CN-121999169-A - High-precision Gaussian modeling method for steel box girder based on laser scanning point cloud

CN121999169ACN 121999169 ACN121999169 ACN 121999169ACN-121999169-A

Abstract

The invention relates to the technical field of laser point cloud modeling, and discloses a high-precision Gaussian modeling method for a steel box girder based on laser scanning point cloud, which comprises the steps of firstly obtaining the point cloud of a component to be detected, calculating the longitudinal extension axis of the component, and counting the spatial projection density distribution along the axial direction; and analyzing the effective visibility rate of stiffening unit spacing and scanning of the component through spatial spectrum analysis, reversely solving the longitudinal filter kernel width which is adapted to the current shielding state based on a correlation model of shielding effect and smooth response factor, generating density equalization weight by utilizing the parameters, constructing an anisotropic Gaussian fitting model by combining the longitudinal filter kernel width, the density equalization weight and the normalized quality weight, and reconstructing local geometric reference elements of a welding plane, a chord axis and a bolt hole center. The method effectively solves the problem of Gaussian smoothing systematic drift caused by periodic shielding, and ensures the accuracy of three-dimensional modeling in a complex environment.

Inventors

YAN ZHENJUN
CHEN XI
JIANG LING
HUANG XIAOLI
DUAN XUELI
LIU TIANBAO
SUN ZHAOJIE
HAN LU
SUO GUOQING

Assignees

滁州学院

Dates

Publication Date: 20260508
Application Date: 20260129

Claims (8)

1. The high-precision Gaussian modeling method for the steel box girder based on the laser scanning point cloud is characterized by comprising the following steps of: acquiring laser scanning point cloud data of a steel structure member to be tested, and resolving a member longitudinal extension axis of the laser scanning point cloud data; Counting the space projection density distribution of the point cloud along the longitudinal extension axis of the component, performing space spectrum analysis on the space projection density distribution, and analyzing the spacing between stiffening units of the component and the scanning effective visibility; Constructing a correlation model of a shielding effect and a smooth response factor, wherein the correlation model is used for representing the residual degree of periodic density fluctuation after smoothing treatment caused by the shielding in a component, and the correlation model takes the longitudinal filter core width, the component stiffening unit spacing and the scanning effective visible rate as parameters; Setting a target residual threshold according to the scanning effective visibility rate, and calculating a longitudinal filter kernel width adapted to the current component shielding state by reversely resolving the association model; Generating a density equalization weight for counteracting data distribution traction caused by periodic shielding by utilizing the component stiffening unit spacing and the scanning effective visible rate; and establishing an anisotropic Gaussian fitting model by combining the longitudinal filter kernel width and the density equalization weight, and reconstructing local geometric reference elements of the laser scanning point cloud data by using the anisotropic Gaussian fitting model to output a three-dimensional geometric model.
2. The method for high-precision gaussian modeling of steel box girders based on laser scanning point clouds according to claim 1, wherein obtaining laser scanning point cloud data of a steel structural member to be tested and resolving a member longitudinal extension axis of the laser scanning point cloud data comprises: Establishing a laser scanning point cloud data set containing three-dimensional space coordinates and original reflection intensity, and acquiring corresponding scanning station center coordinates; Calculating the scanning range and the laser incidence angle of each scanning point relative to the central coordinate of the scanning measuring station, and counting the scanning range median of all the scanning points; Performing optical correction on the original reflection intensity according to the square attenuation relation of the scanning range and the cosine projection relation of the laser incidence angle to generate normalized quality weight; Calculating point cloud centroid coordinates and a covariance matrix of the laser scanning point cloud data, and extracting a feature vector corresponding to a maximum feature value of the covariance matrix as an initial axis vector; calculating an axial projection coordinate of the laser scanning point cloud data on the initial axial vector, and counting a ninety-five decimal place and a five-five decimal place of an axial projection coordinate sequence; If the nine-five-digit number is smaller than the absolute value of the five-digit number, reversing the direction of the initial axis vector to serve as a final component longitudinal extension axis; If the nine-five digits are greater than or equal to the absolute value of the five digits, the direction of the initial axis vector is maintained as the final member longitudinal extension axis.
3. The method for high-precision gaussian modeling of steel box girders based on laser scanning point clouds according to claim 2, wherein the method for carrying out spatial spectrum analysis on the spatial projection density distribution by counting the spatial projection density distribution of the point clouds along the longitudinal extension axis of the component and analyzing the spacing between stiffening units of the component and the scanning effective visibility rate comprises the following steps: Calculating the neighbor average distance of each scanning point in the laser scanning point cloud data, and taking the median of the neighbor average distances of all the scanning points as a reference statistical neighborhood scale; Setting a longitudinal projection box-dividing step length according to the reference statistics neighborhood scale, carrying out equidistant box-dividing statistics on the axial projection coordinates of the laser scanning point cloud data according to the longitudinal projection box-dividing step length, and generating a box-dividing point cloud counting sequence; Dividing the bin-dividing point cloud counting sequence by the longitudinal projection bin-dividing step length, and performing linear trending treatment to generate space projection density distribution; Applying a hanning window function to the space projection density distribution, performing zero padding expansion, and then performing discrete Fourier transform to obtain a density spectrum; Extracting a main frequency index corresponding to the maximum amplitude of the non-direct current component in the density spectrum, and calculating the stiffening unit spacing of the component according to the main frequency index and the longitudinal projection box-dividing step length; and counting the density mean value and the density second moment of the space projection density distribution, calculating the ratio of the square of the density mean value to the density second moment, and carrying out numerical truncation limitation on the ratio to obtain the effective scanning visibility rate.
4. The method for modeling a steel box girder with high precision based on a laser scanning point cloud as claimed in claim 3, wherein a correlation model of a shielding effect and a smooth response factor is constructed, the correlation model is used for representing the residual degree of periodic density fluctuation caused by internal shielding of a component after smoothing, and the correlation model takes longitudinal filter kernel width, component stiffening unit spacing and scanning effective visible rate as parameters, and comprises the following steps: Calculating a sine value of the product of the peripheral rate and the scanning effective visible rate, and dividing the sine value by the product of the peripheral rate and the scanning effective visible rate to obtain a structure inherent modulation ratio; Calculating the ratio of the square of the longitudinal filter core width to the square of the component stiffening unit spacing, multiplying the ratio by the negative value of the square of the double circumference ratio, and performing exponential operation based on a natural constant on the negative value to obtain a smooth frequency domain attenuation coefficient; multiplying the structure inherent modulation ratio by the smooth frequency domain attenuation coefficient to construct a correlation model of the shielding effect and the smooth response factor.
5. The method for high-precision gaussian modeling of steel box girders based on laser scanning point clouds according to claim 4, wherein setting a target residual threshold according to the scanning effective visibility rate, calculating a longitudinal filter kernel width adapted to a current component shielding state by reversely solving the correlation model comprises: calculating the square of the structure-inherent modulation ratio as the target residual threshold; calculating the opposite number of the natural logarithmic value of the inherent modulation ratio of the structure, performing open square operation on the opposite number, multiplying the open square operation result by the spacing between stiffening units of the component, and dividing the product of the circumferential rate and the square root of two to obtain a calculated initial value of the longitudinal filter kernel width; constructing a numerical value cut-off interval, wherein the lower limit of the numerical value cut-off interval is a double standard statistical neighborhood scale, and the upper limit of the numerical value cut-off interval is a quarter of the spacing of the stiffening units of the component; And limiting the calculated initial value of the longitudinal filter kernel width in the numerical value cut-off interval to obtain the final longitudinal filter kernel width.
6. A method of high-precision gaussian modeling of steel box girders based on laser scanning point clouds according to claim 3, characterized in that generating density equalization weights for counteracting data distribution traction due to periodic occlusion using the component stiffening unit spacing and the scanning effective visibility rate comprises: extracting the maximum amplitude of the non-DC component and the amplitude of the DC component according to the density spectrum, and calculating the ratio of the maximum amplitude of the non-DC component to the amplitude of the DC component twice to obtain a first-order harmonic amplitude ratio; extracting a phase angle corresponding to the maximum amplitude of the non-direct current component in the density spectrum as a shielding phase; Constructing a normalized modulation function by combining the first-order harmonic amplitude ratio, the shielding phase and the component stiffening unit spacing and a cosine trigonometric function; calculating the product of the normalized modulation function and a preset safety coefficient, and adding one to the product to obtain the reciprocal to obtain a single-point compensation weight; and calculating the average value of the single-point compensation weights of all the scanning points, and dividing the single-point compensation weight of each scanning point by the average value to obtain the final density equalization weight.
7. The method for high-precision gaussian modeling of steel box girders based on laser scanning point clouds according to claim 6, wherein establishing an anisotropic gaussian fitting model by combining the longitudinal filter kernel width and the density equalization weight comprises: Setting the two times of standard statistics neighborhood scale as the width of the transverse filtering kernel; Constructing a transverse orthogonal base vector by using a fixed reference vector and a longitudinal extension axis of the component through a vector orthogonalization method, so as to establish a local orthogonal coordinate system; taking the weighted centroid of the region to be fitted as a geometric anchor point, and calculating coordinate component differences of each scanning point in the region to be fitted relative to the geometric anchor point under the local orthogonal coordinate system; According to the coordinate component difference, combining the longitudinal filter kernel width and the transverse filter kernel width, and calculating an anisotropic space neighborhood weight by using a Gaussian exponential function; Multiplying the anisotropic spatial neighborhood weight, the density equalization weight and the normalized quality weight to obtain a final fitting weight; And constructing an anisotropic Gaussian fitting model by utilizing the final fitting weight.
8. The method for high-precision gaussian modeling of steel box girders based on laser scanning point clouds according to claim 7, wherein the step of reconstructing local geometric reference elements of the laser scanning point cloud data by using an anisotropic gaussian fitting model to output a three-dimensional geometric model comprises the following steps: when the local geometric reference element is a welding plane, extracting a point set in a preset welding plane distance threshold value, and carrying out weighted principal component analysis on the point set by utilizing the final fitting weight to obtain a normal vector and position parameters of the welding plane; when the local geometric reference element is a chord axis, extracting a point set in a preset axis distance threshold, and carrying out weighted integral least square fitting on the point set by utilizing the final fitting weight to obtain a direction vector of the chord axis and an on-axis point coordinate; When the local geometric reference element is the center of the bolt hole, projecting the point set to a local plane, extracting the point set in a preset hole radius ring belt threshold, carrying out weighted algebraic circle fitting on the point set by utilizing the final fitting weight, analyzing to obtain a circle center parameter, and carrying out back projection on the circle center parameter to a three-dimensional space to obtain the center coordinate of the bolt hole.

Description

High-precision Gaussian modeling method for steel box girder based on laser scanning point cloud Technical Field The invention relates to the technical field of laser point cloud modeling, in particular to a high-precision Gaussian modeling method for a steel box girder based on laser scanning point cloud. Background In the field of large-scale bridge engineering and steel structure manufacturing, a steel box truss is used as a core bearing member, and the manufacturing precision directly determines the assembly quality and the structural safety of a bridge. With the development of digital manufacturing technology, three-dimensional laser scanning has become a standard means for digital pre-assembly, dimension detection and reverse modeling of steel structures. By acquiring the high-density point cloud data of the surface of the component, engineering personnel can reconstruct a three-dimensional geometric model of the component, and further key reference elements such as chord axis, welding plane and bolt hole center are extracted and used for guiding on-site assembly. In an actual data processing flow, raw point cloud data typically contains a large number of discrete noise points due to field environmental interference and measurement noise of the device itself. In order to obtain a smooth and continuous geometric model, the existing processing method widely adopts algorithms such as Gaussian filtering, gaussian weighted least squares fitting and the like. The method utilizes a Gaussian kernel function to carry out weighted smoothing on point clouds in a local neighborhood so as to inhibit the influence of random noise, thereby fitting a smooth geometric surface or line. However, existing general methods face serious challenges when gaussian modeling for complex large components such as steel box trusses. The steel box truss is internally provided with transverse partition plates, stiffening ribs, node connecting plates and other structures in a dense mode. These constructions exhibit a regular, repetitive arrangement in the direction of longitudinal extension of the bridge. The internal configuration described above produces a significant self-blocking effect when performing laser scanning, resulting in a failure of the laser beam to cover all areas. This causes the acquired point cloud data to be distributed not uniformly along the longitudinal direction of the member, but rather to exhibit periodic fluctuations in which high density windows alternate with low density shadows. Existing gaussian modeling methods typically treat the change in the point cloud density as a random distribution, or default data sampling is uniform. When such a point cloud with periodic deletions is smoothed directly by applying a conventional gaussian kernel function, the low-pass filter characteristic of the gaussian kernel will preserve such periodic density fluctuations caused by occlusion and erroneously transfer them as geometric deformations. Fitting the resulting chord axis, weld plane, or bolt hole center position creates systematic drift along the component longitudinal direction that locks with the blanking period. This drift is visually imperceptible, but numerically constitutes a significant baseline deviation, severely affecting the accuracy and reliability of subsequent segment assembly. In addition, the actual sampling quality of the laser scan is significantly affected by the angle of incidence and range. In the internal space with a complex steel structure, a large number of critical areas (such as hole edges and fillet welds) are in a large-angle incidence or long-distance scanning state, so that the echo intensity and coordinate precision of the areas are reduced. The existing method often lacks a weighting strategy for comprehensively considering the optical reflection intensity, the incident angle mechanism and the geometric shielding effect, and is difficult to realize high-precision geometric reconstruction in a complex shielding environment. Disclosure of Invention The invention provides a high-precision Gaussian modeling method for a steel box girder based on laser scanning point cloud, which solves the technical problems in the background technology. The invention provides a high-precision Gaussian modeling method for a steel box girder based on laser scanning point cloud, which comprises the following steps: acquiring laser scanning point cloud data of a steel structure member to be tested, and resolving a member longitudinal extension axis of the laser scanning point cloud data; Counting the space projection density distribution of the point cloud along the longitudinal extension axis of the component, performing space spectrum analysis on the space projection density distribution, and analyzing the spacing between stiffening units of the component and the scanning effective visibility; Constructing a correlation model of a shielding effect and a smooth response factor, wherein the correlation mod