Regional ecological space pattern big data real-time monitoring and protection management system

Abstract

The application relates to the technical field of remote sensing data processing and ecological environment monitoring, and provides a regional ecological space pattern big data real-time monitoring and protection management system, which comprises the following modules: a data alignment module, which is used for acquiring multi-source remote sensing data and determining a benchmark and a real-time spatial resolution; in response to the real-time resolution being higher than the benchmark resolution, performing spatial aggregation on fine ecological patches based on ecological function equivalent conditions and resampling to obtain scale alignment grid data; a scale normalization module, which is used for calculating real-time landscape pattern indexes and scale response coefficients and correcting to obtain scale normalized landscape pattern indexes; and an artifact elimination and alarm module, which is used for identifying resolution artifacts based on the deviation direction of original and normalized indexes relative to historical benchmarks, and generating an alarm when no artifact is identified and the deviation is higher than a threshold value. The application eliminates scale effects and eliminates resolution artifacts, and realizes cross-scale high-precision ecological pattern monitoring.

Description

Technical Field

[0001] This application relates to the fields of remote sensing data processing and ecological environment monitoring technology, specifically to a real-time monitoring and protection management system for regional ecological spatial patterns based on big data. Background Technology

[0002] In regional ecological spatial pattern monitoring, landscape pattern indicators are typically calculated using multi-source remote sensing data to assess ecological status. However, due to differences in spatial resolution between remote sensing images acquired at different times or by different sensors, directly calculated landscape pattern indicators exhibit significant scale effects, meaning the same land cover displays different pattern characteristics at different resolutions. While existing resampling methods can unify resolution, they are prone to losing surface ecological boundary information in high-resolution images or introducing mixed pixel errors. Furthermore, when the resolution of real-time monitoring data changes, fluctuations in indicator values ​​are often spurious changes caused by resolution differences rather than genuine ecological pattern evolution. These resolution artifacts can lead to false alarms or missed alarms in the monitoring system, severely impacting the accuracy of ecological protection decisions. Therefore, there is an urgent need for an ecological pattern monitoring system that can eliminate scale effects and remove resolution artifacts while preserving high-resolution dominant ecological boundaries. Summary of the Invention

[0003] To address the issues of scale effects and resolution artifacts in landscape pattern indicators caused by resolution inconsistencies in existing multi-source remote sensing data, this application utilizes a regional ecological spatial pattern big data real-time monitoring and protection management system to achieve high-precision ecological pattern monitoring and reliable alarms for cross-scale remote sensing data.

[0004] To achieve the above objectives, this application adopts the following technical solution:

[0005] A real-time monitoring and protection management system for regional ecological spatial patterns, comprising: a data alignment module for acquiring multi-source remote sensing data of a target area, determining the reference spatial resolution of a baseline remote sensing image and the real-time spatial resolution of a real-time remote sensing image; responding to a situation where the real-time spatial resolution is higher than the reference spatial resolution, spatially aggregating fragmented ecological patches in the real-time remote sensing image based on ecological function equivalence conditions, and resampling to the reference spatial resolution to obtain scale-aligned raster data; a scale normalization module for calculating real-time landscape pattern indicators based on the scale-aligned raster data, obtaining scale response coefficients characterizing the relationship between the indicators at the real-time spatial resolution and the reference spatial resolution, and correcting the real-time landscape pattern indicators based on the scale response coefficients to obtain scale-normalized landscape pattern indicators; and an artifact removal and alarm module for determining the original landscape pattern indicators of the real-time remote sensing image before spatial aggregation, identifying resolution artifacts based on the first deviation direction between the original landscape pattern indicators and the historical reference pattern indicators, and the second deviation direction between the scale-normalized landscape pattern indicators and the historical reference pattern indicators; responding to a situation where no resolution artifacts are identified, comparing the scale-normalized landscape pattern indicators with the historical reference pattern indicators, and generating an alarm when the deviation exceeds a preset alarm threshold.

[0006] The above scheme uses a data alignment module to perform spatial aggregation based on ecological function equivalence conditions at high real-time resolution, preserving the boundary integrity of dominant ecological patches and avoiding the loss of details caused by traditional resampling. The scale normalization module uses the scale response relationship across resolution levels to correct the indicators, fundamentally eliminating the impact of scale effects on indicator values. The artifact removal and alarm modules compare the trend of indicator changes before and after scale conversion, accurately identifying and filtering false fluctuations caused by resolution changes, thereby ensuring the authenticity of ecological pattern monitoring results and the reliability of alarms.

[0007] Furthermore, the data alignment module performs spatial aggregation on fragmented ecological patches in real-time remote sensing images based on ecological function equivalence conditions. This includes: calculating the actual area of ​​each connected patch in the real-time remote sensing image, and identifying connected patches with actual areas smaller than the functional equivalence area threshold as fragmented ecological patches; determining the patch with the largest area among the adjacent patches of each fragmented ecological patch as the dominant ecological patch; and updating the pixel category of the fragmented ecological patch to the ecological type of the dominant ecological patch, thereby merging the fragmented ecological patches into the dominant ecological patch.

[0008] This alternative implementation filters out fragmented patches that contribute little to the macroscopic pattern by setting a functional equivalent area threshold, and topologically merges them into adjacent dominant patches. This eliminates noise interference caused by excessive fragmentation at high resolution, maintains the overall characterization of regional ecological functions, and provides a semantically consistent raster basis for subsequent scale alignment.

[0009] Furthermore, the process of determining the functional equivalent area threshold includes: calculating the area of ​​a single pixel at the reference spatial resolution as the reference pixel area; obtaining the minimum identifiable area corresponding to each ecological type; taking the larger value between the reference pixel area and the minimum identifiable area of ​​the corresponding ecological type as the functional equivalent area threshold corresponding to that ecological type; updating the pixel category of fragmented ecological patches to the ecological type of the dominant ecological patch, including: determining whether the ecological types of the fragmented ecological patches and the dominant ecological patches are consistent; if consistent, reassigning the pixel category of the fragmented ecological patches to the ecological type of the dominant ecological patch, so that the fragmented ecological patches are incorporated into the dominant ecological patch in terms of connectivity; if inconsistent, reassigning the pixel category of the fragmented ecological patches to the ecological type of the dominant ecological patch, so that the fragmented ecological patches are merged into the dominant ecological patch.

[0010] This alternative implementation combines the physical constraints of the baseline resolution with the semantic constraints of the ecological type to determine the aggregation threshold, and uses the type consistency judgment branch processing merging logic to ensure that the aggregation operation is consistent with the remote sensing imaging mechanism and respects the patch independence in an ecological sense, thus preventing the erroneous fusion of different types of ecological patches.

[0011] Furthermore, the scale normalization module obtains the scale response coefficient, including: generating multiple simulated resolution levels starting from the reference spatial resolution and increasing the scale step size by a preset step size; resampling the real-time remote sensing images to each simulated resolution level and calculating the corresponding simulated landscape pattern index value at each simulated resolution level; constructing a scale response curve based on the simulated resolution level and the simulated landscape pattern index value; reading the first simulated landscape pattern index value corresponding to the real-time spatial resolution and the second simulated landscape pattern index value corresponding to the reference spatial resolution from the scale response curve; and using the ratio of the first simulated landscape pattern index value to the second simulated landscape pattern index value as the scale response coefficient.

[0012] This alternative implementation quantifies the nonlinear response law of landscape pattern indicators as resolution changes by constructing multi-resolution level scale response curves, so that the scale response coefficient can dynamically reflect the theoretical deviation of the current real-time resolution relative to the reference resolution, providing an accurate mathematical basis for the normalization correction of the indicators.

[0013] Furthermore, scale response curves are constructed based on the simulated resolution levels and simulated landscape pattern index values, including: calculating the landscape fragmentation index, patch connectivity index, and Shannon diversity index at each simulated resolution level; performing logarithmic transformations on the landscape fragmentation index, patch connectivity index, and Shannon diversity index, and then performing linear regression fitting with the logarithmic values ​​of the simulated resolution levels as independent variables to obtain the scale response regression equations corresponding to each index; calculating the first predicted value of each scale response regression equation at the real-time spatial resolution and the second predicted value at the baseline spatial resolution, and using the ratio of the first predicted value to the second predicted value of each index as the individual scale response coefficient for each index; and performing a weighted average of each individual scale response coefficient to obtain the comprehensive scale response coefficient as the scale response coefficient; the weights of the weighted average are preset based on the sensitivity of each index to changes in the ecological pattern.

[0014] This alternative implementation captures the scale dependence of pattern characteristics in different dimensions by performing log-linear regression fitting on multiple key landscape indicators, and generates a comprehensive coefficient by combining sensitivity-weighted averaging, thus overcoming the one-sidedness of scale response of a single indicator and improving the ability of normalized correction results to represent changes in complex ecological patterns.

[0015] Furthermore, the artifact removal and alarm module identifies resolution artifacts based on the first deviation direction between the original landscape pattern index and the historical baseline pattern index, and the second deviation direction between the scale-normalized landscape pattern index and the historical baseline pattern index. This includes: calculating the first deviation direction between the original landscape pattern index and the historical baseline pattern index, and the second deviation direction between the scale-normalized landscape pattern index and the historical baseline pattern index; if the first deviation direction and the second deviation direction are opposite, the current index change is determined to be a resolution artifact, the scale-normalized landscape pattern index is corrected to the historical baseline pattern index, and the alarm is skipped; if the first deviation direction and the second deviation direction are the same, the resolution artifact is determined not to be identified, the scale-normalized landscape pattern index is retained, and a deviation comparison is performed.

[0016] This optional implementation uses the consistency of the changing trends of the original and normalized indicators relative to the historical benchmark as an artifact discrimination criterion. When the two trends are opposite, it indicates that the normalization process has introduced reverse disturbances or that the original data itself is only resolution noise, thereby effectively intercepting false alarms and enhancing the robustness of the monitoring system.

[0017] Furthermore, it also includes a dynamic monitoring window determination module, used for: extracting sample pixels from scale-aligned raster data at a preset sampling interval, calculating the semivariogram values ​​of the sample pixels at different hysteresis distances; fitting the semivariogram values ​​to a spherical model, extracting the range value of the fitted curve as the semivariogram range; rounding the semivariogram range up to an integer multiple of the reference spatial resolution to obtain the initial window side length; obtaining the available memory capacity of the current edge computing node, calculating the maximum number of raster pixels that the available memory capacity can support, and determining the memory constraint side length based on the maximum number of raster pixels and the reference spatial resolution; and setting the initial window... The smaller of the side length of the opening and the side length of the memory constraint is used as the window side length of the dynamic monitoring window; the artifact removal and alarm module generates an alarm when the deviation exceeds the preset alarm threshold, including: calculating the absolute deviation value between the scale-normalized landscape pattern index and the historical baseline pattern index; comparing the absolute deviation value with the first-level alarm threshold and the second-level alarm threshold; the first-level alarm threshold is less than the second-level alarm threshold; if the absolute deviation value is greater than the first-level alarm threshold and less than or equal to the second-level alarm threshold, a pattern trend warning is generated; if the absolute deviation value is greater than the second-level alarm threshold, a pattern change alarm is generated, and a dense observation command for the target area is triggered.

[0018] This optional implementation determines a monitoring window that conforms to geoscientific statistical laws through a semi-variogram range, while adaptively adjusting it in conjunction with the memory constraints of edge computing nodes, thus balancing the scientific nature of pattern analysis with the feasibility of engineering implementation; the hierarchical alarm mechanism distinguishes between trend changes and sudden damage based on the degree of deviation, and links encrypted observations to achieve optimized allocation of monitoring resources.

[0019] Furthermore, it also includes a benchmark update module, which is used to: continuously acquire multi-source remote sensing data according to a preset time period, and store the scale-normalized landscape pattern indicators corresponding to each time period into a time-series index database; when the scale-normalized landscape pattern indicators of multiple consecutive time periods show a monotonic change trend, update the historical benchmark pattern indicators to the moving average of the scale-normalized landscape pattern indicators of the most recent preset number of time periods; and use the updated historical benchmark pattern indicators as the deviation comparison benchmark for subsequent time periods.

[0020] This optional implementation triggers dynamic updates to the layout baseline by monitoring the monotonic changing trend of indicators, enabling historical baseline values ​​to adaptively drift with the natural evolution of the ecological pattern. This avoids misjudging normal ecological succession as abnormal changes due to a long-term fixed baseline, thus improving the adaptability of long-term monitoring.

[0021] In addition, this application also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the regional ecological spatial pattern big data real-time monitoring and protection management system as described above.

[0022] The aforementioned computer equipment executes programs stored in memory through a processor, realizing all the functional modules of the aforementioned system, providing a physical hardware carrier for ecological pattern monitoring, and facilitating deployment and implementation in edge computing nodes or cloud servers.

[0023] In addition, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the regional ecological spatial pattern big data real-time monitoring and protection management system as described above.

[0024] The aforementioned computer-readable storage medium carries program instructions for implementing the monitoring system's functions, supporting independent software distribution, version management, and cross-platform deployment.

[0025] Beneficial effects:

[0026] The regional ecological spatial pattern big data real-time monitoring and protection management system provided in this application, through a data alignment module, merges fragmented patches into dominant patches and resamples them based on ecological function equivalence conditions when the real-time resolution is higher than the baseline resolution. This eliminates high-resolution noise while preserving key ecological boundary information, solving the problem of surface detail loss caused by traditional resampling. Furthermore, a scale normalization module uses scale response curves constructed at multiple resolution levels to obtain scale response coefficients, normalizing and correcting real-time landscape pattern indicators. This mathematically decouples the indicator values ​​from spatial resolution, effectively overcoming the scale effect in multi-source data fusion. Further, an artifact removal and alarm module accurately identifies false fluctuations caused by resolution switching by comparing the deviation directions of the original and normalized indicators relative to historical baselines, triggering alarms only after confirming that they are not artifacts, significantly reducing the false alarm rate. The synergistic effect of these three modules enables cross-scale, highly reliable real-time monitoring of regional ecological spatial patterns, providing accurate data support for ecological protection decisions. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the structure of the regional ecological spatial pattern big data real-time monitoring and protection management system according to an embodiment of this application. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application.

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0030] Example 1

[0031] like Figure 1 As shown in the figure, this embodiment provides a real-time monitoring and protection management system for regional ecological spatial patterns using big data. Through a modular architecture, this system achieves a complete closed-loop processing from multi-source heterogeneous data input to reliable alarm output. The system mainly includes a data alignment module, a scale normalization module, and an artifact removal and alarm module. These modules are cascaded through standardized data interfaces to jointly address the scale effect and resolution artifact problems in multi-source remote sensing data monitoring.

[0032] The data alignment module acquires multi-source remote sensing data of the target area, determines the reference spatial resolution of the baseline remote sensing image and the real-time spatial resolution of the real-time remote sensing image. In response to a real-time spatial resolution higher than the reference spatial resolution, it performs spatial aggregation on fragmented ecological patches in the real-time remote sensing image based on ecological function equivalence conditions, and resamples them to the reference spatial resolution to obtain scale-aligned raster data. Specifically, the data alignment module, as the preprocessing front-end of the entire system, plays a crucial role in unifying the spatial reference of the data and preserving key ecological semantics. In practical applications, due to sensor updates or changes in task requirements, real-time acquired remote sensing images often have higher spatial resolutions than historical reference images. If resampling is performed directly, a large amount of detail noise at high resolution will be forcibly smoothed, resulting in blurred surface ecological boundaries; if resampling is not performed, pixel-level comparison with historical data is impossible. Therefore, this embodiment sets a trigger condition of "real-time spatial resolution higher than the reference spatial resolution," initiating enhancement processing only when necessary. The spatial aggregation based on "ecological function equivalence conditions" is not a traditional image filtering or simple morphological operation. Instead, it follows ecological principles to merge fragmented patches that are visible at high resolution but contribute little to macroscopic ecological functions into adjacent dominant ecological patches. This approach eliminates fragmentation noise introduced by high resolution while preserving the geometric boundary integrity of dominant patches. Subsequently, the semantically corrected data is resampled to the baseline spatial resolution. The output scale-aligned raster data is not only consistent with the baseline in terms of mathematical grid but also comparable in ecological semantics, laying a solid foundation for subsequent index calculations.

[0033] The scale normalization module is used to calculate real-time landscape pattern indicators based on scale-aligned raster data. It obtains scale response coefficients, which characterize the relationship between the indicator response at real-time spatial resolution and the baseline spatial resolution. Based on these scale response coefficients, the real-time landscape pattern indicators are corrected to obtain scale-normalized landscape pattern indicators. Specifically, even after spatial alignment, due to the significant scale dependence of landscape pattern indicators, the indicator values ​​for the same feature at different resolutions may still exhibit systematic deviations, i.e., scale effects. The scale normalization module is designed to eliminate this inherent bias. This module first calculates the current real-time landscape pattern indicators based on scale-aligned raster data, but this is only the raw calculated value. To ensure fair comparison with historical benchmarks, the module introduces a correction factor called the "scale response coefficient." This coefficient quantifies the theoretical scaling or offset that the indicator value should undergo when transitioning from the baseline resolution to the real-time resolution. By dividing or multiplying the real-time indicator by this coefficient (the specific calculation method depends on the indicator type), the real-time indicator can be "translated" back to the equivalent value at the baseline resolution, i.e., the scale-normalized landscape pattern indicator. This process essentially separates the numerical fluctuations caused by resolution differences from the monitoring results, so that the final output indicators only reflect the real changes in the ecological pattern, rather than the differences in the physical attributes of the data source.

[0034] The artifact removal and alarm module is used to determine the original landscape pattern indicators of real-time remote sensing images before spatial aggregation. It identifies resolution artifacts based on the first deviation direction between the original landscape pattern indicators and historical baseline indicators, and the second deviation direction between the scale-normalized landscape pattern indicators and historical baseline indicators. In response to the absence of resolution artifacts, it compares the scale-normalized landscape pattern indicators with the historical baseline indicators and generates an alarm when the deviation exceeds a preset alarm threshold. Specifically, although the first two modules strive to eliminate errors, false signals may still remain under extreme data conditions. The artifact removal and alarm module acts as a safety valve for the system. This module innovatively introduces a dual-path comparison mechanism: one path shows the change trend of the original landscape pattern indicators relative to the historical baseline (first deviation direction), and the other path shows the change trend of the indicators after complete normalization relative to the historical baseline (second deviation direction). Ideally, if a real change in the ecological pattern occurs, the change directions of both should be consistent (e.g., both showing an increase in fragmentation). However, if the two deviations are in opposite directions (e.g., the raw metric shows a sharp increase in fragmentation, while the normalized metric shows stable or decreasing fragmentation), it strongly suggests that the current abnormal fluctuations are "resolution artifacts" caused by nonlinear effects during resolution switching or aggregation, rather than real ecological events. In this case, the system will automatically intercept the alarm to avoid false alarms. Only when the two deviations are in the same direction and confirmed to be non-artifacts will the system further determine whether the deviation of the normalized metric exceeds a preset alarm threshold, thereby generating a high-confidence ecological pattern change alarm. This error correction mechanism based on trend consistency verification significantly improves the robustness and reliability of the monitoring system in complex multi-source data environments.

[0035] It should be understood that, although Figure 1 The linear connection between the three modules described above is illustrated. However, in other embodiments, the modules can also be processed in parallel, pipelined, or distributed manner, as long as the aforementioned data flow logic and functional coordination can be achieved. Furthermore, the system architecture described in this embodiment is not only suitable for terrestrial ecosystem monitoring but can also be extended to other types of ecological pattern monitoring scenarios such as marine and wetland areas, requiring only adjustments to the corresponding ecological function equivalence parameters and landscape pattern indicator system.

[0036] Example 2

[0037] Building upon Example 1, this example further details the specific implementation logic of the data alignment module performing spatial aggregation of fragmented ecological patches in real-time remote sensing images based on ecological functional equivalence conditions. Specifically, this process is not a simple image smoothing or morphological filtering, but a topological reconstruction process that preserves ecological semantics, mainly including the following steps: First, the actual area of ​​each connected patch in the real-time remote sensing image is calculated, and connected patches with actual areas smaller than the functional equivalence area threshold are identified as fragmented ecological patches; second, the patch with the largest area among the adjacent patches of each fragmented ecological patch is identified as the dominant ecological patch; finally, the pixel category of the fragmented ecological patch is updated to the ecological type of the dominant ecological patch, thus merging the fragmented ecological patches into the dominant ecological patch. In this process, the system traverses all connected regions in the raster data, calculates the number of pixels in each region and multiplies it by the area of ​​a single pixel at the real-time spatial resolution to obtain the accurate physical area. For the selected fragmented ecological patches, the algorithm searches for adjacent patches within its four-neighbor or eight-neighbor range and selects the one with the largest area as the merging target. For example, in wetland monitoring scenarios, high-resolution imagery may produce numerous scattered "small noise" patches due to water surface reflections or vegetation shadows. These patches do not possess independent habitat functions ecologically and instead interfere with the calculation of landscape fragmentation. By merging them into adjacent large wetland patches, not only is data noise eliminated, but the true connectivity of the wetland ecosystem is also restored. This processing method ensures that the subsequently generated scale-aligned raster data matches the baseline resolution geometrically while maintaining a high degree of fidelity in ecological semantics.

[0038] As an optional implementation method, the process of determining the functional equivalent area threshold includes: calculating the area of ​​a single pixel at the reference spatial resolution as the reference pixel area; obtaining the minimum identifiable area corresponding to each ecological type; and taking the larger value between the reference pixel area and the minimum identifiable area of ​​the corresponding ecological type as the functional equivalent area threshold for that ecological type. This threshold setting mechanism integrates the physical limitations of remote sensing imaging with the objective needs of ecological significance. The reference pixel area represents the spatial sampling limit of historical reference data; any geographic entity smaller than this area cannot be independently represented in the reference data. The minimum identifiable area, on the other hand, is the minimum spatial range required to define a specific ecosystem type in landscape ecology. For example, forest patches typically need to reach a certain size to exert an edge effect or internal habitat function. Taking the larger value of the two as the threshold means that only when a patch can be distinguished at the reference resolution and possesses independent ecological functions will it be retained. Conversely, if a patch is larger than the reference pixel but smaller than the lower limit of ecological function, or possesses ecological functions but is too small to be visible in the reference data, it will be considered a fragmented patch and aggregated. This dual constraint mechanism effectively avoids the problems of over-aggregation or under-aggregation caused by relying solely on resolution or solely on ecological parameters.

[0039] Furthermore, updating the pixel category of the fragmented ecological patches to the ecological type of the dominant ecological patch includes: determining whether the ecological types of the fragmented ecological patches and the dominant ecological patches are consistent; if they are consistent, the pixel category of the fragmented ecological patches is reassigned to the ecological type of the dominant ecological patch, so that the fragmented ecological patches are incorporated into the dominant ecological patch in terms of connectivity; if they are inconsistent, the pixel category of the fragmented ecological patches is reassigned to the ecological type of the dominant ecological patch, so that the fragmented ecological patches are merged into the dominant ecological patch. It should be noted that when the fragmented patches and the dominant patch type are inconsistent, this embodiment still chooses to forcibly merge them into the dominant patch. This may seem to violate the principle of classification consistency, but it is actually a targeted design for the characteristics of high-resolution remote sensing data. In high-resolution images, mixed pixels are common at the boundaries of different land features, resulting in a large number of small fragments with mixed types at the boundaries in the classification results. These fragments are not real ecological transition zones, but classification artifacts caused by spectral mixing. If only similar types are allowed to be merged, these heterogeneous fragments will be retained, leading to a large number of false patch counts and boundary densities in landscape pattern indicators. Therefore, regardless of type consistency, as long as a fragmented patch is below the functional equivalence threshold, it is judged as non-functional noise, and its attributes are forcibly corrected to reflect the dominant type of the surrounding environment. This strategy fundamentally eliminates the mixed pixel noise introduced by high resolution, ensuring that the scale-aligned data can truly reflect the macro-ecological pattern, rather than data artifacts caused by the increase in sensor resolution. It should be understood that although this embodiment describes an aggregation method based on area thresholds, in other embodiments, morphological indicators such as shape index and compactness can also be combined to assist in the judgment of fragmented patches, as long as the core logic still follows the principle of ecological functional equivalence.

[0040] Example 3

[0041] Building upon Example 1, this example further analyzes the specific construction method and comprehensive weighting algorithm for obtaining scale response coefficients using the scale normalization module. Specifically, the process of obtaining scale response coefficients using the scale normalization module includes: generating multiple simulated resolution levels starting from the reference spatial resolution and increasing the step size by a preset scale; resampling the real-time remote sensing images to each simulated resolution level and calculating the corresponding simulated landscape pattern index values ​​at each simulated resolution level; constructing scale response curves based on the simulated resolution levels and simulated landscape pattern index values; reading the first simulated landscape pattern index value corresponding to the real-time spatial resolution and the second simulated landscape pattern index value corresponding to the reference spatial resolution from the scale response curves; and using the ratio of the first simulated landscape pattern index value to the second simulated landscape pattern index value as the scale response coefficient. The core of this process lies in establishing a continuous response model of the index changing with resolution through multi-scale simulation, rather than relying solely on a simple comparison of two discrete points. Since there is usually a complex nonlinear relationship between landscape pattern indicators and spatial resolution, directly calculating the ratio using two measured values ​​of real-time resolution and reference resolution often fails to accurately represent the full picture of the scale effect. Therefore, this embodiment uses the reference spatial resolution as an anchor point and generates a series of simulated resolution levels upwards according to a preset step size (e.g., increasing by 10 meters each time or doubling proportionally), forming a gradient sequence covering from the reference to the real-time resolution and even coarser scales. By resampling the same real-time remote sensing image level by level and calculating the index values ​​at each level, the system can depict a complete scale response curve. This curve reflects the inherent dependence of landscape indicators on the observation scale in a specific area under the current land cover state. Based on this, by reading the theoretical values ​​at the real-time resolution and the reference resolution on the curve and calculating the ratio, the obtained scale response coefficient not only includes the current resolution difference information but also integrates the changing trend characteristics within the entire scale domain, thereby significantly improving the robustness and accuracy of the normalization correction. It should be understood that although this embodiment describes a simulation method based on resampling, in other embodiments, wavelet transform or fractal dimension estimation methods can also be used to construct the scale response relationship, as long as they can quantify the response law of the index with scale changes.

[0042] As an optional implementation, the process of constructing scale response curves based on simulated resolution levels and simulated landscape pattern index values ​​specifically includes: calculating the landscape fragmentation index, patch connectivity index, and Shannon diversity index at each simulated resolution level; performing logarithmic transformations on the landscape fragmentation index, patch connectivity index, and Shannon diversity index, and then performing linear regression fitting with the logarithmic value of the simulated resolution level as the independent variable to obtain the scale response regression equation corresponding to each index; calculating the first predicted value of each scale response regression equation at the real-time spatial resolution and the second predicted value at the baseline spatial resolution, and using the ratio of the first predicted value to the second predicted value of each index as the individual scale response coefficient for each index; and performing a weighted average of each individual scale response coefficient to obtain the comprehensive scale response coefficient as the scale response coefficient; the weight of the weighted average is based on a preset sensitivity of each index to changes in ecological pattern. In this implementation, the selection of three types of indicators—landscape fragmentation, patch connectivity, and Shannon diversity—has clear ecological considerations: fragmentation characterizes the geometric degree of landscape fragmentation, connectivity reflects the topological accessibility of ecological processes, and diversity measures the complexity and information entropy of the system composition. These three aspects comprehensively characterize the ecological pattern from the dimensions of morphology, structure, and function. More importantly, this embodiment employs a fitting strategy of "logarithmic transformation + linear regression." This is because numerous landscape ecology studies have shown that most landscape pattern indicators follow a power-law distribution relationship with spatial resolution, i.e. ,in For indicator values, For resolution, and It is a constant. Taking the logarithm of both sides of this equation transforms it into a linear form. This transformation not only conforms to natural laws but also effectively compresses the dynamic range of the data, reduces the interference of extreme values ​​on the fitting results, and makes the obtained scale response regression equation more statistically significant. The first and second predicted values ​​predicted by this equation are actually theoretical expected values ​​after removing random noise, and the individual scale response coefficients calculated based on them are more reliable than the original measured ratios.

[0043] Furthermore, the weighted average of the individual scale response coefficients to obtain the comprehensive scale response coefficient is used to address the issue of inconsistent sensitivity of different indicators to scale changes. In actual monitoring, landscape fragmentation is typically extremely sensitive to changes in resolution; even small scale changes can lead to drastic fluctuations in the values. In contrast, indicators based on area proportions, such as Shannon diversity, are relatively stable. If a simple arithmetic average is used, the scale effect of highly sensitive indicators will dominate the comprehensive coefficient, leading to overcorrection of low-sensitivity indicators or misjudgment of the overall pattern. Therefore, this embodiment introduces a sensitivity-based preset weighting mechanism. For example, in a set of preferred implementation parameters, the weight of landscape fragmentation can be set to 0.4, patch connectivity to 0.3, and Shannon diversity to 0.3 to reflect the dominant role of fragmentation in scale transformation. Of course, these weights are not fixed. In practical applications, technicians can dynamically adjust them based on the ecological characteristics of the target area or the results of variation coefficient analysis of historical data. For example, in wildlife corridor monitoring with habitat connectivity as the core protection objective, the weight of the patch connectivity indicator can be appropriately increased. This flexible weighting strategy ensures that the comprehensive scale response coefficient can not only keenly capture key scale effects, but also take into account the overall stability of the ecological pattern, so that the final scale-normalized landscape pattern index can truly be comparable across scales.

[0044] Example 4

[0045] Building upon Example 1, this example further elucidates the specific logic of identifying resolution artifacts based on deviation direction comparison in the artifact removal and alarm module. Specifically, the artifact removal and alarm module identifies resolution artifacts based on the first deviation direction between the original landscape pattern index and the historical baseline pattern index, and the second deviation direction between the scale-normalized landscape pattern index and the historical baseline pattern index. This includes: calculating the first deviation direction between the original landscape pattern index and the historical baseline pattern index, and the second deviation direction between the scale-normalized landscape pattern index and the historical baseline pattern index; if the first deviation direction and the second deviation direction are opposite, the current index change is determined to be a resolution artifact, the scale-normalized landscape pattern index is corrected to the historical baseline pattern index, and the alarm is skipped; if the first deviation direction and the second deviation direction are the same, it is determined that no resolution artifact has been identified, the scale-normalized landscape pattern index is retained, and deviation comparison is performed.

[0046] The core of this mechanism lies in using a dual reference system to verify the authenticity of the data. The first bias direction reflects the trend of the raw data, without any scale correction, relative to its historical state. This includes the superposition of genuine ecological change signals and spurious noise introduced by resolution differences. The second bias direction reflects the trend of the data, after scale response coefficient correction, relative to the same historical state. Theoretically, this should be free from scale effects and more closely approximate the true ecological condition. By comparing the consistency of these two directions, the system can effectively distinguish which fluctuations are caused by changes in the physical properties of the data source itself and which are caused by actual changes in land cover.

[0047] When the first deviation direction is opposite to the second deviation direction, the system determines that the current indicator change is a resolution artifact. For example, in a monitoring task, due to the replacement of a higher resolution sensor, the fragmentation value in the original landscape pattern indicator may increase significantly due to the increase in detail information, causing the first deviation direction to show "positive increase." However, after correction by the scale normalization module, the scale normalized landscape pattern indicator falls back to a level close to the historical baseline pattern indicator, causing the second deviation direction to show "basically flat" or "negative decrease." This divergence in trend strongly suggests that the current indicator anomaly mainly stems from the change in data granularity brought about by the resolution increase, rather than a substantial increase in ecosystem fragmentation. At this time, the system will automatically correct the scale normalized landscape pattern indicator to the historical baseline pattern indicator and skip the subsequent alarm process, thereby avoiding false alarms triggered by data source switching.

[0048] Conversely, when the first deviation direction is the same as the second deviation direction, the system determines that no resolution artifact has been identified. This means that both the original observations and the theoretical values ​​after scale correction point to the same trend. For example, when large-scale deforestation occurs in the target area, regardless of the resolution change, the fragmentation index will show a significant upward trend, in which case both deviation directions are positive and of similar magnitude. In this case, the system confirms that the current index fluctuation exceeds the interference range of scale effects and belongs to the actual ecological pattern evolution. Therefore, it retains the scale-normalized landscape pattern index and transmits it to the subsequent alarm comparison stage.

[0049] It should be understood that the "deviation direction" described in this embodiment is not limited to a simple sign judgment. In practical applications, it can also be combined with the ratio of deviation amplitudes or correlation coefficients for comprehensive judgment. For example, when two deviation directions have the same sign but their amplitudes differ by more than a preset multiple, they can also be considered as having potential artifact risks and marked. In addition, for some indicators that are not sensitive to scale changes, the system can be configured to trust their normalized results by default, thus simplifying the comparison process. This error correction mechanism based on trend consistency verification is equivalent to setting up an intelligent safety valve in the automated monitoring link, significantly improving the robustness and alarm reliability of the system in multi-source heterogeneous data environments.

[0050] Example 5

[0051] Building upon Example 4, this example further describes the adaptive determination method for the dynamic monitoring window and the hierarchical alarm triggering mechanism. Specifically, the system also includes a dynamic monitoring window determination module, used for: extracting sample pixels from scale-aligned raster data at preset sampling intervals, calculating the semivariogram values ​​of the sample pixels at different hysteresis distances; fitting the semivariogram values ​​to a spherical model, extracting the range value of the fitted curve as the semivariogram range; rounding the semivariogram range up to an integer multiple of the reference spatial resolution to obtain the initial window side length; obtaining the available memory capacity of the current edge computing node, calculating the maximum number of raster pixels that the available memory capacity can support, and determining the memory constraint side length based on the maximum number of raster pixels and the reference spatial resolution; and using the smaller value between the initial window side length and the memory constraint side length as the window side length of the dynamic monitoring window.

[0052] The core of this dynamic window determination mechanism lies in balancing the effectiveness of geoscientific statistics with the security of hardware resources. Specifically, landscape pattern analysis is not conducted uniformly across the entire map; rather, it needs to be carried out within local areas exhibiting spatial autocorrelation to be statistically significant. The semivariogram is a classic tool for quantifying this spatial dependence. The system first extracts sample pixels from scale-aligned raster data at preset sampling intervals (e.g., every 10 pixels or random uniform sampling) to avoid wasting computational resources on full-scale calculations. Subsequently, the semivariogram values ​​of these sample pixels at different lag distances are calculated, plotting a curve of spatial heterogeneity as a function of distance. By fitting this curve to a spherical model, the range value can be accurately extracted, representing the maximum distance at which ecological elements maintain spatial correlation. The range value is rounded up to an integer multiple of the baseline spatial resolution as the initial window side length, ensuring that the monitoring window can completely cover an independent spatial structural unit and avoiding distortion of pattern index calculations due to window truncation. It should be understood that although a spherical model is preferred for fitting in this embodiment, in other embodiments, an exponential model or a Gaussian model can also be selected according to the land cover characteristics of the target area, as long as it can accurately characterize the range of spatial autocorrelation.

[0053] Meanwhile, considering that actual deployment environments are typically resource-constrained edge computing nodes, the system introduces a memory constraint side length as a safety valve. By acquiring the available memory capacity of the current edge computing node in real time and combining it with the data occupancy of a single pixel, the maximum number of raster pixels that can theoretically be supported is calculated, and the memory constraint side length is then deduced. Finally, the system determines the smaller value between the initial window side length and the memory constraint side length as the actual dynamic monitoring window side length. This dual constraint strategy reflects the defensive thinking in engineering design: when the analysis scale required by the ecological process is smaller than the hardware's carrying capacity, the accuracy of scientific analysis is prioritized; while when the scale required by the ecological process is too large and exceeds the hardware limit, it automatically degrades to operation within the hardware safety boundary to prevent monitoring service crashes due to memory overflow. At this point, although some large-scale spatial correlation information is sacrificed, the continuous availability of the system is guaranteed, and since the previous embodiment has eliminated the impact of resolution differences through scale normalization, the indicators calculated even under a smaller window still have cross-time period comparability.

[0054] Furthermore, the process by which the artifact removal and alarm module generates an alarm when the deviation exceeds a preset alarm threshold includes: calculating the absolute deviation between the scale-normalized landscape pattern index and the historical baseline pattern index; comparing the absolute deviation with the first-level alarm threshold and the second-level alarm threshold; wherein the first-level alarm threshold is less than the second-level alarm threshold; if the absolute deviation is greater than the first-level alarm threshold and less than or equal to the second-level alarm threshold, a pattern trend warning is generated; if the absolute deviation is greater than the second-level alarm threshold, a pattern change alarm is generated, and a denser observation command for the target area is triggered.

[0055] This tiered alarm mechanism optimizes the allocation of monitoring resources, avoiding a one-size-fits-all response to all changes. Specifically, the evolution of ecological patterns typically follows two modes: one is gradual natural succession or slow human interference, and the other is sudden disasters or drastic development activities. The first-level alarm threshold corresponds to the former. When the absolute deviation value falls within this range, the system determines it as a pattern trend warning, only pushing a notification to the monitoring terminal for managers to monitor periodically or include in long-term assessment reports, without requiring immediate emergency action. The second-level alarm threshold corresponds to the latter. When the absolute deviation value exceeds this upper limit, the system determines it as a pattern abrupt change alarm. This not only signifies a qualitative change in the ecological situation but also requires immediate verification of the data's authenticity and the on-site situation. Therefore, while generating an alarm, the system automatically triggers encrypted observation commands for the target area. For example, this command can control remote sensing satellites or drones to acquire images of the area at higher frequencies and resolutions, or activate ground-based IoT sensors to enter high-frequency reporting mode. Through this linkage mechanism, the system can automatically increase the granularity of perception as soon as an anomaly is confirmed, providing more detailed data support for subsequent emergency response or verification. It should be understood that the specific form of encrypted observation commands is not limited to adjusting remote sensing acquisition parameters. In other embodiments, it could also be dispatching nearby mobile inspection equipment to the site for verification, or sending high-priority instant messaging messages to relevant management departments, as long as it can achieve enhanced perception and rapid response to sudden events. This hierarchical response strategy avoids "alarm fatigue" caused by frequent false alarms and ensures that no reports are missed or delayed at critical moments, significantly improving the practical effectiveness of the ecological pattern monitoring system.

[0056] Example 6

[0057] Building upon Embodiment 4 or Embodiment 5, this embodiment further illustrates the dynamic update mechanism of historical baseline pattern indicators. Specifically, the system also includes a baseline update module, used to continuously acquire multi-source remote sensing data according to a preset time period and store the scale-normalized landscape pattern indicators corresponding to each time period in a time-series index database; when the scale-normalized landscape pattern indicators of multiple consecutive time periods show a monotonic change trend, the historical baseline pattern indicators are updated to the moving average of the scale-normalized landscape pattern indicators of the most recent preset number of time periods; the updated historical baseline pattern indicators are used as the deviation comparison benchmark for subsequent time periods.

[0058] This dynamic update mechanism aims to address the problem of fixed benchmarks becoming outdated due to the natural evolution of ecological patterns during long-term monitoring. In traditional static benchmark models, historical benchmark indicators are often set to values ​​from a specific historical moment and remain unchanged. However, ecosystems are complex, dynamically evolving systems. For example, vegetation cover fluctuates seasonally, or urbanization leads to a gradual increase in the proportion of impervious surfaces. These are normal ecological successions or gradual changes within planning expectations, not sudden ecological damage. If benchmark values ​​are not updated for a long time, as the actual pattern gradually deviates from the initial benchmark, the deviation accumulates and eventually triggers false alarms, causing the monitoring system to lose sensitivity to real anomalies. This embodiment introduces a trend-triggered adaptive update strategy, enabling historical benchmark indicators to smoothly adjust to the natural drift of ecological patterns, thereby maintaining the effectiveness of alarm thresholds.

[0059] Among these conditions, the key prerequisite for triggering baseline updates is that "the scale-normalized landscape pattern index exhibits a monotonic trend over multiple consecutive time periods." This condition is designed to significantly mitigate fluctuations. Specifically, the system does not update the baseline every time a change in the index is detected. Instead, it requires the index to maintain a consistent evolutionary trend over several consecutive periods. For example, a stable evolution signal can be defined as the scale-normalized landscape pattern index showing an increasing or decreasing trend over three or five consecutive monitoring periods. This strict monotonicity constraint effectively filters out random oscillations caused by meteorological fluctuations, temporary sensor malfunctions, or short-term human activities. Resetting the baseline based solely on single-period or short-term index fluctuations can easily lead to frequent jumps in the baseline value, masking genuine anomalous signals. It should be understood that in practical engineering implementation, "monotonic change" is not limited to strict mathematical monotonicity; minor reverse fluctuations within a preset tolerance range are also permissible, as long as the overall trend is significant and continuous, to avoid disrupting the normal baseline evolution process due to individual noise points.

[0060] Once the above-mentioned monotonic change condition is met, the system updates the historical baseline pattern index to the moving average of the scale-normalized landscape pattern index over the most recent preset number of time periods. For example, in a preferred embodiment, the preset number can be set to 5, meaning the average of the index over the most recent 5 periods is used as the new baseline value. Using a moving average instead of directly taking the latest value as the baseline is to further smooth out high-frequency noise in the data, ensuring that the new baseline represents a stable state level over a period of time, rather than a snapshot of a moment. Simultaneously, the length of the moving window (i.e., the preset number) can be configured differently according to the ecological type characteristics of the target area. For farmland or grassland ecosystems with significant seasonal changes, the number of moving window periods can be appropriately increased to cover the complete phenological cycle; while for relatively stable urban built-up areas or water bodies, the number of moving window periods can be appropriately decreased to improve the baseline's response speed to slow changes. Furthermore, in addition to simple arithmetic averages, weighted moving averages or exponentially weighted moving averages can also be used in other embodiments, giving higher weight to recent data to improve the timeliness of baseline updates while ensuring stability.

[0061] Through this mechanism, the updated historical baseline pattern indicators are automatically replaced in the comparison reference system of the artifact removal and alarm modules for deviation calculation in subsequent time periods. This gives the entire monitoring system self-evolution capabilities, enabling it to maintain accurate differentiation between "abnormal changes" and "normal evolution" throughout operating cycles of several years or even decades. For example, in the long-term monitoring of a wetland reserve, as ecological restoration projects progress, the wetland area steadily recovers year by year, and the scale-normalized landscape pattern indicators show a monotonous upward trend for several consecutive cycles. At this time, the baseline update module automatically raises the historical baseline value to match the current restoration results. In this way, the system will not falsely report "pattern abrupt changes" due to positive changes brought about by restoration. At the same time, if illegal encroachment suddenly occurs during this period, causing a sharp drop in indicators, the deviation of this drop relative to the new baseline will be more significant because the baseline has been updated to a higher level, making it easier to be captured by the secondary alarm threshold. This adaptive adjustment mechanism significantly improves the robustness and practicality of the ecological pattern monitoring system in long-term applications.

[0062] Example 7

[0063] This embodiment provides a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement a real-time monitoring and protection management system for regional ecological spatial patterns as described in any one of embodiments 1 to 6. Specifically, this computer device constitutes the physical operating carrier of the aforementioned virtual functional modules. In actual deployment scenarios, this computer device can be a high-performance server deployed in a data center for processing large-scale regional ecological pattern historical retrospectives and model training; it can also be an edge computing node integrated into a field monitoring station, utilizing local computing power to perform data alignment and artifact removal logic in real time, thereby reducing dependence on cloud communication bandwidth; or it can be a portable workstation or mobile terminal used by researchers for real-time data analysis during on-site verification. Regardless of the hardware form adopted, as long as its internal processor can complete the complete data processing flow from multi-source remote sensing data acquisition, ecological function equivalence aggregation, scale normalization correction to resolution artifact recognition and hierarchical alarm generation by calling program instructions in the memory, it falls within the protection scope of this application. This hardware-level implementation ensures that the technical solution can be transformed from abstract algorithmic logic into a manufacturable and marketable physical product, providing a clear legal basis for combating hardware infringement.

[0064] This embodiment also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements a real-time monitoring and protection management system for regional ecological spatial patterns as described in any one of embodiments 1 to 6. Specifically, the computer-readable storage medium refers to any non-transient medium capable of persistently storing data and being readable by a computer system. For example, the medium may include read-only memory (ROM), random access memory (RAM), disks (such as hard disk drives and floppy disks), optical disks (such as CD-ROMs and DVDs), flash memory devices (such as USB flash drives and SSDs), etc. It should be noted that the computer-readable storage medium referred to in this application explicitly excludes transient transmission signals such as carrier waves and electromagnetic signals, and only covers media with physical entities or stable storage states. By storing the computer program in the above-mentioned medium, the technical solution can be independently distributed, version-managed, and deployed across platforms without the need for specific hardware devices. For example, software vendors can burn the program containing the core algorithm of the monitoring system onto an optical disc for sale, or download it to the user's local storage device via the Internet for installation and upgrade. When a user connects the medium to a computer device and triggers the program execution, the ecological pattern monitoring function described in the aforementioned embodiments can be reproduced on general-purpose hardware. This implementation method improves the entire chain of intellectual property protection system from methods and systems to software products.

[0065] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A regional ecological spatial pattern big data real-time monitoring and protection management system, characterized in that, include: The data alignment module is used to acquire multi-source remote sensing data of the target area and determine the reference spatial resolution of the reference remote sensing image and the real-time spatial resolution of the real-time remote sensing image. In response to the real-time spatial resolution being higher than the reference spatial resolution, spatial aggregation is performed on the fragmented ecological patches in the real-time remote sensing image based on the ecological function equivalence condition, and the data is resampled to the reference spatial resolution to obtain scale-aligned raster data; The scale normalization module is used to calculate real-time landscape pattern indicators based on the scale-aligned raster data, obtain scale response coefficients that characterize the relationship between the real-time spatial resolution and the index response under the reference spatial resolution, and correct the real-time landscape pattern indicators based on the scale response coefficients to obtain scale-normalized landscape pattern indicators. The artifact removal and alarm module is used to determine the original landscape pattern indicators of the real-time remote sensing image before spatial aggregation, and to identify resolution artifacts based on the first deviation direction between the original landscape pattern indicators and the historical benchmark pattern indicators, and the second deviation direction between the scale-normalized landscape pattern indicators and the historical benchmark pattern indicators. In response to the failure to identify the resolution artifacts, the scale-normalized landscape pattern index is compared with the historical baseline pattern index, and an alarm is generated when the deviation exceeds a preset alarm threshold.

2. The system according to claim 1, characterized in that, The data alignment module performs spatial aggregation of fragmented ecological patches in the real-time remote sensing image based on ecological function equivalence conditions, including: The actual area of ​​each connected patch in the real-time remote sensing image is calculated, and the connected patches whose actual area is less than the functional equivalent area threshold are identified as the fragmented ecological patches. The largest adjacent patch among the fragmented ecological patches is identified as the dominant ecological patch. The pixel category of the fragmented ecological patches is updated to the ecological type of the dominant ecological patch, so that the fragmented ecological patches are merged into the dominant ecological patch.

3. The system according to claim 2, characterized in that, The process of determining the functional equivalent area threshold includes: The area of ​​a single pixel at the reference spatial resolution is calculated as the reference pixel area; Obtain the minimum identifiable area corresponding to each ecological type; The larger value between the reference pixel area and the minimum identifiable area of ​​the corresponding ecological type is used as the functional equivalent area threshold for that ecological type. The step of updating the pixel category of the fragmented ecological patches to the ecological type of the dominant ecological patch includes: Determine whether the ecological type of the fragmented ecological patches is consistent with that of the dominant ecological patches; If they match, the cell category of the fragmented ecological patch is reassigned to the ecological type of the dominant ecological patch, so that the fragmented ecological patch is incorporated into the dominant ecological patch in terms of connectivity. If there is a discrepancy, the pixel category of the fragmented ecological patch is reassigned to the ecological type of the dominant ecological patch, so that the fragmented ecological patch is merged into the dominant ecological patch.

4. The system according to claim 1, characterized in that, The scale normalization module obtains the scale response coefficients, including: Starting from the reference spatial resolution, multiple simulated resolution levels are generated by increasing the step size according to a preset scale; The real-time remote sensing images are resampled to each of the simulated resolution levels, and the corresponding simulated landscape pattern index values ​​are calculated at each of the simulated resolution levels. A scale response curve is constructed based on the simulated resolution level and the simulated landscape pattern index value; Read the first simulated landscape pattern index value corresponding to the real-time spatial resolution and the second simulated landscape pattern index value corresponding to the reference spatial resolution from the scale response curve. The ratio of the first simulated landscape pattern index value to the second simulated landscape pattern index value is used as the scale response coefficient.

5. The system according to claim 4, characterized in that, The construction of the scale response curve based on the simulated resolution level and the simulated landscape pattern index value includes: Calculate the landscape fragmentation index, patch connectivity index, and Shannon diversity index at each of the simulated resolution levels. After performing logarithmic transformations on the landscape fragmentation index, the patch connectivity index, and the Shannon diversity index, linear regression fitting is performed with the logarithmic value of the simulated resolution level as the independent variable to obtain the scale response regression equations corresponding to each index. Calculate the first predicted value and the second predicted value of each scale response regression equation at the real-time spatial resolution and the reference spatial resolution respectively, and use the ratio of the first predicted value to the second predicted value of each index as the single scale response coefficient corresponding to each index. The weighted average of each individual scale response coefficient is used to obtain the comprehensive scale response coefficient, which is then used as the scale response coefficient. The weights of the weighted average are based on the preset sensitivity of each indicator to changes in the ecological pattern.

6. The system according to claim 1, characterized in that, The artifact removal and alarm module identifies resolution artifacts based on the first deviation direction between the original landscape pattern index and the historical baseline pattern index, and the second deviation direction between the scale-normalized landscape pattern index and the historical baseline pattern index, including: Calculate the first deviation direction between the original landscape pattern index and the historical baseline pattern index, and the second deviation direction between the scale-normalized landscape pattern index and the historical baseline pattern index; If the direction of the first deviation is opposite to the direction of the second deviation, the current indicator change is determined to be the resolution artifact, and the scale-normalized landscape pattern indicator is corrected to the historical baseline pattern indicator before the alarm is skipped. If the first deviation direction is the same as the second deviation direction, it is determined that the resolution artifact has not been identified, the scale-normalized landscape pattern index is retained, and the deviation comparison is performed.

7. The system according to claim 6, characterized in that, It also includes a dynamic monitoring window determination module, used for: Sample pixels are extracted from the scale-aligned raster data at a preset sampling interval, and the semivariogram values ​​of the sample pixels are calculated at different hysteresis distances. The semivariogram values ​​are fitted with a spherical model, and the range values ​​of the fitted curve are extracted as the range of the semivariogram. The semi-variogram range is rounded up to an integer multiple of the reference spatial resolution to obtain the initial window side length; Obtain the available memory capacity of the current edge computing node, calculate the maximum number of raster cells that the available memory capacity can support, and determine the memory constraint side length based on the maximum number of raster cells and the reference spatial resolution; The smaller value between the initial window side length and the memory constraint side length is used as the window side length of the dynamic monitoring window; The artifact removal and alarm module generates an alarm when the deviation exceeds a preset alarm threshold, including: Calculate the absolute deviation between the scale-normalized landscape pattern index and the historical baseline pattern index; The absolute deviation value is compared with the first-level alarm threshold and the second-level alarm threshold; the first-level alarm threshold is less than the second-level alarm threshold; If the absolute deviation value is greater than the first-level alarm threshold and less than or equal to the second-level alarm threshold, a pattern trend warning is generated. If the absolute deviation value is greater than the second-level alarm threshold, a pattern change alarm is generated, and an encrypted observation command for the target area is triggered.

8. The system according to claim 6, characterized in that, It also includes a benchmark update module, used for: The multi-source remote sensing data is continuously acquired according to a preset time period, and the scale-normalized landscape pattern index corresponding to each time period is stored in the time series index database. When the scale-normalized landscape pattern index shows a monotonic trend over multiple consecutive time periods, the historical baseline pattern index is updated to the moving average of the scale-normalized landscape pattern index over the most recent preset number of time periods. The updated historical benchmark pattern indicators are used as the deviation comparison benchmark for subsequent time periods.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements a regional ecological spatial pattern big data real-time monitoring and protection management system as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the regional ecological spatial pattern big data real-time monitoring and protection management system as described in any one of claims 1 to 8.

Images