A method for dynamic monitoring of national space planning based on multi-source data fusion

Abstract

This invention discloses a dynamic monitoring method for territorial spatial planning based on multi-source data fusion, belonging to the field of territorial spatial planning technology, and includes the following steps: S1, acquiring multi-source data and preprocessing it; S2, performing time registration on cross-source state segments and screening target monitoring segments; S3, mapping the target monitoring segments to planning monitoring nodes and constructing a planning monitoring correlation diagram; S4, constructing deviation diffusion channels, transmitting node states along the deviation diffusion channels, and performing residual correction on the transmitted node states; S5, compiling effective deviation relationships into planning conflict determination operators, performing state transition matching, and performing conflict marking and early warning sorting; S6, acquiring regulatory feedback data and incrementally updating the planning monitoring correlation diagram. This invention can achieve continuous monitoring of changes in the implementation of territorial spatial planning and improve the timeliness of planning conflict early warning and handling.

Description

Technical Field

[0001] This invention relates to the field of land spatial planning technology, and in particular to a dynamic monitoring method for land spatial planning based on multi-source data fusion. Background Technology

[0002] With the gradual improvement of the national spatial planning system, dynamic monitoring of planning implementation has become an important aspect of spatial governance. Current technologies typically utilize remote sensing imagery, basic geographic information, planning outcome data, land use status data, and manual inspection records to monitor construction activities, land use changes, and the encroachment on planning boundaries. These methods can, to a certain extent, detect surface changes and provide a reference for planning implementation evaluation and regulatory action.

[0003] However, in practical applications, land spatial planning monitoring relies on numerous data sources, which differ in collection periods, spatial scales, update frequencies, and representation methods. Existing methods often employ layer overlay, manual verification, or single change detection, typically only determining whether a change has occurred at a specific point in time. This makes it difficult to accurately reflect the state differences of the same planning object across multiple data sources and to promptly identify misjudgments caused by data time discrepancies.

[0004] Meanwhile, changes in national land space often exhibit continuity and phases. Some construction changes, land use adjustments, or regulatory conflicts do not fully manifest within a single monitoring cycle but gradually form over multiple monitoring cycles. Existing technologies do not adequately address the continuity of changes between different monitoring cycles. They lack a relatively stable dynamic tracking mechanism for continuous deviations, intermittent changes, lagging changes, and changes in status after regulatory intervention, which can easily lead to duplicate, delayed, or inaccurate early warning results.

[0005] Furthermore, after regulatory review, the feedback from existing monitoring results is often only used for manual correction of the current results, making it difficult to update subsequent monitoring rules and related judgment processes. When similar areas, similar changes, or similar control conflicts occur again, the system may still repeatedly generate the same false alarms or omissions, affecting the continuity and adaptability of dynamic monitoring of territorial spatial planning.

[0006] Therefore, how to provide a dynamic monitoring method for land and space planning based on multi-source data fusion is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] One objective of this invention is to propose a dynamic monitoring method for territorial spatial planning based on multi-source data fusion. This invention fully utilizes multi-source data fusion, Graph Laplace diffusion, finite state machines, Viterbi path search, and semi-supervised label propagation technology. It describes in detail the implementation methods for multi-source data collaborative analysis, planning conflict determination, dynamic early warning generation, and feedback incremental updates in the dynamic monitoring process of territorial spatial planning. It has the advantages of strong cross-source data collaboration capability, high accuracy of dynamic monitoring, good stability of early warning results, and strong continuous adaptive update capability.

[0008] A method for dynamic monitoring of land spatial planning based on multi-source data fusion according to an embodiment of the present invention includes the following steps: S1. Acquire multi-source data under the land and space planning scenario and preprocess it to form a planning status sequence; S2. The planned state sequence is divided into sliding window segments, and dynamic time warping is used to perform time registration of cross-source state segments. The target monitoring segments are screened in combination with mutual information measurement. S3. Map the target monitoring segments to the planned monitoring nodes, establish candidate association relationships based on the window connection relationship between the target monitoring segments, and use conditional transition entropy to screen out effective deviation relationships and construct a planned monitoring association diagram. S4. In the planning and monitoring correlation diagram, construct the deviation diffusion channel based on the effective deviation relationship, use the graph Laplace diffusion method to transmit the node status along the deviation diffusion channel, and perform residual correction on the transmitted node status based on the cross-source time registration results. S5. Based on the node status of residual correction, the effective deviation relationship is compiled into a planning conflict determination operator. A finite state machine is used to perform state transition matching on the planning monitoring nodes. The planning monitoring nodes that enter the conflict state are marked for conflict and ordered for early warning, thus forming the monitoring and early warning results. S6. Obtain the regulatory feedback data corresponding to the monitoring and early warning results, and use a semi-supervised label propagation method to incrementally update the planning monitoring correlation diagram.

[0009] Optionally, the preprocessing specifically includes: Extract source markers, spatial markers, and time markers from multi-source data; unify the coordinates of spatial markers; and merge the monitoring cycles of time markers. Field mapping is performed on multi-source data according to a unified field table, and cross-source records under the same planning object are indexed and linked based on the field mapping results; After the index is attached, perform missing item completion, duplicate item merging, and abnormal item removal on the cross-source records, and retain the source tag corresponding to each cross-source record; The cross-source records are sorted and concatenated according to the monitoring cycle and the planning object index to form a planning status sequence with source and cycle markers.

[0010] Optionally, S2 specifically includes: S21. According to the preset window length and window step, the planned state sequence is divided into multiple state segments by sliding window, and cross-source state segments are divided according to the data source and monitoring cycle corresponding to the state segments. S22. Using the state segment corresponding to the planning and control status as the reference segment, establish a registration pair between the cross-source state segment and the reference segment, use dynamic time warping to perform path search on the registration pair, and adjust the arrangement position of the cross-source state segment within the monitoring period according to the path search results. S23. Perform window splicing on the cross-source state segments that have completed time registration, and extract the offset changes of the cross-source state segments relative to the planning and control state. Use mutual information metric to filter the degree of correlation between the offset changes and the planning and control state. S24. Within the continuous monitoring window, retain the state segments with stable correlation as candidate monitoring segments, and remove the state segments that only show offset changes within a single monitoring window, to obtain the target monitoring segments that have stable deviation correlation with the planning and control status.

[0011] Optionally, S22 specifically includes: S221. According to 3 to 12 monitoring cycles, expand the reference segment and cross-source state segment in the registration pair, and map each state item to 1 cycle marker, 1 source marker and 2 to 8 state fields to form a reference state queue and a state queue to be registered. S222. Scale the state fields in the baseline state queue and the state queue to be registered to the range of 0 to 1, and establish a local distance matrix according to the matching range of 1 to 3 state items to be registered for 1 baseline state item. S223. In the local distance matrix, the path cost is accumulated grid by grid according to the three path directions of horizontal progression, vertical progression and diagonal progression, and the number of consecutive horizontal progression or consecutive vertical progression is limited to 2 monitoring cycles to form a cumulative distance matrix. S224. Backtrack backward from the end position of the cumulative distance matrix to select a minimum cost registration path, and mark the state items in the registration queue as early state items, late state items, and overlapping state items according to the minimum cost registration path. S225. For advanced state items, perform a cycle shift of 1 to 2 monitoring cycles; for lagging state items, perform a cycle shift of 1 to 2 monitoring cycles; for cases where two or more consecutive state items to be registered match the same reference state item, perform same-cycle merging; and backfill the registered cross-source state segments into the corresponding sliding window.

[0012] Optionally, S3 specifically includes: S31. According to the spatial markers and monitoring windows corresponding to the target monitoring segments, target monitoring segments with the same spatial markers are grouped and merged, each group of target monitoring segments is mapped to a planned monitoring node, and the planned monitoring nodes are arranged in the order of the monitoring windows. S32. Perform delayed registration of planning monitoring nodes within adjacent monitoring windows, and establish candidate association relationships based on spatial location continuity and planning control matching offset relationships; S33. Using the planning and control status as a condition constraint, the conditional transfer entropy is used to measure the deviation transmission direction between the planning and monitoring nodes before and after the candidate association, and the connection direction of the candidate association is configured according to the deviation transmission direction. S34. Perform sparse screening on the candidate associations after configuring the connection direction, retain the candidate associations that satisfy the deviation transmission direction and planning control constraints as valid deviation relationships, and construct the planning monitoring association diagram based on the valid deviation relationships.

[0013] Optionally, S4 specifically includes: S41. Based on the connection direction of the effective deviation relationship, sort the transmission paths between the planned monitoring nodes, and configure the diffusion step size according to the difference between the forward conditional transfer entropy and the reverse conditional transfer entropy to construct the deviation diffusion channel, wherein: The deviation diffusion channel consists of effective deviation relationships arranged in the connection direction and their corresponding diffusion step size, which is used to limit the transmission order and single-round transmission distance of the planning monitoring node status in the planning monitoring association diagram. S42. Arrange the node status of the planned monitoring nodes according to the monitoring window order, establish a graph Laplace diffusion matrix with a restart term based on the deviation from the diffusion channel, and retain the node status of the planned monitoring nodes before diffusion through the restart term; S43. Use a graph Laplace diffusion matrix with a restart term to propagate the node status of the planned monitoring nodes along the deviation diffusion channel, and suppress the reverse propagation results that do not meet the effective deviation relationship connection direction. S44. Based on the cross-source time registration results, locate the node state offset after propagation, perform residual backflow correction between the node state offset and the node state before diffusion, and continue to participate in the next round of diffusion propagation after correction.

[0014] Optionally, S42 specifically includes: S421. Sort the planned monitoring nodes according to the order of 3 to 12 monitoring windows, decompose each deviation from the diffusion channel into the starting node, the ending node and the diffusion step size, and establish a directed diffusion adjacency matrix based on the arrangement of the starting node and the ending node. S422. In the directed diffusion adjacency matrix, establish a degree matrix according to 1 to 8 outward diffusion relationships for each planned monitoring node, and construct a graph Laplace matrix from the degree matrix and the directed diffusion adjacency matrix to recursively deduce the node state along the direction defined by the effective deviation relationship. S423. In the recursive process of the graph Laplace matrix, a restart coefficient with a value of 0.1 to 0.5 is introduced. The node state before the planning and monitoring node diffusion is taken as the state maintenance term. The node state after each round of diffusion is simultaneously constrained by the deviation from the diffusion channel and the state maintenance term. S424. According to the restart coefficient, perform 2 to 6 rounds of recursive fusion on the state-maintaining term and the graph Laplace diffusion term. Stop the recursion when the change ratio of the result of two consecutive rounds of recursive fusion is less than 5%, and form a graph Laplace diffusion matrix with restart term based on the recursive fusion result.

[0015] Optionally, S5 specifically includes: S51. Traverse the node status of residual correction along the connection direction of the effective deviation relationship, match the node status with the planning control conditions item by item, and convert the failed matching items into the trigger items of the planning conflict judgment operator. S52. Arrange the trigger items according to the monitoring window order, use a finite state machine to perform state transition matching on the planned monitoring nodes, and generate candidate transition chains based on the continuous hit status of the trigger items. S53. Viterbi path search is used to filter candidate transfer chains, retain the target transfer chain that is consistent with the continuous triggering process, and mark the planned monitoring nodes for conflict based on the target transfer chain. S54. Backtrack the planning monitoring nodes with conflict markers along the effective deviation relationship, merge the conflict markers caused by the same trigger item, and sort the planning monitoring nodes according to the merged conflict markers to form the monitoring and early warning results.

[0016] Optionally, S53 specifically includes: S531. Expand the candidate transfer chain according to the monitoring window order, and configure the trigger cost for the candidate transfer chain based on the results of continuous hit, interrupted hit and de-hit of the trigger item. S532. Establish the acceptance cost based on the state transition matching results within adjacent monitoring windows, and apply a jump penalty to the candidate transition chain with short-cycle reciprocating switching to form the path cost. S533. The Viterbi path search with bundled wide pruning is used to recursively calculate the path cost along the monitoring window, and the candidate transition chain with the minimum path cost is retained in each monitoring window. S534. Backtrack the candidate transfer chain with the minimum path cost from the last monitoring window, screen the target transfer chain, and mark the corresponding planning monitoring nodes for conflict based on the target transfer chain.

[0017] Optionally, S6 specifically includes: S61. Match the regulatory feedback data with the planned monitoring nodes in the monitoring and early warning results, configure confirmation tags for conflict markers that confirm feedback, and configure revocation tags for conflict markers that withdraw feedback, wherein: The regulatory feedback data includes: regulatory review results, spatial markers, monitoring window markers, and conflict resolution markers; S62. Using the planning monitoring node with configured confirmation and cancellation labels as the starting point for label propagation, semi-supervised label propagation is carried out along the effective deviation relationship in the planning monitoring association diagram, and the label propagation path is limited according to the connection direction of the effective deviation relationship. S63. Perform consistency screening on the confirmation and cancellation tags received by the same planning monitoring node, retain feedback tags that are consistent with the continuous changes in the status of nodes in adjacent monitoring windows, and suppress feedback tags that are contrary to the state transition matching results. S64. Based on the retained feedback labels, correct the connection weights of effective deviation relationships, the triggering conditions of deviation diffusion channels and planning conflict determination operators, and incrementally update the planning monitoring association diagram.

[0018] The beneficial effects of this invention are: First, this invention preprocesses multi-source data to form a planned state sequence, and uses a sliding window, dynamic time warping, and mutual information measurement to complete the registration and screening of cross-source state segments, enabling data from different sources and periods to participate in monitoring collaboratively, reducing misjudgments and omissions caused by data asynchrony.

[0019] Secondly, this invention maps target monitoring segments to planned monitoring nodes, uses conditional transition entropy to screen effective deviation relationships, and transmits node states through graph Laplace diffusion and residual correction, enabling the monitoring process to reflect cross-cycle deviation continuity relationships and improve the ability to identify continuous changes, lagging changes, and planning conflicts.

[0020] Finally, this invention uses finite state machines and Viterbi path search to generate monitoring and early warning results, and combines semi-supervised label propagation to introduce regulatory feedback data to incrementally update the planning monitoring association graph, thereby improving the stability and adaptability of subsequent monitoring. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is an overall flowchart of a dynamic monitoring method for land spatial planning based on multi-source data fusion proposed in this invention; Figure 2 This is a flowchart illustrating the construction of a Laplace diffusion matrix with restart term graph for a dynamic monitoring method of territorial spatial planning based on multi-source data fusion proposed in this invention. Figure 3 This is a flowchart illustrating the regulatory feedback and incremental update process of a dynamic monitoring method for territorial spatial planning based on multi-source data fusion proposed in this invention. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0023] refer to Figures 1-3 A dynamic monitoring method for territorial spatial planning based on multi-source data fusion includes the following steps: S1. Acquire multi-source data under the land and space planning scenario and preprocess it to form a planning status sequence; S2. The planned state sequence is divided into sliding window segments, and dynamic time warping is used to perform time registration of cross-source state segments. The target monitoring segments are screened in combination with mutual information measurement. S3. Map the target monitoring segments to the planned monitoring nodes, establish candidate association relationships based on the window connection relationship between the target monitoring segments, and use conditional transition entropy to screen out effective deviation relationships and construct a planned monitoring association diagram. S4. In the planning and monitoring correlation diagram, construct the deviation diffusion channel based on the effective deviation relationship, use the graph Laplace diffusion method to transmit the node status along the deviation diffusion channel, and perform residual correction on the transmitted node status based on the cross-source time registration results. S5. Based on the node status of residual correction, the effective deviation relationship is compiled into a planning conflict determination operator. A finite state machine is used to perform state transition matching on the planning monitoring nodes. The planning monitoring nodes that enter the conflict state are marked for conflict and ordered for early warning, thus forming the monitoring and early warning results. S6. Obtain the regulatory feedback data corresponding to the monitoring and early warning results, and use a semi-supervised label propagation method to incrementally update the planning monitoring correlation diagram.

[0024] In this embodiment, the preprocessing specifically includes: Extract source markers, spatial markers, and time markers from multi-source data; unify the coordinates of spatial markers; and merge the monitoring cycles of time markers. Field mapping is performed on multi-source data according to a unified field table, and cross-source records under the same planning object are indexed and linked based on the field mapping results; After the index is attached, perform missing item completion, duplicate item merging, and abnormal item removal on the cross-source records, and retain the source tag corresponding to each cross-source record; The cross-source records are sorted and concatenated according to the monitoring cycle and the planning object index to form a planning status sequence with source and cycle markers; In this embodiment, multi-source data are distinguished according to four types of source labels, and the monitoring period is one calendar month or one quarter; The unified field table includes six categories of fields: source marker, spatial marker, time marker, planning object marker, status field, and control field. Cross-source records with a missing rate exceeding 30% are removed, and fields that have values ​​for the same planning object within two consecutive monitoring periods are filled in. The planning status sequence is formed by assembling the planning object index, monitoring cycle, and source marker in a three-level order.

[0025] In this embodiment, S2 specifically includes: S21. According to the preset window length and window step, the planned state sequence is divided into multiple state segments by sliding window, and cross-source state segments are divided according to the data source and monitoring cycle corresponding to the state segments. S22. Using the state segment corresponding to the planning and control status as the reference segment, establish a registration pair between the cross-source state segment and the reference segment, use dynamic time warping to perform path search on the registration pair, and adjust the arrangement position of the cross-source state segment within the monitoring period according to the path search results. S23. Perform window splicing on the cross-source state segments that have completed time registration, and extract the offset changes of the cross-source state segments relative to the planning and control state. Use mutual information metric to filter the degree of correlation between the offset changes and the planning and control state. S24. Within the continuous monitoring window, retain the state segments with stable correlation as candidate monitoring segments, and remove the state segments that only show offset changes within a single monitoring window, to obtain the target monitoring segments that have stable deviation correlation with the planning and control status.

[0026] In this embodiment, S22 specifically includes: S221. According to 3 to 12 monitoring cycles, expand the reference segment and cross-source state segment in the registration pair, and map each state item to 1 cycle marker, 1 source marker and 2 to 8 state fields to form a reference state queue and a state queue to be registered. S222. Scale the state fields in the baseline state queue and the state queue to be registered to the range of 0 to 1, and establish a local distance matrix according to the matching range of 1 to 3 state items to be registered for 1 baseline state item. S223. In the local distance matrix, the path cost is accumulated grid by grid according to the three path directions of horizontal progression, vertical progression and diagonal progression, and the number of consecutive horizontal progression or consecutive vertical progression is limited to 2 monitoring cycles to form a cumulative distance matrix. S224. Backtrack backward from the end position of the cumulative distance matrix to select a minimum cost registration path, and mark the state items in the registration queue as early state items, late state items, and overlapping state items according to the minimum cost registration path. S225. For advanced state items, perform a cycle shift of 1 to 2 monitoring cycles; for lagging state items, perform a cycle shift of 1 to 2 monitoring cycles; for cases where two or more consecutive state items to be registered match the same reference state item, perform same-cycle merging; and backfill the registered cross-source state segments into the corresponding sliding window.

[0027] In this embodiment, S3 specifically includes: S31. According to the spatial markers and monitoring windows corresponding to the target monitoring segments, target monitoring segments with the same spatial markers are grouped and merged, each group of target monitoring segments is mapped to a planned monitoring node, and the planned monitoring nodes are arranged in the order of the monitoring windows. S32. Perform delayed registration of planning monitoring nodes within adjacent monitoring windows, and establish candidate association relationships based on spatial location continuity and planning control matching offset relationships; S33. Using the planning and control status as a conditional constraint, and employing conditional transition entropy, measure the deviation propagation direction between preceding and subsequent planning monitoring nodes in candidate associations, and configure the connection direction of candidate associations according to the deviation propagation direction, specifically including: According to the order of the monitoring windows, the planned monitoring nodes at both ends of the candidate association are formed into a forward state transition pair, and in the reverse order, they are formed into a reverse state transition pair. Based on the planning and control status, the forward and reverse state transition pairs are conditionally stratified. Within each condition layer, the frequency of state transitions between planning and monitoring nodes is statistically analyzed and normalized to form a forward conditional transition distribution and a reverse conditional transition distribution. Conditional transition entropy is measured for both the forward and reverse conditional transition distributions. The deviation direction of the candidate association is determined based on the difference between the forward and reverse conditional transition entropy. When the forward conditional transition entropy is less than the reverse conditional transition entropy, the candidate association relationship is configured to point from the previous planning monitoring node to the next planning monitoring node. When the reverse conditional transition entropy is less than the forward conditional transition entropy, the candidate association relationship is configured to point from the subsequent planning monitoring node to the previous planning monitoring node. S34. Perform sparse screening on the candidate associations after configuring the connection direction, retain the candidate associations that satisfy the deviation transmission direction and planning control constraints as valid deviation relationships, and construct the planning monitoring association diagram based on the valid deviation relationships.

[0028] In this embodiment, S4 specifically includes: S41. Based on the connection direction of the effective deviation relationship, sort the transmission paths between the planned monitoring nodes, and configure the diffusion step size according to the difference between the forward conditional transfer entropy and the reverse conditional transfer entropy to construct the deviation diffusion channel, wherein: The deviation diffusion channel consists of effective deviation relationships arranged in the connection direction and their corresponding diffusion step size, which is used to limit the transmission order and single-round transmission distance of the status of planning monitoring nodes in the planning monitoring association diagram. S42. Arrange the node status of the planned monitoring nodes according to the monitoring window order, establish a graph Laplace diffusion matrix with a restart term based on the deviation from the diffusion channel, and retain the node status of the planned monitoring nodes before diffusion through the restart term; S43. Use a graph Laplace diffusion matrix with a restart term to propagate the node status of the planned monitoring nodes along the deviation diffusion channel, and suppress the reverse propagation results that do not meet the effective deviation relationship connection direction. S44. Based on the cross-source time registration results, locate the node state offset after propagation, perform residual backflow correction between the node state offset and the node state before diffusion, and continue to participate in the next round of diffusion propagation after correction.

[0029] In this embodiment, S42 specifically includes: S421. Sort the planned monitoring nodes according to the order of 3 to 12 monitoring windows, decompose each deviation from the diffusion channel into the starting node, the ending node and the diffusion step size, and establish a directed diffusion adjacency matrix based on the arrangement of the starting node and the ending node. S422. In the directed diffusion adjacency matrix, establish a degree matrix according to 1 to 8 outward diffusion relationships for each planned monitoring node, and construct a graph Laplace matrix from the degree matrix and the directed diffusion adjacency matrix to recursively deduce the node state along the direction defined by the effective deviation relationship. S423. In the recursive process of the graph Laplace matrix, a restart coefficient with a value of 0.1 to 0.5 is introduced. The node state before the planning and monitoring node diffusion is taken as the state maintenance term. The node state after each round of diffusion is simultaneously constrained by the deviation from the diffusion channel and the state maintenance term. S424. According to the restart coefficient, perform 2 to 6 rounds of recursive fusion on the state-maintaining term and the graph Laplace diffusion term. Stop the recursion when the change ratio of the result of two consecutive rounds of recursive fusion is less than 5%, and form a graph Laplace diffusion matrix with restart term based on the recursive fusion result.

[0030] In this embodiment, S5 specifically includes: S51. Traverse the node status of residual correction along the connection direction of the effective deviation relationship, match the node status with the planning control conditions item by item, and convert the failed matching items into the trigger items of the planning conflict judgment operator. S52. Arrange trigger items according to the monitoring window order, use a finite state machine to perform state transition matching on the planned monitoring nodes, and generate candidate transition chains based on the continuous hits of trigger items, specifically including: According to the monitoring window order, the trigger items corresponding to the same planned monitoring node are connected in series, and a trigger sequence is formed based on the successive relationship of the trigger items in adjacent monitoring windows; Based on the planning and control conditions, the trigger sequence is matched according to rules, and the trigger items in the trigger sequence are divided into first hit, consecutive hit, interrupted hit and de-hit; A finite state machine is used to perform state transition matching window by window along the trigger sequence. For the first hit and consecutive hits, the state is advanced. For interrupted hits, the state is maintained and the chain is paused. For unhit, the state is rolled back. The planning and monitoring nodes that continuously execute state advancement are chained together, and the chain connection is cut off at the point where the state is maintained or rolled back, generating candidate transition chains. S53. Viterbi path search is used to filter candidate transfer chains, retain the target transfer chain that is consistent with the continuous triggering process, and mark the planned monitoring nodes for conflict based on the target transfer chain. S54. Backtrack the planning monitoring nodes with conflict markers along the effective deviation relationship, merge the conflict markers caused by the same trigger item, and sort the planning monitoring nodes according to the merged conflict markers to form the monitoring and early warning results.

[0031] In this embodiment, S53 specifically includes: S531. Expand the candidate transfer chain according to the monitoring window order, and configure the trigger cost for the candidate transfer chain based on the results of continuous hit, interrupted hit and de-hit of the trigger item. S532. Establish the acceptance cost based on the state transition matching results within adjacent monitoring windows, and apply a jump penalty to the candidate transition chain with short-cycle reciprocating switching to form the path cost. S533. The Viterbi path search with bundled wide pruning is used to recursively calculate the path cost along the monitoring window, and the candidate transition chain with the minimum path cost is retained in each monitoring window. S534. Backtrack the candidate transfer chain with the minimum path cost from the last monitoring window, screen the target transfer chain, and mark the corresponding planning monitoring nodes for conflict based on the target transfer chain.

[0032] In this embodiment, S6 specifically includes: S61. Match the regulatory feedback data with the planned monitoring nodes in the monitoring and early warning results, configure confirmation tags for conflict markers that confirm feedback, and configure revocation tags for conflict markers that withdraw feedback, wherein: Regulatory feedback data includes: regulatory review results, spatial markers, monitoring window markers, and conflict resolution markers; S62. Using the planning monitoring node with configured confirmation and cancellation labels as the starting point for label propagation, semi-supervised label propagation is carried out along the effective deviation relationship in the planning monitoring association diagram, and the label propagation path is limited according to the connection direction of the effective deviation relationship. S63. Perform consistency screening on the confirmation and cancellation tags received by the same planning monitoring node, retain feedback tags that are consistent with the continuous changes in the status of nodes in adjacent monitoring windows, and suppress feedback tags that are contrary to the state transition matching results. S64. Based on the retained feedback labels, correct the connection weights of effective deviation relationships, the triggering conditions of deviation diffusion channels and planning conflict determination operators, and incrementally update the planning monitoring association diagram.

[0033] Example 1: To verify the feasibility of this invention in practice, it was applied to a dynamic monitoring scenario for land spatial planning. This scenario includes urban expansion edges, concentrated agricultural land areas, and ecological control areas, with a total of 3260 monitoring objects. Remote sensing change records, planning control records, current land use records, and regulatory feedback records were continuously collected over 12 monitoring cycles. Before actual operation, there were significant asynchrony issues in the raw data. For some spatial objects, surface changes had already appeared in the remote sensing records, but the planning control records remained stuck in the previous cycle. Furthermore, some objects no longer required warnings after regulatory review, but the traditional static overlay method still repeatedly prompted warnings in subsequent cycles, resulting in significant review pressure.

[0034] When applying this invention, the aforementioned multi-source data is first organized into a planning state sequence. The multi-source data is divided into four types of source markers, and the unified field table uses six types of fields: source marker, spatial marker, time marker, planning object marker, status field, and control field. Cross-source records with a missing rate exceeding 30% are no longer processed; for fields with consecutive values ​​within two consecutive monitoring periods, periodic completion is performed.

[0035] After preprocessing, the number of records with inconsistent spatial markers was reduced from 438 to 46, the number of records with inconsistent temporal markers was reduced from 612 to 79, and the integrity rate of cross-source records under the same planning object was increased from 86.4% to 96.1%.

[0036] Subsequently, the planned state sequence was segmented using a sliding window, with a window length of 6 monitoring periods and a window step size of 1 monitoring period. After time registration of cross-source state segments through dynamic time warping, the cross-source state segments that originally differed by an average of 1.8 monitoring periods were shortened to within 0.6 monitoring periods. Then, target monitoring segments with stable deviations from the planned control state were screened out using mutual information measurement. A total of 18,420 state segments were obtained in this scenario, and 3,127 target monitoring segments were ultimately retained, eliminating a large number of temporary change records that only appeared in a single period and did not continue afterward.

[0037] In subsequent processing, target monitoring segments are mapped to planning monitoring nodes, and a planning monitoring correlation graph is established based on window acceptance relationships and planning control constraints. For nodes suspected of continuous deviation, graph Laplace diffusion is used to propagate node status along the effective deviation relationship, and residual correction is performed in conjunction with cross-source time registration results.

[0038] The diffusion process employed a four-round recursive approach, with a restart coefficient set to 0.3. After processing, the number of early warning records that were previously disconnected between adjacent cycles was significantly reduced, the number of cross-cycle breakage early warnings decreased from 126 to 38, and the stability rate of continuous deviation identification increased from 81.6% to 91.8%.

[0039] When generating monitoring and early warning results, the finite state machine first performs transition matching on the node states according to the planning and control conditions, and then uses a Viterbi path search with bundle width pruning to screen the target transition chain, with a bundle width of 5. This process filters out short-lived early warnings caused by false triggering in a single cycle, and prioritizes the retention of continuously triggered deviation nodes.

[0040] After regulatory feedback is received, the confirmation and revocation labels are propagated semi-supervised along the planning and monitoring correlation map. After three rounds of feedback updates, the false alarm rate dropped from 14.7% to 5.9%, the underreporting rate dropped from 9.8% to 4.1%, and the amount of manual review was reduced by about 32.5%.

[0041] Table 1 Comparison of the Application Effects of Multi-Source Fusion Dynamic Monitoring

[0042] As can be seen from the data in Table 1, the present invention can generate more stable dynamic monitoring and early warning results in scenarios where multi-source data are asynchronous, planning changes continue across cycles, and regulatory feedback is delayed.

[0043] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for dynamic monitoring of land spatial planning based on multi-source data fusion, characterized in that, Includes the following steps: S1. Acquire multi-source data under the land and space planning scenario and preprocess it to form a planning status sequence; S2. The planned state sequence is divided into sliding window segments, and dynamic time warping is used to perform time registration of cross-source state segments. The target monitoring segments are screened in combination with mutual information measurement. S3. Map the target monitoring segments to the planned monitoring nodes, establish candidate association relationships based on the window connection relationship between the target monitoring segments, and use conditional transition entropy to screen out effective deviation relationships and construct a planned monitoring association diagram. S4. In the planning and monitoring correlation diagram, construct the deviation diffusion channel based on the effective deviation relationship, use the graph Laplace diffusion method to transmit the node status along the deviation diffusion channel, and perform residual correction on the transmitted node status based on the cross-source time registration results. S5. Based on the node status of residual correction, the effective deviation relationship is compiled into a planning conflict determination operator. A finite state machine is used to perform state transition matching on the planning monitoring nodes. The planning monitoring nodes that enter the conflict state are marked for conflict and ordered for early warning, thus forming the monitoring and early warning results. S6. Obtain the regulatory feedback data corresponding to the monitoring and early warning results, and use a semi-supervised label propagation method to incrementally update the planning monitoring correlation diagram.

2. The method for dynamic monitoring of territorial spatial planning based on multi-source data fusion according to claim 1, characterized in that, The preprocessing specifically includes: Extract source markers, spatial markers, and time markers from multi-source data; unify the coordinates of spatial markers; and merge the monitoring cycles of time markers. Field mapping is performed on multi-source data according to a unified field table, and cross-source records under the same planning object are indexed and linked based on the field mapping results; After the index is attached, perform missing item completion, duplicate item merging, and abnormal item removal on the cross-source records, and retain the source tag corresponding to each cross-source record; The cross-source records are sorted and concatenated according to the monitoring cycle and the planning object index to form a planning status sequence with source and cycle markers.

3. The method for dynamic monitoring of land spatial planning based on multi-source data fusion according to claim 1, characterized in that, S2 specifically includes: S21. According to the preset window length and window step, the planned state sequence is divided into multiple state segments by sliding window, and cross-source state segments are divided according to the data source and monitoring cycle corresponding to the state segments. S22. Using the state segment corresponding to the planning and control status as the reference segment, establish a registration pair between the cross-source state segment and the reference segment, use dynamic time warping to perform path search on the registration pair, and adjust the arrangement position of the cross-source state segment within the monitoring period according to the path search results. S23. Perform window splicing on the cross-source state segments that have completed time registration, and extract the offset changes of the cross-source state segments relative to the planning and control state. Use mutual information metric to filter the degree of correlation between the offset changes and the planning and control state. S24. Within the continuous monitoring window, retain the state segments with stable correlation as candidate monitoring segments, and remove the state segments that only show offset changes within a single monitoring window, to obtain the target monitoring segments that have stable deviation correlation with the planning and control status.

4. The method for dynamic monitoring of territorial spatial planning based on multi-source data fusion according to claim 3, characterized in that, S22 specifically includes: S221. According to 3 to 12 monitoring cycles, expand the reference segment and cross-source state segment in the registration pair, and map each state item to 1 cycle marker, 1 source marker and 2 to 8 state fields to form a reference state queue and a state queue to be registered. S222. Scale the state fields in the baseline state queue and the state queue to be registered to the range of 0 to 1, and establish a local distance matrix according to the matching range of 1 to 3 state items to be registered for 1 baseline state item. S223. In the local distance matrix, the path cost is accumulated grid by grid according to the three path directions of horizontal progression, vertical progression and diagonal progression, and the number of consecutive horizontal progression or consecutive vertical progression is limited to 2 monitoring cycles to form a cumulative distance matrix. S224. Backtrack backward from the end position of the cumulative distance matrix to select a minimum cost registration path, and mark the state items in the registration queue as early state items, late state items, and overlapping state items according to the minimum cost registration path. S225. For advanced state items, perform a cycle shift of 1 to 2 monitoring cycles; for lagging state items, perform a cycle shift of 1 to 2 monitoring cycles; for cases where two or more consecutive state items to be registered match the same reference state item, perform same-cycle merging; and backfill the registered cross-source state segments into the corresponding sliding window.

5. The method for dynamic monitoring of territorial spatial planning based on multi-source data fusion according to claim 1, characterized in that, S3 specifically includes: S31. According to the spatial markers and monitoring windows corresponding to the target monitoring segments, target monitoring segments with the same spatial markers are grouped and merged, each group of target monitoring segments is mapped to a planned monitoring node, and the planned monitoring nodes are arranged in the order of the monitoring windows. S32. Perform delayed registration of planning monitoring nodes within adjacent monitoring windows, and establish candidate association relationships based on spatial location continuity and planning control matching offset relationships; S33. Using the planning and control status as a condition constraint, the conditional transfer entropy is used to measure the deviation transmission direction between the planning and monitoring nodes before and after the candidate association, and the connection direction of the candidate association is configured according to the deviation transmission direction. S34. Perform sparse screening on the candidate associations after configuring the connection direction, retain the candidate associations that satisfy the deviation transmission direction and planning control constraints as valid deviation relationships, and construct the planning monitoring association diagram based on the valid deviation relationships.

6. The method for dynamic monitoring of territorial spatial planning based on multi-source data fusion according to claim 1, characterized in that, S4 specifically includes: S41. Based on the connection direction of the effective deviation relationship, sort the transmission paths between the planned monitoring nodes, and configure the diffusion step size according to the difference between the forward conditional transfer entropy and the reverse conditional transfer entropy to construct the deviation diffusion channel, wherein: The deviation diffusion channel consists of effective deviation relationships arranged in the connection direction and their corresponding diffusion step size, which is used to limit the transmission order and single-round transmission distance of the planning monitoring node status in the planning monitoring association diagram. S42. Arrange the node status of the planned monitoring nodes according to the monitoring window order, establish a graph Laplace diffusion matrix with a restart term based on the deviation from the diffusion channel, and retain the node status of the planned monitoring nodes before diffusion through the restart term; S43. Use a graph Laplace diffusion matrix with a restart term to propagate the node status of the planned monitoring nodes along the deviation diffusion channel, and suppress the reverse propagation results that do not meet the effective deviation relationship connection direction. S44. Based on the cross-source time registration results, locate the node state offset after propagation, perform residual backflow correction between the node state offset and the node state before diffusion, and continue to participate in the next round of diffusion propagation after correction.

7. The method for dynamic monitoring of territorial spatial planning based on multi-source data fusion according to claim 6, characterized in that, S42 specifically includes: S421. Sort the planned monitoring nodes according to the order of 3 to 12 monitoring windows, decompose each deviation from the diffusion channel into the starting node, the ending node and the diffusion step size, and establish a directed diffusion adjacency matrix based on the arrangement of the starting node and the ending node. S422. In the directed diffusion adjacency matrix, establish a degree matrix according to 1 to 8 outward diffusion relationships for each planned monitoring node, and construct a graph Laplace matrix from the degree matrix and the directed diffusion adjacency matrix to recursively deduce the node state along the direction defined by the effective deviation relationship. S423. In the recursive process of the graph Laplace matrix, a restart coefficient with a value of 0.1 to 0.5 is introduced. The node state before the planning and monitoring node diffusion is taken as the state maintenance term. The node state after each round of diffusion is simultaneously constrained by the deviation from the diffusion channel and the state maintenance term. S424. According to the restart coefficient, perform 2 to 6 rounds of recursive fusion on the state-maintaining term and the graph Laplace diffusion term. Stop the recursion when the change ratio of the result of two consecutive rounds of recursive fusion is less than 5%, and form a graph Laplace diffusion matrix with restart term based on the recursive fusion result.

8. The method for dynamic monitoring of territorial spatial planning based on multi-source data fusion according to claim 1, characterized in that, S5 specifically includes: S51. Traverse the node status of residual correction along the connection direction of the effective deviation relationship, match the node status with the planning control conditions item by item, and convert the failed matching items into the trigger items of the planning conflict judgment operator. S52. Arrange the trigger items according to the monitoring window order, use a finite state machine to perform state transition matching on the planned monitoring nodes, and generate candidate transition chains based on the continuous hit status of the trigger items. S53. Viterbi path search is used to filter candidate transfer chains, retain the target transfer chain that is consistent with the continuous triggering process, and mark the planned monitoring nodes for conflict based on the target transfer chain. S54. Backtrack the planning monitoring nodes with conflict markers along the effective deviation relationship, merge the conflict markers caused by the same trigger item, and sort the planning monitoring nodes according to the merged conflict markers to form the monitoring and early warning results.

9. A method for dynamic monitoring of territorial spatial planning based on multi-source data fusion as described in claim 8, characterized in that, S53 specifically includes: S531. Expand the candidate transfer chain according to the monitoring window order, and configure the trigger cost for the candidate transfer chain based on the results of continuous hit, interrupted hit and de-hit of the trigger item. S532. Establish the acceptance cost based on the state transition matching results within adjacent monitoring windows, and apply a jump penalty to the candidate transition chain with short-cycle reciprocating switching to form the path cost. S533. The Viterbi path search with bundled wide pruning is used to recursively calculate the path cost along the monitoring window, and the candidate transition chain with the minimum path cost is retained in each monitoring window. S534. Backtrack the candidate transfer chain with the minimum path cost from the last monitoring window, screen the target transfer chain, and mark the corresponding planning monitoring nodes for conflict based on the target transfer chain.

10. A method for dynamic monitoring of territorial spatial planning based on multi-source data fusion according to claim 1, characterized in that, S6 specifically includes: S61. Match the regulatory feedback data with the planned monitoring nodes in the monitoring and early warning results, configure confirmation tags for conflict markers that confirm feedback, and configure revocation tags for conflict markers that withdraw feedback, wherein: The regulatory feedback data includes: regulatory review results, spatial markers, monitoring window markers, and conflict resolution markers; S62. Using the planning monitoring node with configured confirmation and cancellation labels as the starting point for label propagation, semi-supervised label propagation is carried out along the effective deviation relationship in the planning monitoring association diagram, and the label propagation path is limited according to the connection direction of the effective deviation relationship. S63. Perform consistency screening on the confirmation and cancellation tags received by the same planning monitoring node, retain feedback tags that are consistent with the continuous changes in the status of nodes in adjacent monitoring windows, and suppress feedback tags that are contrary to the state transition matching results. S64. Based on the retained feedback labels, correct the connection weights of effective deviation relationships, the triggering conditions of deviation diffusion channels and planning conflict determination operators, and incrementally update the planning monitoring association diagram.

Images