Adaptive elliptic cylinder filtering method based on tunnel point cloud data

Abstract

The invention discloses a self-adaptive elliptic cylinder filtering method based on tunnel point cloud data, and belongs to the field of tunnel point cloud data processing. According to the method, the RANSAC algorithm is innovatively combined with dynamic threshold estimation, dynamic parameter adjustment and iterative optimization are carried out, adaptive optimization of point cloud characteristics of different regions is realized, a multi-stage optimization process of PCA initialization, RANSAC optimization and iterative least square optimization is fused, the precision and efficiency of model fitting are improved, and the method is suitable for large-scale popularization and application. High-precision point cloud filtering of a complex tunnel environment is realized, and the precision and efficiency of tunnel three-dimensional reconstruction and deformation analysis are improved.

Description

Technical Field

[0001] The invention belongs to the field of tunnel point cloud data processing, and in particular relates to an adaptive elliptical cylinder filtering method based on tunnel point cloud data. Background Art

[0002] In the existing field of tunnel point cloud data processing, point cloud data processing mainly includes filtering methods based on statistical analysis, geometric feature extraction methods and model fitting methods. These methods have some common limitations in practical applications and are difficult to meet the needs of complex and changing tunnel environments.

[0003] Filtering methods based on statistical analysis, such as the Statistical Outlier Removal (SOR) model, usually use fixed parameters to identify and remove noise points. However, such methods are sensitive to parameter selection and are difficult to adapt to the dynamic changes in the shape and size of the tunnel cross section, which can easily lead to insufficient filtering or excessive filtering. Especially when dealing with local deformation of the tunnel or special structures (such as shelters, cross passages, etc.), fixed parameter methods often have difficulty in accurately identifying structural features, resulting in the loss or misjudgment of useful information.

[0004] Geometric feature extraction methods, including those based on normal vector analysis and local curvature estimation, can capture the geometric characteristics of the tunnel surface well, but their computational efficiency is low when processing high-noise, high-density point cloud data. In addition, these methods have high requirements for the uniformity of point cloud density. In the actual collected tunnel point cloud data, the uneven density caused by factors such as occlusion and scanning angle often affects the accuracy of feature extraction.

[0005] Model fitting methods, such as the Random Sample Consensus (RANSAC) algorithm and Hough transform, identify tunnel structures by fitting predefined geometric models (such as cylinders or ellipsoids). Such methods work well under ideal conditions, but when faced with complex tunnel structures (such as double-track tunnels, variable-section tunnels, etc.), a single model is difficult to accurately describe the geometry of the entire tunnel. At the same time, these methods usually adopt global optimization strategies with high computational complexity, making it difficult to meet the real-time processing requirements of large-scale point cloud data.

[0006] In addition, existing technologies generally lack an effective mechanism to dynamically adjust the filtering threshold, and it is difficult to adapt to the changes in the local characteristics of point cloud data in different areas. This leads to unstable filtering effects when processing long tunnels or tunnels under complex geological conditions. At the same time, the recognition ability of ancillary facilities in the tunnel (such as pipelines, signs, monitoring equipment, etc.) is limited, and it is difficult to effectively distinguish these structures from real noise points. Existing technologies also face the problem of balancing accuracy and efficiency. High-precision methods often have high computational complexity and are difficult to meet the needs of real-time processing; while fast processing methods may sacrifice accuracy and cannot meet the requirements of high-precision three-dimensional reconstruction and deformation monitoring. Summary of the invention

[0007] The present invention aims to address the technical problems existing in the prior art and provides an adaptive elliptical cylinder filtering method based on tunnel point cloud data. Through dynamic parameter adjustment and iterative optimization, the RANSAC algorithm is innovatively combined with dynamic threshold estimation to achieve adaptive optimization of point cloud characteristics in different regions. It integrates a multi-stage optimization process of PCA initialization, RANSAC optimization and iterative least squares optimization, improves the accuracy and efficiency of model fitting, achieves high-precision point cloud filtering for complex tunnel environments, and improves the accuracy and efficiency of tunnel 3D reconstruction and deformation analysis.

[0008] The present invention provides an adaptive elliptical cylinder filtering method based on tunnel point cloud data, comprising the following steps: Extract tunnel centerline and set cross section based on tunnel point cloud data; Initialize the elliptical cylinder model parameters on each cross section to form an elliptical cylinder model; The initial screening threshold is set to preliminarily screen outliers in the distance from all points in the cross section to the elliptical cylinder, and the RANSAC algorithm is used to dynamically iterate and estimate the optimal screening threshold; Iteratively update and optimize the elliptical cylinder model parameters according to the current inner point until the elliptical cylinder model parameters converge; Based on the optimized elliptical cylinder model and dynamic threshold, the tunnel point cloud data is filtered, and the filtered point cloud data and elliptical cylinder model parameters are output.

[0009] Preferably, the tunnel centerline is extracted from the preliminary denoised tunnel point cloud data using a least squares method or a principal curve algorithm.

[0010] Preferably, the tunnel centerline is used as a reference axis, and the cross sections are arranged at fixed intervals along the axial direction of the reference axis.

[0011] Preferably, the fixed interval is 1-2 m.

[0012] Preferably, on each cross section, the principal component analysis method is used to initialize the elliptical cylinder model parameters, and the length and direction of the major axis and minor axis of the cross section are estimated.

[0013] Preferably, the principal component analysis method is used to initialize the elliptical cylinder model parameters, and the length and direction of the major axis and minor axis of the cross section are estimated, specifically including: Calculate the covariance matrix of the cross-section point cloud and perform eigenvalue decomposition on the covariance matrix; The directions of the major and minor axes of the elliptical cylinder model are determined using the eigenvectors corresponding to the largest eigenvalue and the second largest eigenvalue, and the lengths of the major and minor axes are estimated according to the ratio of the eigenvalues.

[0014] Preferably, dynamically iteratively estimating the optimal screening threshold using the RANSAC algorithm specifically includes the following steps: The RANSAC iterative process randomly selects the minimum sample set to fit the elliptical cylinder model; Calculate the distances from all points to the elliptical cylinder and count the number of inner points. Repeat this process multiple times and select the elliptical cylinder model with the most inner points as the best fitting result. According to the best fitting result, the screening threshold d is dynamically adjusted as the optimal screening threshold.

[0015] Preferably, the initial screening threshold d is 20 cm, and the 95% quantile of the inner point distance is taken as the optimal screening threshold.

[0016] Preferably, iteratively updating and optimizing the elliptical cylinder model parameters according to the current interior point until the elliptical cylinder model parameters converge specifically includes: The parameters of the elliptic cylinder are optimized using an iterative least square method, wherein the iterative process includes re-estimating the major axis, minor axis, center position and direction of the elliptic cylinder based on the current interior point; The sum of squared distances from a point to an elliptical cylinder is minimized using the Levenberg-Marquardt algorithm. Update the elliptical cylinder model parameters and recalculate the interior points; The above process is repeated until the parameters of the elliptical cylinder model converge.

[0017] Preferably, the convergence criterion of the elliptical cylinder model parameters is that the relative change rate of the parameter value is less than 0.1% or reaches a preset number of iterations.

[0018] Preferably, filtering the tunnel point cloud data based on the optimized elliptical cylinder model and the dynamic threshold comprises: Points within the optimal screening threshold are retained as tunnel structure points, and points beyond the optimal screening threshold are removed as non-structure points or noise; The parameters of adjacent sections are smoothed to ensure the continuity of the elliptical cylinder model.

[0019] Preferably, a super elliptical cylinder or a B-spline surface model is used to replace the elliptical cylinder, or a multi-model fusion method is used to automatically select the most suitable model according to the characteristics of different tunnel sections.

[0020] Preferably, M-estimator Sample Consensus (MSAC) or Progressive Sample Consensus (PROSAC) is used instead of the RANSAC algorithm.

[0021] Preferably, the iterative least squares method is replaced by a genetic algorithm, a particle swarm optimization or a simulated annealing algorithm.

[0022] Preferably, a Bayesian optimization method is used to automatically adjust various parameters in the algorithm, such as the number of iterations of RANSAC, the convergence threshold, etc.

[0023] Compared with the prior art, the present invention has the following beneficial effects: 1. Improve filtering accuracy and adaptability: Compared with conventional technologies, the present invention can filter out noise points more accurately while retaining the detailed features of the tunnel structure. This is mainly achieved by using an elliptical cylinder model instead of a traditional cylindrical model to better adapt to the non-circular cross-section and deformation of the tunnel. PCA is used for initialization, providing a more accurate initial estimate. Combined with RANSAC and dynamic threshold estimation, adaptive optimization of point cloud characteristics in different regions is achieved.

[0024] 2. Improve the stability and reliability of the model: Through multiple optimization strategies, the present invention significantly improves the stability of the model. PCA initialization provides a stable starting point; RANSAC optimization enhances the robustness to outliers; and iterative least squares optimization further optimizes the elliptical cylinder model parameters.

[0025] 3. Improve the separation effect of noise points and structural points: The present invention significantly improves the separation effect of noise points and structural points through dynamic threshold and multi-stage optimization. The initial threshold setting provides a reasonable starting separation standard; the threshold is dynamically adjusted during the RANSAC process to adapt to the characteristics of different regions; the final filtering stage uses the optimized model and threshold for accurate separation. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 The flowchart of the adaptive elliptical cylinder filtering method based on tunnel point cloud data is an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] The present invention provides an adaptive elliptical cylinder filtering method based on tunnel point cloud data, comprising the following steps: Extract tunnel centerline and set cross section based on tunnel point cloud data; Initialize the elliptical cylinder model parameters on each cross section to form an elliptical cylinder model; The initial screening threshold is set to preliminarily screen outliers in the distance from all points in the cross section to the elliptical cylinder, and the RANSAC algorithm is used to dynamically iterate and estimate the optimal screening threshold; Iteratively update and optimize the elliptical cylinder model parameters according to the current inner point until the elliptical cylinder model parameters converge; Based on the optimized elliptical cylinder model and dynamic threshold, the tunnel point cloud data is filtered, and the filtered point cloud data and elliptical cylinder model parameters are output.

[0029] In certain specific embodiments of the present invention, the tunnel centerline is extracted from the preliminary denoised tunnel point cloud data using a least squares method or a principal curve algorithm.

[0030] In certain specific embodiments of the present invention, the tunnel centerline is taken as a reference axis, and the cross sections are arranged at fixed intervals along the axial direction of the reference axis.

[0031] In certain specific embodiments of the present invention, the fixed interval is 1-2 m.

[0032] In certain specific embodiments of the present invention, on each cross section, the principal component analysis method is used to initialize the elliptical cylinder model parameters and estimate the length and direction of the major axis and minor axis of the cross section.

[0033] In certain specific embodiments of the present invention, the principal component analysis method is used to initialize the elliptical cylinder model parameters, and the length and direction of the major axis and the minor axis of the cross section are estimated, which specifically includes: Calculate the covariance matrix of the cross-section point cloud and perform eigenvalue decomposition on the covariance matrix; The directions of the major and minor axes of the elliptical cylinder model are determined using the eigenvectors corresponding to the largest eigenvalue and the second largest eigenvalue, and the lengths of the major and minor axes are estimated according to the ratio of the eigenvalues.

[0034] In certain specific embodiments of the present invention, dynamically iterating and estimating the optimal screening threshold using the RANSAC algorithm specifically includes the following steps: The RANSAC iterative process randomly selects the minimum sample set to fit the elliptical cylinder model; Calculate the distances from all points to the elliptical cylinder and count the number of inner points. Repeat this process multiple times and select the elliptical cylinder model with the most inner points as the best fitting result. According to the best fitting result, the screening threshold d is dynamically adjusted as the optimal screening threshold.

[0035] In certain specific embodiments of the present invention, the initial screening threshold d is 20 cm, and the 95% quantile of the internal point distance is taken as the optimal screening threshold.

[0036] In certain specific embodiments of the present invention, iteratively updating and optimizing the elliptical cylinder model parameters according to the current interior point until the elliptical cylinder model parameters converge specifically includes: The parameters of the elliptic cylinder are optimized using an iterative least square method, wherein the iterative process includes re-estimating the major axis, minor axis, center position and direction in the elliptic cylinder based on the current interior point; The sum of squared distances from a point to an elliptical cylinder is minimized using the Levenberg-Marquardt algorithm. Update the elliptical cylinder model parameters and recalculate the interior points; The above process is repeated until the parameters of the elliptical cylinder model converge.

[0037] In certain specific embodiments of the present invention, the convergence criterion of the elliptical cylinder model parameters is that the relative change rate of the parameter value is less than 0.1% or reaches a preset number of iterations.

[0038] In certain specific embodiments of the present invention, filtering the tunnel point cloud data based on the optimized elliptical cylinder model and the dynamic threshold comprises: Points within the optimal screening threshold are retained as tunnel structure points, and points beyond the optimal screening threshold are removed as non-structure points or noise; The parameters of adjacent sections are smoothed to ensure the continuity of the elliptical cylinder model.

[0039] In certain specific embodiments of the present invention, a super elliptical cylinder or a B-spline surface model is used to replace the elliptical cylinder, or a multi-model fusion method is used to automatically select the most suitable model according to the characteristics of different tunnel sections.

[0040] In certain specific embodiments of the present invention, M-estimator Sample Consensus (MSAC) or Progressive Sample Consensus (PROSAC) is used to replace the RANSAC algorithm.

[0041] In certain embodiments of the present invention, a genetic algorithm, a particle swarm optimization or a simulated annealing algorithm is used instead of the iterative least squares method.

[0042] In certain specific embodiments of the present invention, a Bayesian optimization method is used to automatically adjust various parameters in the algorithm, such as the number of iterations of RANSAC, the convergence threshold, etc.

[0043] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. An adaptive elliptical cylinder filtering method based on tunnel point cloud data, It is characterized in that The steps include: Extract tunnel centerline and set cross section based on tunnel point cloud data; Initialize the elliptical cylinder model parameters on each cross section to form an elliptical cylinder model; The initial screening threshold is set to preliminarily screen outliers in the distance from all points in the cross section to the elliptical cylinder, and the RANSAC algorithm is used to dynamically iterate and estimate the optimal screening threshold; Iteratively update and optimize the elliptical cylinder model parameters according to the current inner point until the elliptical cylinder model parameters converge; Based on the optimized elliptical cylinder model and dynamic threshold, the tunnel point cloud data is filtered, and the filtered point cloud data and elliptical cylinder model parameters are output.

2. The adaptive elliptical cylinder filtering method based on tunnel point cloud data according to claim 1, It is characterized in that The tunnel centerline is extracted from the preliminary denoised tunnel point cloud data using the least squares method or principal curve algorithm.

3. The adaptive elliptical cylinder filtering method based on tunnel point cloud data according to claim 1, It is characterized in that Taking the tunnel centerline as the reference axis, sections are set at fixed intervals along the axial direction of the reference axis, with a fixed interval of 1-2m.

4. The adaptive elliptical cylinder filtering method based on tunnel point cloud data according to claim 1, It is characterized in that On each cross section, the principal component analysis method is used to initialize the parameters of the elliptical cylinder model and estimate the length and direction of the major and minor axes of the cross section.

5. The adaptive elliptical cylinder filtering method based on tunnel point cloud data according to claim 4, It is characterized in that The principal component analysis method is used to initialize the parameters of the elliptical cylinder model, and the length and direction of the major and minor axes of the cross section are estimated, including: Calculate the covariance matrix of the cross-section point cloud and perform eigenvalue decomposition on the covariance matrix; The directions of the major and minor axes of the elliptical cylinder model are determined using the eigenvectors corresponding to the largest eigenvalue and the second largest eigenvalue, and the lengths of the major and minor axes are estimated according to the ratio of the eigenvalues.

6. The adaptive elliptical cylinder filtering method based on tunnel point cloud data according to claim 5, It is characterized in that The dynamic iterative estimation of the optimal screening threshold using the RANSAC algorithm specifically includes the following steps: The RANSAC iterative process randomly selects the minimum sample set to fit the elliptical cylinder model; Calculate the distances from all points to the elliptical cylinder and count the number of inner points. Repeat this process multiple times and select the elliptical cylinder model with the most inner points as the best fitting result. According to the best fitting result, the screening threshold d is dynamically adjusted as the optimal screening threshold.

7. The adaptive elliptical cylinder filtering method based on tunnel point cloud data according to claim 1, It is characterized in that The initial screening threshold d was 20 cm, and the 95% quantile of the inlier distance was taken as the optimal screening threshold.

8. The adaptive elliptical cylinder filtering method based on tunnel point cloud data according to claim 1, It is characterized in that Iteratively updating and optimizing the elliptical cylinder model parameters according to the current inner point until the elliptical cylinder model parameters converge, specifically including: The parameters of the elliptic cylinder are optimized using an iterative least square method, wherein the iterative process includes re-estimating the major axis, minor axis, center position and direction in the elliptic cylinder based on the current interior point; The sum of squared distances from a point to an elliptical cylinder is minimized using the Levenberg-Marquardt algorithm. Update the elliptical cylinder model parameters and recalculate the interior points; The above process is repeated until the parameters of the elliptical cylinder model converge.

9. The adaptive elliptical cylinder filtering method based on tunnel point cloud data according to claim 8, It is characterized in that The parameter convergence criterion of the elliptical cylinder model is that the relative change rate of the parameter value is less than 0.1% or reaches the preset number of iterations.

10. The adaptive elliptical cylinder filtering method based on tunnel point cloud data according to claim 1, It is characterized in that Based on the optimized elliptical cylinder model and dynamic threshold, the tunnel point cloud data is filtered including: Points within the optimal screening threshold are retained as tunnel structure points, and points beyond the optimal screening threshold are removed as non-structure points or noise; The parameters of adjacent sections are smoothed to ensure the continuity of the elliptical cylinder model.

Images