Land space planning data monitoring and evaluation method and system

Abstract

The invention relates to the technical field of territorial space planning, and discloses a monitoring and evaluation method based on multi-source remote sensing images and geographic information vector data. Comprising the following steps: acquiring a multi-source remote sensing image and geographic information vector data of a target area; performing preprocessing and fusion analysis on the multi-source remote sensing image, generating a dynamic earth surface change detection model with high time resolution, and constructing an initial monitoring baseline in combination with geographic information vector data; according to the method, a high-frequency change detection condition and a semantic segmentation condition are loaded on the basis of an initial monitoring baseline to form a comprehensive interpretation model, a real-time early warning evaluation index is output, and a dynamic earth surface change detection model is generated by preprocessing a remote sensing image. And generating an illegal behavior distribution result by using the key identification parameters, establishing a task distribution model to optimize a supervision scheduling scheme, and finally realizing accurate management and control. According to the invention, the efficiency and accuracy of territorial space planning monitoring are improved, and technical support is provided for intelligent supervision.

Description

Technical Field

[0001] This invention belongs to the field of land spatial planning and information technology, specifically a method and system for monitoring and evaluating land spatial planning data. Background Technology

[0002] As the requirements for controlling the "three zones and three lines" bottom line in territorial spatial planning become increasingly stringent, traditional monitoring methods are gradually showing their limitations in detecting and handling illegal land use. Currently, the mainstream monitoring methods mainly rely on low-frequency manual verification or post-event remote sensing comparison. This model is insufficient in terms of timeliness and coverage, often leading to a delay in problem detection and failing to meet the regulatory needs of discovering problems at their initial stage and resolving them in their nascent stages.

[0003] In terms of data acquisition, traditional methods usually employ periodic manual inspections or comparisons of single remote sensing images, failing to fully utilize high temporal resolution multi-source remote sensing data.

[0004] For example, in areas prone to fog and cloud cover, the availability of optical remote sensing data is limited, while the penetrating power of radar data has not been fully explored, affecting the continuity and accuracy of monitoring results. Furthermore, the frequency and scope of manual verification are limited, making it easy to miss highly concealed or small-scale violations.

[0005] The application of change detection technology also has certain limitations. Existing methods are mostly based on simple image difference analysis and lack the ability to deeply identify change attributes.

[0006] For example, traditional algorithms struggle to accurately classify different types of surface changes, such as new buildings, paved surfaces, or vegetation destruction, impacting the scientific rigor and targeted nature of subsequent regulatory decisions. Furthermore, the automation level of the overlay analysis of change detection results with the three-zone, three-line vector boundaries is low, still requiring significant manual intervention and reducing overall efficiency.

[0007] The construction of early warning mechanisms also faces challenges. Current monitoring systems are mostly based on post-event feedback, lacking real-time and proactive capabilities. For example, by the time violations are discovered, irreversible damage to the ecological environment or arable land resources has often already occurred.

[0008] In addition, the generation and dissemination of early warning information are relatively decentralized, failing to form a closed-loop management system, making it difficult for local management personnel to respond and handle the situation in a timely manner.

[0009] Coordination and optimization between different stages also need to be strengthened. From data acquisition to intelligent interpretation, and then to the generation and distribution of early warning work orders, the modules in the existing technology system are relatively independent and lack an efficient integration mechanism. For example, the collaboration between the remote sensing data processing team and the geographic information system analysis team is not close enough, which may lead to the inefficient transmission of key information and affect the overall monitoring effect.

[0010] With the increasing demand for refined management and intelligent supervision in territorial spatial planning, it is urgent to develop a monitoring and evaluation method that can integrate high-frequency remote sensing data, intelligent interpretation models, and real-time early warning mechanisms.

[0011] This method requires the synergistic application of multimodal data fusion and semantic segmentation technologies to improve the timeliness, accuracy, and intelligence of monitoring, help safeguard the bottom line of the three zones and three lines, and promote the development of territorial spatial governance in a more efficient and precise direction. Summary of the Invention

[0012] The purpose of this invention is to provide a method and system for monitoring and evaluating land spatial planning data, so as to solve the problems mentioned in the background art.

[0013] To achieve the above objectives, the present invention provides the following technical solution: a method for monitoring and evaluating land spatial planning data, the method comprising: Acquire multi-source remote sensing images and geographic information vector data of the target area; Multi-source remote sensing images are preprocessed and fused to generate a high temporal resolution dynamic land surface change detection model, and an initial monitoring baseline is constructed by combining geographic information vector data. Based on the initial monitoring baseline, high-frequency change detection conditions and semantic segmentation conditions are added to form a comprehensive interpretation model. Combined with the three-zone, three-line vector boundary data, real-time early warning and evaluation indicators are output. A multi-level intelligent optimization framework is constructed and iteratively trained using real-time early warning and evaluation indicators to obtain a violation patch identification model, and then key identification parameters are extracted. The distribution results of violations are generated by using key identification parameters, comprehensive interpretation models, and geographic information vector data; Based on the distribution of violations and regulatory response parameters, a task allocation model is established and the regulatory response parameters are optimized to obtain the optimal regulatory scheduling scheme. Based on the performance simulation framework, the current stage of regulatory and scheduling plan is simulated to obtain the current stage performance simulation indicators. Combining the current stage performance simulation indicators, the current stage optimal regulatory and scheduling plan, and the current stage actual response parameters, the regulatory response parameters are dynamically adjusted to achieve precise control of territorial spatial planning.

[0014] Preferably, the real-time early warning assessment indicators include at least the proportion of illegal areas, the distribution of change types, and the set of conflict risk levels; the key identification parameters include at least the location of the violation center point, the category of change attributes, and the boundary expansion direction parameters; and the violation behavior distribution results include at least the range of newly added patches, violation frequency statistics, and potential diffusion trends. Preferably, the preprocessing and fusion analysis of multi-source remote sensing images to generate a high temporal resolution dynamic land surface change detection model, and the construction of an initial monitoring baseline in conjunction with geographic information vector data, includes the following steps: Collect high temporal resolution satellite image sequences covering the target area, and simultaneously acquire radar data to compensate for the loss of optical images in cloudy or foggy weather; Radiometric correction, geometric correction and image registration are performed on satellite image sequences and radar data to obtain a standardized image set. The processing includes one or more of the following: denoising, projection transformation, pixel alignment and consistency verification. Based on a standardized image set, a change detection network in deep learning is used to compare the latest images with the baseline images at the pixel level to generate a dynamic land change detection model. The dynamic land surface change detection model is divided into grids and its attributes mapped. Combined with geographic information vector data, an initial monitoring baseline is constructed. The grid division includes at least regular and irregular grid subdivisions. Basic monitoring units are determined through regular grid division, and each unit is assigned corresponding geographic information attributes. Preferably, the step of adding high-frequency change detection conditions and semantic segmentation conditions to the initial monitoring baseline to form a comprehensive interpretation model, and combining the three-zone, three-line vector boundary data to output real-time early warning evaluation indicators, includes the following steps: By applying high-frequency change detection conditions to the initial monitoring baseline and constraining the time dimension, the first interpretation model is obtained; The semantic segmentation conditions are superimposed on the first interpretation model to form a comprehensive interpretation model. The semantic segmentation conditions include new building classification conditions, hardened ground classification conditions, and vegetation destruction classification conditions. Based on the comprehensive interpretation model and the three-zone, three-line vector boundary data, real-time early warning assessment indicators are calculated. The specific process includes: A multidimensional feature coupling equation is constructed, which includes at least an equation for calculating the area of ​​violations, an equation for identifying the type of change, and an equation for assessing the risk of conflict. The equation is then solved using a numerical discretization method in conjunction with a comprehensive interpretation model to obtain the proportion of the area of ​​violations, the distribution of the type of change, and the set of conflict risk levels.

[0015] Preferably, obtaining the conflict risk level set includes the following steps: Based on the distribution of change types, calculate the frequency difference of change types for each monitoring unit; By identifying monitoring units in the integrated interpretation model whose change type frequency difference is not lower than the frequency threshold, a set of conflict risk levels is obtained.

[0016] Preferably, the construction of a multi-level intelligent optimization framework and iterative training using real-time early warning evaluation indicators to obtain a violation patch identification model, and then extraction of key identification parameters, includes the following steps: A multi-level intelligent optimization framework based on convolutional neural networks. The optimization framework was trained and validated multiple times using real-time early warning evaluation indicators to obtain a violation patch recognition model. The real-time early warning assessment indicators to be analyzed are input into the violation patch identification model, which outputs a set of predicted conflict risk levels. Based on the predicted set of conflict risk levels, key identification parameters are extracted, wherein the key identification parameters include at least the location of the violation center point, the category of change attribute, and the boundary expansion direction parameter.

[0017] Preferably, the extraction of key identification parameters based on the predicted conflict risk level set includes the following steps: Determine the geometric centroid of the predicted set of conflict risk levels as the location of the violation center point; The change attribute categories are fitted based on the spatial distribution characteristics of the predicted conflict risk level set to obtain the change attribute category parameters, which include the category confidence and the category distribution density. The boundary expansion direction of the predicted conflict risk level set is analyzed, the weighted average of the expansion direction is calculated, and it is converted into a standard boundary expansion path, which is the boundary expansion direction parameter.

[0018] Preferably, the step of generating the distribution results of violations through key identification parameters, a comprehensive interpretation model, and geographic information vector data includes the following steps: Using the location of the violation center point as a reference point, the changing attribute category parameters are mapped to the integrated interpretation model. The boundary expansion direction parameters are adjusted, and the mesh of the associated monitoring units is refined to complete the update of the integrated interpretation model. The violation evolution rules, violation evolution direction, and violation evolution step size are defined to form a violation evolution analysis model. Based on the violation evolution analysis model, a dynamic iterative method is used to solve the multidimensional feature coupling equation. Based on the results of each iteration stage in the solution process, the distribution results of violation behaviors are generated, specifically including: When the violation evolution rules are met, boundary optimization is performed to obtain the current range of newly added patches, violation evolution step size, violation evolution direction, violation frequency statistics and potential diffusion trend, and the violation evolution analysis model is updated. The multidimensional feature coupling equation is re-solved based on the updated violation evolution analysis model until the termination condition is met, thus ending the violation evolution process. Based on the violation frequency statistics and potential diffusion trends at each iteration stage, calculate the statistical distribution of violation behavior and the set of diffusion trends; The violation evolution rule is defined as triggering boundary optimization if the current cumulative violation area is not lower than the area threshold. The direction of violation evolution is determined based on historical violation trends and the current distribution direction of violations; The range of the newly added patches and the step size of the illegal evolution are determined based on the temporal resolution and the distribution of change types, respectively.

[0019] Preferably, the step of establishing a task allocation model based on the distribution results of violations and regulatory response parameters, and optimizing the regulatory response parameters to obtain the optimal regulatory scheduling scheme, includes the following steps: A task allocation model is established, which includes at least a variable initialization module, a target decision module, and a constraint module. The variable initialization module is defined based on regulatory response parameters, which include at least personnel configuration scale, response time limit, resource allocation ratio, and communication delay time. The target decision module is constructed based on the second performance evaluation value and resource consumption value. The second performance evaluation value is calculated based on the distribution results of violations, and the resource consumption value includes at least human resource cost, equipment maintenance cost, and communication transmission cost. The constraint module includes a regulatory response constraint module and a performance guarantee constraint module. The performance guarantee constraint module is defined based on task integrity, response timeliness, and system coordination. A genetic algorithm is used to initially solve the task allocation model and obtain an initial optimized scheduling scheme. The task allocation model is globally optimized using the particle swarm optimization algorithm to obtain the optimal solution set of the particle swarm as the optimal supervisory scheduling scheme.

[0020] Preferably, the present invention further includes a computer device, the computer device including a memory, a processor and a communication interface, which are connected to each other via a bus; the memory is used to store a set of computer program instructions and data, and to transmit the stored data to the processor, the processor executing the program instructions stored in the memory to perform the land spatial planning data monitoring and evaluation method as described in any of the above claims.

[0021] A land spatial planning data monitoring and evaluation system is provided to implement the aforementioned land spatial planning data monitoring and evaluation method.

[0022] Compared with the prior art, the beneficial effects of the present invention are: 1. By integrating high temporal resolution optical satellite imagery and radar data, the missing information in optical imagery during cloudy and foggy weather was compensated for, enabling continuous monitoring of surface changes under complex meteorological conditions and ensuring data integrity and temporal consistency. Through preprocessing steps such as radiometric correction, geometric correction, and denoising, a standardized image set was formed, laying a high-precision data foundation for subsequent change detection and model building, and reducing the impact of data errors on monitoring results.

[0023] 2. A high-temporal-resolution dynamic land surface change detection model is constructed based on pixel-level comparison using a deep learning-based change detection network. This model can capture small-scale, short-cycle land surface changes (such as new buildings and hardened surfaces), overcoming the limitations of traditional methods in identifying concealed or small-scale violations. By loading semantic segmentation conditions such as new buildings and vegetation destruction, and combining them with vector boundary data from the three zones and three lines, the change types are accurately classified. Simultaneously, the risk level is quantified through a conflict risk level set, making the monitoring results more aligned with the needs of land space management.

[0024] 3. Output real-time early warning indicators such as the proportion of non-compliant areas, the distribution of change types, and the level of conflict risk, transforming traditional post-event feedback into pre-event early warning. This helps to identify and resolve regulatory targets at the initial stage, reducing the irreversible impact of violations. By simulating the potential spread trend of violations through a violation evolution analysis model and combining key identification parameters, it can predict the risk spread path in advance, providing decision support for targeted prevention and control. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the land spatial planning data monitoring and evaluation method provided in an embodiment of the present invention.

[0026] Figure 2 This is a schematic diagram illustrating the process of constructing a dynamic land surface change detection model in an embodiment of the present invention.

[0027] Figure 3 This is a logical structure diagram of the integrated interpretation model and real-time early warning evaluation index generated in the embodiments of the present invention.

[0028] Figure 4 This is a schematic diagram of the framework for optimizing the task allocation model in an embodiment of the present invention. Detailed Implementation

[0029] This invention provides a method and system for monitoring and evaluating land spatial planning data. Its core lies in achieving precise control over land spatial planning through comprehensive analysis of multi-source remote sensing imagery and geographic information vector data, combined with an intelligent optimization framework and task allocation model. The following is in conjunction with the appendix... Figure 1 To be continued Figure 4 The specific embodiments of the present invention will be described in detail below.

[0030] The first step in implementation is to acquire multi-source remote sensing imagery and geographic information vector data of the target area. This data includes high temporal resolution satellite image sequences and radar data. The satellite image sequences are used to capture the dynamic characteristics of surface changes, while the radar data is used to compensate for the loss of optical imagery in cloudy or foggy weather.

[0031] As a preferred approach, the acquired data undergoes radiometric correction, geometric correction, and image registration to form a standardized image set. This process involves denoising, projection transformation, pixel alignment, and consistency verification. Figure 2 As shown, the construction process of the dynamic land change detection model starts with a standardized image set. The latest image is compared with the reference image at the pixel level through a change detection network in deep learning to generate the dynamic land change detection model.

[0032] Subsequently, the dynamic land surface change detection model was divided into grids and its attributes mapped, and an initial monitoring baseline was constructed by combining geographic information vector data. Grid division included regular grid subdivision and irregular grid subdivision. Regular grids were used to determine the basic monitoring units, and each unit was assigned corresponding geographic information attributes.

[0033] Based on the initial monitoring baseline, high-frequency change detection conditions and semantic segmentation conditions are added to form a comprehensive interpretation model. The high-frequency change detection conditions are used to constrain the time dimension, ensuring that the model can capture significant changes that occur within a short period of time; Semantic segmentation conditions include new building classification conditions, hardened ground classification conditions, and vegetation destruction classification conditions, used to refine the accuracy of identifying change types. For example... Figure 3 As shown, the integrated interpretation model combines the three-zone, three-line vector boundary data and calculates real-time early warning assessment indicators through multi-dimensional feature coupling equations.

[0034] The multidimensional feature coupling equation includes at least the equation for calculating the area of ​​violations, the equation for identifying the type of change, and the equation for assessing the risk of conflict. These equations are solved by numerical discretization methods to obtain the proportion of the area of ​​violations, the distribution of the type of change, and the set of conflict risk levels.

[0035] The process of generating the conflict risk level set includes calculating the frequency difference of change types for each monitoring unit based on the distribution of change types, and identifying monitoring units whose frequency difference is not lower than the frequency threshold.

[0036] To further improve the accuracy of violation identification, this invention constructs a multi-level intelligent optimization framework, designed and implemented based on a convolutional neural network. The optimization framework is trained and validated multiple times using real-time early warning evaluation metrics to obtain a violation patch identification model.

[0037] As a preferred approach, the real-time early warning assessment indicators to be analyzed are input into the violation patch identification model, which then outputs a set of predicted conflict risk levels. Based on this set of predicted conflict risk levels, key identification parameters are extracted, including the location of the violation center point, the category of changing attributes, and the direction of boundary expansion.

[0038] The extraction process for key identification parameters is as follows: First, the geometric centroid of the predicted conflict risk level set is determined as the location of the violation center point; second, the change attribute categories are fitted according to the spatial distribution characteristics of the conflict risk level set to obtain the change attribute category parameters, which include category confidence and category distribution density. Finally, the boundary expansion direction of the conflict risk level set is analyzed, the weighted average of the expansion direction is calculated, and it is converted into a standard boundary expansion path.

[0039] The extracted key identification parameters are further combined with the integrated interpretation model and geographic information vector data to generate the distribution results of violations. Using the location of the violation center point as a reference point, the changing attribute category parameters are mapped into the integrated interpretation model. The boundary expansion direction parameters are adjusted, and the associated monitoring units are refined into a grid, thus completing the update of the integrated interpretation model.

[0040] As a preferred approach, violation evolution rules, violation evolution directions, and violation evolution step sizes are defined based on this to form a violation evolution analysis model. The violation evolution rules stipulate that boundary optimization is triggered if the current cumulative violation area is not lower than the area threshold; the violation evolution direction is determined based on historical violation trends and the current violation distribution direction; the range of newly added patches and the violation evolution step size are determined based on the time resolution and the distribution of change types, respectively.

[0041] A dynamic iterative method is used to solve the multidimensional feature coupling equation, and the distribution results of violations are generated based on the results of each iteration stage in the solution process. When the violation evolution rules are met, boundary optimization is performed to obtain the current range of newly added patches, violation evolution step size, violation evolution direction, violation frequency statistics and potential diffusion trend, and the violation evolution analysis model is updated.

[0042] The multidimensional feature coupling equation is resolved based on the updated violation evolution analysis model until the termination condition is met, ending the violation evolution process. Finally, based on the violation frequency statistics and potential diffusion trends at each iteration stage, the statistical distribution of violation behavior and the set of diffusion trends are calculated.

[0043] As a preferred approach, a task allocation model is established based on the distribution of violations and regulatory response parameters. This model includes at least a variable initialization module, a target decision module, and a constraint module. The variable initialization module is based on the regulatory response parameters, which include personnel allocation scale, response time limits, resource allocation ratios, and communication latency. The target decision module is constructed based on the second performance evaluation value and the resource consumption value. The second performance evaluation value is calculated based on the distribution results of violations, and the resource consumption value includes the human cost value, equipment maintenance cost value, and communication transmission cost value. The constraint module includes the regulatory response constraint module and the performance guarantee constraint module, wherein the performance guarantee constraint module is defined based on task integrity, response timeliness, and system coordination.

[0044] like Figure 4 As shown, the genetic algorithm module is used to initially solve the task allocation model and obtain an initial optimized scheduling scheme. Subsequently, the particle swarm optimization module is used to globally optimize the task allocation model and obtain the optimal solution set of the particle swarm optimization as the optimal supervisory scheduling scheme. Finally, the current stage of regulatory scheduling is simulated based on the performance simulation framework to obtain the current stage performance simulation indicators. Combining the current stage performance simulation indicators, the optimal current stage regulatory scheduling plan, and the actual current stage response parameters, the regulatory response parameters are dynamically adjusted to achieve precise control of territorial spatial planning.

[0045] The whole process is as follows Figure 1 As shown, a complete closed-loop system has been formed from data acquisition to supervision, scheduling and optimization, ensuring the efficient management of territorial spatial planning.

[0046] This invention also provides a computer device including a memory, a processor, and a communication interface, which are connected via a bus. The memory stores a set of computer program instructions and data, and transmits the stored data to the processor. The processor executes the program instructions stored in the memory to implement the aforementioned method for monitoring and evaluating land spatial planning data. The communication interface is used for data interaction with other devices or systems, supporting the transmission and sharing of multi-source remote sensing imagery and geographic information vector data.

[0047] IV. Application Extension Scenarios In addition to being applicable to the baseline control of national land space, the method described in this invention can also be extended to multiple dimensions of urban development monitoring and evaluation, including but not limited to the following aspects: 1. Monitor and assess the speed of urban development Based on the dynamic land surface change detection model and initial monitoring baseline constructed by this invention, annual construction land index data can be further integrated to establish a "major engineering project - spatial coordinate" mapping matrix. By analyzing the spatiotemporal correlation between construction land changes and the layout of major projects, it is possible to assess whether construction land indicators are mainly used for the construction of major projects; combined with the characteristics of urban development stages (such as the transportation-driven period, the rapid industrial development period, and the functional quality improvement period), it is possible to determine the source of urban development momentum and the evolution trend of development stages.

[0048] 2. Monitor and evaluate the efficiency of urban structure By comprehensively utilizing annual change survey data, remote sensing imagery, land use approval data, regional economic data, and commuter traffic data, and relying on the comprehensive interpretation model and violation evolution analysis capabilities of this invention, key urban development areas and their functional integration are identified. By assessing the impact of annual construction on the integrity of the urban spatial pattern (such as whether there are phenomena like "playing checkers" or "spreading out like a pancake") and the realization of the dominant functions of each area, data support is provided for optimizing the urban spatial structure.

[0049] 3. Monitor and evaluate the quality of urban development Based on the high-precision change detection and semantic segmentation capabilities of this invention, combined with multi-source big data (such as POI, population heatmap, traffic flow, etc.), it is possible to achieve refined perception of the functions of the 15-minute living circle at the street and community levels. By analyzing the radiation range and coverage blind spots of public service facilities such as education, medical care, sports, elderly care, and commerce, the scientific nature of the public facility layout can be evaluated, and early warnings can be issued for areas with insufficient facility configuration, thus contributing to the improvement of urban quality and refined governance.

[0050] Example 1: Monitoring and Evaluation of Territorial Spatial Planning in a Farmland Protection Area in the Jianghuai Plain This example focuses on a farmland protection area in the Jianghuai Plain, with a target area of ​​approximately 500 km². 2 The focus is on monitoring illegal occupation of arable land within the red line, and the specific implementation process is as follows: Step 1: Data Acquisition Collect multi-source remote sensing data covering the target area: High temporal resolution satellite image sequence: using data from the Gaofen-7 satellite, with a temporal resolution of 8 days and a spatial resolution of 1 meter, covering January to June 2024; Radar data: Sentinel-1 radar data was used to compensate for the problem of cloud and fog obstruction in optical imagery during the rainy season; Geographic information vector data: basic vector data such as the arable land red line vector boundary, administrative village boundaries, and road network of this region.

[0051] Step 2: Preprocessing and Construction of Dynamic Surface Change Detection Model Preprocessing: Radiometric correction, geometric correction, and image registration are performed on satellite imagery and radar data, and noise reduction is carried out simultaneously. Gaussian filtering is used on satellite imagery and Lee filtering is used on radar data to form a standardized image set. Dynamic land surface change detection model: Based on the U-Net change detection network, the latest image in June 2024 is compared with the baseline image in January 2024 at the pixel level to identify areas of land surface change; Initial monitoring baseline: The detection model is divided into 50m×50m regular grids, and each grid is assigned geographical attributes such as cultivated land type and ownership unit. The baseline is constructed by combining the cultivated land red line vector data.

[0052] Step 3: Output of integrated interpretation model and real-time early warning evaluation indicators Load high-frequency change detection conditions: Set the time window to 15 days to capture rapid changes within a short period; Superimposed semantic segmentation conditions: In response to key areas of farmland protection, new building categories are set with a confidence level ≥ 0.85, and hardened ground categories are set, including concrete or asphalt covering with an area ≥ 100㎡; Real-time early warning indicator calculation: Based on the multi-dimensional feature coupling equation, the solution is as follows: Percentage of areas violating regulations: The total area of ​​illegal activities within the red line of arable land accounts for 0.32% of the total area within the red line. Distribution of change types: New buildings accounted for 65%, and paved ground accounted for 35%; Conflict risk level set: Three high-risk monitoring units were identified, including those with a change type frequency difference ≥0.6, meaning that the frequency of violations is significantly higher than the historical average.

[0053] Step 4: Illegal Patch Identification Model and Key Parameter Extraction Multi-level intelligent optimization framework: Based on ResNet-50, the framework was trained for 5 rounds using historical violation data and real-time early warning indicators of the region in 2023. The accuracy of the validation set reached 92%, and a violation patch recognition model was obtained. Key identification parameters: Location of the violation center: the geometric centroid of the three high-risk units Change attribute category: New building, category confidence score 0.91, distribution density 1.2 buildings / grid; Boundary expansion direction parameter: Northwest-Southeast.

[0054] Step 5: Generating the distribution results of violations Violation evolution analysis model: Based on the center point of the violation, the grid is refined to 10m×10m, and the evolution rules are defined. Among them, the boundary optimization is triggered when the cumulative area of ​​the violation is ≥500㎡. The evolution direction is determined to be to expand along the road based on the historical data of the past 3 months, with a step size of 20m / 15 days. Dynamic iterative solution: After 3 rounds of iteration, with the termination condition being that the newly added area is ≤10㎡, the results are generated. New patch area: polygon boundaries of 3 illegal areas, accuracy ±5m; Violation frequency statistics: On average, 1.2 new violations were recorded every 30 days over the past 3 months; Potential spread trend: It is expected to spread approximately 50m northeastward over the next month.

[0055] Step 6: Optimal Supervision and Scheduling Scheme Task allocation model: Regulatory response parameters: Personnel configuration consists of 3 enforcement teams, with a 24-hour response time limit, and resource allocation including 1 drone per team; Target decision: The goal is to maximize the second performance evaluation value (violation handling rate ≥ 90%) and minimize resource consumption (total fuel consumption ≤ 200L); Optimization solution: First, the genetic algorithm is used to obtain the initial solution, including group A being responsible for 2 locations and group B being responsible for 1 location. Then, the particle swarm optimization algorithm is used for global optimization. The final solution reduces the average response time to 18 hours and reduces resource consumption by 15%.

[0056] Step 7: Dynamic Adjustment and Precise Control Performance simulation: The simulated scheduling plan predicts a 95% completion rate and a processing time of 3 days; Following the actual response and based on on-site feedback, one of the violations has been partially removed. The regulatory response parameters were adjusted, and a mobile team was added to form a closed-loop management system.

[0057] Example 2: Monitoring and Evaluation of Territorial Spatial Planning in an Ecological Reserve in Southwest Mountainous Area This embodiment is for an area of ​​800 km². 2 In an ecological protection zone in the southwestern mountainous region, which includes an ecological protection red line, the focus is on monitoring vegetation damage and illegal construction. The implementation process is as follows: Step 1: Data Acquisition Remote sensing data: Landsat-9 imagery with a temporal resolution of 16 days, ALOS-2 radar data, to address cloudy and foggy weather in mountainous areas, covering March to September 2024; Vector data: ecological protection red line vector boundary, forest land ownership data, contour lines (1:10,000).

[0058] Step 2: Preprocessing and Dynamic Model Construction Preprocessing: Radar data undergoes multi-view processing to reduce speckle noise, and satellite imagery is topographically corrected and registered to UTM projection; Dynamic land surface change detection model: Based on the Transformer change detection network, compare images from September 2024 and March 2024 to identify areas with decreased vegetation cover (≥30%). Initial monitoring baseline: Irregular grid subdivision was used, and altitude and vegetation type attributes were assigned.

[0059] Step 3: Comprehensive Interpretation and Early Warning Indicators: Semantic segmentation criteria: Focus on setting vegetation destruction classification, including logging or burning, with an area ≥500㎡; illegal construction classification, non-ecological facilities, such as illegal construction of agritainment facilities; Real-time early warning indicators: Percentage of areas violating regulations: The area within the red line accounts for 0.28% of the total area; Distribution of change types: Vegetation destruction accounted for 70%, and illegal construction accounted for 30%; Conflict risk level: 2 extremely high risk units.

[0060] Steps 4-7: Key Parameters and Regulatory Optimization Key parameters: The center of the violation is located at an altitude of 1200m, the change attribute is concentrated logging, and the boundary extends towards the valley; Monitoring and dispatch: Taking into account the mountainous terrain, the route planning was optimized, and the response time was reduced from 48 hours to 36 hours by using the genetic + particle swarm algorithm. Finally, by dynamically adjusting the resource allocation, including increasing the number of helicopter patrols, the vegetation damage restoration rate was increased to 85%.

[0061] Both embodiments verified the effectiveness of the method under different terrains and different control objectives. Through multi-source data fusion, intelligent interpretation and dynamic scheduling, it achieved accurate monitoring and efficient supervision of land and space planning.

[0062] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring and evaluating land spatial planning data, characterized in that, Includes the following steps: Acquire multi-source remote sensing images and geographic information vector data of the target area; Multi-source remote sensing images are preprocessed and fused to generate a high temporal resolution dynamic land surface change detection model, and an initial monitoring baseline is constructed by combining geographic information vector data. Based on the initial monitoring baseline, high-frequency change detection conditions and semantic segmentation conditions are added to form a comprehensive interpretation model. Combined with the three-zone, three-line vector boundary data, real-time early warning and evaluation indicators are output. A multi-level intelligent optimization framework is constructed and iteratively trained using real-time early warning and evaluation indicators to obtain a violation patch identification model, and then key identification parameters are extracted. The distribution results of violations are generated by using key identification parameters, comprehensive interpretation models, and geographic information vector data; Based on the distribution of violations and regulatory response parameters, a task allocation model is established and the regulatory response parameters are optimized to obtain the optimal regulatory scheduling scheme. Based on the performance simulation framework, the current stage of regulatory scheduling plan is simulated to obtain the current stage performance simulation indicators. Combining the current stage performance simulation indicators, the current stage optimal regulatory scheduling plan, and the current stage actual response parameters, the regulatory response parameters are dynamically adjusted.

2. The method for monitoring and evaluating territorial spatial planning data according to claim 1, characterized in that, The real-time early warning assessment indicators include the proportion of illegal area, the distribution of change types, and the set of conflict risk levels; the key identification parameters include the location of the violation center point, the category of change attributes, and the parameters of boundary expansion direction; the results of the violation behavior distribution include the range of newly added patches, the statistics of violation frequency, and the potential diffusion trend.

3. The method for monitoring and evaluating territorial spatial planning data according to claim 1, characterized in that, The process of preprocessing and fusing multi-source remote sensing images to generate a high temporal resolution dynamic land surface change detection model, and constructing an initial monitoring baseline by combining geographic information vector data, includes the following steps: Collect high temporal resolution satellite image sequences covering the target area, and simultaneously acquire radar data to compensate for the loss of optical images in cloudy or foggy weather; Radiometric correction, geometric correction and image registration are performed on satellite image sequences and radar data to obtain a standardized image set. The processing includes one or more of the following: denoising, projection transformation, pixel alignment and consistency verification. Based on a standardized image set, a change detection network in deep learning is used to compare the latest images with the reference images at the pixel level, generating a dynamic surface change detection model. The dynamic surface change detection model is divided into grids and its attributes are mapped. Combined with geographic information vector data, an initial monitoring baseline is constructed. The grid division includes regular grid subdivision and irregular grid subdivision. The basic monitoring units are determined by regular grid division, and each unit is assigned a corresponding geographic information attribute.

4. The method for monitoring and evaluating territorial spatial planning data according to claim 1, characterized in that, The process of adding high-frequency change detection conditions and semantic segmentation conditions to the initial monitoring baseline to form a comprehensive interpretation model, and combining the three-zone, three-line vector boundary data to output real-time early warning evaluation indicators, includes the following steps: By applying high-frequency change detection conditions to the initial monitoring baseline and constraining the time dimension, the first interpretation model is obtained; The semantic segmentation conditions are superimposed on the first interpretation model to form a comprehensive interpretation model. The semantic segmentation conditions include new building classification conditions, hardened ground classification conditions, and vegetation destruction classification conditions. Based on the comprehensive interpretation model and the three-zone, three-line vector boundary data, real-time early warning assessment indicators are calculated. The specific process includes: A multidimensional feature coupling equation is constructed, which includes an equation for calculating the area of ​​violations, an equation for identifying the type of change, and an equation for assessing the risk of conflict. The equation is solved by numerical discretization method in conjunction with a comprehensive interpretation model to obtain the proportion of the area of ​​violations, the distribution of the type of change, and the set of conflict risk levels.

5. Obtain the aforementioned set of conflict risk levels, including the following steps: Based on the distribution of change types, calculate the frequency difference of change types for each monitoring unit; By identifying monitoring units in the integrated interpretation model whose change type frequency difference is not lower than the frequency threshold, a set of conflict risk levels is obtained.

6. The method for monitoring and evaluating territorial spatial planning data according to claim 1, characterized in that, The process of constructing a multi-level intelligent optimization framework and iteratively training it using real-time early warning evaluation indicators to obtain a violation patch identification model, and then extracting key identification parameters, includes the following steps: A multi-level intelligent optimization framework based on convolutional neural networks; The optimization framework was trained and validated multiple times using real-time early warning evaluation indicators to obtain a violation patch recognition model. Input the real-time early warning assessment indicators to be analyzed into the illegal patch identification model, and output the set of predicted conflict risk levels; Based on the predicted set of conflict risk levels, key identification parameters are extracted, including the location of the violation center point, the category of change attribute, and the boundary expansion direction parameter.

7. The method for monitoring and evaluating territorial spatial planning data according to claim 6, characterized in that, The key identification parameters are extracted from the predicted set of conflict risk levels, including the following steps: Determine the geometric centroid of the predicted set of conflict risk levels as the location of the violation center point; The change attribute categories are fitted based on the spatial distribution characteristics of the predicted conflict risk level set to obtain the change attribute category parameters, which include the category confidence and the category distribution density. The boundary expansion direction of the predicted conflict risk level set is analyzed, the weighted average of the expansion direction is calculated, and it is converted into a standard boundary expansion path, which is the boundary expansion direction parameter.

8. The method for monitoring and evaluating territorial spatial planning data according to claim 1, characterized in that, The process of generating the distribution results of violations by using key identification parameters, a comprehensive interpretation model, and geographic information vector data includes the following steps: Using the location of the violation center point as the benchmark reference point, the parameters of the changing attribute categories are mapped to the comprehensive interpretation model. The parameters of the boundary expansion direction are adjusted and the mesh of the associated monitoring units is refined to complete the update of the comprehensive interpretation model. The violation evolution rules, violation evolution direction and violation evolution step size are defined to form a violation evolution analysis model. Based on the violation evolution analysis model, a dynamic iterative method is used to solve the multidimensional feature coupling equation. Based on the results of each iteration stage in the solution process, the distribution results of violation behaviors are generated, specifically including: When the violation evolution rules are met, boundary optimization is performed to obtain the current range of newly added patches, violation evolution step size, violation evolution direction, violation frequency statistics and potential diffusion trend, and the violation evolution analysis model is updated. The multidimensional feature coupling equation is re-solved based on the updated violation evolution analysis model until the termination condition is met, thus ending the violation evolution process. Based on the violation frequency statistics and potential diffusion trends at each iteration stage, calculate the statistical distribution of violation behavior and the set of diffusion trends.

9. The method for monitoring and evaluating territorial spatial planning data according to claim 1, characterized in that, The process of establishing a task allocation model based on the distribution of violations and regulatory response parameters, optimizing the regulatory response parameters, and obtaining the optimal regulatory scheduling scheme includes the following steps: A task allocation model is established, which includes a variable initialization module, a target decision module, and a constraint module. The variable initialization module is defined based on regulatory response parameters, which include personnel configuration scale, response time limit, resource allocation ratio, and communication delay time. The target decision module is constructed based on the second efficiency evaluation value and resource consumption value. The second efficiency evaluation value is calculated based on the distribution results of violations, and the resource consumption value includes human resource cost, equipment maintenance cost, and communication transmission cost. The constraint module includes a regulatory response constraint module and an efficiency guarantee constraint module. The efficiency guarantee constraint module is defined based on task integrity, response timeliness, and system coordination. A genetic algorithm is used to initially solve the task allocation model and obtain an initial optimized scheduling scheme. The task allocation model is globally optimized using the particle swarm optimization algorithm to obtain the optimal solution set of the particle swarm as the optimal supervisory scheduling scheme.

10. A land spatial planning data monitoring and evaluation system, characterized in that, This method is used to implement the land and space planning data monitoring and evaluation method as described in any one of claims 1 to 9.

Images