An airborne laser point cloud ground target device and a pose solving method

Abstract

The embodiment of the application provides an airborne laser point cloud ground target device and a pose solving method, and belongs to the technical field of measurement, and specifically comprises: a point cloud data target component, a target space pose determination component and a digital marker component; the point cloud data target component is composed of two square plane target plates, the two plane target plates are fixed through rigid support rods, and form an inverted V-shaped structure in a relatively inclined manner, and the upper edges of the two plane target plates are parallel to each other and have a spacing; the target space pose determination component is composed of two coplanar and symmetric isosceles right triangles whose symmetry axes are perpendicular to each other, and the two triangles are arranged at a common reference vertex; the digital marker component is four circular markers with numbers, the four circular markers are coplanar with the two triangles, and are symmetrically arranged in four quadrants divided by the symmetry axes of the triangles with the common reference vertex as the center. Through the scheme of the application, the monitoring accuracy and adaptability are improved.

Description

Technical Field

[0001] This invention relates to the field of metrology, and in particular to an airborne laser point cloud ground target device and a method for solving its pose. Background Technology

[0002] Currently, in the field of high-precision monitoring of geological hazards such as mining subsidence and slope deformation, the layout and measurement of deformation monitoring points are crucial for obtaining surface deformation data. The accuracy and richness of this data directly affect the reliability of disaster assessment and early warning. Existing monitoring technologies can generally be divided into two main categories: point monitoring and area monitoring.

[0003] Among these technologies, point monitoring, represented by the Global Navigation Satellite System (GNSS), can provide high-precision three-dimensional coordinates. However, the high cost of equipment procurement and manual deployment limits the density of monitoring stations, making it difficult to comprehensively and continuously reflect the overall characteristics of the deformation field. More importantly, its measurement information is limited to the three-dimensional displacement of a point. Key deformation parameters such as tilt, curvature, and torsion, which are directly related to the safety of the disaster-bearing body, cannot be directly measured and can only be estimated through multi-point calculations.

[0004] In large-scale area monitoring, airborne lidar (LiDAR) aerial surveying technology is widely used, but it typically relies on traditional ground-based targets as geometric control points. However, these targets are essentially passive geometric positioning markers with limited functionality. Their main role is to provide high-precision geographic references for point cloud data; obtaining deformation data still requires differential comparison of multiple observation results, a cumbersome process. Furthermore, the targets themselves cannot sense changes in attitude, making it impossible to acquire information on surface tilt and torsion, significantly limiting the potential of LiDAR technology in refined deformation mechanism analysis and disaster early warning applications.

[0005] In summary, existing mainstream monitoring technologies have obvious technical shortcomings: GNSS point-based methods are costly and have limited information dimensions, while traditional aerial survey target methods cannot meet the need for direct monitoring of key deformation parameters such as surface attitude.

[0006] It is evident that there is an urgent need for an airborne laser point cloud ground target device that can simultaneously provide pose information and is also economical. Summary of the Invention

[0007] In view of this, embodiments of the present invention provide an airborne laser point cloud ground target device and a pose solving method, which at least partially solves some of the problems existing in the prior art.

[0008] In a first aspect, embodiments of the present invention provide an airborne laser point cloud ground target device, comprising:

[0009] Point cloud data target component, target spatial pose determination component, and digital marker component;

[0010] The point cloud data target component consists of two square planar target plates, which are fixed by rigid support rods and form an inverted V-shaped structure with relative inclination. The upper edges of the two planar target plates are parallel to each other and maintain a gap.

[0011] The target spatial pose determination component consists of two coplanar isosceles acute triangles with mutually perpendicular axes of symmetry, and the two triangles are arranged at a common reference vertex.

[0012] The digital identifier component consists of four circular markers with numbers on them. The four circular markers are coplanar with two triangles and are symmetrically arranged in four quadrants divided by the symmetry axis of the triangles, with a common reference vertex as the center.

[0013] According to one specific implementation of the present invention, the surfaces of the two planar target plates are coated with matte paint, and the two target plates are respectively coated with matte white paint and matte silver-gray paint with different reflective properties.

[0014] According to one specific implementation of the present invention, the two isosceles acute triangles are red and white, respectively.

[0015] According to one specific implementation of the present invention, the circular mark of the digital identifier component has a white bottom surface and a black number sprayed on the center position.

[0016] Secondly, embodiments of the present invention provide a pose calculation method, applied to an airborne laser point cloud ground target device as described in any of the above-disclosed embodiments, the method comprising:

[0017] Step 1: Acquire measured point cloud data of the ground target device using an airborne LiDAR device;

[0018] Step 2: Based on the measured point cloud data, initialize the pose parameters of the target theoretical model in the point cloud coordinate system, wherein the pose parameters include the initial values ​​of the translation vector and the initial values ​​of the rotation component.

[0019] Step 3: Construct an iterative optimization algorithm framework that integrates adaptive weighting mechanism and Bayesian prior to refine the solution of pose parameters.

[0020] According to a specific implementation of an embodiment of the present invention, the step of constructing an iterative optimization algorithm framework that integrates an adaptive weighting mechanism and Bayesian prior includes:

[0021] Each observation point is matched to the plane that is closest to the target in the theoretical model at the current pose, and the residual function is constructed by the distance from the measured point to the theoretical plane;

[0022] Solve the Jacobian matrix based on the residual function and construct the incremental equation;

[0023] An adaptive weight matrix is ​​introduced into the incremental equation, wherein the adaptive weight matrix includes a distance sensitivity weight function and a reflection intensity weight function.

[0024] According to a specific implementation of an embodiment of the present invention, the distance sensitivity weighting function is:

[0025] ;

[0026] in, Input distance, The maximum distance parameter, To control the steepness of the curve, let... For intermediate point parameters;

[0027] The reflection intensity weighting function is a piecewise function, assigning a constant high weight within the reflection intensity range corresponding to the target material, while using exponential decay at the interval boundaries to achieve a smooth transition, specifically expressed as:

[0028] ;

[0029] in, The attenuation coefficient; This is the reflection intensity value. and These represent the lower and upper limits of the reflection intensity range, respectively.

[0030] According to a specific implementation of an embodiment of the present invention, the iterative optimization algorithm framework introduces the empirical prior distribution of rotation and translation parameters, constructing the optimization problem as maximizing the posterior probability estimate, and the corresponding cost function is defined as:

[0031] ;

[0032] in, The pose of the current iteration state. Here is the Huber loss function. For the geometric residuals between all points and the theoretical template, The prior mean of the translation. It is the inverse of the translated prior covariance matrix. Let be the natural parameter matrix of the Matrix Fisher distribution.

[0033] According to a specific implementation of the present invention, before the step of matching each observation point to the plane with the closest distance in the target theoretical model at the current pose, the method further includes:

[0034] Edge point removal: Calculate each measurement point The distance to each boundary of its associated target theoretical surface, wherein the boundary is a set of points. Definition; for boundary points and The edges and points formed Distance to that side Calculated using the following formula:

[0035] ;

[0036] in, Represents the set of points on the target surface boundary; where Characteristic points and pointing to points respectively and The area of ​​the parallelogram formed by the two vectors; This represents the length of the line segment formed by two points;

[0037] If the distance from a certain measurement point to any boundary If the value is less than a preset threshold, the point is identified as an edge point and removed from subsequent iterations.

[0038] According to a specific implementation of an embodiment of the present invention, the edge point removal step further includes:

[0039] Measurement points The projection point is obtained by orthogonally projecting it onto the theoretical surface of the nearest target to which it is classified. :

[0040] ;

[0041] in, Let be any one of the four vertices of this theoretical plane. Let be the normal vector of this theoretical surface;

[0042] Establish a local two-dimensional coordinate system on this theoretical plane, and select a right-angled vertex of the target plate as the origin, so as to... to adjacent vertices The vector is used as a basis vector ,from to another adjacent vertex The vector is used as a basis vector ;

[0043] Projection point and target plate apex By transforming to a local two-dimensional coordinate system, the local coordinates can be solved using the dot product. ;

[0044] Determining the projection point in a two-dimensional local coordinate system Is it inside the target plate?

[0045] If the local coordinates of the projection point If the projection point exceeds the valid range defined by the target surface in the local coordinate system, it is determined that the projection point is located outside the target, and its corresponding original measurement point is... Removed from subsequent iterations.

[0046] The airborne laser point cloud ground target scheme in this embodiment of the invention includes: a point cloud data target component, a target spatial pose determination component, and a digital marker component; the point cloud data target component consists of two square planar target plates, which are fixed by rigid support rods and form an inverted V-shaped structure with relative inclination, and the upper edges of the two planar target plates are parallel to each other and maintain a gap; the target spatial pose determination component consists of two coplanar isosceles acute triangles with mutually perpendicular axes of symmetry, and the two triangles are arranged at a common reference vertex; the digital marker component consists of four circular markers with numbers, which are coplanar with the two triangles and symmetrically arranged in four quadrants divided by the axes of symmetry of the triangles, with the common reference vertex as the center.

[0047] The beneficial effects of the embodiments of the present invention are as follows: by solving the absolute pose of the target in each period of point cloud data and then calculating the difference in absolute pose between adjacent time series, the displacement and rotation changes of the target during the monitoring period can be accurately determined. Attached Figure Description

[0048] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments 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.

[0049] Figure 1 The following are schematic diagrams of the top view, front view, and side view of an airborne laser point cloud ground target device provided in an embodiment of the present invention, wherein (a) is a top view, (b) is a front view, and (c) is a side view;

[0050] Figure 2 This is a partial flowchart illustrating a pose solving method provided in an embodiment of the present invention;

[0051] Figure 3This is a schematic diagram illustrating the comparison between the results obtained by the pose solving method and the high-precision reference values ​​measured by a total station, provided in an embodiment of the present invention. In this diagram, (a) is the difference in displacement between the total station measurement results and the algorithm calculation results of the first simulation experiment; (b) is the difference in rotation between the total station measurement results and the algorithm calculation results of the first simulation experiment; (c) is the difference in displacement between the total station measurement results and the algorithm calculation results of the second simulation experiment; and (d) is the difference in rotation between the total station measurement results and the algorithm calculation results of the second simulation experiment.

[0052] Summary of attached figure labels

[0053] Point cloud data target component 110, target spatial pose determination component 120, digital marker component 130. Detailed Implementation

[0054] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0055] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0056] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this invention, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0057] It should also be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. The illustrations only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0058] Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.

[0059] This invention provides an airborne laser point cloud ground target device, and the method can be applied to the monitoring of surface deformation in geological monitoring scenarios.

[0060] See Figure 1 This is a schematic diagram of the structure of an airborne laser point cloud ground target device provided in an embodiment of the present invention. Figure 1 As shown, the airborne laser point cloud ground target device mainly includes:

[0061] Point cloud data target component 110, target spatial pose determination component 120 and digital marker component 130;

[0062] The point cloud data target component 110 consists of two square planar target plates. The two planar target plates are fixed by rigid support rods and form an inverted V-shaped structure with relative inclination. The upper edges of the two planar target plates are parallel to each other and maintain a distance.

[0063] The target spatial pose determination component 120 is composed of two coplanar isosceles acute triangles with mutually perpendicular axes of symmetry, and the two triangles are arranged at a common reference vertex.

[0064] The digital identifier component 130 consists of four circular markers with numbers. The four circular markers are coplanar with two triangles and are symmetrically arranged in four quadrants divided by the symmetry axis of the triangles, with a common reference vertex as the center.

[0065] In specific implementation, the point cloud data target component 110 consists of two square planar target plates, which are fixed to the three-dimensional target at a specific angle by rigid, immutable support rods. Specifically, the two planar target plates are fixed at relative inclinations, forming an inverted "V" shape structure, creating a highly recognizable geometric feature in the three-dimensional laser point cloud that is unaffected by the scanning angle. The surfaces of the two planar target plates are at a certain angle in space. The orthogonal arrangement of the two target plates creates an open, acute-angle corner reflection structure in space, ensuring that point cloud data of sufficient density can be obtained from different scanning angles. Based on this, to establish a geometric reference that avoids point cloud data interference and can be stably identified as a whole for attitude calculation, the two planar target plates are not directly intersecting. Instead, their upper edges are set to be parallel and spaced apart. The parallel edges and horizontal spacing ensure that the point cloud data of the two target plates are completely separated in space, fundamentally avoiding point cloud noise and feature blurring that may occur at the physical boundary. At the same time, it is also to ensure that the two separated target plates still constitute a compact unit that can be stably identified by the algorithm as a single target rather than two discrete objects.

[0066] To improve the acquisition quality of point cloud data and the algorithmic recognition capability of target features, the surfaces of both planar target plates are coated with matte paint. The use of matte paint enables ideal diffuse reflection of the incident laser, ensuring the effectiveness of laser ranging and the stable reception of effective echo signals by the airborne laser sensor, avoiding problems such as voids in point cloud data or gross errors due to multipath effects caused by specular reflection. The two planar target plates are coated with matte white paint and matte silver-gray paint, respectively. In terms of coating selection, this invention avoids dark colors such as matte black, as their strong absorption of laser energy would result in weak echo signals and low signal-to-noise ratios, easily causing voids in the point cloud or a decrease in ranging accuracy. White and silver-gray not only ensure high-quality laser echoes, but their reflective characteristics also create a strong contrast with common green vegetation and yellowish-brown soil, ensuring that the entire target unit can be reliably segmented from the environmental point cloud. Furthermore, the point cloud data obtained based on the significant difference in reflection intensity between the high-reflectivity matte white paint and the low-reflectivity matte silver-gray paint provides a non-geometric basis, enabling the unique differentiation of the two planar target plates through the point cloud reflection intensity values.

[0067] The target spatial pose determination component 120 consists of two coplanar isosceles acute triangles with mutually perpendicular axes of symmetry, one triangle being red and the other white. These two triangles are arranged at a common reference vertex, which is also the geometric center of the overall structure of the two square target plates. This arrangement ensures that the red and white triangles form two orthogonal direction indicators, whose pointing has a unique and fixed correlation with the target plate's pose.

[0068] The digital identification component 130 consists of four circular markers, each with a white paint coating on its base and a black number painted at its center. These four circular markers, along with two triangles, lie in a single plane and are symmetrically arranged within four quadrants defined by the triangles' common reference vertex, with one marker in each quadrant. This design leverages the capability of an airborne lidar system to simultaneously generate two-dimensional raster images while acquiring three-dimensional point cloud data. To achieve automated and unambiguous identification of each target, optical character recognition (OCR) can be performed on the two-dimensional image to extract a unique identification number for each target. This unique identification number provides a crucial index for subsequent fully automated data processing, ensuring that target data from different time phases can be matched one-to-one.

[0069] The airborne laser point cloud ground target device provided in this embodiment calculates the absolute pose of the target in each period of point cloud data, and then calculates the difference in absolute pose between adjacent time series, so as to accurately determine the displacement and rotation changes of the target during the monitoring period.

[0070] Corresponding to the above device embodiment, see [link to relevant documentation]. Figure 2 This invention also provides a pose calculation method, including:

[0071] Acquire measured point cloud data of ground target devices using airborne LiDAR equipment;

[0072] Based on measured point cloud data, the pose parameters of the target theoretical model in the point cloud coordinate system are initialized, wherein the pose parameters include the initial values ​​of the translation vector and the initial values ​​of the rotation component.

[0073] An iterative optimization algorithm framework integrating adaptive weighting mechanism and Bayesian prior is constructed to refine the solution of pose parameters.

[0074] In practice, an iterative optimization algorithm framework integrating adaptive weighting and Bayesian priors is constructed to refine the solution of the 3D spatial pose derived from target point cloud data. The target theoretical model is defined by eight key vertices of two point cloud data target plates in the target coordinate system. After acquiring the measured point cloud data of the target using an airborne LiDAR device, the pose parameters of the theoretical model in the point cloud coordinate system are first initialized based on the point cloud data of the lower indicative marker and the two upper planes of the target: the initial value of the translation vector is represented by clustering of the point cloud data of the two upper target plates, plane fitting, and intersection of normal vectors. Initial values ​​of rotational components Then through The point cloud data of the triangle markers representing the X and Y axes of the target coordinate system at the bottom, and the plane inversion calculation formed by the two, are used to calculate the plane, where the normal vector of the plane is defined as the initial Z axis of the coordinate system.

[0075] Specifically, the following steps are included:

[0076] First, residual modeling is performed. In each iteration step, each observation point is matched to the plane that is closest to the theoretical target model at the current pose. The residual function is then constructed using the distance from the measured point to the theoretical plane, specifically expressed as:

[0077] ;

[0078] in, The target model coordinate system in the current pose is the first... The unit normal vector of each face ( These represent the two aspects of the target theory. The normal vector in the target model coordinate system is rotated to the global normal vector. In the target model coordinate system at the current pose, the first... Any reference point on the surface. This is measured point cloud data in the global coordinate system.

[0079] Based on residual function Solve the Jacobian matrix , it is The derivative with respect to small changes in pose parameters:

[0080] ;

[0081] in, The number of pairing points used for optimization. Indicates the first The plane index corresponding to each point.

[0082] Then, based on the Jacobian matrix, an incremental equation is constructed using the Gauss-Newton method, which is based on the distance values ​​from all measured points to the theoretical surface of the target at the current pose:

[0083] ;

[0084] in, It is an approximate Hessian matrix. It is the gradient vector. for This represents the changes in rotation and translation. Simultaneously, to handle the uncertainty and outliers in point cloud data, a weight definition is introduced, and the equation becomes:

[0085] ;

[0086] In the above formula, ,in Here is the weighting function for distance sensitivity, used to quantify the spatial consistency between each measurement point and the theoretical model, specifically expressed as:

[0087] ;

[0088] in, Input distance (scalar). The maximum distance parameter (dynamically adjusted based on the distance vector calculated in each iteration); This is a parameter that controls the steepness of the curve. Let... The intermediate point parameter represents the value when the normalized distance equals... When the weight is 0.5.

[0089] The piecewise reflection intensity weighting function is assigned a constant high weight within the reflection intensity range corresponding to the target material, while exponential decay is used at the interval boundaries to achieve a smooth transition, specifically expressed as follows:

[0090] ;

[0091] in, The attenuation coefficient; This represents the reflection intensity value.

[0092] To address this, Bayesian inference methods can be introduced, combining observational information based on measured point cloud data with prior geometric knowledge of the rigid target's lower structure to achieve 3D pose estimation through maximum a posteriori estimation. However, traditional Bayesian methods require subjective specification of the prior distribution, which is difficult to achieve in our application scenario. For complex variables such as local coordinate system parameters, it is difficult to determine a suitable prior distribution form and its hyperparameters based on experience. Subjectively chosen priors may introduce human bias, affecting the objectivity of the final estimation results. In contrast, empirical Bayesian methods effectively solve this problem by driving the construction of the prior distribution through the data itself.

[0093] Six subsets of the target were obtained from the measured point cloud. .in, Divide into four diagonal sections. These are two auxiliary partitions used for orientation. The shape of the sub-regions is not restricted, as long as their respective geometric centers can be fitted. The union of the six point cloud data is fitted to a plane using the RANSAC method, and the plane equation is expressed as:

[0094] ;

[0095] in, For unit normal vector, This represents the distance from the origin to the plane. Subsequently, through least-squares refinement, the accurate distance is obtained. Let be the orientation of the plane normal vector. , Take the Z-axis direction of the global coordinate system.

[0096] Take the geometric centroid of the point cloud As a projection reference, a temporary orthogonal basis is established in the plane. After projecting each sub-region onto this plane, a two-dimensional center is obtained by fitting the data. And projecting it back into the three-dimensional coordinate system, we get:

[0097] ;

[0098] The origin is obtained by averaging the centers of the four diagonal partitions.

[0099] ;

[0100] by Let it be an auxiliary point. After removing its normal component and normalizing it, we get:

[0101] ;

[0102] Take by right-hand rule Then use The sign judgment makes its direction consistent with the expectation, thus obtaining This forms a right-handed orthogonal basis. .

[0103] In the uncertainty modeling and robust estimation process, each sub-region is first independently resampled. Next, the sample set was obtained. For location estimation, assume Follows a Gaussian prior distribution ,in, The mean of the coordinates of the origin of the local coordinate system represents the most likely location of that point. Let be the covariance matrix of the coordinates of the origin of the local coordinate system. For rotation estimation, a matrix Fisheri distribution prior is used. The average rotation matrix of the samples The first-order moment matching is obtained by finding its parametric decomposition. ,in The matrix obtained by decomposition;

[0104] During point cloud data acquisition, heavy-tailed errors and outliers can occur due to occlusion, object edge effects, and laser multipath effects. In such cases, pure quadratic loss allows a small number of outlier observations to dominate the solution and weakens the effects of weighting and priors. To address these issues, this experiment introduces the Huber loss function, which behaves as a quadratic function (equivalent to the Gaussian assumption) when the residuals are small, but transforms into a linear function when the residuals are large.

[0105] ;

[0106] in, is a threshold parameter used to control the sensitivity to outliers. The MAP cost function can then be written as:

[0107] ;

[0108] in, The pose of the current iteration state. Here is the Huber loss function. For the geometric residuals between all points and the theoretical template, The prior mean of the translation. It is the inverse of the translated prior covariance matrix. is the natural parameter matrix of the Matrix Fisher distribution, which is calculated based on rotation axis samples obtained by resampling the target's geometric features, and includes the prior mean orientation and concentration.

[0109] Therefore, the Gauss-Newton incremental equation under the empirical Bayesian prior framework is:

[0110] ;

[0111] In the formula, It is composed of the Jacobian matrices of all effective residuals stacked together; It is a vector composed of all effective residuals. It is a diagonal weight matrix; For the a priori Hessian increment ; gradient increments for priors .

[0112] Furthermore, during actual measurements, lidar points located near the target edge may be affected by interference echoes from the ground or other terrain features, leading to false observations. To suppress such errors, edge distance detection is required for each measurement point. If a point is too close to the edge, it can be inferred that it has a high probability of being a false observation, and therefore it is removed from subsequent calculations. Similarly, if the orthogonal projection of a point on its associated surface falls outside the target surface, that point will also be removed. The algorithm convergence requires at least three iterations (the first two iterations require iteratively bringing the original point cloud data closer to the theoretical target with the initial pose). The point removal mechanism introduced in the third iteration aims to prioritize measurements from the center region of the target surface to improve the reliability of the overall solution. (Any measurement point...) The distance relationship with the theoretical boundary of the target at the current pose can be expressed as:

[0113] ;

[0114] in, Represents the set of points on the target surface boundary; where Characteristic points and pointing to points respectively and The area of ​​the parallelogram formed by the two vectors; This represents the length of the line segment formed by two points. If the distance from a measured point to any boundary... If the value is less than a preset threshold, the point is identified as an edge point and removed from subsequent iterations;

[0115] point The projection onto the nearest target surface in its classification can be expressed as:

[0116] ;

[0117] in, Let be any one of the four vertices of this theoretical plane. Next, to define the normal vector of this theoretical surface, a local two-dimensional coordinate system is established on the plane. A right-angled vertex of the target plate is chosen as the origin of the coordinate system. From... to adjacent vertices The vector is used as a basis vector ,from to another adjacent vertex The vector is used as a basis vector Due to the angle It's a right angle; these two vectors are orthogonal. (A point on any plane) Both can be represented as:

[0118] ;

[0119] Local coordinates can be obtained by solving the dot product. , project point and trapezoidal vertices Transform to this coordinate system and determine the projection point in the two-dimensional local coordinate system. Is it inside the target plate? If the calculated local coordinates... If the projection point exceeds the valid range defined by the target surface in the local coordinate system, it is determined that the projection point is located outside the target, and its corresponding original measurement point... It will be removed from subsequent iterations.

[0120] To better demonstrate the accuracy and reliability of the target device and pose solving method provided in this application in deformation monitoring, the following is an exemplary description with reference to specific implementation examples.

[0121] In the embodiments of this application, two sets of known displacements and rotations are applied to the target to simulate its rigid body pose changes in a real scene. Using the iterative optimization algorithm that integrates adaptive weights and Bayesian priors proposed in this application, the absolute pose of the target in the two periods of point cloud data before and after the change is calculated, and the difference is calculated to obtain the pose change determined by this method.

[0122] To demonstrate the advantages of the method provided in this application, the calculated pose change is compared with the high-precision reference value of the change measured by a total station. The comparison of the calculation results is as follows: Figure 3 As shown.

[0123] from Figure 3 As can be seen, the six-degree-of-freedom pose changes calculated by this method are basically consistent with the reference true values ​​measured by the total station. Regarding translational changes, the calculation error is basically controlled within 2 cm, with the smallest difference in the Z-direction, reaching 0.02 cm. The calculation results for rotational changes are equally accurate, with errors all less than 0.6 degrees, and the differences in rotation around the Z-axis are the smallest, only 0.076° and 0.121°.

[0124] Both sets of experiments consistently demonstrate that the proposed method, which determines the pose change by solving the absolute pose and calculating the difference, exhibits high precision at the centimeter and sub-degree levels in both translation and rotation calculations. It possesses excellent stability and reliability, fully meeting the stringent requirements of applications such as high-precision deformation monitoring.

[0125] The pose determination method provided in this embodiment combines a unique target structure with a robust solution algorithm to achieve high-precision calculation of the absolute pose of the target in three-dimensional space, thereby reliably obtaining its pose changes at different times. Experimental results show that this method can effectively capture displacement and attitude changes, providing an effective technical means for automated monitoring of surface deformation.

[0126] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof.

[0127] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An airborne laser point cloud ground target device, characterized in that, include: Point cloud data target component, target spatial pose determination component, and digital marker component; The point cloud data target component consists of two square planar target plates, which are fixed by rigid support rods and form an inverted V-shaped structure with relative inclination. The upper edges of the two planar target plates are parallel to each other and maintain a gap. The target spatial pose determination component consists of two coplanar isosceles acute triangles with mutually perpendicular axes of symmetry, and the two triangles are arranged at a common reference vertex. The digital identifier component consists of four circular markers with numbers on them. The four circular markers are coplanar with two triangles and are symmetrically arranged in four quadrants divided by the symmetry axis of the triangles, with a common reference vertex as the center.

2. The apparatus according to claim 1, characterized in that, Both flat target plates are coated with matte paint, and the two target plates are coated with matte white paint and matte silver-gray paint with different reflective properties, respectively.

3. The apparatus according to claim 1, characterized in that, The two isosceles acute triangles are red and white, respectively.

4. The apparatus according to claim 1, characterized in that, The circular marker of the digital identifier component has a white bottom and a black number sprayed in the center.

5. A pose determination method, applied to the airborne laser point cloud ground target device according to any one of claims 1 to 4, characterized in that, include: Step 1: Acquire measured point cloud data of the ground target device using an airborne LiDAR device; Step 2: Based on the measured point cloud data, initialize the pose parameters of the target theoretical model in the point cloud coordinate system, wherein the pose parameters include the initial values ​​of the translation vector and the initial values ​​of the rotation component. Step 3: Construct an iterative optimization algorithm framework that integrates adaptive weighting mechanism and Bayesian prior to refine the solution of pose parameters; The steps for constructing an iterative optimization algorithm framework that integrates adaptive weighting mechanisms and Bayesian priors include: Each observation point is matched to the plane that is closest to the target in the theoretical model at the current pose, and the residual function is constructed by the distance from the measured point to the theoretical plane; Solve the Jacobian matrix based on the residual function and construct the incremental equation; An adaptive weight matrix is ​​introduced into the incremental equation, wherein the adaptive weight matrix includes a distance sensitivity weight function and a reflection intensity weight function; The steps for refining the pose parameters include: Based on the iterative optimization algorithm framework, the absolute pose of the target in the two periods of point cloud data before and after the change is calculated, and the difference is calculated to obtain the pose change.

6. The method according to claim 5, characterized in that, The distance sensitivity weighting function is: in, Input distance, The maximum distance parameter, To control the steepness of the curve, let... For intermediate point parameters; The reflection intensity weighting function is a piecewise function, assigning a constant high weight within the reflection intensity range corresponding to the target material, while using exponential decay at the interval boundaries to achieve a smooth transition, specifically expressed as: in, The attenuation coefficient; This is the reflection intensity value. and These represent the lower and upper limits of the reflection intensity range, respectively.

7. The method according to claim 6, characterized in that, The iterative optimization algorithm framework introduces the empirical prior distribution of rotation and translation parameters, constructing the optimization problem as maximizing the posterior probability estimate, and its corresponding cost function is defined as: in, The pose of the current iteration state. Here is the Huber loss function. For the geometric residuals between all points and the theoretical template, The prior mean of the translation. It is the inverse of the translated prior covariance matrix. Let be the natural parameter matrix of the Matrix Fisher distribution.

8. The method according to claim 5, characterized in that, Before the step of matching each observation point to the nearest plane in the target theoretical model at the current pose, the method further includes: Edge point removal: Calculate each measurement point The distance to each boundary of its associated target theoretical surface, wherein the boundary is a set of points. Definition; for boundary points and The edges and points formed Distance to that side Calculated using the following formula: in, Represents the set of points on the target surface boundary; where Characteristic points and pointing to points respectively and The area of ​​the parallelogram formed by the two vectors; This represents the length of the line segment formed by two points; If the distance from a certain measurement point to any boundary If the value is less than a preset threshold, the point is identified as an edge point and removed from subsequent iterations.

9. The method according to claim 8, characterized in that, The edge point removal step also includes: Measurement points The projection point is obtained by orthogonally projecting it onto the theoretical surface of the nearest target to which it is classified. : in, Let be any one of the four vertices of this theoretical plane. Let be the normal vector of this theoretical surface; Establish a local two-dimensional coordinate system on this theoretical plane, and select a right-angled vertex of the target plate as the origin, so as to... to adjacent vertices The vector is used as a basis vector ,from to another adjacent vertex The vector is used as a basis vector ; Projection point and target plate apex By transforming to a local two-dimensional coordinate system, the local coordinates can be solved using the dot product. ; Determining the projection point in a two-dimensional local coordinate system Is it inside the target plate? If the local coordinates of the projection point If the projection point exceeds the valid range defined by the target surface in the local coordinate system, it is determined that the projection point is located outside the target, and its corresponding original measurement point is... Removed from subsequent iterations.

Images