Deformation Monitoring Algorithm for Underground Phosphate Mine Support Structures Based on Laser Point Clouds

Abstract

This invention relates to the field of image data processing technology, specifically to an algorithm for monitoring the deformation of underground phosphate mine support structures based on laser point clouds. The algorithm includes: acquiring laser point cloud data of underground phosphate mine tunnels; identifying a stress reference structure in the point cloud data; determining the stress distribution field of the surrounding rock based on the surface geometric features of the stress reference structure; acquiring the geometric deformation field of the tunnel based on multi-period laser point cloud data; calculating the correlation index between the stress distribution field and the geometric deformation field; determining correction parameters for the geometric deformation field based on the correlation index; and correcting the geometric deformation field based on the correction parameters; and locating the actual deformation area of ​​the underground phosphate mine support structure based on the corrected geometric deformation field. This algorithm, by introducing prior physical information about the stress reference structure, establishes a correlation constraint between the stress distribution field and the geometric deformation, effectively distinguishing and separating registration errors from the actual deformation of the support structure, thus improving the reliability of support structure deformation monitoring.

Description

Technical Field

[0001] This invention relates to the field of image data processing technology, specifically to an algorithm for monitoring the deformation of underground phosphate mine support structures based on laser point clouds. Background Technology

[0002] Laser point cloud technology acquires a set of three-dimensional coordinates of an object's surface through lidar scanning. Compared to traditional total stations, it offers advantages such as higher data acquisition efficiency, greater information completeness, and stronger safety through non-contact measurement, and has been widely applied in the field of deformation monitoring of underground mine support structures. Currently, mine support structure monitoring based on laser point clouds typically employs the cross-sectional method: first, the acquired multi-period point cloud data is registered and stitched together; then, cross-sections are obtained by slicing along the tunnel's extension direction; finally, the cross-sections from multiple periods are compared and analyzed to identify the deformation of the support structure.

[0003] However, existing technologies have significant limitations when processing point cloud data from long-distance tunnels: due to the cumulative effect of point cloud registration errors during long-distance scanning, the geometric deformation field obtained from multi-stage point cloud comparisons inevitably contains registration errors introduced by factors such as registration uncertainty, motion distortion, and sensor errors. These errors exhibit spatial distributions highly similar to the geometric characteristics of minute deformations in the actual support structure (such as steel strip bending caused by surrounding rock convergence, and mesh bulging), making it difficult to effectively distinguish between registration errors and actual deformations in multi-stage point cloud comparisons. This leads to misidentification of registration errors as actual deformations of the support structure, severely impacting the accuracy and reliability of monitoring results. Summary of the Invention

[0004] This invention provides an algorithm for monitoring the deformation of underground phosphate mine support structures based on laser point clouds, in order to solve existing problems.

[0005] The present invention provides an algorithm for monitoring the deformation of underground phosphate mine support structures based on laser point clouds, which employs the following technical solution: One embodiment of the present invention provides a deformation monitoring algorithm for underground phosphate mine support structures based on laser point clouds, comprising: acquiring laser point cloud data of an underground phosphate mine tunnel; identifying a stress reference structure in the point cloud data; and determining the stress distribution field of the surrounding rock based on the surface geometric features of the stress reference structure; acquiring the geometric deformation field of the tunnel based on multiple periods of the laser point cloud data; calculating a correlation index between the stress distribution field and the geometric deformation field; determining a correction parameter for the geometric deformation field based on the correlation index; and correcting the geometric deformation field based on the correction parameter; and locating the actual deformation area of ​​the underground phosphate mine support structure based on the corrected geometric deformation field.

[0006] Further, determining the stress distribution field of the surrounding rock based on the surface geometric features of the stress reference structure includes: extracting the edge point cloud of the stress reference structure, and determining the principal plane and geometric center of the stress reference structure based on the edge point cloud, and determining the normal vector of the principal plane as the reference axis of the stress reference structure; dividing the surface of the stress reference structure into multiple radial planes along the circumference based on the reference axis, and calculating the radial curvature of the sampling points on each radial plane; segmenting the arched stress region of the stress reference structure based on the sign flip position of the radial curvature; determining the relative stress at the location of the stress reference structure according to the degree of change of the radial curvature of the sampling points in the arched stress region relative to the radial curvature of the initial unstressed state, and constructing a sparse stress distribution field.

[0007] Furthermore, it also includes: spatially interpolating the sparse stress distribution field to obtain a dense stress distribution field with the same spatial resolution as the geometric deformation field.

[0008] Further, determining the correction parameters of the geometric deformation field based on the correlation index includes: establishing local neighborhood windows point by point in the geometric deformation field; calculating the local correlation coefficient between the stress distribution field and the geometric deformation field within the local neighborhood window; establishing a negative correlation between the local correlation coefficient and the correction magnitude; wherein the negative correlation results in a larger correction magnitude corresponding to a lower local correlation coefficient; and determining the correction parameters of the center point of the local neighborhood window based on the negative correlation.

[0009] Further, the step of correcting the geometric deformation field based on the correction parameters includes: establishing local neighborhood windows point by point in the geometric deformation field, extracting statistical features of the geometric deformation field and the stress distribution field within the local neighborhood windows, and establishing a correction model based on the statistical features; wherein, the correction model is used to align the geometric deformation field with the statistical characteristics of the stress distribution field; and based on the correction parameters, fusing the output value of the correction model with the original value of the geometric deformation field to obtain the correction value of the center point of the local neighborhood window.

[0010] Further, the step of extracting the statistical features of the geometric deformation field and the stress distribution field within the local neighborhood window, and establishing a correction model based on the statistical features, includes: extracting the central tendency features and dispersion features of the geometric deformation field and the stress distribution field within the local neighborhood window, respectively; establishing a statistical mapping relationship from the stress distribution field to the geometric deformation field based on the degree of deviation of each point in the stress distribution field from its central tendency feature, combined with the dispersion feature of the geometric deformation field; determining the mapping direction of the statistical mapping relationship according to the sign of the local correlation coefficient, and establishing a correction model.

[0011] Furthermore, it also includes: for the same spatial location point in the geometric deformation field, obtaining multiple candidate correction values ​​obtained by correcting it in multiple overlapping local neighborhood windows respectively; fusing the multiple candidate correction values ​​to determine the final correction result of the spatial location point.

[0012] Further, the step of fusing multiple candidate correction values ​​to determine the final correction result for the spatial location point includes: determining the fusion weight corresponding to each candidate correction value based on the interpolation confidence distribution determined during the spatial interpolation of the stress distribution field; wherein the interpolation confidence is negatively correlated with the distance from the spatial location point to the stress reference structure; and performing weighted fusion of the multiple candidate correction values ​​based on the fusion weight to obtain the final correction result.

[0013] Furthermore, the stress reference structure includes an anchor bolt and a tray; The step of identifying stress reference structures in the point cloud data includes: identifying anchor bolts and trays in the point cloud data based on laser reflection intensity characteristics and geometric shape characteristics; and performing local point cloud density enhancement processing on the point cloud data of the identified anchor bolts and trays.

[0014] Furthermore, the step of locating the actual deformation area of ​​the underground phosphate mine support structure based on the corrected geometric deformation field includes: setting a deformation threshold based on the measurement error of the laser point cloud data acquisition device; extracting deformation areas whose absolute deformation exceeds the deformation threshold in the corrected geometric deformation field; and performing morphological processing on the deformation areas to determine the actual deformation area.

[0015] The beneficial effects of the technical solution of the present invention are: In this embodiment of the invention, laser point cloud data of underground phosphate mine tunnels is acquired, stress reference structures are identified in the point cloud data, and the stress distribution field of the surrounding rock is determined based on the surface geometric features of the stress reference structures; the geometric deformation field of the tunnels is acquired based on multi-period laser point cloud data; the correlation index between the stress distribution field and the geometric deformation field is calculated; the correction parameters of the geometric deformation field are determined based on the correlation index, and the geometric deformation field is corrected based on the correction parameters; the actual deformation area of ​​the underground phosphate mine support structure is located based on the corrected geometric deformation field.

[0016] This invention, by introducing prior physical information from a stress reference structure and establishing correlation constraints between the stress distribution field and geometric deformation, effectively distinguishes and separates registration errors from actual support structure deformation, thus improving the reliability of support structure deformation monitoring based on laser point clouds. Furthermore, by analyzing the surface geometric features of the stress reference structure to invert the surrounding rock stress distribution, stress sensors are not required, enabling indirect acquisition of the stress field based on existing support components, reducing the complexity and deployment cost of monitoring hardware. Moreover, by adaptively determining correction parameters using the local correlation index between the stress distribution field and the geometric deformation field, the correction magnitude of the registration error matches the error reliability, avoiding over-correction or under-correction and ensuring the accuracy and rationality of the deformation field correction. Finally, through a positioning strategy combining morphological processing and threshold screening, the actual deformation area of ​​the support structure is accurately extracted based on the corrected geometric deformation field, providing reliable data support for subsequent support reinforcement and engineering treatment, and ensuring the structural safety of underground phosphate mine tunnels. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating the deformation monitoring algorithm for underground phosphate mine support structures based on laser point clouds provided in this application embodiment; Figure 2 This is a flowchart illustrating the deformation monitoring algorithm for underground phosphate mine support structures based on laser point clouds in a specific application scenario provided in this application embodiment. Detailed Implementation

[0019] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following detailed description, in conjunction with the accompanying drawings and preferred embodiments, provides a specific implementation method, structure, features, and effects of the invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] The following description, in conjunction with the accompanying drawings, details a specific solution provided by the present invention.

[0022] like Figure 1 As shown in the figure, this application provides an algorithm for monitoring the deformation of underground phosphate mine support structures based on laser point clouds, including: Step S110: Acquire laser point cloud data of underground phosphate mine tunnels, identify stress reference structures in the point cloud data, and determine the stress distribution field of the surrounding rock based on the surface geometric features of the stress reference structures.

[0023] The aforementioned underground phosphate mine tunnels refer to underground passageways excavated within the underground mining area of ​​phosphate deposits to serve functions such as ore transportation, personnel passage, and ventilation and drainage. Underground phosphate mine tunnels typically employ anchor-mesh support structures for surrounding rock reinforcement. This involves laying metal wire mesh on the roof and sidewalls and covering it with corrugated steel strips. The steel strips are then firmly pressed against the rock surface by evenly distributed anchor bolt trays, with the exposed ends of the anchor bolts and trays clearly visible, forming a typical stress reference structure. In transport tunnels with relatively short service lives, due to mining pressure or the rheological effects of weak surrounding rock, the support structure is prone to pressure-induced deformations such as steel strip bending and mesh bulging. Therefore, periodic deformation monitoring is necessary to ensure structural robustness and safety.

[0024] The aforementioned laser point cloud data for underground phosphate mine roadways refers to the three-dimensional digital representation of the roadway space obtained through lidar scanning. Specifically, it is a set of three-dimensional coordinates of discrete sampling points on the roadway surface (including spatial location information and laser reflection intensity information for each point). The laser point cloud data for underground phosphate mine roadways records the geometric shape and spatial distribution of the roadway support structure (including anchor bolts, support plates, steel strips, and surrounding rock surfaces) in point cloud form. It has millimeter-level measurement accuracy and can completely reflect the roadway cross-sectional contour and detailed geometric features of the support components, providing high-resolution basic data support for subsequent deformation analysis of the support structure. One method for acquiring the aforementioned laser point cloud data includes: fixing a millimeter-precision multi-line rotating mechanical lidar scanning system at the center of the roof of a special explosion-proof battery-powered locomotive used in mining; allowing the scanning system to move at a constant speed along the roadway with the locomotive and complete continuous scanning; simultaneously introducing a wheeled odometer and an inertial navigation unit for auxiliary measurement, and performing real-time calculation of the pose information during the scanning process; based on the pose calculation results, stitching together the continuous scan segments and unifying the coordinates to finally construct a complete three-dimensional point cloud model covering the entire roadway. During the data acquisition process, the original point cloud needs to be denoised to eliminate interference from phosphate rock dust and ensure that the point cloud data quality meets the requirements for fine identification of the support structure.

[0025] The aforementioned stress reference structure refers to a support component installed in the support system of underground phosphate mine roadways, possessing clear mechanical response characteristics and whose geometric shape is physically related to the stress state of the surrounding rock. The stress reference structure is typically a metal plate component, installed between the exposed end of the anchor bolt and the surrounding rock of the roadway. Pre-tightening force is applied by tightening nuts to press the corrugated steel strip and metal mesh firmly against the rock surface, serving as a surface protection component. This ensures that the axial tensile force of the anchor bolt is evenly transferred to the surface of the surrounding rock, preventing rock fragments from loosening and falling off, and maintaining the stability of the surrounding rock in the roadway.

[0026] It is understandable that during the service of the tunnel, the stress reference structure directly bears the convergent compression and shear stress from the surrounding rock, and its surface geometry undergoes regular deformation with changes in load. When the support is subjected to pressure from the surrounding rock, its arched stress area gradually transforms from an initial curved state to a planar state, and the radial curvature approaches zero from its initial value. The degree of curvature change is positively correlated with the magnitude of the load. This deterministic physical mapping relationship between geometric deformation and mechanical state makes the geometric characteristics of the stress reference structure an observable characterization of the surrounding rock stress distribution. The core technical value of setting up the stress reference structure in the above scheme lies in providing physical prior information independent of the point cloud registration process for laser point cloud deformation monitoring. Since the deformation of the stress reference structure is driven by the actual surrounding rock stress, there is a physical correlation between the stress distribution field inferred from its geometric characteristics and the actual deformation field of the support structure; while the point cloud registration error is a technical deviation in the data acquisition and processing process and has no physical correlation with the surrounding rock stress distribution. Therefore, the stress distribution field formed by the stress reference structure can be used as a reference benchmark. By analyzing the correlation between it and the geometric deformation field, the registration errors mixed in the geometric deformation field can be effectively identified and corrected, thereby achieving the separation of real deformation and error and improving the accuracy and reliability of deformation monitoring of the support structure.

[0027] Optionally, the stress reference structure mentioned above includes anchor bolts and trays; the identification of stress reference structures in point cloud data includes: identifying anchor bolts and trays in point cloud data based on laser reflection intensity characteristics and geometric shape characteristics; and performing local point cloud density enhancement processing on the point cloud data of the area where the identified anchor bolts and trays are located.

[0028] The aforementioned anchor bolts and trays together constitute the core force transmission unit of the anchor bolt-tray support system, a key component combination for achieving active support of the surrounding rock in the roadway. The anchor bolt is a slender, rod-shaped metal component, its inner end anchored in a stable stratum deep within the surrounding rock, while its outer end protrudes from the roadway surface. It is mechanically connected to the tray via a threaded structure. The tray is a metal plate component, typically an arched or dished steel plate structure, with a central mounting hole through which the exposed end of the anchor bolt passes, and is pressed against the roadway's surrounding rock surface by the tightening of a nut. In this combined structure, the anchor bolt provides axial preload, and the tray, acting as a force transmission medium, converts the axial tension of the anchor bolt into uniformly distributed pressure on the surrounding rock surface. Simultaneously, it presses the corrugated steel strip and metal mesh support components tightly against the rock surface, preventing rock fragments from loosening and falling off, thus forming active constraint support for the roadway roof and sidewalls. In the laser point cloud data, the exposed end of the anchor bolt exhibits a cylindrical or conical high reflectivity feature, while the tray exhibits a circular or square plate-like high reflectivity feature. The two have clear geometric shapes and fixed spatial positions, forming an identifiable stress reference structure as a whole.

[0029] The stress distribution field of the surrounding rock mentioned above refers to the spatial distribution field characterizing the internal stress state of the rock mass surrounding the underground phosphate mine tunnel. Specifically, it is the collection of the relative magnitudes of stress borne by the surrounding rock at various spatial locations. The stress distribution field of the surrounding rock reflects the spatial pattern of stress transmission and concentration within the surrounding rock around the tunnel under the action of loads such as mining pressure, rock mass self-weight, and tectonic stress. The stress concentration areas usually correspond to the locations where the support structure bears a large load. In the laser point cloud monitoring system, this stress distribution field is indirectly obtained through the mechanical response characteristics of the stress reference structure. Its value adopts relative dimensions (normalized based on the maximum bearing capacity of the stress reference structure) to characterize the relative intensity of the surrounding rock stress at each location relative to the ultimate state. This distribution field has a dual expression form of sparse and dense: at the installation location of the stress reference structure, it presents discrete relative stress values, characterizing the stress state of local anchoring points; after spatial interpolation expansion, it forms a dense distribution field covering the entire tunnel, characterizing the continuous variation trend of surrounding rock stress along the tunnel axis and cross-sectional direction, providing a physical benchmark for judging the authenticity of support structure deformation.

[0030] In the embodiments of this application, the stress distribution field of the surrounding rock can be determined by at least one of the following multiple implementation methods: First implementation method: Based on radial curvature distribution analysis; Optionally, the above-mentioned determination of the stress distribution field of the surrounding rock based on the surface geometric features of the stress reference structure includes: extracting the edge point cloud of the stress reference structure, and determining the principal plane and geometric center of the stress reference structure based on the edge point cloud, and determining the normal vector of the principal plane as the reference axis of the stress reference structure; based on the reference axis, dividing the surface of the stress reference structure into multiple radial planes along the circumference, and calculating the radial curvature of the sampling points on each radial plane; based on the sign flip position of the radial curvature, segmenting to obtain the arched stress region of the stress reference structure; and determining the relative stress at the location of the stress reference structure according to the degree of change of the radial curvature of the sampling points in the arched stress region relative to the radial curvature of the initial unstressed state, thereby constructing a sparse stress distribution field.

[0031] The aforementioned edge point cloud refers to the set of discrete laser scanning points located at the contour boundary of the stress reference structure (pallet). This point cloud data characterizes the geometric transition boundary between the pallet and the surrounding roadway rock or steel strip structure. In practical processing, edge point clouds can be extracted from the pallet area point cloud through threshold segmentation based on laser reflection intensity or edge detection algorithms based on geometric location. Specifically, initial segmentation can be performed using the difference in reflection intensity between the pallet surface and the background rock surface. Subsequently, the outer contour boundary of the pallet can be accurately located by calculating the abrupt change in the point cloud normal vector or the local extremum of curvature, thereby obtaining edge point cloud data distributed around the pallet circumferentially.

[0032] The aforementioned principal plane refers to the approximate plane obtained by fitting the pallet edge point cloud using a principal component analysis algorithm. This plane represents the orientation of the reference plane where the pallet chassis is located. The geometric center refers to the centroid of the geometric shape formed by the pallet edge after projection onto the principal plane. It is usually obtained by fitting the projected edge as a quadrilateral and calculating the intersection of its diagonals. The reference axis refers to the normal vector perpendicular to the principal plane and passing through the geometric center. This axis represents the axial direction of the pallet anchor system and provides a rotational reference for subsequent radial segmentation. In the specific calculation, the edge point cloud is first decomposed into principal components. The eigenvector corresponding to the largest eigenvalue is used as the normal vector of the principal plane, and the other two eigenvectors span the principal plane. Then, the edge point cloud is orthogonally projected onto the principal plane. The projection points are fitted into four straight line segments using the least squares method or the RANSAC algorithm. The coordinates of the two sets of diagonal intersection points are calculated and averaged as the geometric center. Finally, a local cylindrical coordinate system with this geometric center as the origin and the reference axis as the Z-axis is established. It is understood that the aforementioned principal component analysis algorithm, least squares method, RANSAC algorithm, etc. are all mature existing technologies. For their implementation methods and working principles, please refer to the relevant technologies. The embodiments in this application will not be repeated.

[0033] The aforementioned radial surfaces refer to a family of half-planes distributed circumferentially around a reference axis as the rotation axis, with specific angular steps. Each radial surface includes the reference axis and extends radially along the tray, intersecting the tray surface to form a radial cross-section. In practical implementation, the tray surface can be uniformly divided into 360 radial surfaces circumferentially using an angular step of 1 degree. The segmentation process is based on an established local cylindrical coordinate system, generating radial surface equations at equal intervals according to azimuth angles. Subsequently, a spatial intersection algorithm is used to calculate the intersection line between each radial surface and the point cloud of the tray surface, and the discrete sampling points on the intersection line are extracted as the feature point set for that radial direction.

[0034] Radial curvature refers to the degree of curvature of the tray surface along the radial direction (i.e., away from the reference axis), mathematically defined as the local curvature of the radial section. For each sampling point on the radial surface, the radial curvature can be obtained by fitting the local curve in the neighborhood of that point and calculating the reciprocal of its radius of curvature. Specifically, a three-point fitting method or a curve curvature formula based on differential geometry can be used. In numerical processing, considering the discreteness of the point cloud, three to five consecutive sampling points on the radial section are usually selected for quadratic curve fitting. The ratio of the second derivative to the first derivative of the fitted curve at that point is calculated as an approximation of the radial curvature. The sign of the curvature value is determined according to the direction of curvature: when the surface is concave towards the reference axis, it is defined as positive curvature, and vice versa.

[0035] The aforementioned arched stress-bearing area refers to the functional area on the pallet surface that undergoes primary plastic deformation during stress. This area is located around the central mounting hole of the pallet and presents an arched or dish-shaped protrusion. It is the main load-bearing part that bears the surrounding rock pressure and transmits the load to the anchor bolts. The identification of the arched stress-bearing area and the outer base area is based on the sign reversal characteristic of radial curvature: on the radial section, the arched area exhibits a positive curvature characteristic that bulges outward from the center, while the base area, close to the rock surface, exhibits a relatively flat or reverse curvature characteristic. The curvature value at the boundary between the two reverses sign. By detecting the transition position on the radial section where the curvature sign changes from positive to negative, the boundary between the arched area and the base area can be determined. The area with a positive curvature sign closer to the geometric center is classified as the arched stress-bearing area, while the outer area is classified as the base contact area.

[0036] Based on the aforementioned logic, the stress state at corresponding locations on the pallet can be analyzed by examining the curvature distribution of the pallet arch. Let the maximum withstand stress of the pallet be 1. When the actual stress is 1, the entire arched area of ​​the pallet is completely deformed; when the actual stress is 0, the radial curvature is uniform throughout the arched area, indicating no stress. The greater the change in radial curvature (minimum becoming 0) throughout the arched area of ​​the pallet, the greater the stress it bears. Let the relative stress borne by an anchor pallet be... The calculation method is as follows: ; In the formula, This represents the initial radial curvature of the arched region of the tray when it is not subjected to external stress. Represents the surface of the arched area The th uniform sampling point The radial curvature at each point. The curvature distribution of the arched region and the initial radial curvature. The greater the difference, the greater the stress on the tray. The stress magnitude for each anchor tray is calculated. This refers to the stress distribution in the surrounding rock within mine roadways. The distribution of the stress field is denoted as the relative stress field. Thus, by analyzing the curvature distribution of the arched region of the anchor bolt tray, the stress distribution at each anchor bolt location was obtained.

[0037] It is understandable that, in the initial installation state, the arched area of ​​the tray has a uniform initial radial curvature. When subjected to surrounding rock pressure, the arched area gradually transforms into a planar state, with the local radial curvature approaching zero from its initial value. The magnitude of curvature change is positively correlated with the stress level. The above scheme establishes a mapping relationship with the relative stress value by calculating the relative rate of change between the current radial curvature and the initial radial curvature at each sampling point and statistically analyzing the average degree of change at all sampling points within the arched area: when the curvature remains unchanged, the relative stress is zero; when all curvatures approach zero, the relative stress reaches its maximum value (normalized to 1). The relative stress value calculated in this way characterizes the relative intensity of the surrounding rock stress at the location of the stress reference structure. Aggregating the relative stress values ​​at each location according to spatial position constitutes a sparse stress distribution field.

[0038] The second implementation method: based on surface normal vector consistency analysis; In this embodiment, the geometric characteristic of the surface normal vector of the arched area of ​​the pallet changing from a divergent distribution to a uniform distribution during the flattening process under pressure can be utilized. For example, the local normal vector of each sampling point on the pallet surface is first calculated, and its distribution concentration in the hemispherical space is statistically analyzed. In the initial state, the normal vector of the arched area shows a divergent distribution around the axis, while after being compressed, it tends to be oriented parallel to the axis. By calculating the degree of dispersion of the normal vector distribution (such as the ratio of the second and third eigenvalues ​​based on principal component analysis) or calculating the variance of the angle between the normal vector and the pallet axis, the flatness of the pallet surface can be quantitatively evaluated, and then mapped to the relative stress magnitude.

[0039] The third implementation method: based on surface fitting parameter analysis; In this embodiment, the stress state can be characterized by fitting the point cloud of the tray surface to a quadratic surface (such as a parabola or a spherical cap) and utilizing the changes in the fitting parameters. Specifically, the least squares method is used to fit the point cloud of the arched region of the tray to a quadratic surface equation, and the Gaussian curvature or mean curvature of the surface is extracted as a geometric feature parameter; or the quadratic coefficient of the parabolic equation is extracted as a quantitative index of the steepness of the arch. As the tray is compressed, the absolute value of the curvature parameter of the fitted surface decreases or the absolute value of the quadratic coefficient decreases. By establishing a difference model between these parameters and the initial unstressed state, the relative stress distribution is determined.

[0040] Fourth implementation method: based on relative height change analysis; In this embodiment, the stress state can be determined by analyzing the height change of the apex (or center region) of the arched area of ​​the pallet relative to the edge reference plane. Specifically, firstly, a reference plane is determined based on the point cloud of the pallet edge, and the orthogonal distance from each point in the arched area to this reference plane is calculated. Initially, the center point of the arched area has the maximum orthogonal distance (maximum arch height), and the arch height gradually decreases with compression deformation. By calculating the relative rate of change between the current arch height and the initial arch height, or by calculating the variance change of the height of each point in the arched area, a mapping relationship with relative stress is established. This embodiment is computationally simple and suitable for quickly assessing the degree of pressure on the pallet.

[0041] Optionally, the above-mentioned algorithm for monitoring the deformation of underground phosphate mine support structures based on laser point clouds may further include: spatial interpolation of the sparse stress distribution field to obtain a dense stress distribution field with the same spatial resolution as the geometric deformation field.

[0042] It is understandable that the stress distribution field, calculated based on a discrete stress reference structure (anchor bolt tray), only has discrete relative stress values ​​at the anchor bolt installation location, forming spatially sparse stress data. In contrast, the geometric deformation field, obtained through multi-period comparison of laser point clouds, covers the entire tunnel surface and has a high point cloud density, forming spatially dense deformation data. This inconsistency in spatial resolution makes it impossible to directly compare the two fields point-to-point in the spatial domain, making it difficult to assess the degree of matching between the geometric deformation field and the stress distribution at each spatial location. Therefore, spatial interpolation is needed to expand the sparse stress distribution field into a dense stress distribution field with the same spatial resolution as the geometric deformation field, achieving alignment of the two physical fields on the spatial grid. The core purpose of the above scheme in converting the sparse stress distribution field into a dense stress distribution field is to establish a stress reference benchmark covering the entire space, enabling any spatial location in the geometric deformation field to obtain a corresponding stress distribution reference value, thereby supporting point-to-point correlation calculations and correction parameter determination. This transformation extends stress information that was originally only observable at discrete anchor points to the entire surface of the tunnel through reasonable spatial correlation assumptions (the continuity and correlation of stress in space). This provides a basis for judging spatial continuity in the identification of registration errors, ensuring that correction operations can be implemented in every local area of ​​the geometric deformation field, and improving the spatial consistency and integrity of the correction results.

[0043] The aforementioned spatial interpolation technique is based on the assumption of the continuity and gradual change of the stress distribution field in space. It uses the relative stress values ​​at known discrete anchor locations as sample points and predicts the stress value at unknown locations by establishing a mapping function between spatial locations and stress values. Essentially, this technique extrapolates physical observation information from a limited number of measuring points to the entire spatial domain based on spatial correlation laws, constructing a continuously differentiable stress distribution surface. This ensures that the dense stress distribution field maintains consistency with the original measuring points while conforming to the physical law of continuous propagation of the stress field in the rock mass. Gaussian process regression is a nonparametric interpolation method based on a Bayesian framework. This method treats the stress distribution as a function of a Gaussian stochastic process and characterizes the spatial correlation of stress values ​​by defining a covariance function (such as the squared exponential covariance function) between spatial locations. During the interpolation process, this method not only provides the predicted stress value at unknown locations but also provides the prediction variance as a quantitative indicator of uncertainty. This variance is positively correlated with the distance to the nearest anchor location; the greater the distance, the lower the interpolation confidence. This characteristic precisely matches the physical understanding that stress estimation is more reliable around the stress reference structure, while uncertainty increases in areas far from the reference structure. The inverse distance weighted interpolation method is based on the principle of spatial proximity. It assumes that the stress value at an unknown location is correlated with the weighted average of the stress values ​​at surrounding known sample points, with the weights negatively correlated with distance (usually a power function of the inverse of the distance). This method is computationally simple, and the degree of local smoothing can be controlled by adjusting the power exponent: a larger power exponent results in an interpolation closer to the nearest neighbor sample (local sensitivity), while a smaller power exponent considers a larger spatial average (global smoothing), making it suitable for tunnel areas with relatively uniform stress distribution gradients. The Kriging interpolation method is based on regionalized variable theory. It uses a semi-variogram to describe the decay of the correlation between stress values ​​and spatial distance, and employs a semi-variogram model (such as a spherical model, exponential model, or Gaussian model) for spatial structure analysis. This method considers not only the distance between the sample point and the prediction point but also the spatial configuration relationship between the sample points. By solving the Kriging equations, it obtains the optimal unbiased estimate, resulting in a statistically minimal estimation variance. It is suitable for scenarios where the stress distribution has obvious spatial structure (such as stress transmission along the tunnel axis). Radial basis function interpolation uses radially symmetric basis functions (such as Gaussian functions, multiple quadratic surface functions, or thin plate spline functions) to construct a continuous approximation of the stress distribution. This method determines the weighting coefficients of each basis function by solving a system of linear equations, ensuring that the interpolation surface strictly passes through all known sample points and forms a smooth transition between sample points.

[0044] Step S120: Obtain the geometric deformation field of the tunnel based on multi-phase laser point cloud data.

[0045] The geometric deformation field of the aforementioned roadway refers to the spatial geometric difference field obtained by comparing roadway laser point cloud data acquired at different time points. Mathematically, it is characterized as the three-dimensional displacement vector distribution of corresponding points or regions between two point clouds. This field covers the entire surface of the roadway in the form of a scalar or vector, numerically reflecting the degree of geometric morphological change of the roadway support structure and surrounding rock within the detection time interval. Positive values ​​usually indicate deformation converging towards the roadway interior, while negative values ​​indicate expansion or extension outward. As a direct observation indicator of the health status of the support structure, the geometric deformation field provides a quantitative basis for identifying potential structural failure areas. The geometric deformation field consists of two inseparable components: one is the actual structural deformation caused by physical processes such as stress release of surrounding rock, transmission of mining pressure, or yielding of support components; the other is the technical registration error introduced by factors such as cumulative errors in the multi-stage point cloud registration process, motion distortion of scanning equipment, sensor system errors, and environmental interference. These two types of components are highly similar in geometric features, both exhibiting local displacement or shape changes. This makes it difficult to distinguish them using simple geometric threshold criteria in traditional detection processes, thus creating a technical challenge of misidentification.

[0046] One possible implementation of step S120 above for obtaining the roadway geometric deformation field is as follows: Acquire the current laser point cloud data at the current detection time point, and perform overall registration with the baseline (or previous) laser point cloud data acquired at historical detection time points. Spatial alignment of the two point clouds in a unified coordinate system is achieved through an iterative nearest-point algorithm or a feature-based registration method. During the registration process, the rigid structural features of the roadway or manually placed control points are used as constraints to solve for the optimal rigid body transformation matrix (including rotation and translation parameters) between the two data periods, minimizing the global registration error of the current point cloud relative to the baseline point cloud. Based on the overall registration, geometric differences are calculated and the field is constructed. Specifically, when using the cross-section method, cross-sectional point cloud slices perpendicular to the roadway axis are extracted along the roadway extension direction at specific intervals (e.g., every 0.5 meters or 1 meter). Contour extraction and center alignment are performed on the corresponding cross-sections of the current and baseline periods, and the shortest distance or normal projection distance from each sampling point on the cross-section to the corresponding period's cross-sectional point cloud is calculated as the deformation at that location. When using the point-by-point comparison method, the nearest neighbor corresponding point in the baseline point cloud is directly searched for each point in the current point cloud under a unified coordinate system. The Euclidean distance or the distance difference along the normal direction of the surface between the two points is calculated, and the calculation result is assigned to the corresponding spatial coordinate position in the current point cloud. Through the above spatial difference calculation, the discrete point-by-point or cross-sectional difference values ​​are continuously interpolated to construct a dense geometric deformation field covering the entire tunnel, where each spatial location is assigned a deformation value that characterizes the degree of change in geometric shape relative to the baseline period.

[0047] The aforementioned cross-section method is a support structure deformation detection method based on laser point cloud data. Its technical process includes spatial slicing and cross-sectional contour comparison of registered multi-period point cloud data. First, the cross-section method vertically cuts the registered current point cloud and the historical baseline point cloud along the roadway extension direction at specific intervals to extract cross-sectional point cloud slices reflecting the roadway's cross-sectional contour characteristics. Then, it performs geometric registration and difference calculation on cross-sections from different periods at the same spatial location. By measuring the normal distance or spatial displacement between the two cross-sectional contours, the local deformation at that cross-sectional location is obtained. Finally, the deformation results of each cross-section are continuously processed along the roadway axis to construct a geometric deformation field covering the entire roadway, thereby locating the deformation area of ​​the support structure.

[0048] Step S130: Calculate the correlation index between the stress distribution field and the geometric deformation field.

[0049] The aforementioned correlation index is a statistical metric used to quantify the degree of coordination between the numerical changes of the stress distribution field and the geometric deformation field at corresponding spatial locations. Its value characterizes the linear or nonlinear correlation strength between the two fields within a local region, reflecting the degree of matching between the actual support deformation caused by the release of surrounding rock stress and the observed values ​​of the geometric deformation field. The correlation index is constructed based on the core assumption that "actual deformation and stress distribution have a physical correlation, while registration error and stress distribution have no physical correlation." By calculating the statistical dependence of the two fields within a local neighborhood window, it provides a quantitative basis for assessing the reliability of data at various spatial locations of the geometric deformation field and determining the correction magnitude. A higher correlation index value indicates that the geometric deformation in that local region is more likely to be driven by actual stress, with less interference from registration error; conversely, a lower correlation index indicates a significant proportion of registration error, requiring a larger correction.

[0050] Correlation indices can be measured using various statistical correlation methods: (1) The Pearson product-moment correlation coefficient is used to quantify the linear correlation between two fields, with a value range of [-1, 1]. The closer the absolute value is to 1, the stronger the linear correlation. (2) The Spearman rank correlation coefficient is used to evaluate the monotonic correlation based on the rank information of the data, which is suitable for scenarios where there is a nonlinear but monotonic mapping between stress and deformation. (3) Mutual information is used to measure the nonlinear statistical dependence between two fields, capturing more complex probability distribution correlations based on information entropy theory. (4) Normalized covariance or cosine similarity is used to characterize the trend strength of the coordinated change of the two fields or the consistency of the spatial vector direction, respectively. These indices can be applied independently or in combination to the correlation calculation of local neighborhood windows to adapt to different stress-deformation physical response relationships and data distribution characteristics.

[0051] In the above scheme, the correlation index between the stress distribution field and the geometric deformation field aims to establish a quantitative mapping verification mechanism between the physical stress state and geometric observation data. By using statistical correlation metrics, prior physical information about the mechanical behavior of the surrounding rock is introduced into the geometric deformation analysis process. This calculation process essentially constructs a data reliability assessment framework with stress distribution as a reference benchmark. This allows data at each spatial location in the geometric deformation field to be graded according to its degree of conformity with the physical stress release law, thus providing an objective spatial decision-making basis for subsequent targeted correction operations. The correlation index provides a quantitative foundation for the implementation of adaptive correction strategies, enabling the correction magnitude to be dynamically adjusted based on the consistency between local data and physical laws. By treating the stress distribution field as an independent physical quantity reflecting the actual stress state of the support structure, the correlation between the geometric deformation field and this reference field directly characterizes the proportion of the actual deformation component in the observation data: high correlation regions indicate a high degree of consistency between geometric changes and stress-driven forces, indicating high data reliability and the need for retention; low correlation regions indicate that the observed values ​​deviate from physical expectations, indicating registration error interference and requiring significant correction. This spatial discrimination mechanism based on physical correlation effectively decouples the real deformation information and technical errors in the geometric deformation field, ensuring that the corrected deformation field can more accurately reflect the real mechanical response of the support structure under the stress of the surrounding rock, and providing accurate data support for the reliable assessment of the health status of the support structure.

[0052] Step S140: Based on the correlation index, determine the correction parameters of the geometric deformation field, and correct the geometric deformation field based on the correction parameters.

[0053] The aforementioned correction parameter for the geometric deformation field is a quantitative control factor used to regulate the correction magnitude of the geometric deformation field at specific spatial locations. Its value characterizes the degree to which the original geometric deformation field data is adjusted towards the correction target value inferred based on physical correlation. This parameter exhibits a continuous spatial distribution and establishes a mapping relationship with the correlation index, enabling differentiated correction treatment for regions with different confidence levels in the geometric deformation field: lower correction parameters are assigned to regions with high correlation to preserve the characteristics of the original observation data to the greatest extent; higher correction parameters are assigned to regions with low correlation to further reduce the impact of registration errors and move closer to the physically expected deformation pattern. Through the spatial distribution of this correction parameter, an adaptive and non-uniform correction operation is achieved for the geometric deformation field, ensuring that the corrected geometric deformation field conforms to the stress-deformation physical correlation law in its overall distribution while maintaining the basic spatial structure characteristics of the original data.

[0054] In the embodiments of this application, the correction parameters can be determined through at least one of the following multiple implementation methods: The first implementation method: a method for determining correction parameters based on continuous function mapping; Optionally, the above-mentioned determination of correction parameters for the geometric deformation field based on correlation indices includes: establishing local neighborhood windows point by point in the geometric deformation field, calculating the local correlation coefficient between the stress distribution field and the geometric deformation field within the local neighborhood window; establishing a negative correlation between the local correlation coefficient and the correction magnitude; wherein, the negative correlation means that the lower the local correlation coefficient, the greater the corresponding correction magnitude; and determining the correction parameters for the center point of the local neighborhood window based on the negative correlation.

[0055] A local neighborhood window refers to a local spatial region in a geometric deformation field that extends outward from a specific spatial location. It is typically rectangular or circular, and its spatial scale determines the degree of localization in the correlation analysis. In engineering implementation, considering computational efficiency and the reasonableness of the local stationarity assumption, a 5×5 or other odd-dimensional square window can be used, ensuring that the center point is located at the geometric center of the window. The point cloud data within this window are assumed to be under the same local deformation mode and stress state, thus supporting a robust estimation of the correlation characteristics between the two fields. The selection of the window size requires a trade-off between spatial resolution and statistical stability: a window that is too small may lead to insufficient samples and increased statistical fluctuations, while a window that is too large may smooth out the true local deformation gradient changes, resulting in a loss of spatial positioning accuracy.

[0056] The local correlation coefficient is a statistical measure characterizing the degree of linear correlation between the numerical changes of the stress distribution field and the geometric deformation field within a local neighborhood window. Mathematically, it is usually quantified using the Pearson product-moment correlation coefficient. This coefficient is calculated by dividing the covariance of the two fields within the window by their respective standard deviations, yielding a value between -1 and +1. A positive value indicates that the two fields change in the same direction (increased stress corresponds to increased deformation), while a negative value indicates opposite changes. The closer the absolute value is to 1, the stronger the linear correlation. In the specific calculation, the numerical sequences of the stress distribution field and the corresponding numerical sequences of the geometric deformation field at each spatial location within the window are first extracted. The sample mean and sample standard deviation of the two sequences are calculated, and then the normalized inner product value after centering is solved to obtain the local correlation coefficient at the center point of the window. This coefficient, as the core input variable for determining the subsequent correction magnitude, directly reflects the reliability level of the geometric deformation data at the center point.

[0057] This implementation establishes a negatively correlated continuous mapping relationship between the correction magnitude and the local correlation coefficient, and realizes the transformation from the correlation index space to the correction intensity space through a monotonically decreasing mathematical function. A typical implementation uses a logistic sigmoid function to establish the mapping, the mathematical expression of which is: ; In the formula, This indicates the correction magnitude (i.e., correction parameter) for the center point of the local neighborhood window. This represents the preset maximum correction range, characterizing the maximum degree of intervention allowed by the correction operation on the original data in extreme cases. It is usually less than 1 (e.g., 0.9) to preserve a certain degree of the original data characteristics. This represents the local correlation coefficient calculated within the local neighborhood window; This indicates a correction of 50%. The corresponding local correlation coefficient threshold is used to control the position of the symmetry center of the mapping curve (e.g., take 0.5). This is a natural constant. The above calculation formula can ensure that the local correlation coefficient... Much higher At that time, the correction range Approaching 0 indicates that the highly correlated region undergoes almost no correction; when far below hour, Approaching This indicates that near-maximum corrections are implemented in the low-correlation region; when In When nearby, Follow The intensity decreases gradually but increases smoothly, achieving a continuous transition in the correction strength.

[0058] In addition to the Sigmoid function mapping mentioned above, other forms of continuously decreasing functions can also be used to establish negative correlation: a linearly decreasing function can be used, setting an upper and lower cutoff correlation coefficient, establishing a mapping between the two thresholds where the correction magnitude increases linearly as the correlation coefficient decreases, and fixing the minimum and maximum correction magnitudes respectively outside the two thresholds; an exponentially decaying function can be used, establishing a mapping relationship where the correction magnitude decreases exponentially as the correlation coefficient increases, so that the correction magnitude in the high correlation region quickly converges to zero; a piecewise polynomial function can also be used, splicing function segments with different curvatures in different correlation intervals to adapt to specific correction strategy requirements.

[0059] In the above scheme, the local correlation coefficient As a spatial index of geometric deformation field data quality, its value directly indicates the degree of consistency between the observed deformation and the actual stress-driven deformation at the center point. When When the value is in the high range, it indicates that the geometric deformation field is highly coordinated with the stress distribution field at that location, the deformation mode conforms to physical expectations, and the data has high reliability. Therefore, a low correction magnitude is assigned to respect the original observation value; when When the value is in the low range, it indicates that the geometric deformation field lacks coordination with the stress distribution field at that location, the deformation characteristics deviate from the physical expectations, and there is significant registration error interference. Therefore, a high correction magnitude is assigned to allow for a greater replacement of the original observations with inferred values ​​based on physical correlations. Through this adaptive spatial allocation strategy of correction parameters, spatial selective suppression of registration errors in the geometric deformation field and preservation of true deformation information are achieved, ensuring that the corrected geometric deformation field conforms to the physical laws of stress-deformation while maintaining reasonable spatial continuity.

[0060] The second implementation method: a segmented mapping method based on threshold intervals; In this embodiment, multiple correlation index thresholds can be set, dividing the correlation index value range into high correlation intervals, medium correlation intervals, and low correlation intervals. A very low correction parameter is assigned to the high correlation interval, indicating that the geometric deformation field data in this region has high reliability and does not require significant correction. A moderate correction parameter is assigned to the medium correlation interval, implementing appropriate data adjustments. A high correction parameter is assigned to the low correlation interval, correcting the original geometric deformation field to suppress registration errors. Through a segmented, step-like mapping relationship, the discretized hierarchical management of the correction strategy is achieved, simplifying the calculation process and clearly defining the processing intensity for different reliability regions.

[0061] The third implementation method: determination of correction parameters based on statistical significance test; In this embodiment, the correlation index can be combined with the statistical significance level. Hypothesis testing can be used to determine whether the correlation between the stress distribution field and the geometric deformation field within the local neighborhood window is significantly different from the random, uncorrelated state. For highly correlated regions that pass the significance test, a lower correction parameter is assigned to indicate that the observed data reflects the real physical process. For low-correlation regions that fail the significance test, a higher correction parameter is assigned to indicate that the observed data is dominated by registration error. The specific value of the correction parameter can be continuously adjusted according to the magnitude of the test statistic, so that the correction magnitude and the degree of statistical significance form a continuous negative correlation mapping.

[0062] The fourth implementation method: membership mapping based on fuzzy logic; In this implementation, fuzzy membership functions for "high confidence" and "low confidence" can be defined to fuzzify the correlation index values ​​into the degree of membership to the "high confidence" and "low confidence" categories. The correction parameter is determined through fuzzy rule reasoning. For example, when the correlation index has a high degree of membership to "high confidence", a low correction parameter is output, and when it has a high degree of membership to "low confidence", a high correction parameter is output. For intermediate states, a correction parameter value that takes into account both types of membership is obtained through defuzzification operation. This allows the determination of the correction parameter to flexibly handle the transitional state and boundary uncertainty of the correlation index, thereby improving the robustness of the correction strategy.

[0063] In the embodiments of this application, the geometric deformation field can be modified through at least one of the following multiple implementation methods: The first implementation method: direct replacement based on hard threshold switching; In this embodiment, a correction parameter threshold can be set to divide the spatial region into a high correction region and a low correction region. For regions where the correction parameter exceeds the threshold, the deformation estimate obtained based on the stress distribution field is directly used as the correction result, completely discarding any suspicious registration error components in the original geometric deformation field. For regions where the correction parameter is below the threshold, the original geometric deformation field value is directly retained, and it is determined that the observed data at this location reflects the true deformation. For transition regions where the correction parameter is near the threshold, simple linear interpolation can be used to achieve a smooth transition between the two values.

[0064] The second implementation method: a weighted fusion method based on Bayesian inference; In this embodiment, the geometric deformation field can be considered as a measurement containing observation noise, the deformation value inferred from the stress distribution field can be considered as a priori expectation, and the correction parameter can be mapped to the relative weight of observation confidence and prior confidence. The posterior estimate is calculated using Bayes' theorem, for example, using the correction parameter... As prior weights, As observation weights, a weighted average is calculated as a posterior estimate, or the continuity assumption of the spatial prior is further considered, and a Markov random field model is introduced to make a global optimal inference of the correction results for the entire field.

[0065] The third implementation method: a local window correction scheme based on statistical standardization alignment; Optionally, the above-mentioned correction of the geometric deformation field based on the correction parameters includes: establishing local neighborhood windows point by point in the geometric deformation field, extracting statistical features of the geometric deformation field and stress distribution field within the local neighborhood windows, and establishing a correction model based on the statistical features; wherein, the correction model is used to align the statistical characteristics of the geometric deformation field with the stress distribution field; based on the correction parameters, fusing the output value of the correction model with the original value of the geometric deformation field to obtain the correction value of the center point of the local neighborhood window.

[0066] A local neighborhood window refers to a local spatial region in a geometric deformation field that extends outward from a specific spatial location. It is typically square or circular in shape, and its spatial scale determines the degree of localization and robustness of statistical feature extraction. In practice, considering computational efficiency and the assumption of local statistical stationarity, a 5×5 or other odd-dimensional square window is usually used, ensuring that the center point is located at the geometric center of the window. The point cloud data within this window are assumed to be under the same local deformation mode and stress state, thus supporting robust estimation of the statistical correlation characteristics between the geometric deformation field and the stress distribution field, while providing sufficient sample support for subsequent statistical standardization alignment.

[0067] The correction model refers to a mathematical mapping relationship established based on the above statistical characteristics, used to align the numerical values ​​of the geometric deformation field to the statistical characteristics of the stress distribution field. The core mechanism of the correction model lies in statistical standardization alignment: it is assumed that in the local neighborhood, the actual deformation of the support structure should have a similar statistical distribution pattern to the stress distribution. Therefore, the values ​​in the stress distribution field are standardized relative to their own statistical center (subtract the mean and divide by the standard deviation), and then reconstructed according to the statistical scale of the geometric deformation field (multiplied by the standard deviation of the geometric field and added to the mean of the geometric field), thereby achieving consistency in the statistical characteristics of the two fields.

[0068] Fusion based on correction parameters refers to the weighted fusion of the output value of the correction model with the original observation value of the geometric deformation field to determine the final correction value of the center point of the local neighborhood window. This fusion process uses the correction parameters as control weights, tending to retain the original geometric deformation field values ​​in high confidence regions (low correction parameters) and tending to adopt the physical inference values ​​of the correction model in low confidence regions (high correction parameters).

[0069] Optionally, the above-mentioned extraction of statistical features of the geometric deformation field and stress distribution field within the local neighborhood window, and the establishment of a correction model based on the statistical features, includes: extracting the central tendency features and dispersion features of the geometric deformation field and stress distribution field within the local neighborhood window respectively; establishing a statistical mapping relationship from the stress distribution field to the geometric deformation field based on the degree of deviation of each point in the stress distribution field from its central tendency features, combined with the dispersion features of the geometric deformation field; determining the mapping direction of the statistical mapping relationship according to the sign of the local correlation coefficient, and establishing a correction model.

[0070] Central tendency is a statistical measure that characterizes the central location of data distribution within a local neighborhood window, typically implemented using the arithmetic mean. For geometrically deformable fields, central tendency... It reflects the overall average level of deformation values ​​within this local area, representing the benchmark displacement of the support structure deformation in this neighborhood; for the stress distribution field, it exhibits a concentration trend characteristic. This reflects the average stress level within the local area and represents the baseline strength of the surrounding rock stress in this neighborhood. These two central tendency features are extracted by taking the arithmetic mean of the values ​​at all spatial locations within the local neighborhood window. The calculation formula is the sum of the values ​​at each point within the window divided by the number of points within the window, and the resulting mean serves as the reference benchmark for subsequent statistical standardization and alignment.

[0071] Discreteness characteristics refer to statistical measures that characterize the degree of dispersion or variation in the distribution of data within a local neighborhood window, typically expressed as standard deviation. For geometrically deformable fields, discreteness characteristics... This reflects the degree of dispersion of deformation values ​​around the mean within the local region, characterizing the intensity and spatial heterogeneity of the local deformation gradient; for the stress distribution field, it reflects the degree of dispersion. This reflects the degree of dispersion of stress values ​​around the mean within the local area, characterizing the non-uniformity of local stress distribution. These two dispersion features are extracted by taking the square root of the average of the sum of the squares of the differences between the values ​​at each point within the window and their corresponding mean values. The resulting standard deviation is used as the scaling factor for subsequent statistical standardization alignment.

[0072] Establishing a statistical mapping from the stress distribution field to the geometric deformation field hinges on converting the deviation of the stress distribution field's values ​​from its own central tendency characteristics to the discreteness scale of the geometric deformation field. In practice, this involves first calculating the values ​​at each point in the stress distribution field. Compared to its central tendency characteristics deviation This deviation represents the relative position of the stress value at that point with respect to the local average stress; subsequently, this deviation is divided by the dispersion characteristic of the stress distribution field. Standardization is performed to obtain the dimensionless relative deviation. This value indicates the position of the stress value at that point within a certain number of standard deviations in the stress distribution field; finally, this relative deviation is multiplied by the dispersion characteristic of the geometric deformation field. And according to the central tendency characteristics of the geometric deformation field Translation is performed to achieve the conversion from the stress distribution field scale to the geometric deformation field scale.

[0073] The mapping direction is determined based on the sign of the local correlation coefficient to ensure that the statistical mapping relationship reflects the correct direction of coordinated change between the two fields. The local correlation coefficient β characterizes the direction of linear correlation between the geometric deformation field and the stress distribution field within a local neighborhood window: when... When the two fields are positively correlated, i.e., increased stress corresponds to increased deformation (or decreased stress corresponds to decreased deformation), the mapping direction should maintain the standard mapping relationship; when When the two fields are negatively correlated, meaning that an increase in stress corresponds to a decrease in deformation (or a decrease in stress corresponds to an increase in deformation), the mapping direction should be reversed. This can be addressed by introducing a sign function. The statistical mapping relationship is corrected in direction, and the corrected statistical mapping relationship expression is as follows: ; In the formula, The output value after statistical mapping represents the value of the th element in the stress distribution field. The stress state at a point should be mapped to the deformation value on the geometric deformation field scale; and These represent the mean (central tendency characteristic) and standard deviation (dispersion characteristic) of the geometric deformation field within a local neighborhood window, respectively. For the stress distribution field, the first Stress values ​​at each point; and These represent the mean (central tendency characteristic) and standard deviation (dispersion characteristic) of the stress distribution field within the local neighborhood window, respectively. Local correlation coefficient The sign function, when The value is +1 when... When the value is -1, The value can be 0 or processed according to preset rules. The introduction of this sign function ensures that when two fields are negatively correlated, the statistical mapping relationship will reverse the deviation of the stress distribution field to the geometric deformation field scale, maintaining the correctness of the physical correlation.

[0074] The above scheme extracts the central tendency and dispersion characteristics of the two fields within a local window, respectively. It first normalizes the values ​​of each point in the stress distribution field relative to its own distribution center, eliminating differences in the dimensions and numerical ranges of the two physical fields. Then, it reconstructs the data according to the distribution scale of the geometric deformation field, achieving consistency and alignment of the two fields in their statistical distribution patterns. Simultaneously, it determines the mapping direction based on the actual correlation direction (positive or negative) between the two fields, ensuring the correct expression of the physical correlation. The resulting correction model can convert the physical laws inherent in the stress distribution field (the actual deformation should be consistent with the statistical stress distribution) into specific numerical inferences at the geometric deformation field scale, providing a physically reasonable reference value for subsequent fusion correction based on correction parameters.

[0075] Optionally, the above-mentioned algorithm for monitoring the deformation of underground phosphate mine support structures based on laser point clouds may further include: for the same spatial location point in the geometric deformation field, obtaining multiple candidate correction values ​​obtained by correcting it in multiple overlapping local neighborhood windows respectively; fusing multiple candidate correction values ​​to determine the final correction result of the spatial location point.

[0076] It is understandable that, since the correction operation in the geometric deformation field is based on point-by-point local neighborhood windows, and there is significant spatial overlap between the local neighborhood windows of adjacent spatial locations, the same spatial location point will inevitably fall within the coverage of multiple adjacent windows. During the correction process of each window, a candidate correction value for that point will be generated. This multi-window overlap mechanism results in multiple independent correction estimates for a single spatial location point. If not effectively fused, this will cause discontinuities and local contradictions in the correction results in the spatial domain, undermining the overall smoothness and physical rationality of the geometric deformation field. By fusing the candidate correction values ​​generated by multiple overlapping windows, we can comprehensively utilize the statistical information provided by different neighborhood perspectives, effectively smooth the numerical transition at the window boundaries, suppress the estimation fluctuations caused by limited statistical samples or local outliers in a single local window, thereby improving the spatial continuity and robustness of the correction results and ensuring that the final geometric deformation field exhibits physically consistent deformation distribution characteristics across the entire domain.

[0077] In the embodiments of this application, multiple candidate correction values ​​can be fused through at least one of the following methods: The first method: a fusion method based on interpolation confidence weighting; Optionally, the above-mentioned fusion of multiple candidate correction values ​​to determine the final correction result of the spatial location point includes: determining the fusion weight corresponding to each candidate correction value based on the interpolation confidence distribution determined during the spatial interpolation of the stress distribution field; wherein, the interpolation confidence is negatively correlated with the distance from the spatial location point to the stress reference structure; and weighting and fusing multiple candidate correction values ​​based on the fusion weight to obtain the final correction result.

[0078] The interpolation confidence distribution refers to a set of quantitative representations of the confidence level of the interpolation results at each spatial location during the spatial interpolation process of extending a sparse stress distribution field into a dense stress distribution field. This distribution is closely related to the uncertainty quantification mechanism within the spatial interpolation algorithm, and is typically expressed as the prediction variance. Or its derivatives, expressed numerically. Prediction variance This statistical measure characterizes the uncertainty of the spatial interpolation model's estimates at various locations. Its magnitude directly reflects the degree to which the interpolation result at that location depends on the constraints of the observed data: at sampling points close to the stress reference structure (anchor plate), the interpolation model exhibits high determinism due to the direct support of physical observation data, resulting in a small prediction variance. As the distance from the stress reference structure increases, the interpolation result gradually becomes dependent on spatial correlation assumptions and model extrapolation, leading to increased uncertainty and a corresponding increase in prediction variance. Therefore, the interpolation confidence distribution essentially constitutes a confidence map in the spatial domain, indicating the reliability of the values ​​at each location within a dense stress distribution field.

[0079] The fusion weight refers to the weighting coefficient assigned to the candidate correction values ​​generated by each overlapping local neighborhood window, used to regulate the contribution ratio of different window correction results in the final fusion process. This weight is determined based on the interpolation confidence distribution, achieved by establishing a negative correlation between prediction variance and weight values: the smaller the prediction variance (higher interpolation confidence), the larger the fusion weight at the center of the corresponding window; the larger the prediction variance (lower interpolation confidence), the smaller the fusion weight. A typical implementation uses a normalized inverse square mapping, whose mathematical expression is: ;

[0080] In the formula, Indicates the first The fusion weights corresponding to the center points of overlapping local neighborhood windows; For the first The prediction variance of the spatial interpolation process at the center point of an overlapping local neighborhood window; and These represent the minimum and maximum values ​​in the global prediction variance distribution, respectively, and are used to normalize the prediction variance to the [0,1] interval. The squaring operation enables the weight distribution to have a non-linear response to changes in the prediction variance, amplifying the weight decay effect in high variance (low confidence) regions.

[0081] Weighted fusion refers to a mathematical operation that, based on determined fusion weights, averages the candidate correction values ​​obtained from multiple overlapping local neighborhood windows for the same spatial location point. This operation generates the final correction result for that spatial location point by comprehensively considering the correction estimates provided from different window perspectives and their corresponding confidence levels. The mathematical expression for weighted fusion is: ; In the formula, Represents a spatial location point in a geometric deformation field. The final revised result; This represents the total number of overlapping local neighborhood windows that this spatial location point falls into; Indicates the first Candidate correction values ​​generated for this point within a local neighborhood window; Let be the fusion weight corresponding to the j-th window; in the above calculation formula, the numerator is the weighted sum of all candidate correction values, and the denominator is the normalization factor of all weights, to ensure that the fusion result satisfies the mathematical consistency of the weighted average.

[0082] The above scheme utilizes the uncertainty quantification information in the spatial interpolation process to establish a positive correlation mechanism between data credibility and the degree of adoption of the correction results. Since the spatial interpolation of the stress distribution field has higher certainty near the stress reference structure (the stress value at the anchor position is directly calculated from the tray's geometric features, representing a strong constraint), the candidate correction values ​​generated in this region are based on a more reliable physical reference benchmark, and therefore are given higher weights to ensure that the final correction result is faithful to the high-credibility data. However, in regions far from the stress reference structure, the uncertainty of the interpolation results increases, and the candidate correction values ​​may contain more speculative components; therefore, they are given lower weights to suppress the excessive influence of unreliable extrapolation on the final result. Through this adaptive weighting mechanism based on the strength of physical data support, the credibility-weighted smoothing of the correction results in the spatial domain is achieved, ensuring that the final geometric deformation field maintains high reliability in regions with sufficient stress data constraints and transitions robustly in regions with sparse data, thus improving the overall spatial continuity and physical rationality of the corrected deformation field.

[0083] The second method: a weighted fusion method based on spatial distance attenuation; In this approach, based on the principle of geometric proximity, candidate correction values ​​provided by the center point of a local neighborhood window that is closer to the spatial location point should have higher credibility and influence. For example, the Euclidean distance or geodesic distance along the roadway surface from the center point of each overlapping window to the spatial location point can be calculated, and a weighting function (such as a Gaussian decay function or a reciprocal distance weighting function) that monotonically decreases with increasing distance can be established, so that adjacent windows are given high weights and distant windows are given low weights. The working principle of this implementation is based on the spatial correlation assumption: the statistical characteristics of the neighboring region and the correction estimate have higher spatial continuity, so the correction results of neighboring windows should be given greater confidence; through distance-weighted fusion, the local discontinuities caused by the window boundary truncation effect are effectively smoothed, improving the spatial smoothness and visual continuity of the correction results.

[0084] The third approach: an adaptive weighted fusion method based on the reliability of the corrected parameters; In this approach, the correction parameters (or correlation indices) calculated within each overlapping local neighborhood window can be used as a measure of the window's correction reliability. Windows with lower correction parameters (i.e., higher correlation and reliable original data) are assigned higher fusion weights, while windows with higher correction parameters (i.e., lower correlation and larger correction magnitude) are assigned lower weights. Alternatively, the reverse logic can be used, assigning high correction parameter windows high weights (assuming that this region needs the intervention of physical correction values ​​more). In addition to weighted averaging, a median filtering fusion strategy can also be used, selecting the median of multiple candidate correction values ​​as the final correction result. This nonlinear fusion method is robust to extreme correction values ​​generated by local anomalies in individual windows and can effectively suppress the influence of outlier noise on the final correction field. It is suitable for scenarios where there is measurement noise or registration anomalies in the point cloud data.

[0085] Step S150: Based on the corrected geometric deformation field, locate the actual deformation area of ​​the underground phosphate mine support structure.

[0086] Optionally, step S150 may include: setting a deformation threshold based on the measurement error of the laser point cloud data acquisition device; extracting deformation regions whose absolute deformation exceeds the deformation threshold in the corrected geometric deformation field; and performing morphological processing on the deformation regions to determine the actual deformation regions.

[0087] The aforementioned deformation threshold is set based on the inherent measurement error characteristics of the laser point cloud data acquisition equipment, and is typically based on the measurement error value. The multiple is used as the boundary standard for judging the true deformation and measurement uncertainty. Using a deformation threshold can statistically effectively distinguish between random measurement fluctuations and systematic structural changes, ensuring that the extracted deformation region has a higher reliability than the measurement noise level. In the corrected geometric deformation field, the absolute deformation degree is calculated point-by-point and numerically compared with this deformation threshold to extract the deformation region that satisfies the specified deformation threshold. The region with the given conditions is used as the initial deformation region, in which This represents the deformation value of the geometric deformation field at the corresponding spatial location. This extraction operation essentially completes the first layer of filtering of the actual structural changes and residual measurement uncertainties in the geometric deformation field.

[0088] Morphological processing is performed on the initially extracted deformation regions. Opening operations remove isolated discrete points or tiny patches smaller than a specific scale; these isolated regions typically correspond to residual registration errors or local measurement anomalies rather than actual structural deformation. Subsequently, closing operations connect spatially adjacent but slightly gapped deformation regions, filling in gaps caused by local occlusion or missing point clouds, ensuring spatial connectivity and integrity of the deformation regions. The resulting set of connected regions after these morphological operations constitutes the actual deformation region. This region physically corresponds to the actual support structure response caused by the release of surrounding rock stress or the transmission of mining pressure, providing precise spatial positioning and quantified deformation parameters for subsequent engineering measures such as support reinforcement, steel strip replacement, or secondary anchor bolt tightening.

[0089] To facilitate understanding of the working principle of the above-mentioned algorithm for monitoring the deformation of underground phosphate mine support structures based on laser point clouds, this application also provides specific application examples of the algorithm in a certain application scenario. For example... Figure 2 As shown, in this application scenario, the above-mentioned algorithm for monitoring the deformation of underground phosphate mine support structures based on laser point clouds mainly includes: Step 1: Obtain laser point cloud data of tunnels in underground phosphate mines and perform preprocessing. A special explosion-proof battery-powered locomotive used in mining was employed as the mobile platform. A millimeter-level multi-line rotating mechanical lidar scanning system was fixed at the center of the locomotive's roof and traveled at a constant speed along the tunnel to collect data. Simultaneously, a wheeled odometer and an inertial navigation unit were introduced for auxiliary measurements. The scanning results were used to calculate pose and stitch point clouds together to construct a complete 3D point cloud model of the tunnel. The acquired raw point cloud data underwent preprocessing: firstly, statistical filtering or radius filtering was applied to remove discrete noise points caused by phosphate dust; then, using a high reflectivity feature recognition algorithm based on laser reflection intensity, combined with the geometric features of the anchor bolt tray, the anchor bolts and trays were identified from the point cloud background as stress reference structures; the point cloud data of the identified anchor bolt and tray areas were subjected to local point cloud density enhancement processing to provide a data foundation for subsequent detailed geometric analysis.

[0090] Step 2: Analyze the stress distribution of the surrounding rock in the tunnel through stress reference structural characteristics, and use this to correct the results of the cross-section method detection. After obtaining the current laser point cloud, the geometric deformation field is derived by comparing it with the previous point cloud using the cross-sectional method. This geometric deformation field represents the difference between the laser point clouds obtained from the two detections, reflecting the deformation of the mine support structure, but it includes actual deformation and registration errors. To eliminate misidentification caused by registration errors, it is necessary to correct for misidentification by utilizing the physical correlation between the surrounding rock stress distribution and the actual deformation.

[0091] First, the stress distribution field of the surrounding rock is determined based on the surface geometric features of the stress reference structure. Edge point clouds of the anchor plate region are extracted, and principal component analysis is used to calculate the principal plane of the edge point clouds. The edge points are orthogonally projected onto this principal plane, and the projected edges are fitted as four straight line segments. The intersection of the diagonals is calculated as the geometric center of the plate, and the normal vector of the principal plane is determined as the reference axis of the plate. Based on this reference axis, the plate surface is divided into multiple radial surfaces along the circumference at specific angular steps, and the radial curvature of the sampling points on each radial surface is calculated. Based on the sign reversal of the radial curvature, the arched stress-bearing region of the plate is obtained. According to the degree of change of the radial curvature of the sampling points within the arched stress-bearing region relative to the radial curvature of the initial unstressed state, the relative stress magnitude at each anchor plate location is determined, constructing a sparse stress distribution field. (Relative stress) The calculation formula is: ; Subsequently, spatial interpolation is performed on the sparse stress distribution field to obtain a dense stress distribution field with the same spatial resolution as the geometric deformation field, and the prediction variance during the spatial interpolation process is recorded as the interpolation confidence distribution.

[0092] Next, the correlation index between the stress distribution field and the geometric deformation field is calculated, and the correction parameters for the geometric deformation field are determined based on the correlation index, thereby correcting the geometric deformation field. Local neighborhood windows are established point-by-point in the geometric deformation field, and the correlation index between the dense stress distribution field and the geometric deformation field within these local neighborhood windows is calculated. (e.g., Pearson correlation coefficient). Establish correlation indicators. With correction parameters The negative correlation between them means that the lower the correlation index, the larger the corresponding correction parameter. Correction parameter The calculation formula is: ; When correcting the geometric deformation field based on the correction parameters, the central tendency and dispersion characteristics (i.e., mean and standard deviation) of the geometric deformation field and the stress distribution field are first extracted within a local neighborhood window. Based on the deviation of each point in the stress distribution field from its central tendency characteristics, combined with the dispersion characteristics of the geometric deformation field, a statistical mapping relationship from the stress distribution field to the geometric deformation field is established. Based on the correlation index... The sign of the symbol determines the mapping direction of the statistical mapping relationship, and a corrected model is established. The corrected model applies to the first element within the window. Output value of each point The calculation formula is: ; Based on correction parameters The output value of the corrected model is fused with the original value of the geometric deformation field to obtain the corrected value of the center point of the local neighborhood window. : ; In the formula, This represents the original value of the geometric deformation field at the center point.

[0093] For the same spatial location point in the geometric deformation field, multiple candidate correction values ​​are obtained by correcting it in multiple overlapping local neighborhood windows. Based on the interpolation confidence distribution determined during the spatial interpolation process, the fusion weight corresponding to each candidate correction value is determined. This fusion weight is negatively correlated with the prediction variance. The formula for calculating the fusion weight is: ; The final corrected value is obtained by weighting and fusing multiple candidate corrected values ​​based on the fusion weights. : .

[0094] Step 3: Identify and process the actual deformation area of ​​the support structure; The actual deformation region is located based on the corrected geometric deformation field. Measurement errors are based on the laser point cloud data acquisition equipment. Set the deformation threshold to Extract regions whose absolute deformation exceeds the deformation threshold from the corrected geometric deformation field; perform morphological processing on the extracted regions to remove isolated noise points and connect minor breaks to determine the connected actual deformation regions.

[0095] After obtaining the actual deformation area, engineering measures are taken according to the deformation situation: for areas with significant deformation, support reinforcement is carried out; if the steel strip yields, breaks, or is severely twisted, the steel strip is replaced to ensure that the surface protection components can evenly transmit the preload of the anchor bolts; if obvious delamination displacement occurs near the anchor bolts, the support plate is tightened again; if minor tension cracks, weathering spalling, or shallow surface loosening and breakage occur, the surface is solidified by small-scale shotcreting. Through the above measures, the structural stability and safety of the underground phosphate mine tunnels are ensured.

[0096] This invention is now complete.

[0097] In summary, in this embodiment of the invention, laser point cloud data of underground phosphate mine tunnels is acquired; stress reference structures are identified in the point cloud data; and the stress distribution field of the surrounding rock is determined based on the surface geometric features of the stress reference structures. Based on multi-period laser point cloud data, the geometric deformation field of the tunnel is acquired. The correlation index between the stress distribution field and the geometric deformation field is calculated. Based on the correlation index, correction parameters for the geometric deformation field are determined, and the geometric deformation field is corrected based on these correction parameters. Based on the corrected geometric deformation field, the actual deformation area of ​​the underground phosphate mine support structure is located.

[0098] This invention introduces prior physical information from a stress reference structure to establish correlation constraints between the stress distribution field and geometric deformation, effectively distinguishing and separating registration errors from actual support structure deformation, thus improving the reliability of support structure deformation monitoring based on laser point clouds. Furthermore, by analyzing the surface geometric features of the stress reference structure to invert the surrounding rock stress distribution, stress sensors are not required, enabling indirect acquisition of the stress field based on existing support components, reducing the complexity and deployment cost of monitoring hardware. Additionally, the local correlation index between the stress distribution field and the geometric deformation field is used to adaptively determine correction parameters, ensuring that the correction magnitude of the registration error matches the error reliability, avoiding over-correction or under-correction, and guaranteeing the accuracy and rationality of the deformation field correction. Finally, through a positioning strategy combining morphological processing and threshold screening, the actual deformation area of ​​the support structure is accurately extracted based on the corrected geometric deformation field, providing reliable data support for subsequent support reinforcement and engineering treatment, and ensuring the structural safety of underground phosphate mine tunnels.

[0099] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An algorithm for monitoring the deformation of underground phosphate mine support structures based on laser point clouds, characterized in that, include: Laser point cloud data of underground phosphate mine tunnels is acquired, stress reference structures are identified in the point cloud data, and the stress distribution field of the surrounding rock is determined based on the surface geometric features of the stress reference structures. Based on the laser point cloud data from multiple phases, the geometric deformation field of the tunnel is obtained; Calculate the correlation index between the stress distribution field and the geometric deformation field; Based on the correlation index, the correction parameters of the geometric deformation field are determined, and the geometric deformation field is corrected based on the correction parameters. Based on the corrected geometric deformation field, the actual deformation area of ​​the underground phosphate mine support structure is located. Determining the correction parameters of the geometric deformation field based on the correlation index includes: Local neighborhood windows are established point by point in the geometric deformation field, and the local correlation coefficient between the stress distribution field and the geometric deformation field within the local neighborhood window is calculated. A negative correlation is established between the local correlation coefficient and the correction magnitude; wherein, the negative correlation results in a larger correction magnitude corresponding to a lower local correlation coefficient. Based on the negative correlation, the correction parameters for the center point of the local neighborhood window are determined.

2. The algorithm for monitoring deformation of underground phosphate mine support structures based on laser point clouds according to claim 1, characterized in that, The determination of the stress distribution field of the surrounding rock based on the surface geometric features of the stress reference structure includes: Extract the edge point cloud of the stress reference structure, and determine the principal plane and geometric center of the stress reference structure based on the edge point cloud. Then, determine the normal vector of the principal plane as the reference axis of the stress reference structure. Based on the reference axis, the surface of the stress reference structure is divided into multiple radial surfaces along the circumferential direction, and the radial curvature of the sampling points on each radial surface is calculated. Based on the sign flip position of the radial curvature, the arched stress region of the stress reference structure is segmented and obtained; Based on the degree of change in the radial curvature of the sampling points within the arched stress region relative to the radial curvature of the initial unstressed state, the relative stress at the location of the stress reference structure is determined, and a sparse stress distribution field is constructed.

3. The algorithm for monitoring deformation of underground phosphate mine support structures based on laser point clouds according to claim 2, characterized in that, Also includes: Spatial interpolation is performed on the sparse stress distribution field to obtain a dense stress distribution field with the same spatial resolution as the geometric deformation field.

4. The algorithm for monitoring deformation of underground phosphate mine support structures based on laser point clouds according to claim 1, characterized in that, The step of correcting the geometric deformation field based on the correction parameters includes: Local neighborhood windows are established point by point in the geometric deformation field. Statistical features of the geometric deformation field and the stress distribution field within the local neighborhood windows are extracted, and a correction model is established based on the statistical features. The correction model is used to align the geometric deformation field with the statistical characteristics of the stress distribution field. Based on the correction parameters, the output value of the correction model is fused with the original value of the geometric deformation field to obtain the correction value of the center point of the local neighborhood window.

5. The algorithm for monitoring deformation of underground phosphate mine support structures based on laser point clouds according to claim 4, characterized in that, The step of extracting statistical features of the geometric deformation field and the stress distribution field within the local neighborhood window, and establishing a correction model based on the statistical features, includes: The concentration trend characteristics and dispersion characteristics of the geometric deformation field and the stress distribution field within the local neighborhood window are extracted respectively. Based on the degree of deviation of each point in the stress distribution field from its concentration trend characteristics, and combined with the degree of dispersion characteristics of the geometric deformation field, a statistical mapping relationship from the stress distribution field to the geometric deformation field is established. The mapping direction of the statistical mapping relationship is determined based on the sign of the local correlation coefficient, and a correction model is established.

6. The algorithm for monitoring deformation of underground phosphate mine support structures based on laser point clouds according to claim 4, characterized in that, Also includes: For the same spatial location point in the geometric deformation field, obtain multiple candidate correction values ​​obtained by correcting it in multiple overlapping local neighborhood windows respectively; By integrating multiple candidate correction values, the final correction result for the spatial location point is determined.

7. The algorithm for monitoring deformation of underground phosphate mine support structures based on laser point clouds according to claim 6, characterized in that, The process of fusing multiple candidate correction values ​​to determine the final correction result for the spatial location point includes: Based on the interpolation confidence distribution determined during the spatial interpolation of the stress distribution field, the fusion weight corresponding to each candidate correction value is determined; wherein, the interpolation confidence is negatively correlated with the distance from the spatial location point to the stress reference structure; The multiple candidate correction values ​​are weighted and fused based on the fusion weights to obtain the final correction result.

8. The algorithm for monitoring deformation of underground phosphate mine support structures based on laser point clouds according to any one of claims 1-7, characterized in that, The stress reference structure includes an anchor bolt and a tray; The step of identifying the stress reference structure in the point cloud data includes: Based on laser reflection intensity characteristics and geometric shape characteristics, anchor rods and trays are identified in the point cloud data; The point cloud data of the identified anchor bolt and the area where the pallet is located are subjected to local point cloud density enhancement processing.

9. The algorithm for monitoring deformation of underground phosphate mine support structures based on laser point clouds according to any one of claims 1-7, characterized in that, The method of locating the actual deformation area of ​​the underground phosphate mine support structure based on the corrected geometric deformation field includes: Deformation thresholds are set based on the measurement errors of laser point cloud data acquisition equipment; Extract the deformation regions whose absolute deformation degree exceeds the deformation threshold from the corrected geometric deformation field; The deformed region is subjected to morphological processing to determine the actual deformed region.

Images