CN-122023266-A - Method and system for detecting surface flatness of high-speed railway vehicle body

CN122023266ACN 122023266 ACN122023266 ACN 122023266ACN-122023266-A

Abstract

The invention discloses a method and a system for detecting the surface flatness of a high-speed railway vehicle body. The method comprises five steps of point cloud preprocessing, self-adaptive parameter optimization and dual-strategy plane fitting, projection distance calculation, statistical analysis and stable flatness index construction, result output and visualization, wherein a dual-level self-adaptive parameter optimization mechanism and a dual-strategy fitting method are designed based on a RANSAC framework, and the fast, accurate and non-contact detection of the surface flatness of the high-speed railway vehicle body is realized by combining the robust statistical analysis of Median Absolute Deviation (MAD). According to the invention, by introducing a plane fitting strategy with self-adaptive parameter optimization, a double-strategy robust fitting mechanism and robust statistical analysis of projection distance, high-reliability flatness detection of a large-area and weak-feature surface point cloud of a high-speed rail vehicle body is realized, manual parameter adjustment is not needed, fitting stability and noise immunity are strong, detection efficiency is high, and the result visualization degree is high, so that the method can be widely applied to surface flatness detection scenes in high-speed rail manufacturing and maintenance.

Inventors

ZHANG XIAOJIAN
LIAN ZENGWEI
Qi Bosong
ZHANG HAIYANG
YAN SIJIE
DING HAN

Assignees

华中科技大学
华中科技大学无锡研究院

Dates

Publication Date: 20260512
Application Date: 20251231

Claims (10)

1. The method for detecting the surface flatness of the high-speed railway vehicle body is characterized by comprising the following steps of: S100, optimizing original high-speed rail train body point cloud data through a noise reduction unit, a region clipping and ROI limiting unit and a voxel grid uniform downsampling unit, improving quality and regularity of the original high-speed rail train body point cloud data, and outputting regular point cloud data; s200, constructing a double-layer-level self-adaptive parameter space of a distance threshold value and an angle threshold value through the cooperative operation of a parameter generation unit, a double-strategy fitting unit and a model screening unit, and acquiring an optimal reference plane model with the interior point proportion meeting a preset threshold value by combining double-strategy RANSAC fitting with angle constraint and angle-free constraint; S300, calculating the vertical projection distances from all points in the regular point cloud to the plane based on the optimal reference plane model, and generating a projection distance sequence; S400, based on the cooperative operation of a basic statistics unit, an abnormal point removing unit and a comprehensive index construction unit, performing multi-dimensional statistics feature calculation on the projection distance sequence, and constructing a comprehensive flatness index after removing abnormal points; and S500, outputting a global flatness score and a multi-dimensional visual result by a result output and visual module, and providing a basis for quality detection.
2. The method for detecting the flatness of a vehicle body surface of a high-speed railway according to claim 1, wherein in step S100, the point cloud preprocessing includes: S101, identifying and removing isolated points and abnormal noise points based on statistical characteristics of neighborhood point distribution, and adopting a statistical filtering algorithm to reduce noise of input point cloud so as to improve accuracy and stability of point cloud data; s102, combining structural morphological characteristics of the surface of a high-speed railway vehicle body, performing region clipping and limiting a region of interest (ROI) on a point cloud to exclude irrelevant components and interference regions and ensure that a data range is consistent with an actual detection position; And S103, improving the distribution uniformity of the point cloud, reducing the complexity of subsequent operation, uniformly downsampling the cut point cloud by using a Voxel Grid (Voxel Grid), and effectively compressing the number of the point cloud and adjusting the density by dividing voxels of a space region and replacing points in the voxels by using Voxel center points, thereby laying a data foundation for the stable construction of a subsequent plane fitting model and the accurate execution of a flatness evaluation algorithm.
3. The method for detecting the flatness of a car body surface of a high-speed railway according to claim 2, wherein the point cloud executing region clipping and the defining and sampling expression of a region of interest (ROI) are as follows: , Wherein, the P represents the whole set of original high-speed railway vehicle body point clouds, and P represents any three-dimensional point in the original point clouds P; m represents the number of origin points contained in the jth voxel; representing the three-dimensional coordinates of the kth origin within the jth voxel.
4. A method for detecting the flatness of a surface of a high-speed railway car body according to any of claims 1-3, wherein in step S200, the adaptive parameter optimization and the dual strategy plane fitting comprise: S201, constructing a self-adaptive parameter space, generating a distance threshold sequence in a geometric multiplication mode, generating an angle threshold sequence in a linear growth mode, and simultaneously presetting the minimum interior point proportion to form a double-layer-level parameter combination adapting to different point cloud characteristics; S202, executing a RANSAC fitting strategy with angle constraint, performing iterative fitting on all parameter combinations one by one, and calculating a sequence meeting the condition that the point-to-plane distance does not exceed a distance threshold value And the normal included angle is not larger than the inner point set and the proportion of the angle threshold sequence, if the proportion of certain combined inner points reaches the lowest inner point proportion Triggering an early termination mechanism; s203, a second strategy, namely if the first strategy does not meet the inner point proportion requirement, automatically switching to an adaptive RANSAC fitting strategy without angle constraint, judging the inner points only by a distance threshold value, and if the first strategy does not meet the lowest inner point proportion Then expand the distance threshold sequence Searching a range, and continuously performing iterative optimization; s204, terminating the mechanism in advance, if the candidate model interior point proportion corresponding to a certain parameter combination reaches the lowest interior point proportion And immediately triggering a termination flow, and outputting the model as a final fitting optimal reference plane model.
5. The method for detecting the flatness of a vehicle body surface of a high-speed railway according to claim 4, characterized in that the minimum interior point ratio is required In each set of parameter combinations Then, fitting the obtained candidate plane model Calculate its interior point set ; The interior points are defined as satisfying at the same time that the point-to-plane distance is not more than And the normal included angle is not more than Point set of (a), namely: , And requires the normal deviation of the model itself Satisfy the following requirements 。
6. The method for detecting the flatness of a vehicle body surface of a high-speed rail according to claim 5, characterized in that the number of inner points And the proportion thereof Is used as a basis for judging the fitting quality; the proportion of the corresponding internal points to the internal points is as follows: , wherein the distance threshold Using geometric multiplication The mode is set as The method is used for carrying out hierarchical expansion on the allowable range of the point-to-plane distance; Angle threshold By linear growth The mode is set as So as to gradually relax the constraint of the included angle between the normal direction of the plane and the reference direction, thereby adapting to the local posture change of the surface of the high-speed railway vehicle body.
7. The method for detecting the surface flatness of a high-speed railway vehicle body according to claim 6, wherein the two RANSAC fitting strategies are executed, a first type of strategy of RANSAC plane fitting with angle constraint is executed first, and if the inner point proportion requirement is not met in the strategy with angle constraint, the system is automatically switched to a second strategy of adaptive RANSAC fitting without angle constraint; In the first strategy, for all 、 Fitting is performed by parameter combinations of (a) to obtain a plurality of model sets ; For each candidate model Calculation of Selecting an optimal model The criteria of (2) may be defined as: Simultaneously recording the optimal interior point proportion as follows: , If present So that Triggering early termination mechanism and outputting the model To finally fit the model, if at all All in combination have Entering a second strategy; in a second strategy, the normal angle constraint is cancelled based only on the distance threshold Judging the inner point, and comprising: , the same is chosen in the second strategy: , If it is Output then If it is still not satisfied, continue to increase If the inlier ratios of the multiple models are very close, using the sum of the residuals as a secondary criterion, the model residuals are defined as: , Final selection of the The smallest model serves as the optimal model.
8. The method for detecting the flatness of a vehicle body surface of a high-speed railway according to any one of claims 1-7, characterized in that in step S300, based on a final reference plane model obtained by adaptive plane fitting, the accurate geometric deviation quantization is performed on the point cloud surface morphology, and the parameter coefficients of the fitting plane are extracted first The planar model may be expressed as: ; Will input any point in the point cloud Mapping to the reference plane, and calculating the vertical projection distance, wherein the calculation formula is as follows: , by executing the distance operation on all points, a complete projection distance sequence is obtained ; The sequence comprehensively reflects the local protrusion, depression and overall deformation characteristics of the surface to be measured relative to the reference plane, and provides a core quantitative basis for flatness index calculation and abnormal region identification.
9. The method for detecting the flatness of a vehicle body surface of a high-speed railway according to claim 1, wherein the statistical analysis and the quantitative evaluation of the flatness in step S400 include: S401, extracting basic statistical features, calculating multi-dimensional core statistics aiming at projection distance sequences, wherein the multi-dimensional core statistics comprise a mean value mu and a median med reflecting a distance distribution concentration trend, and a standard deviation sigma and a Median Absolute Deviation (MAD) representing discrete features, wherein the MAD is defined as the median of absolute values of projection distances and median differences of all points, and comprehensively describing the overall distribution characteristics of distance data; S402, robust outlier rejection, namely setting an outlier judgment threshold value based on MAD, accurately identifying and rejecting outliers caused by weld residue, local pitting, pits and the like, avoiding interference of extreme deviation on flatness evaluation, and ensuring validity of analysis data; s403, correcting statistical feature calculation, and re-calculating mean value of effective projection distance sequence after eliminating abnormal points Standard deviation of The core statistics are equal, the influence of outliers on the original statistics result is eliminated, and the fit degree of the statistics data and the surface true flatness is improved; S404, constructing a comprehensive flatness index, namely constructing a comprehensive flatness index F taking the overall fluctuation trend and the local fluctuation feature into account based on the corrected statistical features, and realizing comprehensive and accurate quantitative evaluation on the surface flatness by controlling the influence of mean deviation and the influence of the scattering of beta-control data through a weight coefficient alpha.
10. A high-speed railway automobile body surface smoothness detecting system, characterized by comprising: the point cloud preprocessing module optimizes the original high-speed railway vehicle body point cloud data through the noise reduction unit, the region clipping and ROI limiting unit and the voxel grid uniform downsampling unit, improves the quality and regularity of the original high-speed railway vehicle body point cloud data, and outputs regular point cloud data; The self-adaptive parameter optimization and dual-strategy plane fitting module constructs a double-layer-level self-adaptive parameter space of a distance threshold value and an angle threshold value through the cooperative operation of a parameter generation unit, a dual-strategy fitting unit and a model screening unit, and combines dual-strategy RANSAC fitting with angle constraint and non-angle constraint to acquire an optimal reference plane model with the interior point proportion meeting a preset threshold value; the projection distance calculation module comprises a plane parameter extraction unit and a vertical distance calculation unit, and calculates the vertical projection distances from all points in the regular point cloud to the plane based on the optimal reference plane model to generate a projection distance sequence; The statistical analysis and flatness quantitative evaluation module is used for cooperatively operating a basic statistical unit, an abnormal point removing unit and a comprehensive index construction unit, carrying out multi-dimensional statistical feature calculation on the projection distance sequence, and constructing a comprehensive flatness index after removing abnormal points; And the result output and visualization module outputs a global flatness score and a multi-dimensional visualization result, and provides a basis for quality detection.

Description

Method and system for detecting surface flatness of high-speed railway vehicle body The invention belongs to the technical field of high-speed rail manufacturing and detection, and particularly relates to a method and a system for detecting the surface flatness of a high-speed rail vehicle body. Background The surface flatness is a core index for measuring the geometric accuracy and the processing quality of industrial components, and has an irreplaceable effect in the fields of traffic equipment manufacturing, engineering structure detection and the like. The high-speed railway car body is used as key equipment for high-speed operation, the outer covering part of the high-speed railway car body is formed by welding, shaping and assembling large-area metal plates, and the surface flatness not only directly influences the appearance quality of the whole car, but also is closely related to the aerodynamic characteristics and structural stability in the operation process, so that the high-speed railway car body surface flatness is rapidly, reliably and non-contact detected, and the high-speed railway car body is a key technical requirement for high-speed railway manufacturing and maintenance links. At present, a three-dimensional point cloud-based detection method has become a main flow direction of surface flatness detection, related schemes exist in the prior art, a three-dimensional point cloud-based surface flatness detection method disclosed in a patent document CN112233248B is used for constructing a reference curved surface by utilizing local window and natural adjacent point interpolation through converting the point cloud into an ordered lattice so as to generate a flatness image and position defects, and a prefabricated Liang Duanbiao surface flatness detection method based on a three-dimensional point cloud model is disclosed in a patent document CN114037706B, and flatness calculation is realized through point cloud calibration, normal vector estimation and iterative reference surface fitting. However, aiming at the typical characteristics of wide point cloud characteristic area, high noise interference and large point cloud density change caused by high surface reflectivity of the surface of the high-speed railway vehicle, the prior art still has the obvious defects that firstly, the conventional fixed-parameter plane fitting algorithm is easy to generate unstable fitting on the weak characteristic and local irregular surface, the whole surface trend is difficult to be accurately reflected by a reference plane, secondly, the fitting result is highly sensitive to parameters such as distance threshold value, angle setting and the like, manual repeated parameter adjustment is needed under different point cloud scenes, the degree of automation is low, the field detection requirement is difficult to be adapted, thirdly, the flatness evaluation index is single, the interference of local abnormal points such as weld residue and pitting is easy to be caused, and the real surface state is difficult to be comprehensively and accurately reflected. In summary, the existing detection method has defects in noise robustness, parameter self-adaption and reference surface fitting stability, and cannot fully meet the actual requirements of high-speed railway vehicle body surface flatness detection, and a detection method capable of adaptively adjusting parameters, stable fitting and reliable evaluation is needed. Disclosure of Invention Aiming at the defects or improvement demands of the prior art, the invention provides a high-reliability flatness detection method for the surface point cloud of the large-area and weak-feature surface of the high-speed railway vehicle body by introducing a plane fitting strategy, a double-strategy robust fitting mechanism and robust statistical analysis of projection distances optimized by self-adaptive parameters. In order to achieve the above object, the present invention provides a method for detecting the flatness of the surface of a high-speed rail car body, comprising: S100, optimizing original high-speed rail train body point cloud data through a noise reduction unit, a region clipping and ROI limiting unit and a voxel grid uniform downsampling unit, improving quality and regularity of the original high-speed rail train body point cloud data, and outputting regular point cloud data; s200, constructing a double-layer-level self-adaptive parameter space of a distance threshold value and an angle threshold value through the cooperative operation of a parameter generation unit, a double-strategy fitting unit and a model screening unit, and acquiring an optimal reference plane model with the interior point proportion meeting a preset threshold value by combining double-strategy RANSAC fitting with angle constraint and angle-free constraint; S300, calculating the vertical projection distances from all points in the regular point cloud to the plane based on the optimal reference pla