Dynamic monitoring method and system for national space planning based on remote sensing image

Abstract

This invention relates to the field of land and space planning monitoring technology, specifically disclosing a method and system for dynamic monitoring of land and space planning based on remote sensing imagery. The method involves acquiring multi-temporal optical and radar images, extracting temporary land use units, their reclamation periods, and target land types; extracting the temporal spectral and temporal deformation characteristics of each temporary land use unit, and fusing them to construct a multi-dimensional temporal feature vector; constructing a feature evolution benchmark trajectory for similar land use within a spatial neighborhood, using the reclamation period as a constraint; dynamically comparing the actual feature segments after the reclamation period with the benchmark trajectory, calculating the degree of continuous deviation, and identifying candidate abnormal units; and overlaying and analyzing the candidate abnormal units with the planned use zones to generate compliance monitoring results. This invention, by fusing optical and radar temporal features, constructing a benchmark trajectory, and calculating the degree of continuous deviation, achieves proactive and accurate monitoring of situations where temporary land use expires and is not reclaimed, or where changes that should have occurred have not.

Description

Technical Field

[0001] This invention relates to the field of land spatial planning monitoring technology, specifically to a method and system for dynamic monitoring of land spatial planning based on remote sensing imagery. Background Technology

[0002] Dynamic monitoring after the implementation of territorial spatial planning is a crucial means to ensure the effective implementation of the plan and to promptly detect violations. With the rapid development of high-resolution remote sensing satellite technology, land cover change detection using remote sensing imagery has become the mainstream technical approach for dynamic monitoring of territorial spatial planning. Existing technologies typically employ dual-temporal or multi-temporal optical remote sensing imagery. By extracting the spectral differences between images from different time periods, land cover changes are identified. These changed land cover patches are then spatially overlaid with the territorial spatial planning data to determine whether they comply with planning control requirements, thereby enabling the monitoring of explicit changes such as illegal construction and the conversion of arable land to non-agricultural uses.

[0003] The current dynamic monitoring method for land spatial planning based on remote sensing imagery relies on the detection of explicit changes in images from different time periods. Its core logic is to "discover changes that have already occurred." However, for planning control elements with clear reclamation periods, such as temporary land use, the real violation is not "construction changes have occurred," but rather the implicit state of "failure to change from construction land to the original land type after the reclamation period expires." Because there is no obvious spectral difference between images from different time periods in this state, existing technologies cannot proactively identify such violations through conventional change detection methods, resulting in serious lag in monitoring results and creating regulatory blind spots. This invention introduces the joint inversion of temporal radar deformation features and optical spectral features to construct a feature evolution benchmark trajectory within the spatial neighborhood. It also calculates the continuous deviation between the actual feature segment after the reclamation period and the benchmark trajectory, transforming compliance judgment from "static change detection" to "temporal benchmark comparison," thereby achieving proactive discovery and accurate judgment of the implicit violation pattern of "failure to change as required." Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for dynamic monitoring of land spatial planning based on remote sensing imagery, so as to solve the problems mentioned above.

[0005] The objective of this invention can be achieved through the following technical solutions: The method for dynamic monitoring of land spatial planning based on remote sensing imagery includes the following steps: S1 acquires multi-temporal optical and radar images covering the monitoring area, as well as land use planning data, and extracts temporary land use units and their associated reclamation periods and target land categories from the land use planning data; S2, extract the temporal spectral features of each temporary land use unit based on multi-temporal optical images, extract the temporal deformation features of each temporary land use unit based on multi-temporal radar images, and fuse the temporal spectral features and temporal deformation features into a multi-dimensional temporal feature vector for each temporary land use unit; S3 uses the reclamation period as a time constraint to perform spatiotemporal correlation anomaly inference on the feature segments before the reclamation period in the multidimensional time series feature vector. By constructing the feature distribution map of temporary land use units in the spatial neighborhood, it learns the feature evolution benchmark trajectory of similar land use before and after the reclamation period. S4. Dynamically compare the feature segments after the reclamation period in the multidimensional time series feature vector with the feature evolution baseline trajectory, calculate the continuous deviation of the actual feature trajectory relative to the feature evolution baseline trajectory, and identify candidate abnormal units in the state of being reclaimed but not reclaimed based on the continuous deviation. S5 performs spatial overlay analysis on the spatial boundaries of candidate anomaly units and land spatial planning data to generate compliance monitoring results that include information on the location, area, and continuous deviation of candidate anomaly units.

[0006] As a further aspect of the present invention: S2 specifically includes: The temporal spectral features and temporal deformation features of the same temporary land use unit are aligned according to the acquisition time to construct the original feature pairs at each time sampling point; For each time sampling point, the confidence level of the deformation feature of the sampling point is determined based on the coherence coefficient of the radar image. When the coherence coefficient is higher than the preset threshold, the temporal deformation feature is used as the dominant component. When the coherence coefficient is lower than the preset threshold, the temporal spectral feature is used as the dominant component. The feature pairs selected by the dominant component at each time sampling point are spliced ​​together in chronological order to form a multidimensional temporal feature vector for each temporary land use unit.

[0007] As a further aspect of the present invention: the process of obtaining the confidence level of the deformation feature is as follows: Acquire the coherence coefficient values ​​of each pixel in the radar image within the boundary of the temporary land use unit, and calculate the spatial mean and spatial variance of the coherence coefficient; When the spatial mean is higher than the first threshold and the spatial variance is lower than the second threshold, the confidence level of the deformation feature is set to high confidence level; when the spatial mean is lower than the third threshold or the spatial variance is higher than the fourth threshold, the confidence level of the deformation feature is set to low confidence level. High or low confidence levels are used as the deformation feature confidence levels of the sampling points for subsequent selection of dominant components.

[0008] As a further aspect of the present invention: S3 specifically includes: Taking each temporary land use unit as the center, extract adjacent units that are less than a preset distance threshold and have the same land use type as the spatial neighborhood set; The multidimensional temporal feature vectors of all temporary land use units within the spatial neighborhood set at each time sampling point before the reclamation deadline are grouped according to the time sampling point, and the feature value distribution interval at each time sampling point is calculated. Extract the feature value corresponding to the preset quantile from the feature value distribution interval at each time sampling point, and use it as the benchmark feature value of the time sampling point; By connecting the baseline characteristic values ​​of each sampling point before the reclamation deadline in chronological order, a baseline trajectory of the characteristic evolution of land use of the same type is formed.

[0009] As a further aspect of the present invention: the process of obtaining the benchmark feature value is as follows: Obtain the feature values ​​of each temporary land use unit within the spatial neighborhood set at the time sampling point, and arrange them in ascending order according to the size of the feature values ​​to form a feature value sequence; Calculate the difference between every two adjacent feature values ​​in the feature value sequence, identify the breakpoints where the difference is greater than a preset jump threshold, and divide the feature value sequence into several feature value subsets based on the breakpoints. The subset of features containing the largest number of units is selected as the principal feature subset, and the feature values ​​corresponding to the preset quantiles are extracted from the principal feature subset as the reference feature values ​​of the time sampling points.

[0010] As a further aspect of the present invention: S4 specifically includes: Arrange the sampling points after the reclamation period in chronological order, and calculate the difference between the actual feature value at each sampling point and the benchmark feature value at the same sampling point to obtain the instantaneous deviation value of each sampling point. Starting from the first time sampling point after the reclamation period, the instantaneous deviation value of each time sampling point is accumulated. Whenever the accumulated value exceeds the preset deviation threshold, the time sampling point is recorded as the starting point of continuous deviation, and the accumulated value is reset. The number of time sampling points from the starting point to the current time sampling point is counted, and the ratio of the number of time sampling points to the preset duration threshold is used as the degree of continuous deviation. When the continuous deviation exceeds the preset continuous threshold, the temporary land use unit will be marked as a candidate abnormal unit.

[0011] As a further aspect of the present invention: S5 specifically includes: The spatial boundaries of each candidate anomaly unit are spatially overlapped with the planning use zones in the land spatial planning data to determine the planning use zone to which each candidate anomaly unit falls. Candidate abnormal units falling within the same planning use zone are sorted from high to low according to their degree of continuous deviation, and each candidate abnormal unit is divided into a priority verification level and a general verification level based on the ranking results. Generate compliance monitoring records for each candidate abnormal unit, including its spatial boundaries, spatial location, unit area, continuous deviation, planning use zoning, and priority verification level; All compliance monitoring records are grouped and aggregated according to their respective planned use zones to form compliance monitoring results indexed by the planned use zones.

[0012] As a further aspect of the present invention: the step of spatially overlapping the spatial boundaries of each candidate anomaly unit with the planned use zones in the land spatial planning data to determine the planned use zone to which each candidate anomaly unit falls specifically includes: Obtain the boundaries of each planned use zone in the land use planning data, and establish attribute indexes for each planned use zone according to its permitted land use types; For the spatial boundary of each candidate abnormal unit, calculate the overlap area between the spatial boundary and each planned use zone, and select the planned use zones with an overlap area greater than a preset area threshold as candidate zones. When there are multiple candidate zones, the permitted land use types of each candidate zone in the overlapping area are extracted. The permitted land use types are compared with the target reclaimed land type of the temporary land use unit corresponding to the candidate abnormal unit. The candidate zone with the highest matching degree between the permitted land use type and the target reclaimed land type is selected as the planned use zone to which the candidate abnormal unit falls.

[0013] A dynamic monitoring system for land spatial planning based on remote sensing imagery includes: The data acquisition module is used to acquire multi-temporal optical and radar images covering the monitoring area, as well as land use planning data, and to extract temporary land use units and their associated reclamation periods and target land types from the land use planning data. The multi-dimensional temporal feature vector construction module extracts the temporal spectral features of each temporary land use unit based on multi-temporal optical images, extracts the temporal deformation features of each temporary land use unit based on multi-temporal radar images, and fuses the temporal spectral features and temporal deformation features into a multi-dimensional temporal feature vector for each temporary land use unit. The feature evolution baseline trajectory construction module uses the reclamation period as a time constraint to perform spatiotemporal correlation anomaly inference on the feature segments before the reclamation period in the multi-dimensional temporal feature vector. By constructing the feature distribution map of temporary land use units in the spatial neighborhood, it learns the feature evolution baseline trajectory of similar land use before and after the reclamation period. The candidate anomalous unit identification module dynamically compares the feature segments after the reclamation period in the multidimensional time-series feature vector with the feature evolution baseline trajectory, calculates the continuous deviation of the actual feature trajectory from the feature evolution baseline trajectory, and identifies candidate anomalous units in the state of being reclaimed but not reclaimed based on the continuous deviation. The compliance monitoring results generation module performs spatial overlay analysis on the spatial boundaries of candidate abnormal units and land spatial planning data to generate compliance monitoring results that include information on the location, area, and continuous deviation of candidate abnormal units.

[0014] The beneficial effects of this invention are: (1) This invention introduces time-series radar images to extract deformation features of temporary land use units and fuses them with spectral features extracted from optical images. Based on the confidence level of the coherence coefficient, the dominant component is adaptively selected to construct a multi-dimensional time-series feature vector. This solves the limitation of traditional optical remote sensing, which relies solely on spectral change detection to identify the state of "should have changed but has not changed". It realizes the proactive discovery and accurate identification of temporary land use that has expired and has not been reclaimed, thus improving the timeliness and accuracy of monitoring.

[0015] (2) This invention constructs a feature distribution map of similar land use in a spatial neighborhood and extracts the feature evolution baseline trajectory after removing outliers. It dynamically compares the actual features after the reclamation period with the baseline trajectory, calculates the continuous deviation by accumulating instantaneous deviation values, and divides the verification level according to the continuous deviation, thereby realizing the refined hierarchical management of candidate abnormal units, effectively reducing the workload of grassroots verification, and avoiding the waste of regulatory resources caused by a large number of false reports. Attached Figure Description

[0016] The invention will now be further described with reference to the accompanying drawings.

[0017] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 As shown, this invention is a method for dynamic monitoring of land spatial planning based on remote sensing imagery, comprising the following steps: S1 acquires multi-temporal optical and radar images covering the monitoring area, as well as land use planning data, and extracts temporary land use units and their associated reclamation periods and target land categories from the land use planning data; S2, extract the temporal spectral features of each temporary land use unit based on multi-temporal optical images, extract the temporal deformation features of each temporary land use unit based on multi-temporal radar images, and fuse the temporal spectral features and temporal deformation features into a multi-dimensional temporal feature vector for each temporary land use unit; S3 uses the reclamation period as a time constraint to perform spatiotemporal correlation anomaly inference on the feature segments before the reclamation period in the multidimensional time series feature vector. By constructing the feature distribution map of temporary land use units in the spatial neighborhood, it learns the feature evolution benchmark trajectory of similar land use before and after the reclamation period. S4. Dynamically compare the feature segments after the reclamation period in the multidimensional time series feature vector with the feature evolution baseline trajectory, calculate the continuous deviation of the actual feature trajectory relative to the feature evolution baseline trajectory, and identify candidate abnormal units in the state of being reclaimed but not reclaimed based on the continuous deviation. S5 performs spatial overlay analysis on the spatial boundaries of candidate anomaly units and land spatial planning data to generate compliance monitoring results that include information on the location, area, and continuous deviation of candidate anomaly units.

[0020] In S1, multi-temporal optical and radar images covering the monitoring area are acquired, along with land use planning data. Temporary land use units and their associated reclamation periods and target land categories are extracted from the land use planning data, specifically including: Multi-temporal optical and radar images covering the monitoring area were acquired. The multi-temporal optical images were acquired from Gaofen-2, Gaofen-6, or Sentinel-2 satellites, downloaded via satellite ground receiving stations. These images have a spatial resolution better than two meters, spanning from the date of temporary land use approval to the current monitoring cycle, with sampling intervals of once per month or quarter. The multi-temporal radar images were acquired from Sentinel-1 or LandScan-1 satellites, downloaded via satellite ground receiving stations. These images have a spatial resolution better than twenty meters, with the same time span and sampling interval as the optical images. The acquired multi-temporal optical images underwent radiometric calibration, atmospheric correction, and geometric fine correction sequentially. Scene-by-scene registration was then performed using the initial imagery as a reference to ensure precise spatial alignment of all optical images. The acquired multi-temporal radar images underwent registration, interferogram generation, phase unwrapping, and atmospheric phase contribution removal. Coherence coefficients and surface deformation information for each radar image period were extracted.

[0021] Obtain land and space planning data. Land and space planning data comes from the "One Map" database of land and space planning released by the natural resources authorities. This database is stored in vector format and includes planning use zoning layers, construction land approval layers, temporary land approval layers, and attribute information for each layer.

[0022] Temporary land use units and their associated reclamation periods and target land categories are extracted from territorial spatial planning data. Specifically, from the temporary land use approval layer of territorial spatial planning data, temporary land use units with valid and uncancelled approval status are selected using spatial queries. The spatial boundaries, approved area, reclamation period, and target reclamation land category of each temporary land use unit are extracted. The reclamation period is the deadline for reclamation completion determined at the time of temporary land use approval, and the target reclamation land category is the land use type that should be restored after the expiration of the temporary land use period. The extracted spatial boundaries, reclamation periods, and target reclamation land categories of the temporary land use units are stored in a structured format as processing objects for subsequent steps, and a spatiotemporal correlation index is established between each temporary land use unit and multi-temporal optical and radar imagery.

[0023] In S2, the temporal spectral features of each temporary land use unit are extracted based on multi-temporal optical images, and the temporal deformation features of each temporary land use unit are extracted based on multi-temporal radar images. The temporal spectral features and temporal deformation features are then fused into a multi-dimensional temporal feature vector for each temporary land use unit, specifically including: Temporal spectral features of each temporary land use unit were extracted based on multi-temporal optical imagery. For each temporary land use unit, the spectral reflectance values ​​of all pixels within the unit's spatial boundary were obtained. The pixel mean values ​​for the red band and near-infrared band were calculated separately, and then the normalized vegetation index (NVI) value was calculated. This NVI value is equal to the difference between the mean reflectance values ​​of the near-infrared band and the red band, divided by the sum of the mean reflectance values ​​of the near-infrared band and the red band. The NVI values ​​calculated at each sampling point were arranged according to the order of the time sampling points to form the temporal spectral features of the temporary land use unit. This feature is represented by a sequence of NVI values ​​arranged in chronological order.

[0024] Temporal deformation features of each temporary land use unit are extracted based on multi-temporal radar imagery. For each temporary land use unit, the deformation displacement values ​​of all pixels within the unit's spatial boundary are obtained from radar imagery across different periods. These deformation displacement values ​​are obtained through interferometric phase transformation after phase unwrapping, representing the cumulative vertical displacement of the land surface relative to the first image. The mean value of the deformation displacement values ​​of all pixels within the unit's spatial boundary is calculated. The mean deformation displacement values ​​calculated at each sampling point are arranged according to the order of the time sampling points to form the temporal deformation feature of the temporary land use unit. This feature is represented as a sequence of mean deformation displacement values ​​arranged in chronological order.

[0025] The temporal spectral and temporal deformation features of the same temporary land use unit are aligned according to the collection time to construct original feature pairs for each time sampling point. Specifically, for each time sampling point, the normalized vegetation index value corresponding to that sampling point is extracted from the temporal spectral features, and the mean deformation displacement corresponding to that sampling point is extracted from the temporal deformation features. The two are combined into an original feature pair, which contains spectral and deformation information from the same time sampling point.

[0026] For each time sampling point, the deformation feature confidence level is determined based on the coherence coefficient of the radar image. For each time sampling point, the corresponding radar image is acquired, and the coherence coefficient values ​​of all pixels within the spatial boundary of the temporary land use unit are extracted. The coherence coefficient values ​​are between 0 and 1, reflecting the stability of the ground object scattering characteristics between two radar images. The spatial mean of the coherence coefficient values ​​within the unit is calculated, which is the sum of the coherence coefficients of all pixels divided by the total number of pixels; simultaneously, the spatial variance of the coherence coefficient values ​​is calculated, which is the sum of the squares of the differences between the coherence coefficient values ​​of each pixel and the spatial mean divided by the total number of pixels. When the spatial mean is higher than a pre-set first threshold and the spatial variance is lower than a pre-set second threshold, the deformation feature confidence level of the sampling point is set to high confidence; when the spatial mean is lower than a pre-set third threshold or the spatial variance is higher than a pre-set fourth threshold, the deformation feature confidence level of the sampling point is set to low confidence. The first threshold is 0.7, the second threshold is 0.05, the third threshold is 0.3, and the fourth threshold is 0.10.

[0027] Based on the confidence level of deformation features, the temporal deformation features or temporal spectral features are used as the dominant components for each sampling point. When the confidence level of the deformation features of a sampling point is high, the mean deformation displacement in the temporal deformation features of that sampling point is used as the dominant component. This dominant component is retained together with the normalized vegetation index value in the temporal spectral features of the same sampling point to form the filtered feature pair for that sampling point. When the confidence level of the deformation features of a sampling point is low, the normalized vegetation index value in the temporal spectral features of that sampling point is used as the dominant component. This dominant component is retained together with the mean deformation displacement in the temporal deformation features of the same sampling point to form the filtered feature pair for that sampling point.

[0028] The feature pairs selected by the dominant component at each time sampling point are concatenated in chronological order to form a multidimensional temporal feature vector for each temporary land use unit. Specifically, the selected feature pairs at each sampling point are arranged sequentially from earliest to latest, with each feature pair containing two numerical components. All feature pairs are concatenated in chronological order to obtain a one-dimensional sequence. The length of this sequence is twice the number of time sampling points. This sequence is the multidimensional temporal feature vector for the temporary land use unit, where the dominant component at each sampling point reflects the more reliable monitoring information at that sampling point.

[0029] In S3, using the reclamation period as a time constraint, anomaly inferences are made regarding the spatiotemporal correlation of feature segments before the reclamation period in the multidimensional temporal feature vector. By constructing a feature distribution map of temporary land use units within their spatial neighborhood, the baseline trajectory of feature evolution of similar land use before and after the reclamation period is learned, specifically including: Taking each temporary land use unit as the center, adjacent units with the same land use type and a spatial distance less than a preset distance threshold are extracted as a spatial neighborhood set. Specifically, the spatial boundaries of all temporary land use units are obtained from the land spatial planning data. Using the geometric center point of each temporary land use unit as a reference, the Euclidean distance between this center point and the geometric center points of other temporary land use units is calculated. Temporary land use units with a distance less than a preset distance threshold are selected, with the preset distance threshold set to 500 meters. Among the selected temporary land use units, units with the same land use type as the current temporary land use unit are further extracted. The land use type is matched according to the land use classification code in the temporary land use approval layer. All temporary land use units that meet the conditions of spatial distance and the same land use type constitute the spatial neighborhood set of the current unit. If the number of temporary land use units in the spatial neighborhood set is less than 3, the preset distance threshold is gradually increased to 1000 meters, 1500 meters, until the number of temporary land use units in the spatial neighborhood set reaches 3 or more.

[0030] The multidimensional temporal feature vectors of all temporary land use units within the spatial neighborhood set at each time sampling point before the reclamation deadline are grouped according to the time sampling point, and the feature value distribution interval at each time sampling point is calculated. Specifically, the feature values ​​of all temporary land use units within the spatial neighborhood set at each time sampling point before the reclamation deadline are obtained. The feature segment before the reclamation deadline refers to the portion of the time sampling point whose corresponding date is earlier than the reclamation deadline date. For each time sampling point, the feature values ​​of all spatial neighborhood units at that time sampling point are aggregated into a feature value set. The minimum value in this feature value set is calculated as the lower limit of the distribution interval, and the maximum value is calculated as the upper limit of the distribution interval, forming the feature value distribution interval for that time sampling point.

[0031] The feature value corresponding to the preset quantile is extracted from the feature value distribution interval at each time sampling point, and used as the benchmark feature value for that time sampling point. The preset quantile is set to 0.5, that is, the median of the feature value distribution interval is extracted as the benchmark feature value. When extracting the median, the feature values ​​of each temporary land use unit in the spatial neighborhood set at that time sampling point are first arranged in ascending order of numerical value to form a feature value sequence. Let the total number of feature values ​​in the feature value sequence be... ,like If the number is odd, then take the first number in the sequence. The eigenvalues ​​are used as the median; if If the number is even, then take the th element in the sequence. The eigenvalue and the th eigenvalue The arithmetic mean of the eigenvalues ​​is used as the median.

[0032] To improve the robustness of the baseline feature values ​​and avoid the influence of outlier feature values ​​on the distribution range, outlier removal is performed on the feature value sequence before extracting the baseline feature values. Specifically, the feature values ​​of each temporary land use unit within the spatial neighborhood set at the sampling point at that time are obtained and arranged in ascending order of feature value magnitude to form a feature value sequence. The difference between any two adjacent feature values ​​in the feature value sequence is calculated, and when the difference is greater than a preset jump threshold, the adjacent position is marked as a breakpoint. Let the feature value sequence be... , , ..., Sort in ascending order, for any two adjacent eigenvalues and Calculate the difference .when When the threshold value is greater than the preset threshold value T, and The points between each breakpoint are marked as breakpoints, and the preset transition threshold T is set to twice the average difference between all adjacent feature values. Based on the breakpoint locations, the feature value sequence is divided into several feature value subsets. Each feature value subset consists of consecutively arranged feature values ​​whose differences between adjacent feature values ​​are not greater than the preset transition threshold. The number of temporary land use units contained in each feature value subset is counted, and the feature value subset containing the largest number of units is selected as the principal feature value subset. If multiple feature value subsets contain the same maximum number of units, the subset with the smallest average feature value is selected as the principal feature value subset. Within the principal feature value subset, the feature value corresponding to the preset quantile is extracted using the median extraction method described above, and this feature value is used as the baseline feature value for that time sampling point.

[0033] By connecting the baseline characteristic values ​​of each sampling point before the reclamation deadline in chronological order, a baseline trajectory of characteristic evolution for similar land uses is formed. Specifically, according to the chronological order of the sampling points, the baseline characteristic values ​​corresponding to each sampling point are arranged sequentially, forming a curve with time as the horizontal axis and baseline characteristic values ​​as the vertical axis. This curve represents the baseline trajectory of characteristic evolution for similar land uses before the reclamation deadline. This baseline trajectory reflects the typical pattern of characteristic value changes over time for similar temporary land uses under normal reclamation processes.

[0034] In S4, the feature segments after the reclamation period in the multi-dimensional time-series feature vector are dynamically compared with the feature evolution baseline trajectory. The persistence deviation of the actual feature trajectory relative to the feature evolution baseline trajectory is calculated, and candidate anomalous units in the state of not being reclaimed when they should be reclaimed are identified based on the persistence deviation. Specifically, this includes: The sampling points after the reclamation period are arranged in chronological order. The difference between the actual feature value at each sampling point and the baseline feature value at the same sampling point is calculated to obtain the instantaneous deviation value of each sampling point. Specifically, for each temporary land use unit, the feature value in the multidimensional temporal feature vector at each sampling point after the reclamation period is obtained as the actual feature value. At the same time, the baseline feature value at the corresponding sampling point in the baseline trajectory of the evolution of similar land use features constructed in step S3 is obtained. The actual feature value is subtracted from the baseline feature value, and the calculation result is used as the instantaneous deviation value of that sampling point.

[0035] Starting from the first time sampling point after the reclamation period, the instantaneous deviation values ​​of each time sampling point are accumulated. Whenever the accumulated value exceeds a preset deviation threshold, that time sampling point is recorded as the starting point of continuous deviation, and the accumulated value is reset. Specifically, the instantaneous deviation value corresponding to the first time sampling point after the reclamation period is set as the first instantaneous deviation value, which is used as the initial accumulated value, and the instantaneous deviation values ​​of subsequent time sampling points are accumulated sequentially. When the accumulated value is greater than the preset deviation threshold, the current time sampling point is marked as the starting point of continuous deviation, and the accumulated value is reset to zero, starting the accumulation process again from the next time sampling point. The preset deviation threshold is set to 0.5. If the accumulated value never exceeds the preset deviation threshold during the accumulation process, there is no starting point of continuous deviation.

[0036] The number of time sampling points elapsed from the starting point to the current time sampling point is counted, and the ratio of this number of time sampling points to a preset duration threshold is used as the persistence deviation. Specifically, using the marked persistence deviation starting point as a baseline, starting from that starting point, the number of time sampling points elapsed for each subsequent time sampling point is recorded, and this number is divided by the preset duration threshold to obtain the persistence deviation. The preset duration threshold is set to 6 time sampling points, corresponding to a monitoring period of 6 months. For example, if 3 time sampling points have elapsed from the starting point to the current time sampling point, the persistence deviation is 3 divided by 6, resulting in a value of 0.5.

[0037] When the sustained deviation exceeds a preset sustained threshold, the temporary land use unit is marked as a candidate anomalous unit. The preset sustained threshold is set to 0.5. When the sustained deviation exceeds 0.5, it indicates that the actual feature value of the temporary land use unit has continuously deviated from the baseline trajectory for more than 3 consecutive time sampling points after the reclamation period, and the temporary land use unit is identified as a candidate anomalous unit in a state of being reclaimed but not reclaimed.

[0038] In S5, the spatial boundaries of candidate anomaly units are spatially overlaid with land and space planning data to generate compliance monitoring results that include information on the location, area, and persistent deviation of the candidate anomaly units. Specifically, this includes: The spatial boundaries of each candidate anomaly unit are spatially overlapped with the planned use zones in the land spatial planning data to determine the planned use zone to which each candidate anomaly unit falls. Specifically, boundary vector data of all planned use zones are obtained from the land spatial planning data, and each planned use zone is associated with permitted land use type attributes. For each candidate anomaly unit, its spatial boundary is spatially superimposed with the boundaries of each planned use zone to calculate the overlap area between the spatial boundary of the candidate anomaly unit and each planned use zone. The method for calculating the overlap area is as follows: the closed area formed by the intersection of the spatial boundary of the candidate anomaly unit and the boundary of the planned use zone is extracted, and the area of ​​this closed area is taken as the overlap area. Planned use zones with an overlap area greater than a preset area threshold are selected as candidate zones, and the preset area threshold is set to 10% of the total area of ​​the candidate anomaly units. When there is only one candidate zone, the candidate zone is determined as the planned use zone to which the candidate anomaly unit falls. When there are multiple candidate zones, the permitted land use types within the overlapping areas of each candidate zone are extracted. The target reclaimed land type of the temporary land use unit corresponding to the candidate anomaly unit is compared with each permitted land use type one by one. The matching degree between the target reclaimed land type and the permitted land use type is calculated as follows: if the target reclaimed land type and the permitted land use type are completely identical, the matching degree is 100%; if the target reclaimed land type belongs to a subclass of the permitted land use type, the matching degree is 80%; if the permitted land use type belongs to a subclass of the target reclaimed land type, the matching degree is 60%; if there is partial overlap but they are not mutually exclusive, the matching degree is 30%; if there is no correlation between the two, the matching degree is zero. The candidate zone with the highest matching degree is selected as the planned use zone to which the candidate anomaly unit falls.

[0039] Candidate anomaly units falling within the same planning use zone are sorted from highest to lowest according to their persistent deviation, and then divided into priority verification levels and general verification levels based on the sorting results. Specifically, for each planning use zone, all candidate anomaly units falling within that zone and their persistent deviations are obtained and sorted in descending order of persistent deviation values. The top 30% of candidate anomaly units in terms of persistent deviation are selected as the priority verification level, and the remaining candidate anomaly units are selected as the general verification level.

[0040] For each candidate anomalous unit, a compliance monitoring record is generated, including its spatial boundary, spatial location, unit area, continuous deviation, planned use zoning, and priority verification level. Specifically, the spatial boundary is recorded in the form of a vector coordinate string, the spatial location is recorded in the longitude and latitude coordinates of the geometric center point, the unit area is recorded in square meters, the continuous deviation is recorded in the value calculated in step S4, the planned use zoning is recorded in the name or code of the zoning, and the priority verification level is recorded in the priority or general identifier. The above information is integrated into a structured monitoring record.

[0041] All compliance monitoring records are grouped and aggregated according to their respective planning use zones to form compliance monitoring results indexed by the planning use zones. Specifically, using the name or code of the planning use zone as the grouping key, the compliance monitoring records of all candidate abnormal units belonging to the same planning use zone are gathered together to form a monitoring record list corresponding to that zone. The monitoring record lists of all zones are then arranged in alphabetical order by the first letter of the zone name in pinyin, constituting the final compliance monitoring results.

[0042] Please see Figure 2 As shown, the land spatial planning dynamic monitoring system based on remote sensing imagery includes: The data acquisition module is used to acquire multi-temporal optical and radar images covering the monitoring area, as well as land use planning data, and to extract temporary land use units and their associated reclamation periods and target land types from the land use planning data. The multi-dimensional temporal feature vector construction module extracts the temporal spectral features of each temporary land use unit based on multi-temporal optical images, extracts the temporal deformation features of each temporary land use unit based on multi-temporal radar images, and fuses the temporal spectral features and temporal deformation features into a multi-dimensional temporal feature vector for each temporary land use unit. The feature evolution baseline trajectory construction module uses the reclamation period as a time constraint to perform spatiotemporal correlation anomaly inference on the feature segments before the reclamation period in the multi-dimensional temporal feature vector. By constructing the feature distribution map of temporary land use units in the spatial neighborhood, it learns the feature evolution baseline trajectory of similar land use before and after the reclamation period. The candidate anomalous unit identification module dynamically compares the feature segments after the reclamation period in the multidimensional time-series feature vector with the feature evolution baseline trajectory, calculates the continuous deviation of the actual feature trajectory from the feature evolution baseline trajectory, and identifies candidate anomalous units in the state of being reclaimed but not reclaimed based on the continuous deviation. The compliance monitoring results generation module performs spatial overlay analysis on the spatial boundaries of candidate abnormal units and land spatial planning data to generate compliance monitoring results that include information on the location, area, and continuous deviation of candidate abnormal units.

[0043] The working principle of this invention is as follows: First, multi-temporal optical and radar images, as well as land use planning data, covering the monitoring area are acquired. Temporary land use units and their associated reclamation periods and target land types are extracted from the planning data. Then, the temporal spectral features of each temporary land use unit are extracted based on the multi-temporal optical images, and the temporal deformation features of each temporary land use unit are extracted based on the multi-temporal radar images. The two are then fused after dominant component selection based on the confidence level determined by the coherence coefficient of the radar images to construct a multi-dimensional temporal feature vector for each temporary land use unit. Next, using the reclamation period as a time constraint, the feature segments before the reclamation period are selected by constructing a feature vector within the spatial neighborhood. After analyzing the distribution map and removing outliers, the baseline trajectory of the characteristic evolution of similar land use before the reclamation deadline is extracted. Then, the actual characteristic segments after the reclamation deadline are dynamically compared with the baseline trajectory. The continuous deviation is calculated by accumulating instantaneous deviation values, and candidate abnormal units in the state of not being reclaimed are identified based on the continuous deviation. Finally, the spatial boundaries of the candidate abnormal units are spatially overlaid with the planning use zoning in the land spatial planning data. The zoning to which they belong is determined by screening the overlapping area and the land use type matching degree. After dividing the verification level according to the continuous deviation, compliance monitoring results containing location, area and continuous deviation information are generated.

[0044] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for dynamic monitoring of land spatial planning based on remote sensing imagery, characterized in that, Includes the following steps: S1 acquires multi-temporal optical and radar images covering the monitoring area, as well as land use planning data, and extracts temporary land use units and their associated reclamation periods and target land categories from the land use planning data; S2, extract the temporal spectral features of each temporary land use unit based on multi-temporal optical images, extract the temporal deformation features of each temporary land use unit based on multi-temporal radar images, and fuse the temporal spectral features and temporal deformation features into a multi-dimensional temporal feature vector for each temporary land use unit; S3 uses the reclamation period as a time constraint to perform spatiotemporal correlation anomaly inference on the feature segments before the reclamation period in the multidimensional time series feature vector. By constructing the feature distribution map of temporary land use units in the spatial neighborhood, it learns the feature evolution benchmark trajectory of similar land use before and after the reclamation period. S4. Dynamically compare the feature segments after the reclamation period in the multidimensional time series feature vector with the feature evolution baseline trajectory, calculate the continuous deviation of the actual feature trajectory relative to the feature evolution baseline trajectory, and identify candidate abnormal units in the state of being reclaimed but not reclaimed based on the continuous deviation. S5 performs spatial overlay analysis on the spatial boundaries of candidate anomaly units and land spatial planning data to generate compliance monitoring results that include information on the location, area, and continuous deviation of candidate anomaly units.

2. The method for dynamic monitoring of land spatial planning based on remote sensing imagery according to claim 1, characterized in that, S2 specifically includes: The temporal spectral features and temporal deformation features of the same temporary land use unit are aligned according to the acquisition time to construct the original feature pairs at each time sampling point; For each time sampling point, the confidence level of the deformation feature of the sampling point is determined based on the coherence coefficient of the radar image. When the coherence coefficient is higher than the preset threshold, the temporal deformation feature is used as the dominant component. When the coherence coefficient is lower than the preset threshold, the temporal spectral feature is used as the dominant component. The feature pairs selected by the dominant component at each time sampling point are spliced ​​together in chronological order to form a multidimensional temporal feature vector for each temporary land use unit.

3. The method for dynamic monitoring of land spatial planning based on remote sensing imagery according to claim 2, characterized in that, The process of obtaining the confidence level of the deformation feature is as follows: Acquire the coherence coefficient values ​​of each pixel in the radar image within the boundary of the temporary land use unit, and calculate the spatial mean and spatial variance of the coherence coefficient; When the spatial mean is higher than the first threshold and the spatial variance is lower than the second threshold, the confidence level of the deformation feature is set to high confidence level; when the spatial mean is lower than the third threshold or the spatial variance is higher than the fourth threshold, the confidence level of the deformation feature is set to low confidence level. High or low confidence levels are used as the deformation feature confidence levels of the sampling points for subsequent selection of dominant components.

4. The method for dynamic monitoring of land spatial planning based on remote sensing imagery according to claim 1, characterized in that, S3 specifically includes: Taking each temporary land use unit as the center, extract adjacent units that are less than a preset distance threshold and have the same land use type as the spatial neighborhood set; The multidimensional temporal feature vectors of all temporary land use units within the spatial neighborhood set at each time sampling point before the reclamation deadline are grouped according to the time sampling point, and the feature value distribution interval at each time sampling point is calculated. Extract the feature value corresponding to the preset quantile from the feature value distribution interval at each time sampling point, and use it as the benchmark feature value of the time sampling point; By connecting the baseline characteristic values ​​of each sampling point before the reclamation deadline in chronological order, a baseline trajectory of the characteristic evolution of land use of the same type is formed.

5. The method for dynamic monitoring of land spatial planning based on remote sensing imagery according to claim 4, characterized in that, The process for obtaining the benchmark feature value is as follows: Obtain the feature values ​​of each temporary land use unit within the spatial neighborhood set at the time sampling point, and arrange them in ascending order according to the size of the feature values ​​to form a feature value sequence; Calculate the difference between every two adjacent feature values ​​in the feature value sequence, identify the breakpoints where the difference is greater than a preset jump threshold, and divide the feature value sequence into several feature value subsets based on the breakpoints. The subset of features containing the largest number of units is selected as the principal feature subset, and the feature values ​​corresponding to the preset quantiles are extracted from the principal feature subset as the reference feature values ​​of the time sampling points.

6. The method for dynamic monitoring of land spatial planning based on remote sensing imagery according to claim 1, characterized in that, S4 specifically includes: Arrange the sampling points after the reclamation period in chronological order, and calculate the difference between the actual feature value at each sampling point and the benchmark feature value at the same sampling point to obtain the instantaneous deviation value of each sampling point. Starting from the first time sampling point after the reclamation period, the instantaneous deviation value of each time sampling point is accumulated. Whenever the accumulated value exceeds the preset deviation threshold, the time sampling point is recorded as the starting point of continuous deviation, and the accumulated value is reset. The number of time sampling points from the starting point to the current time sampling point is counted, and the ratio of the number of time sampling points to the preset duration threshold is used as the degree of continuous deviation. When the continuous deviation exceeds the preset continuous threshold, the temporary land use unit will be marked as a candidate abnormal unit.

7. The method for dynamic monitoring of land spatial planning based on remote sensing imagery according to claim 1, characterized in that, S5 specifically includes: The spatial boundaries of each candidate anomaly unit are spatially overlapped with the planning use zones in the land spatial planning data to determine the planning use zone to which each candidate anomaly unit falls. Candidate abnormal units falling within the same planning use zone are sorted from high to low according to their degree of continuous deviation, and each candidate abnormal unit is divided into a priority verification level and a general verification level based on the ranking results. Generate compliance monitoring records for each candidate abnormal unit, including its spatial boundaries, spatial location, unit area, continuous deviation, planning use zoning, and priority verification level; All compliance monitoring records are grouped and aggregated according to their respective planned use zones to form compliance monitoring results indexed by the planned use zones.

8. The method for dynamic monitoring of land spatial planning based on remote sensing imagery according to claim 7, characterized in that, The step of spatially overlapping the spatial boundaries of each candidate anomaly unit with the planned use zones in the land spatial planning data to determine the planned use zone to which each candidate anomaly unit falls specifically includes: Obtain the boundaries of each planned use zone in the land use planning data, and establish attribute indexes for each planned use zone according to its permitted land use types; For the spatial boundary of each candidate abnormal unit, calculate the overlap area between the spatial boundary and each planned use zone, and select the planned use zones with an overlap area greater than a preset area threshold as candidate zones. When there are multiple candidate zones, the permitted land use types of each candidate zone in the overlapping area are extracted. The permitted land use types are compared with the target reclaimed land type of the temporary land use unit corresponding to the candidate abnormal unit. The candidate zone with the highest matching degree between the permitted land use type and the target reclaimed land type is selected as the planned use zone to which the candidate abnormal unit falls.

9. A dynamic monitoring system for land spatial planning based on remote sensing imagery, characterized in that, The method for dynamic monitoring of land spatial planning based on remote sensing imagery as described in any one of claims 1-8 includes: The data acquisition module is used to acquire multi-temporal optical and radar images covering the monitoring area, as well as land use planning data, and to extract temporary land use units and their associated reclamation periods and target land types from the land use planning data. The multi-dimensional temporal feature vector construction module extracts the temporal spectral features of each temporary land use unit based on multi-temporal optical images, extracts the temporal deformation features of each temporary land use unit based on multi-temporal radar images, and fuses the temporal spectral features and temporal deformation features into a multi-dimensional temporal feature vector for each temporary land use unit. The feature evolution baseline trajectory construction module uses the reclamation period as a time constraint to perform spatiotemporal correlation anomaly inference on the feature segments before the reclamation period in the multi-dimensional temporal feature vector. By constructing the feature distribution map of temporary land use units in the spatial neighborhood, it learns the feature evolution baseline trajectory of similar land use before and after the reclamation period. The candidate anomalous unit identification module dynamically compares the feature segments after the reclamation period in the multidimensional time-series feature vector with the feature evolution baseline trajectory, calculates the continuous deviation of the actual feature trajectory from the feature evolution baseline trajectory, and identifies candidate anomalous units in the state of being reclaimed but not reclaimed based on the continuous deviation. The compliance monitoring results generation module performs spatial overlay analysis on the spatial boundaries of candidate abnormal units and land spatial planning data to generate compliance monitoring results that include information on the location, area, and continuous deviation of candidate abnormal units.

Images