An open-pit mine terrain precision evaluation algorithm

Abstract

The present application relates to the technical field of mine surveying and mapping and terrain modeling precision evaluation, and particularly relates to a kind of open pit mine terrain precision evaluation algorithm, comprising S1: the preprocessing of multi-source data, selects standard test field, calculates the systematic error and random error of each data source;S2: establish three levels index system including basic precision index layer, feature precision index layer and regional precision index layer;S3: based on index system, respectively calculate each level error index, adopt analytic hierarchy process to combine entropy weight method to determine the weight of each index, and weightedly calculate comprehensive precision index;S4: construct dynamic error threshold, adopt the dual mechanism of point-by-point check and local area fitting check to carry out precision determination and locate defect area;S5: output quantification index report, visualized achievement and point deviation list, and generate targeted optimization suggestion according to the cause of defect area;Construct multi-dimensional stereoscopic evaluation system of elevation precision and geometric feature precision, comprehensively reflect the overall quality of terrain modeling.

Description

Technical Field

[0001] This invention relates to the field of accuracy evaluation technology for mine surveying and terrain modeling, specifically to an algorithm for evaluating the accuracy of open-pit mine terrain. Background Technology

[0002] The accuracy of mine topography directly determines the reliability of core aspects such as mine planning, ramp design, slope stability analysis, and safe production. As a key means of controlling topographic data quality, the scientific rigor and accuracy of topographic accuracy assessment are paramount. Currently, with the widespread application of surveying technologies such as lidar, UAV aerial surveying, and total stations in the mining field, mine topographic data acquisition has formed a multi-source fusion pattern. However, existing methods for assessing mine topographic accuracy mainly suffer from the following shortcomings: (1) The evaluation dimensions are singular. Traditional methods are mostly based on single-point elevation error for evaluation, and only the "absolute deviation between the measured value and the model value" is used as the core indicator. This ignores the accuracy differences of key geometric features of the mining area, such as slope, aspect, and roughness, and cannot fully reflect the overall quality of terrain modeling. When the single-point elevation error is qualified but the terrain geometry is distorted, traditional evaluation methods cannot effectively identify it, which may lead to errors in subsequent engineering decisions.

[0003] (2) Poor adaptability of multi-source data: Existing algorithms are insufficient in their ability to fuse and process multi-source data such as lidar point clouds, UAV aerial surveys, and traditional total station measurements. Lidar point clouds are prone to elevation deviations due to vegetation obstruction, UAV aerial surveys suffer from accuracy degradation in edge areas due to the influence of the photography angle, and while total station measurements have high accuracy, the amount of data is small and the distribution is uneven. The error characteristics of different data sources are not considered differently, resulting in large accuracy deviations when evaluating multi-source data collaboratively.

[0004] (3) Insufficient sensitivity to feature regions: The terrain of key feature regions such as steep slopes, faults, and mining benches in mining areas varies drastically, and the accuracy requirements are much higher than those for ordinary flat areas. However, traditional algorithms use a uniform fixed error threshold for evaluation of the entire region without differentiating the specific characteristics of feature regions. If the threshold is set too wide, it will miss the accuracy defects of feature regions, leading to engineering risks; if the threshold is set too strict, it will misjudge normal errors in ordinary areas as defects, increasing unnecessary modeling and correction costs.

[0005] (4) The evaluation results are not practical. The evaluation results output by the existing methods are mostly single error values. They lack auxiliary information such as error distribution visualization, defect area location and accuracy optimization suggestions, which makes it difficult to directly guide the correction of measurement data or the adjustment of modeling parameters. The evaluation results are out of touch with the actual engineering application.

[0006] Furthermore, a search revealed that existing related patents, such as "A Method for Safety Inspection and Production Assessment of Open-Pit Coal Mines Based on UAV Mapping," only focus on a single data source—UAVs—and their core technology is safety inspection and mining volume calculation, without addressing multi-dimensional terrain accuracy assessment. "A Method for Terrain Accuracy Inspection in Open-Pit Mines" is designed only for open-pit coal mines, cannot adaptively adjust weight allocation, and lacks an iterative update mechanism. Papers such as "Research on Accuracy Assessment Methods for Terrain Modeling in Open-Pit Mines" propose assessment schemes that only cover two features: elevation and slope, limiting the assessment dimensions and failing to address differentiated assessments of key areas.

[0007] To address the shortcomings of existing technologies, this invention provides an algorithm for evaluating the accuracy of open-pit mine terrain. This algorithm enables differentiated processing of multi-source data, hierarchical accuracy analysis of terrain features, and precise verification of key areas. It also outputs visualized evaluation results and optimization suggestions, thereby improving the comprehensiveness, accuracy, and engineering applicability of mine terrain accuracy evaluation. Summary of the Invention

[0008] To overcome the aforementioned technical problems in existing technologies, this invention provides an algorithm for evaluating the accuracy of open-pit mine terrain. It constructs a multi-dimensional, three-dimensional evaluation system encompassing both elevation accuracy and geometric feature accuracy, breaking through the limitations of traditional single-point errors and comprehensively reflecting the overall quality of terrain modeling. A multi-source data differentiation processing mechanism is designed to achieve full-process adaptation from error tracing to weighted modeling to differentiated evaluation, improving the accuracy of multi-source data collaborative evaluation. Through a combined strategy of "intelligent feature area recognition + dynamic threshold evaluation," it achieves precise focused evaluation of key areas such as steep slopes, faults, and mining benches, enhancing sensitivity to risk areas. Visualized evaluation results and targeted optimization suggestions are output, enabling the evaluation results to directly guide the correction of mine surveying data and the adjustment of modeling parameters, thus improving engineering practicality.

[0009] To achieve the above objectives, this invention provides an algorithm for evaluating the accuracy of open-pit mine terrain, comprising the following steps: S1: Multi-source data preprocessing and error tracing, performing targeted preprocessing on lidar point clouds, UAV aerial survey data, and total station measurement data, selecting a standard test field, and calculating the systematic and random errors of each data source; S2: Constructing a hierarchical evaluation index system, establishing a three-level index system including a basic accuracy index layer, a feature accuracy index layer, and a regional accuracy index layer; S3: Multi-dimensional error calculation and weight allocation, based on the index system established in step S2, calculating the error index of each level, using analytic hierarchy process combined with entropy weighting to determine the weight of each index, and weighted calculating the comprehensive accuracy index; S4: Accuracy level determination and defect location, identifying key feature areas of the mining area, constructing a dynamic error threshold based on terrain complexity and engineering risk level, and using a dual mechanism of point-by-point verification and local area fitting verification to determine accuracy and locate defect areas; S5: Evaluation result output and optimization suggestion generation, outputting a quantitative index report, visualization results, and a list of point deviations, and generating targeted optimization suggestions based on the causes of defect areas.

[0010] Preferably, the error tracing in S1 specifically includes: selecting a flat area of ​​100m×100m within the mining area as a standard test field, uniformly selecting 50 test points within the test field, and calculating the systematic error of each data source: ,in Let n be the error of the i-th test point, and n be the number of test points; and calculate the random error of each data source: The targeted preprocessing in S1 specifically includes: using statistical filtering to remove noise points from the lidar point cloud, and using elevation thresholding combined with slope filtering to separate ground points from non-ground points; using motion reconstruction technology to generate dense point clouds from UAV aerial survey data, and selecting ground control points for adjustment optimization; using multiple observations and taking the average value from the total station measurement data, and removing out-of-limit data.

[0011] Preferably, the basic accuracy index layer in S2 includes single-point error index, density accuracy index, and overall consistency index; the single-point error index includes elevation mean error and plane mean error, calculated using the Bessel formula; the density accuracy index includes effective point density of the point cloud and DEM grid integrity; the overall consistency index includes mean deviation of terrain elevation and mean deviation of plane position; the feature accuracy index layer includes slope accuracy index, aspect accuracy index, and roughness accuracy index; the slope accuracy index includes slope calculation error ∆α, and slope classification is: gentle slope (<15°), gentle slope (15°-30°), steep slope (30°-60°), and very steep slope (>60°). The slope accuracy index includes slope aspect fit K, which calculates the cosine similarity of the slope aspect between the data to be evaluated and the benchmark data; the roughness accuracy index includes terrain roughness deviation ∆R, which calculates the root mean square error deviation of the terrain surface undulation based on a 3m×3m window; the regional accuracy index layer divides regions and designs indices according to the differences in terrain function and characteristics of the mining area, specifically including key feature areas, engineering operation areas and ordinary areas; the key feature areas include the slope mean error of the slope area and the plane position mean error of the fault area; the engineering operation area includes the elevation mean error of the mining bench and the plane mean error of the slope planning area; the ordinary area includes the comprehensive mean error of non-feature and non-operation areas.

[0012] Preferably, in step S3, the weight allocation employs the analytic hierarchy process (AHP) combined with engineering requirements to determine subjective weights, and the entropy weight method to calculate objective weights based on real error data from the data source. The final evaluation weights are obtained through combined weighting. The AHP constructs a third-order judgment matrix A as follows: When the consistency ratio CR < 0.1, the matrix is ​​considered to be of good consistency; the entropy weight method includes: standardizing the error index, and the standardization formula is: ; Calculate the weight matrix: ; Calculate information entropy: ,in ; Calculate objective weights: Calculate the combined weights: .

[0013] Preferably, identifying key feature areas of the mining area in S4 includes: slope area identification, through elevation gradient amplitude. Quantifying changes in terrain undulation, when When a slope candidate area is identified, it is classified into gentle slopes, medium slopes, and steep slopes based on gradient amplitude; fault areas are identified using Gaussian curvature. Quantify the curvature characteristics of the terrain surface, when When a fault is identified as a candidate area, it is considered a valid fault if the fault width W ≥ 2m after linear fitting. For the identification of the bench area in the mining area, the actual slope angle α = arctan(G(x,y)) × (180 / π) is calculated by traversing the layers at intervals of the designed bench height h. When α meets the design bench slope angle ±5° and the width W of the adjacent gentle area is ≥ 3m, it is considered a bench area in the mining area.

[0014] Preferably, the construction of the dynamic error threshold in S4 includes: calculating terrain complexity. Where σ is the regional elevation standard deviation, and θ is the regional weighted average slope; dynamic error threshold ,in R represents the basic threshold for ordinary areas, and R represents the engineering risk level. , This is the threshold correction coefficient; the threshold for high-risk areas is stricter than that for medium-risk areas, and the threshold for medium-risk areas is stricter than that for low-risk areas.

[0015] Preferably, the threshold correction coefficient , By combining measured samples from multiple working conditions in the mining area with multivariate nonlinear least squares regression fitting, the following sample set was constructed: This sample set includes working conditions in leveled mining areas, gentle slope transition zones, steep slopes, vegetated areas, and boundary distortion zones. The optimal allowable threshold for each area was determined using high-precision total station measured data as the true benchmark. Establish a nonlinear regression model The optimization objective is to minimize the sum of squared residuals. The solution is obtained through gradient descent iterative optimization, with constraints. , .

[0016] Preferably, the dual verification mechanism in S4 includes: point-by-point verification, where if the absolute elevation error E(x,y)≤T(x,y), it is determined as a qualified point; if E(x,y)>T(x,y), it is marked as a defective point; and local window fitting verification, where a 5×5 neighborhood window is used to perform least squares plane fitting on the defective point, and the root mean square error within the window is calculated. If RMSE ≤ k×T, it is determined to be a random error; if RMSE > k×T, it is determined to be a persistent defect region and the region boundary is marked, where k is the allowable deviation coefficient of the characteristic region.

[0017] Preferably, the weight allocation in S3 adopts an incremental online weight update mechanism: the system caches error samples of newly collected data in real time and maintains the error sample set using a sliding window mechanism; after a new sample is added, only the new sample is standardized, its weight is calculated, and its information entropy is incrementally corrected, and the entropy value and objective weight are iteratively updated based on the historical benchmark entropy value; when the error characteristics of multiple consecutive sets of new data from a certain type of data source deviate significantly, the analytic hierarchy process (AHP) is automatically adjusted to determine the matrix scale; after each data update, the latest combined weights are automatically iteratively output.

[0018] Preferably, the visualization results in S5 include an error distribution heatmap, a feature area accuracy comparison map, and radar maps of various indicators; the optimization suggestions include: supplementing UAV oblique photogrammetry for areas with point cloud obstruction on steep slopes, increasing the density of lidar scanning for areas with insufficient data density, and re-performing GNSSRTK field measurement verification for areas with concentrated gross error points.

[0019] Accordingly, the present invention also provides an open-pit mine terrain accuracy assessment system, which uses the open-pit mine terrain accuracy assessment algorithm provided by the present invention to assess terrain accuracy.

[0020] The present invention has at least the following technical effects through the technical solution provided by the present invention: This invention constructs a three-tiered indicator system of "foundation-feature-region," comprehensively covering single-point accuracy, terrain feature accuracy (slope, aspect, roughness, etc.), and overall regional accuracy. Compared to traditional single-indicator evaluation methods, the evaluation dimensions have been expanded from one to more than four, improving comprehensiveness. For key mining feature areas such as steep slopes, faults, and mining benches, differentiated threshold design and multi-criteria decision models can accurately identify accuracy defects, significantly improving the accuracy rate of identifying accuracy defects in key areas and effectively avoiding engineering risks such as slope collapse and mining planning deviations caused by local accuracy defects. For the error characteristics of different data sources such as lidar point clouds (mainly random errors), UAV aerial surveys (errors are positively correlated with slope), and total station measurements (mainly systematic errors), a dedicated differentiated preprocessing process and error calculation method are designed. Through targeted operations such as statistical filtering, ground control point correction, and instrument error compensation, efficient fusion of multi-source data is achieved. The collaborative evaluation error of multi-source data can be stably controlled within ±0.05m, improving accuracy compared to traditional unified processing methods and fully leveraging the advantages of each data source. The complementary advantages of the sources make it suitable for the mainstream surveying and mapping mode of multi-source collaborative data acquisition in current mines. The evaluation results output by this invention not only include conventional error values, but also come with error distribution heatmaps, accuracy level judgments, and targeted optimization suggestions. It can directly guide the correction of on-site measurement data, adjustment of modeling parameters, and supplementary measurements in key areas. The evaluation report can be directly adapted to the needs of different stages such as mine planning, slope design, and slope stability analysis. The setting of a comprehensive error threshold of 0.3m for the basic area and a comprehensive error threshold of 0.25m for the key area provides a clear basis for the selection of terrain data in each stage, greatly improving the engineering application value of accuracy evaluation. Through the weight adaptive adjustment model, the weight allocation of each error index can be automatically adjusted according to the terrain characteristics and surveying needs of different types of open-pit mines (coal mines, iron mines, non-ferrous metal mines, etc.) without major modifications to the core algorithm logic. Combined with the incremental online weight update mechanism, it can track the dynamic changes of terrain in the mining process in real time, continuously adapting to the accuracy evaluation needs of the entire life cycle of mine exploration, construction, production, and closure. It has good adaptability and scalability and high reusability. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of the open-pit mine terrain accuracy evaluation algorithm provided in the embodiments of the present invention. Detailed Implementation

[0022] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0023] In this invention, the terms "system" and "network" are used interchangeably. "Multiple" refers to two or more; therefore, in this invention, "multiple" can also be understood as "at least two." "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / ", unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, it should be understood that in the description of this invention, terms such as "first" and "second" are used only for descriptive purposes and should not be construed as indicating or implying relative importance or order.

[0024] Please see Figure 1 This invention provides an algorithm for evaluating the accuracy of open-pit mine terrain, comprising the following steps: S1: Multi-source data preprocessing and error tracing: Targeted preprocessing is performed on lidar point cloud, UAV aerial survey data, and total station measurement data. A standard test field is selected, and the systematic and random errors of each data source are calculated. S2: Construct a hierarchical evaluation index system, and establish a three-tier index system including a basic accuracy index layer, a feature accuracy index layer, and a regional accuracy index layer; S3: Multi-dimensional error calculation and weight allocation. Based on the indicator system established in step S2, the error indicators of each level are calculated respectively. The weight of each indicator is determined by the combination of hierarchical analysis and entropy weight method. The comprehensive accuracy index is calculated by weighting. S4: Accuracy level determination and defect location, identify key feature areas of the mining area, construct dynamic error thresholds based on terrain complexity and engineering risk level, and use a dual mechanism of point-by-point verification and local area fitting verification to determine accuracy and locate defect areas. S5: Output of evaluation results and generation of optimization suggestions, output quantitative indicator reports, visualization results and point deviation list, and generate targeted optimization suggestions based on the causes of defective areas.

[0025] In this embodiment of the invention, step S1 specifically performs multi-source data preprocessing and error tracing, and performs targeted preprocessing on lidar point clouds, UAV aerial survey data, and total station measurement data respectively. For lidar point clouds, statistical filtering (setting the number of neighboring points to 10 and the standard deviation factor to 1.5) is used to remove noise points, and ground points and non-ground points are separated by elevation thresholding combined with slope filtering. For UAV aerial survey data, structure of motion reconstruction (SfM) technology is used to generate dense point clouds, and at least four evenly distributed ground control points (GCPs) are selected for adjustment optimization. For total station measurement data, multiple observations (≥2) are performed and the average value is taken, and out-of-limit data (closed traverse closure error) are removed. (L is the traverse length, in km). A flat area of ​​100m × 100m within the mining area was selected as the standard test field (real terrain parameters were obtained through high-precision measurement using a total station, with an accuracy ≤ 0.01m). 50 test points were evenly selected within the test field, and the systematic error ε and random error ∆ of each data source were calculated. , ,in, Let n be the error of the i-th test point, and n be the number of test points.

[0026] In one feasible implementation, the acquisition of multi-source data includes three sources of data to be evaluated: (1) LiDAR point cloud data, which is collected by an airborne LiDAR system with a point cloud density of 2 points / m² and a nominal elevation accuracy of ±0.15m; (2) UAV aerial survey data, which is collected by a fixed-wing UAV equipped with a five-lens tilt camera with a ground resolution of 0.05m, a heading overlap of 80%, and a lateral overlap of 60%; (3) Total station measurement data, which is measured by a high-precision total station (angle accuracy ±1″, distance accuracy ±1mm+1ppm) to measure control points. A total of 32 control points are set up and evenly distributed in various functional areas of the mining area. The reference data uses high-precision total station measurements, averaged after multiple observations (≥2), with an accuracy ≤0.01m. For the lidar point cloud, statistical filtering removes noise points, with a neighborhood of 10 points and a standard deviation factor of 1.5. Ground and non-ground points are separated using an elevation threshold method combined with slope filtering, with an elevation threshold of 2m (relative to the lowest elevation) and a slope threshold of 15°. After filtering, there are 1,286,543 ground points, and non-ground points (vegetation, buildings, etc.) have been effectively removed. For UAV aerial survey data, Structure of Motion (SfM) technology is used to generate a dense point cloud. Six evenly distributed ground control points (GCPs) are selected for adjustment optimization. The root mean square error of the reprojection after adjustment is 0.8 pixels, meeting the accuracy requirements. For the total station measurement data, multiple observations (≥2) are performed, and the average value is taken. The closure error of the closed traverse is +8mm, less than the tolerance limit. (L=1.2km, tolerance ±13.1mm), data valid. A flat area of ​​100m×100m within the mining area was selected as the standard test field, and real terrain parameters were obtained through high-precision measurement using a total station. Fifty test points were evenly selected within the test field, and the systematic and random errors of each data source were calculated. Calculations showed: lidar point cloud system error ε_L=0.08m, random error Δ_L=0.12m; UAV aerial survey system error ε_U=0.10m, random error Δ_U=0.15m; total station measurement system error ε_T=0.01m, random error Δ_T=0.008m. The total station measurement data showed the highest accuracy and was used as the reference benchmark.

[0027] In this embodiment of the invention, step S2 constructs a hierarchical evaluation index system, establishing a three-tiered index system comprising a basic accuracy index layer, a feature accuracy index layer, and a regional accuracy index layer. The basic accuracy index layer includes: single-point error indices—elevation mean error (M_z) and plane mean error (M_xy), calculated using the Bessel formula; density accuracy indices—effective point density of the point cloud and DEM grid integrity; and overall consistency indices—mean deviation of terrain elevation (ΔH) and mean deviation of plane position (ΔS). The feature accuracy index layer includes: slope accuracy indices—slope calculation error (Δα), with slope grading into gentle slope (<15°), gentle slope (15°-30°), steep slope (30°-60°), and very steep slope (>0°); aspect accuracy indices—aspect fit (K), calculating the cosine similarity of the aspect between the data to be evaluated and the benchmark data; and roughness accuracy indices—terrain roughness deviation (ΔR), calculating the root mean square error deviation of terrain surface undulations based on a 3m×3m window. The regional accuracy index layer is divided into regions and designed with indicators based on the differences in the topographic function and characteristics of the mining area, including: key feature area - slope error of the slope area and horizontal position error of the fault area; engineering operation area - elevation error of the mining bench and horizontal error of the slope planning area; ordinary area - comprehensive error of non-feature and non-operation areas.

[0028] In one feasible implementation, a three-tiered index system is constructed as follows: Basic accuracy index layer: Elevation mean square error M_z (target value ≤ 0.15m), Planar mean square error M_xy (target value ≤ 0.20m), Effective point density of point cloud (target value ≥ 1.5 points / m²), DEM grid integrity (target value ≥ 95%), Mean deviation of terrain elevation ΔH (target value ≤ 0.10m), Mean deviation of planar position ΔS (target value ≤ 0.15m). Feature accuracy index layer: Slope calculation error Δα (target value: gentle slope ≤ 2°, mild slope ≤ 3°, steep slope ≤ 4°, very steep slope ≤ 5°), Aspect fit K (target value ≥ 0.90), Terrain roughness deviation ΔR (target value ≤ 0.05m). Regional accuracy index layer: Key feature area - slope error of slope area (target value ≤ 3°, fault area plane position error (target value ≤ 0.15m); Engineering operation area - elevation error of mining bench (target value ≤ 0.10m), plane error of ramp planning area (target value ≤ 0.20m); General area - comprehensive error (target value ≤ 0.25m).

[0029] In this embodiment of the invention, step S3 performs multi-dimensional error calculation and weight allocation, specifically including: 3.1: Layered error calculation: Matching the coordinates of the reference benchmark data with the corresponding coordinates of the data to be evaluated, and using Bessel's formula to calculate the elevation error and plane error; constructing a local terrain window (5m×5m) based on the matched points, calculating the slope, aspect, and roughness of the data to be evaluated and the benchmark data respectively, and calculating ∆α, K, and ∆R through the difference; combining the feature region vector layer, extracting the matched points in each region, and calculating the corresponding accuracy index according to the region type. 3.2: Dynamic weight allocation: Using the analytic hierarchy process (AHP) combined with engineering requirements to determine the subjective weights of each index, using the entropy weight method to calculate the objective weights based on the real error data of the data source, and obtaining the final evaluation weights through combined weighting. The analytic hierarchy process constructs a third-order judgment matrix A as follows: Matrix interpretation: Total station is slightly more important than lidar (scale 3), and significantly more important than UAV aerial surveying (scale 5); lidar is slightly more important than UAV aerial surveying (scale 3). The consistency ratio CR = 0.034 < 0.1, indicating the matrix consistency is satisfactory. The entropy weight method includes the following steps: standardizing the error index (error is a negative index where smaller is better), the standardization formula is: ,in Given the error values ​​of the i-th sample and the j-th data source, calculate the weight matrix. ; Calculate information entropy ,in , m is the total number of error samples; calculate the objective weights. Combined weights The weights for the basic accuracy index layer are as follows: 0.4 (0.15 for elevation error, 0.15 for plane error, and 0.1 for density and integrity); 0.3 for the characteristic accuracy index layer (0.12 for slope calculation error, 0.08 for aspect fit, and 0.1 for terrain roughness deviation); and 0.3 for the regional accuracy index layer (0.15 for key feature areas, 0.1 for engineering operation areas, and 0.05 for ordinary areas). The weights can be dynamically adjusted for different mining area types. 3.3: Comprehensive Accuracy Calculation: The comprehensive accuracy index (P) is calculated using a weighted summation method, with the following formula: in Let i be the weight of the i-th indicator. This is the normalized value of the i-th indicator (ranging from 0 to 1; the larger the value, the higher the accuracy).

[0030] In one implementation, the coordinates of the reference data points were matched with the corresponding coordinates of the data to be evaluated, resulting in 1245 valid matching points. The Bessel formula was used to calculate the elevation error M_z = 0.12m and the plane error M_xy = 0.18m. A 5m × 5m local terrain window was constructed based on the matching points, and the slope, aspect, and roughness of the data to be evaluated and the reference data were calculated respectively. The calculated slope calculation error ∆α was 4.2° in steep slope areas and 1.8° in gentle slope areas; the aspect fit K = 0.92; and the terrain roughness deviation ∆R = 0.04m. Combining the feature region vector layer, matching points were extracted from each region, and corresponding accuracy indicators were calculated according to region type. The accuracy indicators for key feature areas (slopes, fault areas) were all within the target value range. The weight allocation was determined using the analytic hierarchy process (AHP) combined with the entropy weight method. The consistency ratio CR of the third-order judgment matrix constructed by the AHP was 0.034 < 0.1, indicating satisfactory consistency. After weighting, the weights for the basic accuracy index layer are 0.4, the characteristic accuracy index layer is 0.3, and the regional accuracy index layer is 0.3. The comprehensive accuracy index P is calculated using a weighted summation method. The normalized values ​​for each index are calculated as follows: normalized elevation error P_z = 0.80 (measured 0.12m / target 0.15m), normalized plane error P_xy = 0.90 (measured 0.18m / target 0.20m), normalized slope calculation error P_α = 0.84 (measured 4.2° / target 5°), normalized aspect fit P_K = 0.92, and normalized terrain roughness deviation P_R = 0.80 (measured 0.04m / target 0.05m). The normalized values ​​for all regional accuracy indices are above 0.80. The overall accuracy index P = Σw_i·P_i = 0.4×0.85 + 0.3×0.85 + 0.3×0.82 ≈ 0.84, and the overall accuracy level is rated as "good".

[0031] In this embodiment of the invention, step S4, which performs accuracy level determination and defect location, specifically includes: 4.1 Intelligent Identification of Feature Regions: Three methods are comprehensively adopted: elevation gradient analysis, terrain curvature calculation, and engineering prior parameter matching to locate three core feature regions: slopes, faults, and mining benches. (1) Slope Region Identification: The changes in terrain undulation are quantified by elevation gradient amplitude to locate and classify candidate slope regions. The formula for calculating elevation gradient amplitude is: Where ∂Z / ∂x and ∂Z / ∂y are the first-order partial derivatives of the elevation at coordinates (x,y) in the x and y directions, respectively (calculated using the central difference method with a 3×3 neighborhood window, grid resolution ∆d=1m). ( When the gradient threshold is calibrated to 0.3~0.5, it is determined to be a candidate area for slope; according to the gradient amplitude, it is classified as: gentle slope (0.3≤G<0.4), medium slope (0.4≤G<0.5), and steep slope (G≥0.5). (2) Fault area identification: the bending characteristics of the terrain surface are quantified by Gaussian curvature to locate the candidate fault area. The formula for calculating Gaussian curvature is: Where Z_xx, Z_yy, and Z_xy are second-order partial derivatives, calculated using the central difference method with a 3×3 neighborhood window. ( When the curvature threshold is calibrated to 0.05~0.1, it is determined to be a candidate fault area; the fault strike is obtained by least squares straight line fitting of the candidate area, and the length of the points that continuously meet the conditions in the strike direction is used as the fault width W. When W≥2m, it is determined to be a valid fault. (3) Identification of the bench area in the mining area: Combine the engineering prior parameters of the mining area design to match and locate the bench area. With the design bench height h as the interval, the elevation of the entire area is traversed layer by layer starting from the lowest elevation of the mining area; within the range of ±0.5h of each elevation layer, the gradient amplitude G(x,y) is calculated using a 3×3 neighborhood window, and the actual slope angle is calculated by α= arctan(G(x,y))×(180 / π); when the actual slope angle meets the design bench slope angle ±5° range, it is marked as a candidate bench slope; the adjacent gentle area (G(x,y)<0.2) is searched along the vertical direction of the slope. If the width W of the gentle area is ≥3m, it is determined to be a complete mining bench area. (4) Region boundary optimization: Morphological dilation (using a 3×3 rectangular structuring element, iterating once) and erosion (using the same structuring element, iterating once) operations are performed on the above three types of candidate regions to eliminate the interference of isolated noise points on the region boundary and obtain smooth and accurate feature region boundaries.

[0032] 4.2: Dynamic Threshold Calculation: A dynamic error threshold is constructed based on the terrain complexity and engineering risk level of the feature region, realizing a differentiated evaluation logic of "the higher the complexity and the higher the risk level, the stricter the threshold." The formula for calculating terrain complexity is: Where σ is the regional elevation standard deviation (normalized), and θ is the regional weighted average slope. The formula for calculating the dynamic error threshold is: in, Let R be the basic threshold for ordinary areas (0.1-0.2m), R be the engineering risk level (high risk R=3, medium risk R=2, low risk R=1), and k1 and k2 be threshold correction coefficients. Steep slopes and fault areas are considered high risk (R=3), mining bench areas are considered medium risk (R=2), gentle slope areas are considered low risk (R=1), and ordinary areas are considered risk-free (R=1). Threshold correction coefficient , By combining measured samples from multiple working conditions in the mining area with multivariate nonlinear least squares regression fitting, the following results were obtained: the mining area was divided into gridded zones, covering typical working conditions such as flat mining areas, gentle slope transition zones, steep slopes, vegetated areas, and boundary distortion zones; using high-precision total station measured data as the true benchmark, the error distribution of the terrain model in each zone was statistically analyzed to determine the optimal allowable threshold for each area. Simultaneously, match the normalized terrain complexity of the corresponding partition. With normalized risk level Construct a fitted sample set; establish a nonlinear regression model. The optimization objective is to minimize the sum of squared residuals. The solution is obtained through gradient descent iterative optimization, with constraints. , Finally, the fitting result was obtained .

[0033] 4.3: Dual-precision verification: A dual mechanism of "point-by-point preliminary verification + local area fitting verification" is employed to accurately distinguish between random errors and persistent defects. Point-by-point verification: If the absolute elevation error E(x,y) ≤ T(x,y), it is judged as a qualified point; if E(x,y) > T(x,y), it is marked as a defective point. Local window fitting verification: A 5×5 neighborhood window is used to perform least squares plane fitting on the defective points. The plane equation is Ax + By + Cz + D = 0. After fitting, the root mean square error of all points within the window is calculated. Where m is the total number of points within the window (m=25). Let RMSE be the absolute elevation error of the i-th point within the window. If RMSE ≤ k×T (where k is the allowable deviation coefficient for the feature region, k = 1.1), it is determined to be a random error (caused by measurement noise or small local disturbances, and no correction is needed); if RMSE > k×T, it is determined to be a persistent defect region (caused by deviations in terrain modeling, and needs to be marked and corrected). The convex hull algorithm is used to mark the boundary of the region corresponding to the window.

[0034] In one implementation, (1) Intelligent identification of feature areas: The slope area is located using elevation gradient analysis. With a grid resolution of ∆d=1m, the elevation gradient amplitude at each point is calculated using the central difference method with a 3×3 neighborhood window. A gradient threshold is set. =0.4, identifying three steep slope areas (G≥0.5) with areas of 1200m², 850m², and 620m² respectively; and five medium slope areas (0.4≤G<0.5). Gaussian curvature analysis was used to locate fault areas. Curvature thresholds were set. =0.08, two candidate fault areas were identified. After least squares linear fitting, the fault strikes were NE45° and NW30°, and the fault widths W were 3.2m and 2.8m (both ≥2m), respectively, and were determined to be valid fault areas. Engineering prior parameters were used to match and locate the mining bench area. The designed bench height h=10m and bench slope angle α=65°. After traversing the elevation of the entire area in layers, a total of 7 mining bench areas were identified, and the bench platform width was ≥3m, which was determined to be valid. Morphological expansion (3×3 rectangular structural element, 1 iteration) and erosion (same structural element, 1 iteration) operations were performed on the above three types of candidate areas to obtain smooth feature area boundaries. (2) Dynamic threshold calculation Taking a high and steep slope area as an example, the area R=3 (high risk), after removing 3σ outliers, the elevation standard deviation σ=2.5m (after normalization σ=2.5 / 5=0.5), and the weighted average slope θ=72°. =0.15m, =0.3, =0.15. Substituting into the formula for calculating terrain complexity, we get C=0.4×0.5+0.6×(72 / 90)=0.2+0.48=0.68. Substituting into the dynamic threshold T=0.15×[1-0.3×0.68-0.15×2] =0.15×0.496≈0.0744m. That is, the allowable error value for this high and steep slope area is 0.0744m, which is stricter than the 0.15m of the ordinary area. Similarly, the dynamic threshold for the medium-risk area (mining bench, R=2) is about 0.106m, and the dynamic threshold for the low-risk area (gentle slope, R=1) is about 0.119m. (3) Double precision verification: The high and steep slope area was verified point by point. It was found that the absolute elevation error of a certain point E(x,y)=0.09m, which is greater than the dynamic threshold T=0.0744m of the area, and it was marked as a defect point. A 5×5 neighborhood window was used to perform least-squares plane fitting on the defect point. The fitted plane equation was Ax + By + Cz + D = 0 (specific coefficients: A = 0.12, B = 0.08, C = 0.99, D = -125.6). The sum of squared absolute errors of the 25 points within the fitted window, ΣE² = 0.18m², was then substituted into the formula. k×T = 1.1×0.0744 ≈ 0.0818m. Since 0.0848m > 0.0818m, it is determined to be a persistent defect area. The convex hull algorithm is used to mark the boundary of the area corresponding to the 5×5 window, with a marking range of approximately 25m². After double verification, this implementation method marks a total of 12 persistent defect areas, including 5 steep slope areas, 2 fault areas, 3 mining bench areas, and 2 other areas. There are a total of 23 random error points, none of which are included in the defect marking.

[0035] In this embodiment of the invention, step S5 executes the evaluation result output and optimization suggestion generation, specifically including: 5.1 Multi-form result output: outputting three types of core results—quantitative indicator report (including the specific values, weights, and comprehensive accuracy levels of each level of indicator), visualization results (including error distribution heatmap, feature area accuracy comparison map, and radar chart of each indicator), and point deviation list (including coordinate deviations and gross error point information of all matching points). 5.2 Precise optimization suggestion generation: based on the causes of defective areas and shortcomings in accuracy indicators, generating targeted optimization suggestions: for areas with point cloud obstruction on steep slopes, it is recommended to supplement with UAV oblique photogrammetry; for areas with insufficient data density, it is recommended to increase the density of LiDAR scanning; for areas with concentrated gross error points, it is recommended to re-perform GNSS RTK field measurement verification.

[0036] In one implementation, a quantitative index report is output: the comprehensive accuracy index P=0.84, and the comprehensive accuracy level is "good". Specific values ​​for each level of index are detailed in the report. Visual results are output: an error distribution heatmap shows that errors in steep slope areas are concentrated between 0.08 and 0.12 m, errors in fault areas are concentrated between 0.06 and 0.10 m, errors in mining bench areas are concentrated between 0.05 and 0.09 m, and errors in ordinary areas are concentrated between 0.02 and 0.06 m. A characteristic area accuracy comparison chart shows that the accuracy in steep slope areas is lower than that in ordinary areas, but both are within the dynamic threshold range. Radar charts for each index show that the slope accuracy index is relatively weak (normalized value 0.84), while other indices perform well. A list of point deviations is output: a total of 1245 matching point coordinate deviations are listed, including 8 gross errors (deviations exceeding 3 times the mean square error), all of which are marked. Based on the causes of defects and the shortcomings of accuracy indicators, the following targeted optimization suggestions are generated: (1) For the five persistent defects in the high and steep slope area, it is recommended to supplement the UAV oblique photogrammetry and increase the image acquisition in the area; (2) For the two persistent defects in the fault area, it is recommended to use ground lidar for supplementary scanning to increase the point cloud density in the area; (3) For the areas with insufficient data density (three areas with effective point cloud density < 1.5 points / m²), it is recommended to increase the lidar scanning; (4) For the areas with concentrated gross error points (two areas), it is recommended to re-perform GNSS RTK field measurement verification.

[0037] Furthermore, this invention provides an incremental online weight update mechanism: the system caches error samples of newly collected data in real time, and uses a sliding window mechanism to maintain the error sample set (retaining the most recent N sets of valid error data and removing expired data); after a new sample is added, only the new sample is standardized, its weight is calculated, and its information entropy is incrementally corrected, and the entropy value and objective weight are iteratively updated based on the historical benchmark entropy value; when the error magnitude and stability of multiple sets of newly added data from a certain type of data source deviate significantly, the scale of the AHP judgment matrix is ​​automatically fine-tuned; after each data update, the latest combined weights are automatically iteratively output. This invention also provides an open-pit mine terrain accuracy assessment system, which uses the above algorithm for terrain accuracy assessment.

[0038] In one feasible implementation, after the initial evaluation, three more batches of new measurement data were collected (each batch 15 days apart). The system uses a sliding window mechanism (window size N=100 groups) to maintain the error sample set. After new samples are added, only the new samples are standardized, their weights are calculated, and their information entropy increments are corrected. After three iterations, the objective weights of each data source changed as follows: the objective weight of the total station was slightly adjusted from 0.42 to 0.44, the objective weight of the lidar was slightly adjusted from 0.35 to 0.33, and the objective weight of the UAV aerial survey was slightly adjusted from 0.23 to 0.23. The combined weights were updated accordingly, with the total station weight slightly increased (due to its consistently good error stability) and the lidar weight slightly decreased (due to increased error fluctuations caused by seasonal vegetation changes). The entire update process does not require a full recalculation of historical data, resulting in a significant improvement in computational efficiency compared to a full recalculation.

[0039] Based on the same inventive concept, this invention provides an open-pit mine terrain accuracy assessment system, which uses the open-pit mine terrain accuracy assessment algorithm provided by this invention to assess terrain accuracy.

[0040] The open-pit mine terrain accuracy evaluation algorithm and system provided by this invention can be widely applied to various open-pit mines (coal mines, iron mines, non-ferrous metal mines, etc.) for terrain survey data quality evaluation, terrain modeling accuracy verification, and mine engineering safety assessment. This invention overcomes the limitations of traditional single-elevation error evaluation, achieving differentiated processing of multi-source data, hierarchical accuracy analysis of terrain features, and precise verification of key areas. The evaluation results can directly guide the correction of mine surveying data and the optimization of modeling parameters, demonstrating significant industrial practical value and broad application prospects.

[0041] Furthermore, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the present invention.

[0042] The optional embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the scope of the technical concept of the embodiments of the present invention, various simple modifications can be made to the technical solutions of the embodiments of the present invention, and these simple modifications all fall within the protection scope of the embodiments of the present invention.

[0043] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the embodiments of the present invention will not describe the various possible combinations separately.

[0044] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a microcontroller, chip, or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0045] Furthermore, various different implementations of the present invention can be combined arbitrarily, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed in the present invention.

Claims

1. An algorithm for evaluating the accuracy of terrain in open-pit mines, characterized in that, Includes the following steps: S1: Multi-source data preprocessing and error tracing, targeted preprocessing of lidar point cloud, UAV aerial survey data and total station measurement data, selection of standard test field, and calculation of systematic error and random error of each data source; S2: Construct a hierarchical evaluation index system, and establish a three-tier index system including a basic accuracy index layer, a feature accuracy index layer, and a regional accuracy index layer; S3: Multi-dimensional error calculation and weight allocation. Based on the indicator system established in step S2, the error indicators of each level are calculated respectively. The weight of each indicator is determined by the combination of hierarchical analysis and entropy weight method. The comprehensive accuracy index is calculated by weighting. S4: Accuracy level determination and defect location, identify key feature areas of the mining area, construct dynamic error thresholds based on terrain complexity and engineering risk level, and use a dual mechanism of point-by-point verification and local area fitting verification to determine accuracy and locate defect areas. S5: Output of evaluation results and generation of optimization suggestions, output quantitative indicator reports, visualization results and point deviation list, and generate targeted optimization suggestions based on the causes of defective areas.

2. The open-pit mine terrain accuracy evaluation algorithm according to claim 1, characterized in that, The error tracing in S1 specifically includes: selecting a flat area of ​​100m×100m within the mining area as a standard test field, uniformly selecting 50 test points within the test field, and calculating the systematic error of each data source: ,in Let n be the error of the i-th test point, and n be the number of test points; and calculate the random error of each data source: ; The targeted preprocessing in S1 specifically includes: using statistical filtering to remove noise points from the lidar point cloud, and using elevation thresholding combined with slope filtering to separate ground points from non-ground points; using motion reconstruction technology to generate dense point clouds from UAV aerial survey data, and selecting ground control points for adjustment optimization; and using multiple observations and averaging the total station measurement data to remove out-of-limit data.

3. The open-pit mine terrain accuracy evaluation algorithm according to claim 1, characterized in that, The basic accuracy index layer in S2 includes single-point error index, density accuracy index, and overall consistency index; the single-point error index includes elevation mean error and plane mean error, calculated using the Bessel formula; the density accuracy index includes effective point density of point cloud and DEM grid integrity; the overall consistency index includes mean deviation of terrain elevation and mean deviation of plane position. The feature accuracy index layer includes slope accuracy index, aspect accuracy index, and roughness accuracy index. The slope accuracy index includes slope calculation error ∆α, and slope classification as: gentle slope (<15°), gentle slope (15°-30°), steep slope (30°-60°), and very steep slope (>60°). The aspect accuracy index includes aspect fit K, which calculates the cosine similarity of the aspect between the data to be evaluated and the benchmark data. The roughness accuracy index includes terrain roughness deviation ∆R, which calculates the root mean square error deviation of the terrain surface undulation based on a 3m×3m window. The regional accuracy index layer is divided into regions and indexes are designed according to the differences in the topographic function and characteristics of the mining area. Specifically, it includes key feature areas, engineering operation areas and ordinary areas. The key feature areas include the slope error of the slope area and the planar position error of the fault area. The engineering operation areas include the elevation error of the mining bench and the planar error of the slope planning area. The ordinary areas include the combined error of non-feature and non-operation areas.

4. The open-pit mine terrain accuracy assessment algorithm according to claim 1, characterized in that, In step S3, the weight allocation employs the analytic hierarchy process (AHP) combined with engineering requirements to determine subjective weights, and the entropy weight method to calculate objective weights based on real error data from the data source. The final evaluation weights are obtained through combined weighting. The AHP constructs a third-order judgment matrix A as follows: When the consistency ratio CR < 0.1, the matrix is ​​considered to be of good consistency. The entropy weight method includes: standardizing the error index, where the standardization formula is: ； Calculate the weight matrix: ; Calculate information entropy: ,in ; Calculate objective weights: Calculate the combined weights: .

5. The open-pit mine terrain accuracy assessment algorithm according to claim 1, characterized in that, The key feature regions of the mining area identified in S4 include: Slope area identification is achieved through elevation gradient magnitude. Quantifying changes in terrain undulation, when The area was identified as a candidate slope and classified into gentle slope, medium slope, and steep slope according to the gradient amplitude. Fault region identification using Gaussian curvature Quantify the curvature characteristics of the terrain surface, when When a fault is identified as a candidate fault region, and after linear fitting, a fault width W ≥ 2m is identified as a valid fault. The identification of the bench area in the mining area involves traversing the layers at intervals of the designed bench height h, and calculating the actual slope angle α = arctan(G(x,y)) × (180 / π). When α satisfies the design bench slope angle within ±5° and the width W of the adjacent gentle area is ≥3m, it is determined to be a bench area in the mining area.

6. The open-pit mine terrain accuracy assessment algorithm according to claim 1, characterized in that, The construction of the dynamic error threshold in S4 includes: calculating terrain complexity. , where σ is the regional elevation standard deviation and θ is the regional weighted average slope; Dynamic error threshold ,in R represents the basic threshold for ordinary areas, and R represents the engineering risk level. , This is the threshold correction coefficient; the threshold for high-risk areas is stricter than that for medium-risk areas, and the threshold for medium-risk areas is stricter than that for low-risk areas.

7. The open-pit mine terrain accuracy evaluation algorithm according to claim 6, characterized in that, The threshold correction coefficient , By combining measured samples from multiple working conditions in the mining area with multivariate nonlinear least squares regression fitting, the following sample set was constructed: This sample set includes working conditions in leveled mining areas, gentle slope transition zones, steep slopes, vegetated areas, and boundary distortion zones. The optimal allowable threshold for each area was determined using high-precision total station measured data as the true benchmark. Establish a nonlinear regression model The optimization objective is to minimize the sum of squared residuals. The solution is obtained through gradient descent iterative optimization, with constraints. , .

8. The open-pit mine terrain accuracy assessment algorithm according to claim 1, characterized in that, The dual verification mechanism in S4 includes: point-by-point verification. If the absolute elevation error E(x,y)≤T(x,y) is determined to be a qualified point, and if E(x,y)>T(x,y) is marked as a defective point. Local window fitting verification: A 5×5 neighborhood window is used for least squares plane fitting of the defect points, and the root mean square error within the window is calculated. If RMSE ≤ k×T, it is determined to be a random error; if RMSE > k×T, it is determined to be a persistent defect region and the region boundary is marked, where k is the allowable deviation coefficient of the characteristic region.

9. The open-pit mine terrain accuracy assessment algorithm according to claim 1, characterized in that, The weight allocation in S3 adopts an incremental online weight update mechanism: the system caches error samples of newly collected data in real time and maintains the error sample set using a sliding window mechanism; After new samples are added, only the new samples are standardized, their weights are calculated, and their information entropy increments are corrected. The entropy value and objective weights are updated iteratively based on the historical benchmark entropy value. When the error characteristics of multiple sets of new data from a certain type of data source shift significantly, the analytic hierarchy process (AHP) is automatically adjusted to determine the matrix scale. After each data update, the latest combined weights are automatically output iteratively.

10. The open-pit mine terrain accuracy assessment algorithm according to claim 1, characterized in that, The visualization results in S5 include an error distribution heatmap, a feature area accuracy comparison map, and radar maps of various indicators; the optimization suggestions include: supplementing UAV oblique photogrammetry for areas with point cloud obstruction on steep slopes, increasing the density of lidar scanning for areas with insufficient data density, and re-performing GNSS RTK field measurements for areas with concentrated gross error points.

Images