmland functional fragmentation that can deeply integrate SAR time-series roughness and optical time-series phenology, and truly reflect the three-dimensional spatial resistance and dynamic utilization status of farmland.

[0070] Reference Figure 2 , Figure 2 This is a structural block diagram of the farmland functional fragmentation analysis system based on multi-source remote sensing time-series data provided in this embodiment; in the diagram: 10, data cube construction module; 20, multi-dimensional feature fragmentation index set construction module; 30, functional fragmentation analysis module.

[0071] Secondly, such as Figure 2 As shown, this embodiment also discloses a farmland functional fragmentation analysis system based on multi-source remote sensing time-series data, which applies the farmland functional fragmentation analysis method based on multi-source remote sensing time-series data as described above, including: The data cube construction module 10 is used to acquire multi-source remote sensing time-series data and digital elevation models of the target area and adjacent areas of the target area to be analyzed for functional fragmentation of cultivated land. The multi-source remote sensing time-series data includes optical data with time series and radar data with time series from the same period as the optical data. The multidimensional feature fragmentation index set construction module 20 is used to analyze the physical geometric fragmentation of the target area, the phenological heterogeneity fragmentation determined based on the time-series remote sensing data, and the micro-topography and operation resistance fragmentation used to characterize the continuity of farmland operations, based on multi-source remote sensing time-series data, to obtain a multidimensional feature fragmentation index set for the target area. The functional fragmentation analysis module 30 is used to obtain the functional fragmentation degree of the target area based on a multidimensional feature fragmentation index set and combined with nonlinear machine learning for functional fragmentation analysis. The functional fragmentation degree is used to characterize the functional fragmentation status of the quantified target area.

[0072] It should be noted that the farmland functional fragmentation analysis system based on multi-source remote sensing time-series data in this embodiment corresponds to the aforementioned farmland functional fragmentation analysis method based on multi-source remote sensing time-series data. Therefore, any content not specifically described in the farmland functional fragmentation analysis system based on multi-source remote sensing time-series data in this embodiment, including but not limited to functional definitions, working principles, and technical effects, can be referred to the description in the aforementioned farmland functional fragmentation analysis method based on multi-source remote sensing time-series data, and will not be repeated here.

[0073] The functional fragmentation analysis method and system for cultivated land based on multi-source remote sensing time-series data in this embodiment will now be described in conjunction with specific application scenarios.

[0074] Specific application scenarios include: dynamic monitoring of farmland fragmentation and assessment of abandonment potential in a hilly agricultural area over a four-year period.

[0075] Specific implementation steps: Using the Google Earth Engine (GEE) platform, Sentinel-2 (10m spatial resolution, 5-day revisit period) and Sentinel-1 SAR data covering hilly agricultural areas were batch-screened. Cloud masking was performed using the QA band, and the maximum NDVI values ​​for each month over four years were extracted to create a composite time series. Lee filtering was used to remove speckles from the SAR data, and co-polarized backscattering coefficients were extracted. SRTM 30m DEM data was then resampled to 10m.

[0076] Using Python parallel computing scripts: Extract the second / third land survey mask for a region. Calculate the geometric edge density within a 5×5 pixel window to obtain... Raster image.

[0077] For each adjacent 10m farmland pixel pair, an NDVI time-series vector of length 12 (by month) is extracted. The DTW distance is calculated using the dtaidistance library, and the vector is averaged to generate phenological heterogeneity. Raster image.

[0078] The slope map extracted from the DEM and the annual variance map of Sentinel-1 are weighted according to the set weights, such as... , Superimposed, generating terrain resistance Raster image.

[0079] In suburban area A and mountainous area B, 1500 known abandoned and 2000 continuously cultivated farmland samples were drawn. The extracted... , and The corresponding numerical input is used to classify the XGBoost classifier. The learning rate is 0.05, and the tree depth is 6. After training convergence (AUC reaches 0.92), inference is performed on the feature map of the entire hilly agricultural area, and the output range is within... High precision between Spatial distribution map of functional fragmentation. The area was marked as a high-energy fracture zone with a high risk of abandonment.

[0080] Execute the SHAP spatial attribution output, run the SHAP interpreter toolkit, and calculate the contribution matrix for each cell. The final output consists of four images: one comprehensive image. The prediction map, along with three feature-dominant maps. For example, the implementation results show that the highly fragmented area of ​​region A is mainly composed of... (Phenological heterogeneity) dominates, indicating frequent land transfers and scattered planting structures in the suburbs; while the highly fragmented area of ​​region B is identified as The dominant factor is (terrain resistance), indicating that it is due to limitations imposed by objective natural geographical conditions.

[0081] In summary, the method and system for analyzing the functional fragmentation of cultivated land based on multi-source remote sensing time-series data have brought at least the following technical benefits: A functional fragmentation measurement framework integrating time-series DTW phenological distance and radar terrain resistance is proposed, which breaks through the limitations of two-dimensional geometry, incorporates physical inaccessibility and operational heterogeneity into the fragmentation consideration, and achieves breakthroughs in multi-dimensional physical and time constraints. By using high-frequency multi-source time series, we can effectively identify physically connected but fragmented farmland units in actual planting structures, improve the accuracy of change detection in complex hilly areas, and capture hidden signs of land abandonment. The SHAP framework not only outputs quantitative functional fragmentation, but also accurately indicates to agricultural managers whether the fragmentation of a plot is due to steep terrain or complex ownership and management, providing direct evidence for comprehensive land consolidation and the construction of high-standard farmland, and realizing white-box decision support.

[0082] In the embodiments provided in this application, it should be understood that the embodiments described herein can be implemented in hardware, software, firmware, middleware, code, or any suitable combination thereof. For hardware implementation, the processor may be implemented in one or more of the following: application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, other electronic units designed to implement the functions described herein, or combinations thereof. For software implementation, some or all of the processes of the embodiments may be performed by a computer program instructing the associated hardware. During implementation, the program may be stored in a computer-readable storage medium or transmitted as one or more instructions or code on a computer-readable storage medium. Computer-readable storage media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of a computer program from one place to another. Storage media may be any available medium accessible to a computer. Computer-readable storage media may include, but are not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code having the form of instructions or data structures and accessible to a computer.

[0083] Finally, it should be noted that the above description is only a preferred embodiment of this application and is not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for functional fragmentation analysis of cultivated land based on multi-source remote sensing time-series data, characterized in that, include: Acquire multi-source remote sensing time-series data and digital elevation models of the target area and its adjacent areas to be analyzed for functional fragmentation of arable land. The multi-source remote sensing time-series data includes optical data with time series and radar data with time series from the same period as the optical data. Based on multi-source remote sensing time-series data, the physical geometric fragmentation of the target area, the phenological heterogeneity fragmentation determined based on time-series trajectories, and the micro-topography and operational resistance fragmentation used to characterize the continuity of farmland operations are analyzed to obtain a multi-dimensional feature fragmentation index set for the target area. Based on a multidimensional feature fragmentation index set and combined with nonlinear machine learning, the functional fragmentation degree of the target region is obtained for functional fragmentation analysis. This functional fragmentation degree is used to characterize the functional fragmentation status of the quantified target region. The analysis of the physical and geometric fracture conditions includes: A sliding window is constructed centered on the target area and extends to adjacent areas; The physical geometric fragmentation is calculated based on the length of the cultivated land boundary and the proportion of centrally connected cultivated land patches within the sliding window. This physical geometric fragmentation is used to quantify the physical geometric fragmentation situation. The analysis of the phenological heterogeneity fragmentation is specifically an analysis of the hidden functional fragmentation that reflects the physical connection but operational heterogeneity between the target area and the adjacent areas. The analysis of micro-topography and operational resistance fragmentation specifically involves analyzing the micro-resistance of agricultural machinery operating between the target area and adjacent areas in three-dimensional space.

2. The method for functional fragmentation analysis of cultivated land based on multi-source remote sensing time-series data according to claim 1, characterized in that, The fragmentation of the hidden function is quantified by the phenological heterogeneity fragmentation degree, which is determined based on the dynamic time warping algorithm.

3. The method for functional fragmentation analysis of cultivated land based on multi-source remote sensing time-series data according to claim 1, characterized in that, The micro-resistance is quantified by micro-topography and operational resistance fragmentation, which is determined by the surface roughness variation coefficient based on radar data and the slope based on a digital elevation model.

4. The method for functional fragmentation analysis of cultivated land based on multi-source remote sensing time-series data according to claim 3, characterized in that, The phenological heterogeneity fragmentation is determined based on a dynamic time warping algorithm. Specifically, the similarity of vegetation index trajectories between the target area and adjacent areas in the time series is determined by the dynamic time warping algorithm, and the phenological heterogeneity fragmentation is determined based on the similarity of vegetation index trajectories.

5. The method for functional fragmentation analysis of cultivated land based on multi-source remote sensing time-series data according to claim 4, characterized in that, The radar data includes co-polarized data and cross-polarized data; the determination of the surface roughness variation coefficient includes determination based on co-polarized data; the determination of the surface roughness variation coefficient also includes determination based on co-polarized data and cross-polarized data.

6. The method for functional fragmentation analysis of cultivated land based on multi-source remote sensing time-series data according to claim 1, characterized in that, Based on the multidimensional feature fragmentation index set and the functional fragmentation degree, and combined with the SHAP framework, local attribution analysis is performed on the functional fragmentation analysis, and a driving force distribution map for the target region is output.

7. A system for analyzing the functional fragmentation of cultivated land based on multi-source remote sensing time-series data, employing the method for analyzing the functional fragmentation of cultivated land based on multi-source remote sensing time-series data as described in any one of claims 1 to 6, characterized in that, include: The data cube construction module is used to acquire multi-source remote sensing time-series data and digital elevation models of the target area and adjacent areas of the target area to be analyzed for functional fragmentation of arable land. The multi-source remote sensing time-series data includes optical data with time series and radar data with time series from the same period as the optical data. The multidimensional feature fragmentation index set construction module is used to analyze the physical geometric fragmentation of the target area based on multi-source remote sensing time series data, the phenological heterogeneity fragmentation determined based on time series trajectory, and the micro-topography and operation resistance fragmentation used to characterize the continuity of farmland operations, so as to obtain a multidimensional feature fragmentation index set for the target area. The functional fragmentation analysis module is used to obtain the functional fragmentation degree of the target area based on a multidimensional feature fragmentation index set and combined with nonlinear machine learning for functional fragmentation analysis. This functional fragmentation degree is used to characterize the functional fragmentation status of the quantified target area.

Images