A method and system for evaluating the effectiveness of ecological restoration applied to open-pit mines in karst areas

Abstract

The application discloses a kind of applied karst area open-pit mine ecological restoration effect evaluation method and system, it is related to ecological restoration technical field, including dividing multiple succession stages, matching the specific regional position corresponding to each succession stage.The stage time node corresponding to different succession stages is reasonably set, so that succession stage division is more accurate, and the specific position of the region is determined considering dynamic evolution, to provide a reliable basis for subsequent regional boundary processing and spatial heterogeneity analysis.According to the image features, the initial fuzzy boundary of different landform types is refined, and the reliability of different landform boundary division is improved.A heterogeneity correction model is constructed to ensure the reliability of the multi-dimensional evaluation system.Horizontal and vertical comparisons are made to determine the ecological restoration effectiveness index, taking into account the evaluation differences in different landform regions and the evaluation differences in different succession stages.

Description

Technical Field

[0001] This invention relates to the field of ecological restoration technology, and in particular to a method and system for evaluating the effectiveness of ecological restoration in open-pit mines in karst areas. Background Technology

[0002] Open-pit mining activities in karst areas cause severe damage to the ecological environment. The unique geological conditions lead to particularly prominent problems such as land degradation, vegetation destruction, soil erosion, and decreased biodiversity. Due to the abundance of groundwater resources but scarcity of surface water in karst areas, mining activities further exacerbate the uneven distribution of water resources and pollution problems. At the same time, steep slopes and waste dumps are prone to triggering geological disasters such as landslides and debris flows, threatening regional ecological security.

[0003] Ecological restoration effectiveness assessment is a core component of ecological restoration management in karst open-pit mines. It is used to quantify key indicators such as vegetation restoration, soil stability, and hydrological improvement, providing a scientific basis for optimizing restoration strategies and accurately allocating resources, and ensuring the effective achievement of ecological goals.

[0004] In existing technologies, the assessment of the effectiveness of ecological restoration is based solely on the current ecological situation, without taking into account the past ecological foundation and the complex spatial heterogeneity effects under karst areas. This results in poor accuracy and adaptability of the assessment of the effectiveness of ecological restoration, and fails to meet the planning needs of ecological restoration.

[0005] Therefore, improving the accuracy and adaptability of ecological restoration effectiveness assessment is a technical problem that needs to be solved. Summary of the Invention

[0006] The purpose of this invention is to address the problem of poor accuracy and adaptability in the evaluation of ecological restoration effectiveness in existing technologies that fail to consider the prior ecological foundation and the complex spatial heterogeneity effects in karst areas. Therefore, this invention proposes a method for evaluating the effectiveness of ecological restoration in open-pit mines in karst areas, comprising:

[0007] Determine the geographical location of the open-pit mine in the karst area, and based on the geographical location of the open-pit mine in the karst area, divide it into multiple successional stages, and match the specific geographical location corresponding to each successional stage;

[0008] According to the specific regional locations corresponding to different succession stages, landform types, geographical features and image features are obtained. The specific regional locations are decomposed based on landform types and geographical features to obtain the initial fuzzy boundaries of different landform types. The initial fuzzy boundaries of different landform types are refined based on image features to obtain the refined boundaries of different landform types.

[0009] Based on the refined boundary analysis of different landform types, the spatial heterogeneity effect under the refined boundary of different landform types is analyzed to construct a heterogeneity correction model and to construct a multi-dimensional evaluation system for open-pit mines in karst areas.

[0010] Based on multiple succession stages, horizontal and vertical comparisons are conducted to determine indicators of ecological restoration effectiveness, thereby quantitatively describing the results of ecological restoration of open-pit mines in karst areas.

[0011] In some embodiments of this application, multiple successional stages are divided based on the geographical location of the open-pit mine in the karst area, including:

[0012] The multiple succession stages, in chronological order, include the original stage of karst areas, the mining stage, the mine abandonment stage, and the ecological restoration stage.

[0013] Obtain information on mining engineering and ecological restoration projects within the geographical location of the mine in the karst area, and use this information to determine the mining timeline, mine closure timeline, and ecological restoration start timeline.

[0014] A timeline for the succession stages of open-pit mines in karst areas was constructed, and the time nodes for mine operation, mine closure, and ecological restoration commencement were marked on the timeline. The NDVI of the continuous time series on the timeline was statistically analyzed. An NDVI threshold was set using the Mann-Kendall test. Based on the NDVI threshold, NDVI mutation points near each of the mine operation, mine closure, and ecological restoration commencement time nodes were screened. The NDVI mutation points near each of the mine operation, mine closure, and ecological restoration commencement time nodes were combined to determine the corresponding mutation time nodes for each time node.

[0015] A lag analysis was performed on the NDVI mutation points near the mining start date, mine closure date, and ecological restoration start date to generate lag durations. The mutation dates were adjusted based on the lag durations. By combining the mining start date, mine closure date, ecological restoration start date, and their corresponding adjusted mutation dates, new mining start dates, mine closure dates, and ecological restoration start dates were determined. The stage time points for the original stage, mining stage, mine closure stage, and ecological restoration stage of the karst area were also determined.

[0016] In some embodiments of this application, the specific regional location corresponding to each successional stage is matched, including,

[0017] Within the geographical range of the open-pit mine in the karst area, the edge corresponding to each successional stage is determined. According to the different NDVI ranges, the edge corresponding to each successional stage is divided into multiple vegetation cover zones of different levels. The migration rate corresponding to different vegetation cover zones under different successional stages is calculated. The edge is dynamically adjusted by combining the migration rates of all vegetation cover zones to determine the specific geographical location corresponding to each successional stage.

[0018] In some embodiments of this application, specific regional locations are segmented based on landform type and geographical features to obtain initial blurred boundaries for different landform types. These initial blurred boundaries are then refined based on image features to obtain refined boundaries for different landform types.

[0019] In GIS software, spatial partitioning is performed on a specific location based on thresholds of multiple geographic features and landform types to obtain multiple landform type regions and their corresponding initial fuzzy boundaries. Image features are then matched with these multiple landform type regions and their corresponding initial fuzzy boundaries according to geographic location.

[0020] Edge extraction is performed on the initial blurred boundary and the image features of the surrounding area using the HED model, and a probability boundary map is output.

[0021] The initial fuzzy boundary and its surrounding area are marked as the first region, and the area on the landform type zone other than the first region is marked as the second region. The differences in geographical features within the first region and the differences in geographical features between the second regions on both sides of the initial fuzzy boundary are statistically analyzed and recorded as the first geographical feature difference and the second geographical feature difference, respectively. Based on the landform type on both sides of the initial fuzzy boundary, the first geographical feature difference and the second geographical feature difference, different probability thresholds for the initial fuzzy boundary are set, and the segmentation boundary is preserved in the probability boundary map based on the probability thresholds.

[0022] Multiple karst landforms are selected for each landform type, and karst landform templates are configured according to the parameter ranges of the previous geographical features and image features of each karst landform. The karst landform templates are used to traverse the segmentation boundary and aggregate the matching points into a continuous boundary.

[0023] Then, multi-evidence chain fusion verification is performed on the continuous boundary to complete the boundary refinement process and obtain refined boundaries for different landform types. Among them, the multi-evidence chain includes verification dimensions such as LiDAR elevation, rock index and temporal consistency.

[0024] In some embodiments of this application, a heterogeneity correction model is constructed by analyzing the spatial heterogeneity effect under the refined boundaries of different landform types, based on the refined boundary analysis of different landform types.

[0025] We obtain geomorphic indicators, ecological indicators, and geomorphic gradients for different landform types. We calculate the geomorphic influence intensity coefficient based on the geomorphic indicators and calculate the geomorphic gradient sensitivity coefficient by fitting the correlation between ecological indicators and geomorphic gradients.

[0026] In some embodiments of this application, a heterogeneity correction model is constructed, including...

[0027] A heterogeneity correction model is constructed based on the geomorphic influence intensity coefficient, geomorphic gradient sensitivity coefficient, and geomorphic gradient.

[0028] In some embodiments of this application, a multi-dimensional evaluation system for open-pit mines in karst areas is constructed, including,

[0029] Based on the logical stratification of ecological processes in karst areas, multiple directional dimensions are determined. Evaluation indicators are selected under each directional dimension to construct a dedicated indicator library for the ecological restoration of open-pit mines in karst areas. The evaluation indicators are dynamically updated through a heterogeneity correction model, thereby updating the dedicated indicator library.

[0030] In some embodiments of this application, horizontal and vertical comparisons are performed, including...

[0031] Horizontal comparison refers to the differences in the same evaluation index between different landform types at the same successive stage, generating horizontal difference values.

[0032] Longitudinal comparison generates longitudinal difference values ​​by comparing the differences of the same evaluation index between different successional stages within the same geomorphic type region.

[0033] In some embodiments of this application, indicators for determining the effectiveness of ecological restoration include,

[0034] Under a multi-dimensional evaluation system, the effectiveness indicators of ecological restoration are determined by combining the horizontal and vertical differences of multiple evaluation indicators.

[0035] Correspondingly, this application also provides an effectiveness evaluation system for ecological restoration of open-pit mines in karst areas, including,

[0036] The first module is used to determine the geographical location of open-pit mines in karst areas, and to divide the geographical location of open-pit mines in karst areas into multiple successional stages, matching the specific geographical location corresponding to each successional stage;

[0037] The second module is used to obtain landform types, geographical features and image features according to the specific regional locations corresponding to different succession stages. It uses landform types and geographical features to decompose the specific regional locations to obtain the initial fuzzy boundaries of different landform types. Based on the image features, it performs boundary refinement processing on the initial fuzzy boundaries of different landform types to obtain the refined boundaries of different landform types.

[0038] The third module is used to analyze the spatial heterogeneity effect under the refined boundary of different landform types based on the refined boundary of different landform types, to construct a heterogeneity correction model, and to construct a multi-dimensional evaluation system for open-pit mines in karst areas.

[0039] The fourth module is used to conduct horizontal and vertical comparisons based on multiple succession stages to determine the indicators of ecological restoration effectiveness, thereby quantitatively describing the results of ecological restoration of open-pit mines in karst areas.

[0040] Compared with the prior art, the beneficial effects of this invention are as follows:

[0041] 1. Based on the geographical location of the open-pit mine in the karst area, multiple successional stages were defined, and the specific geographical location corresponding to each successional stage was matched. The time nodes corresponding to different successional stages were reasonably set to make the successional stage division more accurate. Dynamic evolution was considered to determine the specific location of the region, providing a reliable basis for subsequent regional boundary processing and spatial heterogeneity analysis. Initial blurred boundaries of different landform types were refined based on image features, improving the reliability of distinguishing different landform types and increasing the resolution of boundaries and landform types.

[0042] 2. Based on the refined boundary analysis of different landform types, a spatial heterogeneity correction model is constructed to address the spatial heterogeneity effects under these refined boundaries. This model corrects for the nonlinear impact of the spatial heterogeneity of complex landforms in karst areas on the evaluation indicators, ensuring the reliability of the multi-dimensional evaluation system. Horizontal and vertical comparisons are conducted to determine ecological restoration effectiveness indicators. Combining evaluation differences across different landform regions and successional stages, and using the ecology of previous successional stages as a foundation, ecological restoration effectiveness indicators are established to improve the accuracy and adaptability of ecological restoration effectiveness assessment and meet the planning needs of ecological restoration. Attached Figure Description

[0043] Figure 1 This is a flowchart illustrating the effectiveness evaluation method for ecological restoration of open-pit mines in karst areas proposed in this invention.

[0044] Figure 2 This is a schematic diagram of the structure of an effectiveness evaluation system for ecological restoration of open-pit mines in karst areas, as proposed in this invention. Detailed Implementation

[0045] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0046] Reference Figure 1An effectiveness evaluation method for ecological restoration of open-pit mines in karst areas, comprising the following steps:

[0047] Step S101: Determine the geographical location of the open-pit mine in the karst area, and divide the geographical location of the open-pit mine in the karst area into multiple successional stages, and match the specific geographical location corresponding to each successional stage.

[0048] In this embodiment, a general area for the open-pit mine in the karst region is determined, with the area being as large as possible to avoid omissions. The boundary of the karst region is determined based on the definition of karst landforms, using geological maps and karst development maps overlaid to extract the boundary (e.g., limestone distribution areas). The initial mine area is determined by combining historical remote sensing imagery and mining permits to define the original geomorphological extent before mining operations. The coordinate system is uniformly adopted using the WGS84 coordinate system to ensure spatial analysis accuracy.

[0049] Different successional stages exhibit varying regional variations, potentially altering their specific locations. Previously, defining successional stages solely based on engineering time or NDVI abrupt change time was insufficient to pinpoint an accurate stage node. A more representative stage node can be derived by combining both engineering time and NDVI abrupt change time. Furthermore, considering the variations within each successional stage allows for the identification of the specific regional location corresponding to each stage.

[0050] In some embodiments of this application, multiple successional stages are divided based on the geographical location of the open-pit mine in the karst area, including:

[0051] The multiple succession stages, in chronological order, include the original stage of karst areas, the mining stage, the mine abandonment stage, and the ecological restoration stage.

[0052] Obtain information on mining engineering and ecological restoration projects within the geographical location of the mine in the karst area, and use this information to determine the mining timeline, mine closure timeline, and ecological restoration start timeline.

[0053] A timeline for the succession stages of open-pit mines in karst areas was constructed, and the time nodes for mine operation, mine closure, and ecological restoration commencement were marked on the timeline. The NDVI of the continuous time series on the timeline was statistically analyzed. An NDVI threshold was set using the Mann-Kendall test. Based on the NDVI threshold, NDVI mutation points near each of the mine operation, mine closure, and ecological restoration commencement time nodes were screened. The NDVI mutation points near each of the mine operation, mine closure, and ecological restoration commencement time nodes were combined to determine the corresponding mutation time nodes for each time node.

[0054] A lag analysis was performed on the NDVI mutation points near the mining start date, mine closure date, and ecological restoration start date to generate lag durations. The mutation dates were adjusted based on the lag durations. By combining the mining start date, mine closure date, ecological restoration start date, and their corresponding adjusted mutation dates, new mining start dates, mine closure dates, and ecological restoration start dates were determined. The stage time points for the original stage, mining stage, mine closure stage, and ecological restoration stage of the karst area were also determined.

[0055] In this embodiment, the sequence is: Karst area initial stage - (new mining timeline) - mining stage - (new mine closure timeline) - mine closure stage - (new ecological restoration start timeline) - ecological restoration stage. Mine engineering information and ecological restoration engineering information include, but are not limited to, the following:

[0056] Starting point of mining phase: from the date of issuance of the mining license (e.g., March 15, 2005), combined with the start time of mining activities recorded in local chronicles.

[0057] The starting point for the mine closure phase is the date of the mine closure announcement (e.g., December 1, 2015), combined with the company's production stoppage report.

[0058] Starting point of the ecological restoration phase: from the date of approval of the ecological restoration plan (e.g., June 1, 2016), combined with official announcements.

[0059] NDVI (Normalized Difference Vegetation Index) is one of the most commonly used vegetation indices in remote sensing, used to quantify land cover and vegetation growth status. NDVI calculations for continuous time series using ENVI or QGIS, and abrupt change detection are also performed.

[0060] Mann-Kendall test: to identify significant changes in the NDVI time series (e.g., α=0.05).

[0061] Sliding window method: Calculate the NDVI change rate within a time window. A change rate > 0.05 is marked as a mutation point. The mutation point marks the beginning of change reflected in NDVI. However, NDVI is lagging, and this lag directly affects the accuracy of stage boundary time points in the ecological restoration assessment of open-pit mines in karst areas. Causes of this lag include vegetation growth cycles, remote sensing image acquisition cycles, and the interaction between engineering activities and vegetation responses. For example, in karst areas, soil is infertile and water is easily lost (soil thickness in caves and depressions < 0.5m), resulting in slow vegetation recovery (e.g., herbaceous plants require 3-6 months, shrubs require more than 12 months). The lag time range is 1.5-3 months for NDVI decrease due to mining and 4-8 months for NDVI increase due to restoration.

[0062] Lag duration calculation: Based on the hydrological-vegetation response patterns in karst areas (such as the delayed impact of rainy season precipitation on NDVI), historical data (2010-2015) were analyzed.

[0063] During the mining phase: NDVI decline lags behind the start of mining by 2 months (because soil erosion takes time to manifest).

[0064] Abolition phase: NDVI recovery lags behind pit closure by 3 months (due to the time required for natural vegetation recovery).

[0065] Repair phase: NDVI recovery lags behind the start of repair by 1 month (due to the buffer required for engineering measures to take effect).

[0066] The NDVI mutation points near the respective mining start date, mine closure date, and ecological restoration start date are comprehensively analyzed (multiple NDVI mutation points are compared, and representative points or intermediate points are selected as mutation time points) to determine the corresponding mutation time points for each time point. Based on this, the mutation time points are adjusted to achieve reasonable time adjustments. Combining the mining start date, mine closure date, ecological restoration start date, and their corresponding adjusted mutation time points, new mining start dates, mine closure dates, and ecological restoration start dates are determined. Combining the project time point and the mutation time point, a weighted average is calculated between the two time points, and the midpoint of this weighted average is taken as the new time point.

[0067] In some embodiments of this application, the specific regional location corresponding to each successional stage is matched, including,

[0068] Within the geographical range of the open-pit mine in the karst area, the edge corresponding to each successional stage is determined. According to the different NDVI ranges, the edge corresponding to each successional stage is divided into multiple vegetation cover zones of different levels. The migration rate corresponding to different vegetation cover zones under different successional stages is calculated. The edge is dynamically adjusted by combining the migration rates of all vegetation cover zones to determine the specific geographical location corresponding to each successional stage.

[0069] In this embodiment, based on the succession stage time points determined in the preceding steps (e.g., original stage: 2010.01–2012.04; mining stage: 2012.04–2015.08, etc., the time can be further specified), monthly average images (cloud cover <10%, resolution 10m) for the corresponding time period are extracted from the Sentinel-2 remote sensing database. In a GIS (e.g., QGIS), the images for each stage are preprocessed (atmospheric correction, cloud masking) to generate stage-specific NDVI raster layers. For example (some reference values ​​are below):

[0070] Original phase: Average NDVI chart from 2010 to 2012 (baseline value 0.55 ± 0.05)

[0071] Mining phase: Average NDVI chart from 2012 to 2015 (average value 0.08 ± 0.03)

[0072] Abolishing phase: Average NDVI chart from 2015 to 2016 (mean 0.15 ± 0.04)

[0073] Repair phase: Average NDVI plot from 2016 to present (average 0.30±0.05)

[0074] Karst area adaptation: The NDVI threshold is set based on empirical data of karst areas (e.g., a high coverage threshold of 0.4 in the original stage and a low coverage threshold of 0.2 in the mining stage) to avoid misjudgment caused by a general threshold.

[0075] For each stage of the NDVI raster layer, it is classified into different levels according to the karst area specialization threshold:

[0076] High coverage area (vegetation coverage area) ≥0.4 (NDVI range) Karst plain (soil thickness >0.5m, NDVI stable) (typical characteristics of karst area).

[0077] Medium cover area: 0.2–0.4 peak forest steep slopes / karst cave depressions (vegetation fluctuates greatly, NDVI seasonal variation ±0.15).

[0078] Low cover area <0.2 karst depression (soil exposed, NDVI <0.1).

[0079] In GIS, use a "reclassification" tool (such as ArcGIS's Reclassify) to map NDVI values ​​to the above range and generate a coverage raster (e.g., low coverage areas during the mining phase account for 65% of the total area).

[0080] The migration rate of coverage areas between adjacent stages is calculated using spatial cross-analysis (migration rate = transferred area / original area × 100%).

[0081] Example (Mining Phase → Abandonment Phase):

[0082] Area of ​​low-coverage zone (NDVI < 0.2) during the mining phase: 500 km²

[0083] Area of ​​low-coverage zone from the mining phase within the decommissioning phase (0.2–0.4) of the covered area: 200 km²

[0084] Mobility = (200 / 500) × 100% = 40%

[0085] Mobility matrix: Generates a mobility table between each stage (e.g., mining → abandonment: low → medium coverage 40%; abandonment → repair: medium → high coverage 35%).

[0086] Karst region adaptation: When calculating the migration rate, karst landform interference is eliminated (e.g., the migration rate of areas with a cave density > 3 caves / km² is calculated separately) to avoid errors caused by landform gradients.

[0087] Dynamic weight allocation: Migration rate weights are assigned to areas with different coverage rates (0.3 for high coverage areas, 0.5 for medium coverage areas, and 0.2 for low coverage areas), based on the ecological response intensity of karst areas (e.g., high weight for low coverage areas due to faster migration).

[0088] Boundary adjustment logic:

[0089] If the migration rate of low-coverage areas in a certain stage is >30% (e.g., 40% in the mining → abandonment stage), then the boundary is extended towards the low-coverage areas (the boundary of the abandonment stage extends 10% towards the low-coverage areas in the mining stage).

[0090] If the migration rate of high coverage areas is less than 10% (e.g., 15% in the abolishment → repair phase), then the boundary will shrink towards the high coverage areas (the boundary in the repair phase will shrink by 5% towards the high coverage areas in the original phase).

[0091] Operation: In GIS, use "Buffer Analysis" and "Spatial Overlay" to dynamically update the boundaries according to the weight formula:

[0092] New boundary = Original boundary + ∑(mobility × weight) × buffer distance.

[0093] Step S102: Obtain landform types, geographical features, and image features according to the specific regional locations corresponding to different succession stages. Decompose the specific regional locations based on landform types and geographical features to obtain the initial fuzzy boundaries of different landform types. Refine the initial fuzzy boundaries of different landform types based on image features to obtain the refined boundaries of different landform types.

[0094] In this embodiment, the establishment of a clear boundary for different landform types is necessary to ensure the accuracy and reliability of data analysis in the subsequent spatial heterogeneity analysis and evaluation system. However, the boundary obtained solely from geographical features is blurry and lacks sufficient resolution to support subsequent boundary refinement. Image features (such as UAV imagery (0.1m) and airborne LiDAR point clouds (accuracy ±0.05m)) need to be added for analysis. However, image features encounter two problems when analyzing and processing boundaries: Problem 1: The contradiction between "small-scale landforms" and "large-scale images" exists, as the scale of caves / sinkholes is much smaller than the image resolution. Problem 2: The conflict between "algorithm generality" and "karst specificity" exists, as general edge detection models have a high false detection rate in karst areas.

[0095] In some embodiments of this application, specific regional locations are segmented based on landform type and geographical features to obtain initial blurred boundaries for different landform types. These initial blurred boundaries are then refined based on image features to obtain refined boundaries for different landform types.

[0096] In GIS software, spatial partitioning is performed on a specific location based on thresholds of multiple geographic features and landform types to obtain multiple landform type regions and their corresponding initial fuzzy boundaries. Image features are then matched with these multiple landform type regions and their corresponding initial fuzzy boundaries according to geographic location.

[0097] Edge extraction is performed on the initial blurred boundary and the image features of the surrounding area using the HED model, and a probability boundary map is output.

[0098] The initial fuzzy boundary and its surrounding area are marked as the first region, and the area on the landform type zone other than the first region is marked as the second region. The differences in geographical features within the first region and the differences in geographical features between the second regions on both sides of the initial fuzzy boundary are statistically analyzed and recorded as the first geographical feature difference and the second geographical feature difference, respectively. Based on the landform type on both sides of the initial fuzzy boundary, the first geographical feature difference and the second geographical feature difference, different probability thresholds for the initial fuzzy boundary are set, and the segmentation boundary is preserved in the probability boundary map based on the probability thresholds.

[0099] Multiple karst landforms are selected for each landform type, and karst landform templates are configured according to the parameter ranges of the previous geographical features and image features of each karst landform. The karst landform templates are used to traverse the segmentation boundary and aggregate the matching points into a continuous boundary.

[0100] Then, multi-evidence chain fusion verification is performed on the continuous boundary to complete the boundary refinement process and obtain refined boundaries for different landform types. Among them, the multi-evidence chain includes verification dimensions such as LiDAR elevation, rock index and temporal consistency.

[0101] In this embodiment, the following solutions are adopted to address the two problems involved in the image feature boundary refinement process in step S101.

[0102] Boundary oversegmentation extraction (following the principle of "better to oversegment than undersegment") includes two steps: general edge detection to generate initial boundaries and karst prior knowledge embedding optimization.

[0103] Step 1: Generate initial boundaries using general edge detection

[0104] The HED (Holistically-Nested Edge Detection) model (deep learning edge detection) is used to extract edges from drone images and output a probability boundary map (0-1 continuous values).

[0105] Key operations:

[0106] Retain all regions where the edge probability is greater than the threshold (even if they are suspected noise), and generate oversegmented boundaries (such as cave edges that are identified as multiple discontinuous line segments).

[0107] For example, in karst areas: for cave and depression areas, all points with an edge probability > 0.2 are forcibly retained (because the edges of karst landforms are often covered by vegetation, resulting in low probability values).

[0108] Step 2: Optimization of Karst Prior Knowledge Embedding

[0109] Superimpose karst landform priors into the HED output:

[0110] For example, to identify "ring shadow features" (a typical feature of caves): search for a ring-shaped dark area with a radius of <5m centered on the coordinates of the sinkhole in the image (low NIR band value + shadow index SI <0.2).

[0111] Use Hough Circle Transform to aggregate discontinuous line segments into a complete circular boundary.

[0112] ;

[0113] in, For the first The probability threshold of an initial fuzzy boundary. For the first The base probability corresponding to each initial fuzzy boundary. For the first Landform types on both sides of the initial fuzzy boundary , For the first The compensation probability corresponding to the landform type of the initial fuzzy boundary. This represents the different compensation probabilities obtained by mapping landform types. For example, the compensation probability for a cave-depression-karst plain is -0.1. The edges of cave-depressions are often covered by vegetation, resulting in a low probability value, and the threshold needs to be lowered to protect the boundary. For the first An initial fuzzy boundary involves the compensation probability of determining geographic feature differences (first geographic feature difference and second geographic feature difference, i.e., considering geographic feature differences within the boundary area and geographic feature differences between two adjacent areas). For the first The probability transformation coefficients corresponding to the initial fuzzy boundaries , The number of geographical features in the first and second regions. For the first The first region within the initial fuzzy boundary. The weight of each geographical feature For the first The first region within the initial fuzzy boundary. The differences in geographical features (the changes in geographical features within the boundary area) are standardized to facilitate summation. Similarly, For the first A preset constant for an initial fuzzy boundary, This represents the correction of the average case of the second difference feature to the average case of the first difference feature. The constant is used to control the range of the correction function.

[0114] Based on statistical analysis of field samples from karst areas (100+ typical samples), the parameter range covers 95% of karst landform features, avoiding misjudgment using general templates. Karst landform templates include cave-depression templates, peak-forest steep slope templates (vertical fissure features), and karst plain templates (smooth transition features). For example, the cave-depression template (ring-shaped dark area feature) has a ring radius, r. min =1m, r max =5m, 95% of the karst caves have a radius of 2-4m (statistical data from karst areas in Guizhou).

[0115] ;

[0116] in, For the number of testing points, For the first The shadow index of each detection point, shadow index =0.2, Cave shadow <0.2 (low value in NIR band) The expression AND logic means that both conditions must be met simultaneously. For the first Near-infrared reflectance at each detection point The near-infrared reflectance of the background area surrounding the ring is the average value, and the ring contrast is... =0.3, the contrast between the dark area of ​​the cave and its surroundings is >0.3 (NIR value difference >30%), for example, when the matching threshold is met. In the results with a resolution >0.7, the planar positioning error between LiDAR and RTK ground truth values ​​was < 2 m for 94% of the boundary points.

[0117] The core logic is to preset feature templates for different karst landforms (cave depressions, peak forests and steep slopes, karst plains), and automatically identify the boundary shape through matching degree calculation, avoiding misjudgment by general algorithms.

[0118] Engineering implementation process for formwork configuration in karst areas:

[0119] 1. Construction of a karst landform sample bank (key step):

[0120] Collected 100+ karst area samples (50 caves and depressions, 30 peak forests and steep slopes, and 20 karst plains).

[0121] Statistical parameter distribution:

[0122] Cave radius: 95% are within 2-4m → set r min =1m, r max =5m, maintaining a slightly larger coverage principle.

[0123] Peak forest steep slope angle: 90% at 87 degrees Celsius ° -93 ° →Set θ min =85 ° θ max =93 ° Maintain a slightly larger coverage area.

[0124] 2. Automatic template matching:

[0125] For each point that exceeds the split boundary in the HED output, traverse the points and calculate the matching degree with the three templates.

[0126] Decision-making rules:

[0127] If Matchcave >0.7Match cave >0.7 → Boundary of the karst depression

[0128] If Match peak >0.7Match peak >0.7 → Peak Forest Steep Slope Boundary

[0129] If Match plain >0.7Match plain >0.7 → Karst plain boundary

[0130] 3. Boundary polymerization forming:

[0131] Cave depressions: Use Hough Circle Transform to aggregate them into ring boundaries.

[0132] Peak forest steep slope: Vertical crack boundaries are aggregated using linear fitting.

[0133] Karst plains: The transition boundary is fitted with a smooth curve.

[0134] Multi-evidence chain fusion verification involves retaining only boundary regions that simultaneously meet all three conditions (LiDAR elevation, rock index, and temporal consistency), while removing over-segmentation points that do not meet the conditions (such as the edges of vegetation-covered areas). See Table 1 below.

[0135]

[0136] Table 1

[0137] The comparison of treatment process effects (actual test cases in karst areas) is shown in Table 2 below.

[0138]

[0139] Table 2

[0140] The beneficial effects of this step:

[0141] 1. All suspected boundaries are preserved by oversegmentation + karst prior (ring shadow), and then noise is removed by LiDAR elevation to avoid missing karst caves due to insufficient resolution.

[0142] 2. Embed prior knowledge of karst landforms (ring shadows, elevation changes) into the algorithm process, instead of using general edge detection (such as Canny).

[0143] Step S103: Based on the refined boundary analysis of different landform types, the spatial heterogeneity effect under the refined boundary of different landform types is analyzed to construct a heterogeneity correction model and construct a multi-dimensional evaluation system for open-pit mines in karst areas.

[0144] In this embodiment, spatial heterogeneity refers to the differentiated impact of the spatial distribution of different landform types (karst caves and depressions, peak forests and steep slopes, karst plains) on ecological indicators (such as NDVI and soil erosion rate) in karst areas. This impact is not a simple classification difference, but rather a nonlinear response of ecological indicators caused by the continuous change of landform gradients (such as cave density and slope). Nonlinear correction is applied by constructing a continuous function based on the landform gradient (cave density and slope) to reflect the true ecological response of karst areas.

[0145] In some embodiments of this application, a heterogeneity correction model is constructed by analyzing the spatial heterogeneity effect under the refined boundaries of different landform types, based on the refined boundary analysis of different landform types.

[0146] We obtain geomorphic indicators, ecological indicators, and geomorphic gradients for different landform types. We calculate the geomorphic influence intensity coefficient based on the geomorphic indicators and calculate the geomorphic gradient sensitivity coefficient by fitting the correlation between ecological indicators and geomorphic gradients.

[0147] In some embodiments of this application, a heterogeneity correction model is constructed, including...

[0148] A heterogeneity correction model is constructed based on the geomorphic influence intensity coefficient, geomorphic gradient sensitivity coefficient, and geomorphic gradient.

[0149] In this embodiment, the geomorphic influence intensity coefficient is determined through regression analysis. Geomorphic indicators include karst plain indicators and cave-depression indicators. Karst plain indicators: the baseline value of ecological indicators (i.e., ideal recovery rate) for karst plain areas (low geomorphic gradient), and the average annual growth rate of NDVI for karst plain areas (cavity density < 1 / km²). Cave-depression indicators: the measured value of ecological indicators (i.e., actual recovery rate after being affected by karst) for cave-depression areas (high geomorphic gradient), and the average annual growth rate of NDVI for cave-depression areas (cavity density > 3 / km²). Geomorphic influence intensity coefficient = (karst plain indicator - cave-depression indicator) / karst plain indicator.

[0150] The geomorphic gradient sensitivity coefficient reflects the sensitivity of geomorphic gradients (such as cave density) to ecological indicators. It needs to be fitted using empirical data from karst areas, rather than based on assumptions. Specific steps (taking the NDVI growth rate of cave depressions as an example):

[0151] Step 1: Collect data on the cave depression areas of 10 karst mines (Guizhou and Guangxi), record the cave density (number of caves / km2) and the average annual growth of NDVI (year-1) under multiple samples.

[0152] Step 2: Calculate the linear relationship between ln (index) and geomorphic gradient.

[0153] Take the natural logarithm of the indicator (NDVI growth): ln(NDVI growth)

[0154] Using cave density (geomorphic gradient) as the independent variable and ln(NDVI growth) as the dependent variable, a linear regression was performed:

[0155] ln(NDVI growth) = γ - δρ (cavity density).

[0156] Regression analysis results (least squares fitting):

[0157] Intercept γ = 2.25

[0158] The slope δ = 0.38 (i.e., the geomorphic gradient sensitivity coefficient).

[0159] The correlation coefficient R² = 0.89 (high fit, indicating a significant nonlinear relationship).

[0160] Understandably, the above is just an example of NDVI. Other similar ecological indicators can be calculated in the same way. This geomorphic gradient sensitivity coefficient is only applicable to karst areas, or more accurate geomorphic gradient sensitivity coefficients can be obtained by distinguishing different geomorphic types for different regions.

[0161] The heterogeneity correction model (correction targets: ecological damage / restoration indicators, such as NDVI and soil erosion rate) is as follows:

[0162] ;

[0163] in, For the first A revised indicator, For the first One original indicator, The influence intensity coefficient of landform. The sensitivity coefficient for terrain gradient is... The topographic gradient is calculated as follows: (cavity density (cavity depressions) / slope (peak forest steep slopes)).

[0164] In some embodiments of this application, a multi-dimensional evaluation system for open-pit mines in karst areas is constructed, including,

[0165] Based on the logical stratification of ecological processes in karst areas, multiple directional dimensions are determined. Evaluation indicators are selected under each directional dimension to construct a dedicated indicator library for the ecological restoration of open-pit mines in karst areas. The evaluation indicators are dynamically updated through a heterogeneity correction model, thereby updating the dedicated indicator library.

[0166] In this embodiment, the dimensional design must closely adhere to the driving role of karst landform gradients on ecological processes, rather than simply applying a general framework. Ecological damage and restoration are non-linearly affected by landform gradients (cavity density, slope), requiring the separation of the dynamic "damage-restoration-stabilization" stages. Traditional assessment systems (such as "vegetation coverage") ignore landform gradients, leading to distorted assessments of karst areas (e.g., overestimating the restoration rate of cavity depressions by 25%). The final four core dimensions (covering the entire chain of ecological damage, restoration, stabilization, and resource allocation) are thus determined.

[0167] Screening principles:

[0168] Karst areas can be quantified: indicators must be obtainable through remote sensing (Sentinel-2), GIS (DEM), or ground surveys.

[0169] Geomorphic gradient driven: The index needs to be embedded with geomorphic gradient (cavity density / slope) as a correction factor.

[0170] Dynamic response: The indicators need to reflect the continuous changes in the repair phase (mining → repair → stabilization).

[0171] Ecological damage intensity (mining stage) includes soil erosion rate and hydrological damage index; ecological restoration efficiency (restoration stage) includes vegetation recovery index (VRI) and soil thickness recovery rate; ecological stability (long-term post-restoration) includes NDVI fluctuation index and erosion rate persistence; and restoration resource optimization (decision support dimension) includes resource usage. A heterogeneity correction model is then used to dynamically update and correct some of the above assessment indicators to ensure the reliability of the assessment system.

[0172] Step S104: Based on multiple succession stages, conduct horizontal and vertical comparisons to determine ecological restoration effectiveness indicators, thereby quantitatively describing the results of ecological restoration of open-pit mines in karst areas.

[0173] In some embodiments of this application, horizontal and vertical comparisons are performed, including...

[0174] Horizontal comparison refers to the differences in the same evaluation index between different landform types at the same successive stage, generating horizontal difference values.

[0175] Longitudinal comparison generates longitudinal difference values ​​by comparing the differences of the same evaluation index between different successional stages within the same geomorphic type region.

[0176] In this embodiment, a horizontal comparison is conducted to analyze the differences between different landform types within the same stage. At the same successional stage (e.g., mining stage), ecological indicators of different landform types (karst depressions, peak forests, steep slopes, karst plains) are compared. This reveals the amplifying effect of landform gradients on ecological damage (e.g., the erosion rate of karst depressions is 3 times that of karst plains). The specialized value of karst areas is highlighted: avoiding a "one-size-fits-all" assessment and identifying high-risk landforms (e.g., areas with a cave density > 3 caves / km²). A vertical comparison is also conducted to examine the changes of the same landform type at different stages. For the same landform type (e.g., karst depressions), ecological indicators at different successional stages (original → mining → abandonment → restoration) are compared. The dynamic response of the restoration process is quantified (e.g., the recovery rate of NDVI from 0.05 to 0.15). Key restoration thresholds are identified (e.g., accelerated restoration when NDVI > 0.1 in karst depressions). The specialized value of karst areas is also highlighted: capturing the nonlinear characteristics of karst area restoration (e.g., slow initial recovery → stable later recovery).

[0177] In some embodiments of this application, indicators for determining the effectiveness of ecological restoration include,

[0178] Under a multi-dimensional evaluation system, the effectiveness indicators of ecological restoration are determined by combining the horizontal and vertical differences of multiple evaluation indicators.

[0179] In this embodiment, multiple relevant assessment indicators are selected under a multi-dimensional evaluation system. The ecological restoration effectiveness indicator is determined by combining the horizontal and vertical differences of these indicators. Horizontal differences at different stages are compared to obtain the "theoretical restoration potential" for different landform types. The (original stage indicator - mining stage indicator) is determined by comparing horizontal differences, resulting in a horizontal baseline value. The vertical difference value represents the actual restoration amount and directly reflects the restoration effectiveness. The ratio of the vertical difference value to the horizontal baseline value is the ecological restoration effectiveness indicator for a single assessment indicator. The ecological restoration effectiveness indicator is determined by comprehensively considering all relevant assessment indicators.

[0180] It should be noted that the above is only a specific way to determine the ecological restoration effectiveness indicators by combining the horizontal and vertical difference values ​​of multiple evaluation indicators. It can also be extended and expanded according to the definitions of the two. For example, the indicators under a single stage can be determined first, and then combined. SRE = (restoration stage indicators - mining stage indicators) / (original stage indicators - mining stage indicators).

[0181] Correspondingly, this application also provides an effectiveness evaluation system for ecological restoration of open-pit mines in karst areas, such as... Figure 2 As shown, including,

[0182] The first module is used to determine the geographical location of open-pit mines in karst areas, and to divide the geographical location of open-pit mines in karst areas into multiple successional stages, matching the specific geographical location corresponding to each successional stage;

[0183] The second module is used to obtain landform types, geographical features and image features according to the specific regional locations corresponding to different succession stages. It uses landform types and geographical features to decompose the specific regional locations to obtain the initial fuzzy boundaries of different landform types. Based on the image features, it performs boundary refinement processing on the initial fuzzy boundaries of different landform types to obtain the refined boundaries of different landform types.

[0184] The third module is used to analyze the spatial heterogeneity effect under the refined boundary of different landform types based on the refined boundary of different landform types, to construct a heterogeneity correction model, and to construct a multi-dimensional evaluation system for open-pit mines in karst areas.

[0185] The fourth module is used to conduct horizontal and vertical comparisons based on multiple succession stages to determine the indicators of ecological restoration effectiveness, thereby quantitatively describing the results of ecological restoration of open-pit mines in karst areas.

[0186] Through the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented in hardware or by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0187] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing the present invention.

[0188] Those skilled in the art will understand that the modules in the apparatus of the implementation scenario can be distributed within the apparatus of the implementation scenario as described, or they can be located in one or more apparatuses different from this implementation scenario, with corresponding changes. The modules of the above-described implementation scenario can be combined into one module, or they can be further divided into multiple sub-modules.

[0189] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for evaluating the effectiveness of ecological restoration of open-pit mines in karst areas, characterized in that, include, Determine the geographical location of the open-pit mine in the karst area, and based on the geographical location of the open-pit mine in the karst area, divide it into multiple successional stages, and match the specific geographical location corresponding to each successional stage; According to the specific regional locations corresponding to different succession stages, landform types, geographical features and image features are obtained. The specific regional locations are decomposed based on landform types and geographical features to obtain the initial fuzzy boundaries of different landform types. The initial fuzzy boundaries of different landform types are refined based on image features to obtain the refined boundaries of different landform types. Based on the refined boundary analysis of different landform types, the spatial heterogeneity effect under the refined boundary of different landform types is analyzed to construct a heterogeneity correction model and to construct a multi-dimensional evaluation system for open-pit mines in karst areas. Based on multiple succession stages, horizontal and vertical comparisons are conducted to determine indicators of ecological restoration effectiveness, thereby quantitatively describing the results of ecological restoration of open-pit mines in karst areas.

2. The method for evaluating the effectiveness of ecological restoration of open-pit mines in karst areas according to claim 1, characterized in that, Based on the geographical location of open-pit mines in karst areas, multiple successional stages are identified, including: The multiple succession stages, in chronological order, include the original stage of karst areas, the mining stage, the mine abandonment stage, and the ecological restoration stage. Obtain information on mining engineering and ecological restoration projects within the geographical location of the mine in the karst area, and use this information to determine the mining timeline, mine closure timeline, and ecological restoration start timeline. A timeline for the succession stages of open-pit mines in karst areas was constructed, and the time nodes for mine operation, mine closure, and ecological restoration commencement were marked on the timeline. The NDVI of the continuous time series on the timeline was statistically analyzed. An NDVI threshold was set using the Mann-Kendall test. Based on the NDVI threshold, NDVI mutation points near each of the mine operation, mine closure, and ecological restoration commencement time nodes were screened. The NDVI mutation points near each of the mine operation, mine closure, and ecological restoration commencement time nodes were combined to determine the corresponding mutation time nodes for each time node. A lag analysis was performed on the NDVI mutation points near the mining start date, mine closure date, and ecological restoration start date to generate lag durations. The mutation dates were adjusted based on the lag durations. By combining the mining start date, mine closure date, ecological restoration start date, and their corresponding adjusted mutation dates, new mining start dates, mine closure dates, and ecological restoration start dates were determined. The stage time points for the original stage, mining stage, mine closure stage, and ecological restoration stage of the karst area were also determined.

3. The method for evaluating the effectiveness of ecological restoration of open-pit mines in karst areas according to claim 2, characterized in that, Match the specific regional location corresponding to each successional stage, including, Within the geographical range of the open-pit mine in the karst area, the edge corresponding to each successional stage is determined. According to the different NDVI ranges, the edge corresponding to each successional stage is divided into multiple vegetation cover zones of different levels. The migration rate corresponding to different vegetation cover zones under different successional stages is calculated. The edge is dynamically adjusted by combining the migration rates of all vegetation cover zones to determine the specific geographical location corresponding to each successional stage.

4. The method for evaluating the effectiveness of ecological restoration of open-pit mines in karst areas according to claim 1, characterized in that, By segmenting specific regional locations based on landform type and geographical features, initial fuzzy boundaries for different landform types are obtained. These initial fuzzy boundaries are then refined based on image features to obtain refined boundaries for each landform type, including... In GIS software, spatial partitioning is performed on a specific location based on thresholds of multiple geographic features and landform types to obtain multiple landform type regions and their corresponding initial fuzzy boundaries. Image features are then matched with these multiple landform type regions and their corresponding initial fuzzy boundaries according to geographic location. Edge extraction is performed on the initial blurred boundary and the image features of the surrounding area using the HED model, and a probability boundary map is output. The initial fuzzy boundary and its surrounding area are marked as the first region, and the area on the landform type zone other than the first region is marked as the second region. The differences in geographical features within the first region and the differences in geographical features between the second regions on both sides of the initial fuzzy boundary are statistically analyzed and recorded as the first geographical feature difference and the second geographical feature difference, respectively. Based on the landform type on both sides of the initial fuzzy boundary, the first geographical feature difference and the second geographical feature difference, different probability thresholds for the initial fuzzy boundary are set, and the segmentation boundary is preserved in the probability boundary map based on the probability thresholds. Multiple karst landforms are selected for each landform type, and karst landform templates are configured according to the parameter ranges of the previous geographical features and image features of each karst landform. The karst landform templates are used to traverse the segmentation boundary and aggregate the matching points into a continuous boundary. Then, multi-evidence chain fusion verification is performed on the continuous boundary to complete the boundary refinement process and obtain refined boundaries for different landform types. Among them, the multi-evidence chain includes verification dimensions such as LiDAR elevation, rock index and temporal consistency.

5. The method for evaluating the effectiveness of ecological restoration of open-pit mines in karst areas according to claim 1, characterized in that, Based on the analysis of refined boundaries for different landform types, spatial heterogeneity effects under refined boundaries of different landform types are analyzed to construct a heterogeneity correction model, including, We obtain geomorphic indicators, ecological indicators, and geomorphic gradients for different landform types. We calculate the geomorphic influence intensity coefficient based on the geomorphic indicators and calculate the geomorphic gradient sensitivity coefficient by fitting the correlation between ecological indicators and geomorphic gradients.

6. The method for evaluating the effectiveness of ecological restoration of open-pit mines in karst areas according to claim 5, characterized in that, Constructing a heterogeneity correction model, including, A heterogeneity correction model is constructed based on the geomorphic influence intensity coefficient, geomorphic gradient sensitivity coefficient, and geomorphic gradient.

7. The method for evaluating the effectiveness of ecological restoration of open-pit mines in karst areas according to claim 4, characterized in that, And construct a multi-dimensional evaluation system for open-pit mines in karst areas, including, Based on the logical stratification of ecological processes in karst areas, multiple directional dimensions are determined. Evaluation indicators are selected under each directional dimension to construct a dedicated indicator library for the ecological restoration of open-pit mines in karst areas. The evaluation indicators are dynamically updated through a heterogeneity correction model, thereby updating the dedicated indicator library.

8. The method for evaluating the effectiveness of ecological restoration of open-pit mines in karst areas according to claim 7, characterized in that, Perform horizontal and vertical comparisons, including: Horizontal comparison refers to the differences in the same evaluation index between different landform types at the same successive stage, generating horizontal difference values. Longitudinal comparison generates longitudinal difference values ​​by comparing the differences of the same evaluation index between different successional stages within the same geomorphic type region.

9. The method for evaluating the effectiveness of ecological restoration of open-pit mines in karst areas according to claim 8, characterized in that, Determine the indicators of ecological restoration effectiveness, including: Under a multi-dimensional evaluation system, the effectiveness indicators of ecological restoration are determined by combining the horizontal and vertical differences of multiple evaluation indicators.

10. An effectiveness evaluation system for ecological restoration of open-pit mines in karst areas, characterized in that, include, The first module is used to determine the geographical location of open-pit mines in karst areas, and to divide the geographical location of open-pit mines in karst areas into multiple successional stages, matching the specific geographical location corresponding to each successional stage; The second module is used to obtain landform types, geographical features and image features according to the specific regional locations corresponding to different succession stages. It uses landform types and geographical features to decompose the specific regional locations to obtain the initial fuzzy boundaries of different landform types. Based on the image features, it performs boundary refinement processing on the initial fuzzy boundaries of different landform types to obtain the refined boundaries of different landform types. The third module is used to analyze the spatial heterogeneity effect under the refined boundary of different landform types based on the refined boundary of different landform types, to construct a heterogeneity correction model, and to construct a multi-dimensional evaluation system for open-pit mines in karst areas. The fourth module is used to conduct horizontal and vertical comparisons based on multiple succession stages to determine the indicators of ecological restoration effectiveness, thereby quantitatively describing the results of ecological restoration of open-pit mines in karst areas.

Images