A Tunnel Crack Identification Method and System Based on Fine Point Cloud Data

Abstract

The present invention discloses a tunnel crack identification method based on refined point cloud data, including: S1, collecting three-dimensional point cloud data and preprocessing the three-dimensional point cloud data; S2, performing feature enhancement on the preprocessed three-dimensional point cloud data; S3, automatically identifying the tunnel wall based on the three-dimensional point cloud data after feature enhancement; S4, identifying cracks based on the point cloud data of the tunnel wall surface; S5, after completing crack identification, outputting and analyzing the results of the crack area. A corresponding system, electronic device, and computer-readable storage medium are also disclosed. By obtaining high-precision point cloud data of the tunnel wall, using feature enhancement technology and region growing algorithm based on feature enhancement, the detection accuracy and identification efficiency of tunnel cracks are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of railway infrastructure detection and tunnel crack identification, belongs to the technical application field of tunnel structure service status detection and three-dimensional point cloud data processing, and particularly relates to a tunnel crack identification method and system based on refined point cloud data. Background Art

[0002] As an important infrastructure in the railway transportation network, the service performance of the tunnel structure directly affects the running safety of trains. During railway operation, the tunnel structure is affected by multiple factors such as train loads, environmental temperature and humidity changes, and geological stresses for a long time, resulting in diseases such as cracks on the tunnel wall surface. These cracks may not only weaken the structural strength of the tunnel, but also cause secondary problems such as leakage and frost heaving. In severe cases, it may even lead to major safety accidents such as tunnel collapse. The detection of tunnel wall cracks is an important part of tunnel structure health monitoring.

[0003] At present, the detection of tunnel cracks mainly adopts manual inspection and optical image detection methods. Manual inspection requires professional personnel to enter the tunnel interior and detect by visual observation or simple measuring tools. This method not only has a harsh working environment and high labor intensity, but also has subjective judgment factors, which are prone to missed detection or misdetection. With the increase in the length and number of tunnels, the manual detection method is inefficient and requires a large amount of manpower, unable to meet the needs of large-scale tunnel detection, and the efficiency and reliability of manual inspection are difficult to meet the requirements of modern railway maintenance.

[0004] The optical image detection method uses a camera device to obtain the image information of the tunnel wall surface and detects cracks through image processing and pattern recognition technology. Although this method improves the detection efficiency to a certain extent, it is limited by lighting conditions, wall surface reflection characteristics, and the resolution of the camera device, making it difficult to accurately identify fine cracks in a complex tunnel environment, with insufficient recognition ability for small cracks, and prone to missed detection and misdetection. In addition, optical images can only provide two-dimensional plane information and cannot obtain the depth and three-dimensional spatial form of cracks, limiting the assessment of crack development trends and the impact on the structure. The detection accuracy is not high: traditional image detection methods are limited by lighting conditions.

[0005] With the improvement of sensing technology and data processing capabilities, Light Detection and Ranging (LiDAR) technology has gradually been applied to the field of tunnel detection due to its advantages such as high precision, high density, and insensitivity to environmental lighting. LiDAR can obtain high-precision three-dimensional point cloud data of the tunnel wall surface, providing the possibility for the precise positioning and three-dimensional shape analysis of cracks. However, directly using the original point cloud data for crack detection still faces many challenges, as follows:

[0006] (1) Huge data volume and low processing efficiency. The high-precision point cloud data of the tunnel full section is huge. When traditional data processing algorithms process such a large amount of data, the computing efficiency is low and it is difficult to meet the requirements of real-time detection.

[0007] (2) Noise interference and incomplete data. The tunnel environment is complex. Factors such as equipment vibration and dust interference will cause noise and missing in the point cloud data, affecting the accuracy of crack identification. Moreover, the crack characteristics are not obvious and it is difficult to identify. Cracks are manifested as tiny geometric and reflection characteristic changes in the point cloud data. Traditional feature extraction methods are difficult to effectively distinguish cracks from other fine structures on the tunnel wall.

[0008] (3) Lack of effective crack identification algorithms. Existing point cloud processing algorithms are mostly used for the identification and classification of macroscopic targets, lacking special algorithms for microscopic features such as tiny cracks.

[0009] (4) The existing automated detection systems have insufficient adaptability to complex tunnel environments, low automation level, and are difficult to cope with the challenges of tunnel structure diversity and environmental complexity.

[0010] Therefore, there is an urgent need for a technical method that combines high-precision point cloud data processing and intelligent crack identification, which can efficiently and accurately identify tunnel wall cracks in a complex tunnel environment, obtain the three-dimensional spatial information and morphological characteristics of the cracks; at the same time, this method should have efficient data processing capabilities, be able to quickly process large-scale point cloud data, and meet the real-time detection requirements of tunnel cracks. This is of great significance for improving the safety of tunnel structures, extending the service life of tunnels, and reducing maintenance costs. Summary of the Invention

[0011] The object of the present invention is to provide a tunnel crack identification method and system based on refined point cloud data, aiming to solve the problems of the accuracy and efficiency of tunnel crack detection in the prior art. Specifically, the method and system of the present invention effectively improve the detection accuracy and identification efficiency of tunnel cracks by acquiring high-precision tunnel wall point cloud data and using feature enhancement technology and region growing algorithm based on feature enhancement. Specifically, first, a three-dimensional point cloud data of the tunnel wall surface is collected by using a lidar (LiDAR) system, and the three-dimensional coordinates and reflection intensity information of each sampling point are obtained. Then, the collected point cloud data is denoised, filtered, coordinate registered and smoothed to eliminate noise and anomalies and improve the data quality. Then, in the feature enhancement stage, the features of the crack area are enhanced by calculating the normal vector, curvature, surface roughness and reflection intensity of each point, highlighting the difference between it and the surrounding normal area. Based on the point cloud density and normal vector consistency, the tunnel wall surface is automatically identified and the target area is accurately extracted. Then, for crack identification, a region growing algorithm based on feature enhancement is used. Starting from the initial seed points, the crack area is gradually expanded by measuring the feature similarity of adjacent points, and growth termination conditions are set to prevent false detection. Finally, the three-dimensional spatial position, length, width and morphological features of the cracks are output, and a crack distribution map and a detection report are generated, providing a reliable basis for tunnel safety assessment and maintenance. The tunnel crack identification method and system proposed by the present invention can automatically and accurately identify small cracks on the tunnel wall surface, provide detailed three-dimensional spatial information, significantly improve the detection accuracy and identification efficiency of cracks, and ensure the good service state of the tunnel structure.

[0012] The first aspect of the present invention lies in providing a tunnel crack identification method based on refined point cloud data, including:

[0013] S1, collecting three-dimensional point cloud data and preprocessing the three-dimensional point cloud data;

[0014] S2, performing feature enhancement on the preprocessed three-dimensional point cloud data;

[0015] S3, automatically identifying the tunnel wall based on the three-dimensional point cloud data after the feature enhancement;

[0016] S4, identifying cracks based on the point cloud data of the tunnel wall surface;

[0017] S5, after the crack identification is completed, outputting and analyzing the crack area.

[0018] Preferably, the S1 includes:

[0019] S11, collecting three-dimensional point cloud data of the tunnel internal structure based on a high-precision lidar system, and the three-dimensional point cloud data is used to characterize the three-dimensional coordinates and reflection intensity information of each sampling point;

[0020] S12, preprocess the three-dimensional point cloud data, and the preprocessing includes:

[0021] (1) Remove noise from the collected three-dimensional point cloud data based on a statistical analysis-based noise reduction method to eliminate abnormal points generated by device noise and environmental interference; wherein, the statistical analysis-based noise reduction method calculates the average distance between the points corresponding to each data in the three-dimensional point cloud data and the points corresponding to its neighborhood data to calculate the neighborhood average distance, and determines whether the neighborhood average distance exceeds a preset threshold to identify abnormal points;

[0022] (2) Filter the three-dimensional point cloud data after noise removal based on a Gaussian filter to remove local fluctuations to enhance continuity and maintain the geometric features of tunnel cracks;

[0023] (3) Collect three-dimensional point cloud data multiple times through the iterative closest point algorithm to form a data set, and perform precise registration through the data set to ensure that all three-dimensional point cloud data are aligned under a unified coordinate system; the ICP algorithm continuously iteratively adjusts the rotation matrix and the translation vector , to minimize the distance between the source point cloud and the target point cloud;

[0024] (4) Smooth the three-dimensional point cloud data after the precise registration based on the bilateral filtering algorithm; the calculation formula of the bilateral filtering is formula (4):

[0025] (4);

[0026] wherein, represents a Gaussian function based on spatial distance, represents a weight function based on the difference in reflection intensity, and respectively represent the reflection intensities of the points in the source point cloud and the target point cloud;

[0027] (5) Based on the known geometric information of the tunnel structure, globally correct the three-dimensional point cloud data to ensure the consistency of the point cloud data during the crack detection process.

[0028] Preferably, the S2 includes:

[0029] S21, perform normal vector calculation, including: calculating the normal vector of each point, and enhancing the normal vector characteristics of the crack area by analyzing the normal vector change rate; the covariance matrix corresponding to the normal vector calculation is shown in formula (5):

[0030] (5);

[0031] wherein, Denotes the centroid of the neighborhood points, Denotes all k points in the neighborhood except The eigenvector corresponding to the minimum eigenvalue of the covariance matrix Is the normal vector of point ; ;

[0032] (2) Perform normal vector enhancement, including: after calculating the normal vector of each point, enhance the crack area by analyzing the change in the angle between the normal vectors of the points in the neighborhood. The calculation formula of the normal vector is shown in Equation (6):

[0033] (6);

[0034] In the formula, Denotes the change in the angle between the normal vectors of the points in the neighborhood, Denotes the normal vector of point ; Denotes the normal vector of the neighborhood point ;

[0035] S22, perform local curvature calculation and curvature enhancement, including: calculate the local curvature of each point to enhance the geometric difference between the crack area and other areas;

[0036] The said S22 includes:

[0037] (1) Local curvature calculation

[0038] For point , calculate the change rate of the normal vector in its neighborhood to obtain the local curvature κ. The calculation formula of the local curvature κ is Equation (7):

[0039] (7);

[0040] Among them, And Are the normal vectors of point And the neighborhood point respectively, and k is the number of neighborhood points;

[0041] (2) Curvature enhancement

[0042] Enhance the local curvature through non-linear transformation. The enhanced curvature formula is Equation (8):

[0043] (8);

[0044] In the formula, Denotes the minimum local curvature, Denotes the maximum local curvature;

[0045] S23. Perform surface roughness calculation and surface roughness enhancement, including: by calculating the height difference between the point cloud surface and neighboring points, enhancing the surface roughness characteristics of the crack area; the S23 includes:

[0046] (1) Surface roughness calculation:

[0047] For a point , describe the surface roughness by calculating the height variance of the points within its neighborhood , as shown in Equation (9):

[0048] (9);

[0049] Among them, represents the height of a certain point within the neighborhood of point , represents the average height of all points within the neighborhood of point ;

[0050] (2) Surface roughness enhancement:

[0051] The formula for the enhanced surface roughness is Equation (10):

[0052] (10);

[0053] Among them, represents the minimum surface roughness, represents the maximum surface roughness;

[0054] S24. Perform color intensity or reflection intensity enhancement, including: analyzing the reflection intensity information of the lidar, and using multi-scale analysis and adaptive fusion methods to enhance the reflection intensity characteristics of the crack area.

[0055] Preferably, the specific implementation of the S24 includes the following process:

[0056] (1) Perform multi-scale color intensity analysis;

[0057] Based on the multi-scale analysis method, calculate the reflection intensity difference between each point and the points within its neighborhood at different scales , and the formula is as shown in Equation (11):

[0058] (11);

[0059] Among them, is the neighborhood point set at scale s, and are the reflection intensities of point and neighborhood point respectively;

[0060] (2) Perform adaptive enhancement coefficient fusion;

[0061] Calculate the enhancement coefficient at each scale and perform adaptive fusion; the enhancement coefficient is determined by the standard deviation of the reflection intensity and the calculation formula of the enhancement coefficient is shown in the following formula (12):

[0062] (12);

[0063] In the formula, represents the enhancement coefficient; represents the standard deviation of the reflection intensity; represents the maximum standard deviation of the reflection intensity;

[0064] (3) Calculate the enhanced color intensity or reflection intensity;

[0065] The enhanced color intensity or reflection intensity is calculated by the following formula (13):

[0066] (13);

[0067] Among them, represents the color intensity or reflection intensity of the point before enhancement; represents the minimum value of the color intensity or reflection intensity of all points; represents the minimum color intensity value or reflection intensity value in the current point cloud dataset.

[0068] Preferably, the S3 includes:

[0069] S31, based on the point cloud density and the normal vector consistency, preliminarily identify the tunnel wall surface and each part of the infrastructure, and distinguish the tunnel wall from the interference area based on the spatial density of the tunnel wall point cloud and the change characteristics of its surface normal vector. The interference area includes one or more of the track, equipment, and ground;

[0070] (1) The spatial density of the tunnel wall point cloud is obtained based on the point cloud density analysis, and the point cloud density analysis includes: screening out the tunnel wall surface area by analyzing the density difference of the tunnel point cloud;

[0071] (2) The change characteristics of the surface normal vector are obtained based on the normal vector consistency analysis, and the normal vector consistency analysis includes: further distinguishing the tunnel wall surface area by calculating the change of the normal vector. The calculation formula of the normal vector is shown in the following formula (14):

[0072] (14);

[0073] In the formula, represents the change in the angle between the normal vectors of points in the neighborhood, Indicate point The normal vector of Represents neighborhood points The normal vector of

[0074] S32, based on the characteristics of tunnel point cloud, combined with the spatial position constraints of tunnel point cloud data, using height constraints, curvature constraints and symmetry analysis, accurately extracts the tunnel wall area;

[0075] S33, extracting and outputting accurate tunnel wall point cloud data for subsequent crack detection.

[0076] Preferably, S4 includes:

[0077] S41, using a region growing algorithm based on feature enhancement, automatically selects seed points of the crack region from points with large normal vector change rate, high curvature, and large surface roughness;

[0078] S42, construct a multidimensional similarity measurement model, combine the characteristics of normal vector, curvature, surface roughness and reflection intensity, and calculate the similarity between adjacent points and seed points; the multidimensional similarity measurement formula is shown in the following formula (16):

[0079] (16);

[0080] S43, based on the similarity metric, the fracture area is expanded along the fracture main axis direction to maintain the linear characteristics of the fracture;

[0081] S44, setting the termination condition of crack growth, when the similarity metric value is lower than the preset threshold or the crack width reaches the maximum threshold, stop the regional growth, and prevent misdetection caused by excessive expansion by limiting the crack length;

[0082] S45, smoothing and correcting the crack boundary based on the post-processing algorithm;

[0083] S46, removing redundant small regions identified during the growth process based on morphological analysis.

[0084] Preferably, S5 includes:

[0085] S51, result output, includes generating the spatial position, size and morphological characteristics of the crack through the analysis of the three-dimensional point cloud data. The specific steps are as follows:

[0086] (1) Crack spatial position output, including: recording the three-dimensional spatial coordinates of the crack and generating a crack distribution map to intuitively display the precise location of the crack on the tunnel wall;

[0087] (2) Crack size calculation, including: calculating the length of the crack based on the crack identification area by accumulating the distances between adjacent points in the main axis direction of the crack; and calculating the width of the crack by the distance between the left and right boundary points of the crack;

[0088] (3) Outputting crack morphology information, where the crack morphology features are used to provide detailed crack description information, including the extension direction and geometric features of the crack;

[0089] (4) Performing three-dimensional visualization display on the crack identification results, including: displaying the crack in the tunnel point cloud model with marked colors or lines;

[0090] S52. After the crack identification results are output, classify the types of cracks and analyze the development trend, evaluate the impact of the cracks on the tunnel structure, and generate a complete crack detection report; including:

[0091] (1) Crack type analysis: Classify the cracks according to the length, width, and morphology features of the cracks, and mark the high-risk cracks;

[0092] (2) Crack development trend analysis: Combine multiple detection data to analyze the change trend of the cracks and evaluate the risk level of the cracks;

[0093] (3) Crack impact assessment: Evaluate the impact of the cracks on the tunnel structure according to the location and size of the cracks, and provide early warning information.

[0094] The second aspect of the present invention also provides a tunnel crack identification system based on refined point cloud data for implementing the method of the first aspect, including:

[0095] A data acquisition module for acquiring three-dimensional point cloud data and preprocessing the three-dimensional point cloud data;

[0096] A feature enhancement module for enhancing the features of the preprocessed three-dimensional point cloud data;

[0097] A tunnel wall extraction module for automatically identifying the tunnel wall based on the feature-enhanced three-dimensional point cloud data;

[0098] A crack identification module for identifying cracks based on the tunnel wall point cloud data;

[0099] A result output module for outputting and analyzing the results of the crack area after the crack identification is completed.

[0100] The third aspect of the present invention provides an electronic device, including a processor and a memory, where the memory stores multiple instructions, and the processor is used to read the instructions and execute the method described in the first aspect.

[0101] The fourth aspect of the present invention provides a computer-readable storage medium storing multiple instructions that can be read and executed by a processor to perform the method described in the first aspect.

[0102] Advantages of the method and system of the present invention:

[0103] It can efficiently and accurately identify tunnel wall cracks in a complex tunnel environment, obtain the three-dimensional spatial information and morphological characteristics of the cracks. At the same time, it has efficient data processing capabilities, can quickly process large-scale point cloud data, and meet the real-time detection requirements of tunnel cracks. This is of great significance for improving the safety of tunnel structures, extending the service life of tunnels, and reducing maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0104] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the related art, the following will briefly introduce the drawings required for use in the description of the specific embodiments or the related art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0105] Figure 1 FIG. is the overall flowchart of tunnel crack identification based on refined point cloud data provided by an embodiment of the present invention.

[0106] Figure 2 FIG. is the effect diagram of point cloud data preprocessing provided by an embodiment of the present invention, including steps such as noise reduction, filtering, coordinate registration, and smoothing processing.

[0107] Figure 3 FIG. is the flowchart of feature enhancement provided by an embodiment of the present invention, showing the processing process of normal vector, curvature, surface roughness, and reflection intensity enhancement.

[0108] Figure 4 FIG. is a schematic diagram of the identification of each part of the tunnel cross-section provided by an embodiment of the present invention, showing the original point cloud data and the identification results of each part in the tunnel.

[0109] Figure 5 FIG. is a schematic diagram of the region growing algorithm during crack identification provided by an embodiment of the present invention, showing the growth process of the crack region and the similarity measurement.

[0110] Figure 6 FIG. is the final identification result diagram after the crack region growth ends provided by an embodiment of the present invention, showing the identification results after boundary correction and small region removal.

[0111] Figure 7Structural diagram of an electronic device provided according to an embodiment of the present invention. Detailed implementation manners

[0112] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0113] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and should not be construed as indicating or implying relative importance.

[0114] In the description of the present invention, it should be noted that unless otherwise clearly defined and limited, the terms "mounted", "connected", and "connected" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected, or indirectly connected through an intermediate medium, and it may be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations. Embodiment 1

[0115] As Figure 1 shown, the first aspect of the present invention in this embodiment is to provide a tunnel crack identification method based on refined point cloud data, including:

[0116] S1. Collect three-dimensional point cloud data and preprocess the three-dimensional point cloud data, including:

[0117] S11. Collect three-dimensional point cloud data of the internal structure of the tunnel based on a high-precision lidar (LiDAR) system, and the three-dimensional point cloud data is used to represent the three-dimensional coordinates and reflection intensity information of each sampling point;

[0118] In this embodiment, to achieve high-precision identification of cracks in railway tunnels, the present invention adopts a three-dimensional point cloud data acquisition system based on Light Detection and Ranging (LiDAR). This system obtains high-resolution three-dimensional coordinates and reflection intensity information through non-contact scanning of the tunnel wall surface, and can comprehensively and accurately record the geometric morphological characteristics of the tunnel wall surface, providing reliable data support for subsequent crack detection.

[0119] As Figure 2 shown, S12, preprocess the three-dimensional point cloud data, and the preprocessing includes:

[0120] 1. Use a noise reduction method based on statistical analysis to remove noise from the collected three-dimensional point cloud data to eliminate abnormal points generated by equipment noise and environmental interference; among them, the noise reduction method based on statistical analysis calculates the neighborhood average distance by calculating the average distance between the points corresponding to each data in the three-dimensional point cloud data and the points corresponding to its neighborhood data, and determines whether the neighborhood average distance exceeds a preset threshold to identify abnormal points;

[0121] In this embodiment, to remove noise points and isolated points generated by equipment vibration or external environmental interference in the original point cloud data, a noise reduction method based on statistical analysis is adopted. Vibration and environmental interference in railway tunnels will cause invalid points to appear in the data. As the first step in the preprocessing process, noise reduction processing aims to remove these abnormal points to ensure that subsequent processing is not interfered.

[0122] The noise reduction method based on statistical analysis calculates the neighborhood average distance through the average distance between each point and its neighborhood points, and determines whether the neighborhood average distance exceeds a preset threshold to identify abnormal points. The neighborhood average distance D i is calculated as shown in the following formula (1):

[0123] (1);

[0124] Among them, represents the neighborhood average distance of point , represents the total number of the nearest points in the neighborhood, represents all points in the neighborhood except ; if the neighborhood average distance i of point P exceeds the preset threshold, then this point is considered a noise point and is removed.

[0125] 2. Use a Gaussian filter to filter the three-dimensional point cloud data after noise removal to remove local fluctuations to enhance continuity and maintain the geometric characteristics of the tunnel cracks;

[0126] In this embodiment, after noise reduction processing, there may still be some local fluctuations in the point cloud data caused by device jitter or external vibration, which will affect the accurate identification of cracks. To enhance the smoothness and continuity of the data, the embodiment of the present invention uses Gaussian filtering to process the point cloud data.

[0127] The Gaussian filter smooths by weighting the spatial distances between points and their neighboring points, preserves geometric features, and eliminates high-frequency noise. Its filtering function is shown in Equation (2):

[0128] (2);

[0129] where σ represents the standard deviation of the Gaussian distribution. By adjusting the value of σ, it is ensured that the filter can effectively remove noise when processing tunnel wall data without affecting the retention of crack features; x, y, z represent the distances between adjacent points and the current point.

[0130] 3. Multiple three-dimensional point cloud data are collected through the Iterative Closest Point (ICP) algorithm to form a data set, and precise registration is performed uniformly through the data set to ensure that all three-dimensional point cloud data are aligned under a unified coordinate system;

[0131] In this embodiment, since the detection of railway tunnels usually needs to be carried out in segments, there may be attitude and coordinate deviations in the three-dimensional point cloud data collected in different batches. This embodiment uses the Iterative Closest Point (ICP) algorithm for data registration to ensure that all data sets can be unified under the same coordinate system.

[0132] The ICP algorithm continuously iteratively adjusts the rotation matrix and the translation vector of the point cloud to minimize the distance between the source point cloud and the target point cloud. The objective function of the ICP algorithm is Equation (3):

[0133] (3);

[0134] where, and respectively represent the points in the source point cloud and the target point cloud, and are the rotation matrix and the translation vector respectively.

[0135] 4. The three-dimensional point cloud data after the precise registration is smoothed based on the bilateral filtering algorithm to maintain crack features and enhance the boundary;

[0136] In this embodiment, after registration is completed, the bilateral filtering method is used to further smooth the point cloud data to preserve boundary features such as cracks. Bilateral filtering not only considers the spatial distance of points but also combines the reflection intensity information, enabling the retention of fine structures while smoothing the noise. The calculation formula of bilateral filtering is shown in Equation (4):

[0137] (4);

[0138] wherein, represents the Gaussian function based on the spatial distance, represents the weight function based on the reflection intensity difference, and respectively represent the reflection intensities of the points in the source point cloud and the target point cloud.

[0139] 5. Based on the known geometric information of the tunnel structure, perform global correction on the three-dimensional point cloud data to ensure the consistency of the point cloud data during the crack detection process.

[0140] As Figure 3 shown in, S2, perform feature enhancement on the preprocessed three-dimensional point cloud data;

[0141] In this embodiment, after the collected point cloud data undergoes noise reduction, filtering, coordinate registration, and smoothing processing, the complex geometric structure of the tunnel wall still makes the geometric features in the crack area relatively subtle, especially in areas with uneven curvature, where the features of the cracks are difficult to clearly show. Therefore, to improve the detection accuracy of cracks, this embodiment adopts a variety of feature enhancement techniques, including normal vector enhancement, curvature enhancement, surface roughness enhancement, and reflection intensity enhancement. These techniques can significantly improve the distinguishability between the crack area and other tunnel walls, thereby improving the accuracy of crack identification.

[0142] As a preferred implementation manner, the S2 includes:

[0143] S21, perform normal vector calculation and normal vector enhancement, including: calculating the normal vector of each point and enhancing the normal vector features of the crack area by analyzing the normal vector change rate;

[0144] In this embodiment:

[0145] (1) Normal vector calculation

[0146] The normal vector is the surface direction of each point in the point cloud data. The crack area usually shows a sharp change in the normal vector. Therefore, the normal vector change is one of the important features for crack detection. Calculate the normal vector of each point by the method of local plane fitting, and enhance the features of the crack area by analyzing the normal vector change rate. The covariance matrix corresponding to the normal vector calculation is shown in Equation (5):

[0147] (5);

[0148] Wherein, represents the centroid of the neighborhood points, represents all k points in the neighborhood except . The eigenvector corresponding to the minimum eigenvalue of the covariance matrix is the normal vector of the point .

[0149] (2) Normal vector enhancement

[0150] After calculating the normal vector of each point, the crack area is enhanced by analyzing the change in the angle between the normal vectors of the points in the neighborhood. The calculation formula of the normal vector is shown in Equation (6):

[0151] (6);

[0152] In the formula, represents the change in the angle between the normal vectors of the points in the neighborhood, represents the normal vector of the point , represents the normal vector of the neighborhood point ;

[0153] S22, perform local curvature calculation and curvature enhancement, including: calculating the local curvature of each point and enhancing the geometric difference between the crack area and other areas.

[0154] In this embodiment: The curvature of the crack area is significantly different from that of other areas of the tunnel wall, and the curvature change is also an important feature for crack detection. The present invention calculates the local curvature of each point and performs enhancement processing to highlight the geometric feature change of the crack area.

[0155] (1) Local curvature calculation

[0156] For the point , calculate the change rate of the normal vector in its neighborhood to obtain the local curvature κ. The calculation formula of the local curvature κ is shown in Equation (7):

[0157] (7);

[0158] Wherein, and are the normal vectors of the point and the neighborhood point respectively, and k is the number of neighborhood points.

[0159] (2) Curvature enhancement

[0160] ​Enhance the local curvature through non - linear transformation to better distinguish the curvature changes in the crack area. The enhanced curvature formula is Equation (8):

[0161] (8);

[0162] In the formula, represents the minimum local curvature, represents the maximum local curvature.

[0163] S23. Perform surface roughness calculation and surface roughness enhancement, including: By calculating the height difference between the point cloud surface and neighboring points, enhance the surface roughness characteristics of the crack area.

[0164] In this embodiment, the surface of the crack area is usually relatively rough, while the non - crack area is relatively smooth. Therefore, the characteristics of the crack area can be further enhanced by calculating the surface roughness of the point cloud data. The surface roughness R s reflects the height difference between a point and other points in its neighborhood.

[0165] (1) Surface roughness calculation:

[0166] For point , describe the surface roughness by calculating the height variance of points in its neighborhood , as shown in Equation (9):

[0167] (9);

[0168] Among them, represents the height of a certain point in the neighborhood of point , represents the average height of all points in the neighborhood of point .

[0169] (2) Surface roughness enhancement:

[0170] In order to better distinguish the crack area, this embodiment performs enhancement processing on the roughness. The enhanced surface roughness formula is Equation (10):

[0171] (10);

[0172] Among them, represents the minimum surface roughness, represents the maximum surface roughness.

[0173] S24. Perform color intensity or reflection intensity enhancement, including: Analyze the reflection intensity information of the lidar, and use multi - scale analysis and adaptive fusion methods to enhance the reflection intensity characteristics of the crack area.

[0174] In this embodiment, the reflected intensity information of the lidar provides additional features of the tunnel wall material. In the crack area, due to the change of the surface material or structure, the reflected intensity is different from that of the normal area. To improve the detection accuracy of cracks, this embodiment proposes a color intensity enhancement method based on multi-scale segmentation and adaptive fusion, which can adapt to the complex material and illumination conditions of the tunnel and enhance the reflected intensity characteristics of the crack area. In this embodiment, the specific implementation of step S24 includes the following processes:

[0175] (1) Conduct multi-scale color intensity analysis;

[0176] The present invention introduces a multi-scale analysis method to calculate the reflected intensity difference between each point and the points in its neighborhood at different scales , and the formula is shown in Equation (11):

[0177] (11);

[0178] where, is the neighborhood point set at scale s, and are the reflected intensities of point and neighborhood point respectively.

[0179] (2) Conduct adaptive enhancement coefficient fusion;

[0180] In this embodiment, to enhance the reflected intensity difference, the enhancement coefficient is calculated at each scale and adaptive fusion is performed. The enhancement coefficient is determined by the standard deviation of the reflected intensity. The calculation formula of the enhancement coefficient is shown in the following Equation (12):

[0181] (12);

[0182] In the formula, represents the enhancement coefficient; represents the standard deviation of the reflected intensity; represents the maximum standard deviation of the reflected intensity.

[0183] (3) Calculate the enhanced color intensity or reflected intensity;

[0184] In this embodiment, the enhanced color intensity or reflected intensity is calculated by the following formula (13):

[0185] (13);

[0186] where, represents the color intensity or reflection intensity of the point before enhancement ; represents the minimum value of the color intensity or reflection intensity of all points represents the minimum color intensity value or reflection intensity value in the current point cloud dataset

[0187] For example Figure 4 as shown in, S3, automatically identify the tunnel wall based on the three-dimensional point cloud data enhanced by the described features

[0188] As a preferred embodiment, the S3 includes

[0189] S31, based on the point cloud density and the consistency of the normal vectors, preliminarily identify the tunnel wall surface and each part of the infrastructure, and distinguish the tunnel wall from the interference regions based on the spatial density of the tunnel wall point cloud and the change characteristics of its surface normal vector, where the interference regions include one or more of the track, equipment, and ground

[0190] In this embodiment, step S31 is different from the traditional geometric fitting method and adopts an automatic recognition method based on the point cloud density and the normal vector. This method relies on the spatial density of the tunnel wall point cloud and the change characteristics of its surface normal vector to distinguish the tunnel wall from the interference regions such as the track, equipment, and ground

[0191] (1) The spatial density of the tunnel wall point cloud is obtained based on the point cloud density analysis, and the point cloud density analysis includes: screening out the tunnel wall surface area by analyzing the density difference of the tunnel point cloud

[0192] (2) The change characteristics of the surface normal vector are obtained based on the normal vector consistency analysis, and the normal vector consistency analysis includes: further distinguishing the tunnel wall surface area by calculating the change of the normal vector. The region with higher normal vector consistency is often the smooth tunnel wall surface. The calculation formula of the normal vector is as shown in formula (14):

[0193] (14);

[0194] In the formula represents the change of the angle between the normal vectors of the points in the neighborhood represents the point 's normal vector represents the neighborhood point 's normal vector

[0195] S32, based on the characteristics of the tunnel point cloud, combined with the spatial position constraint of the tunnel point cloud data, use height constraint, curvature constraint, and symmetry analysis to accurately extract the tunnel wall surface area

[0196] In this embodiment, in order to further improve the accuracy of tunnel wall extraction and in combination with the characteristics of tunnel point clouds, spatial position constraints are added to the extraction process in this embodiment, especially considering the curvature and symmetry characteristics of tunnel point cloud data.

[0197] Tunnel point cloud data has the following remarkable characteristics: The tunnel wall is usually symmetrically distributed along the tunnel axis, and the height and curvature are relatively regular in the tunnel wall area.

[0198] In view of these characteristics, the following constraint methods are proposed in this embodiment to materialize the spatial position constraints for accurately extracting the tunnel wall area:

[0199] (1) Height constraint: The height of the tunnel point cloud has obvious upper and lower boundaries in the tunnel wall area. Generally, the point cloud of the tunnel wall is distributed within a certain height range, while the point cloud heights in areas such as the ground, track, and tunnel top equipment are different. In order to accurately extract the tunnel wall, the embodiment of the present invention uses the known structural information of the tunnel to set a reasonable height threshold range h min and h max to screen the point cloud data located within the wall height range;

[0200] (2) Curvature constraint: The point cloud of the tunnel wall is usually distributed along the tunnel surface, with a small and uniform curvature, while other areas in the tunnel (such as the track and equipment) have a larger curvature or complex geometric shapes. Therefore, by calculating the curvature of the tunnel point cloud, the point cloud that does not belong to the wall can be further excluded.

[0201] (3) Symmetry analysis: The tunnel wall usually has a certain symmetry, and the tunnel point cloud distribution has good symmetry along the tunnel axis. By analyzing the distribution of the point cloud relative to the tunnel central axis, the embodiment of the present invention further enhances the recognition of the tunnel wall. Specifically, by fitting the tunnel axis L(t), calculating the distance from the point to the axis , and screening out the point cloud that conforms to the symmetric distribution, as shown in Equation (15):

[0202] (15);

[0203] where, is a point on the axis, is the direction vector of the axis. By setting a reasonable distance threshold , the point cloud of non-tunnel walls can be further filtered out.

[0204] S33, extract and output the accurate tunnel wall point cloud data for subsequent crack detection.

[0205] In this embodiment, the overall method is embodied as an automatic recognition process for tunnel wall surfaces. Based on the above analysis, the automatic recognition process for tunnel walls proposed in the embodiments of the present invention is as follows:

[0206] Point cloud preprocessing: Denoise, filter, and register the original point cloud data to ensure data quality;

[0207] Point cloud density analysis: Screen out high-density areas by calculating the local point cloud density to initially identify the tunnel wall surface;

[0208] Normal vector analysis: Calculate the normal vector consistency to further accurately distinguish the tunnel wall surface from other structures;

[0209] Spatial position constraint and trend analysis: Combine the height, curvature, and symmetry characteristics of the tunnel point cloud to accurately extract the tunnel wall point cloud;

[0210] Output tunnel wall surface data: Extract and output accurate tunnel wall point cloud data for subsequent crack detection.

[0211] S4. Identify cracks based on the tunnel wall point cloud data;

[0212] As a preferred implementation, in view of the characteristics of tunnel cracks, a region growing method based on feature enhancement is proposed. This method is particularly suitable for the identification of tunnel cracks, especially those that are small, slender, and difficult to accurately detect by conventional methods. The specific steps are as follows:

[0213] The said S4 includes:

[0214] As shown in Figure 5 S41. Adopt a region growing algorithm based on feature enhancement to automatically select seed points for the crack region from points with large normal vector change rate, high curvature, and large surface roughness;

[0215] In this embodiment, seed points for the crack region are selected from the point cloud data with large normal vector change rate, high curvature, and large surface roughness.

[0216] S42. Construct a multi-dimensional similarity measurement model, combine the features of normal vector, curvature, surface roughness, and reflection intensity, and calculate the similarity between adjacent points and seed points;

[0217] In this embodiment, a multi-dimensional similarity measurement model is constructed to ensure the accurate expansion of the crack region. The multi-dimensional similarity measurement formula is shown as formula (16) below:

[0218] (16);

[0219] This formula (16) combines the normal vector, curvature, surface roughness, and reflection intensity to judge the similarity between adjacent points and seed points, thereby controlling the expansion of the region.

[0220] S43. Expand the crack area along the main axis direction of the crack based on similarity measurement, and maintain the linear characteristics of the crack;

[0221] In this embodiment, based on similarity measurement, the expansion of the crack area first follows the geometric characteristics of the main axis direction of the crack. The present invention makes special constraints on the slender shape of the crack to ensure that the linear characteristics of the crack area are maintained during the expansion process and to avoid expansion into non-crack areas.

[0222] S44. Set the termination condition for crack growth. When the similarity measurement value is lower than the preset threshold or the crack width reaches the maximum threshold, stop the region growth, and prevent misdetection caused by over-expansion by limiting the crack length.

[0223] In this embodiment, the termination condition for crack growth is: when the expanded area no longer meets the similarity condition, or the width of the crack reaches the maximum threshold, the growth process stops. In addition, by limiting the crack length, misdetection caused by over-growth is prevented.

[0224] As Figure 6 shown, S45. Smoothly correct the crack boundary based on the post-processing algorithm.

[0225] In this embodiment, the expanded crack area may have the problem of irregular boundaries. The embodiment of the present invention smoothly corrects the crack boundary through the post-processing algorithm to ensure the smoothness and accuracy of the boundary.

[0226] S46. Eliminate the redundant small areas identified during the growth process based on morphological analysis;

[0227] In this embodiment, for some redundant small areas identified during the growth process, they are eliminated through morphological analysis to ensure that the finally identified crack area has significant physical meaning.

[0228] S5. After completing the crack identification, output and analyze the results of the crack area.

[0229] In this embodiment, by describing in detail the spatial position, size, and morphological characteristics of the crack, it is ensured that the crack detection results can provide a reliable basis for tunnel safety assessment.

[0230] As a preferred embodiment, the said S5 includes:

[0231] S51. Result output, including generating the three-dimensional spatial position, size (length, width), and morphological characteristics of the crack through the analysis of three-dimensional point cloud data. The specific steps are as follows:

[0232] (1)Output of crack spatial position, including: recording the three-dimensional spatial coordinates of the crack and generating a crack distribution map to visually display the exact position of the crack on the tunnel wall surface;

[0233] In this embodiment, the identified cracks are calibrated by three-dimensional spatial coordinates. The system records the specific positions of the cracks and generates a crack distribution map. This distribution map visually displays the exact position of the cracks on the tunnel wall surface, which is helpful for subsequent maintenance and repair.

[0234] (2)Calculation of crack size, including: calculating the length of the crack based on the crack identification area by accumulating the distances between adjacent points in the main axis direction of the crack; and calculating the width of the crack by the distance between the left and right boundary points of the crack.

[0235] The system automatically calculates the length and width of the crack according to the crack identification area. The crack length is measured by consecutive points in the main axis direction, as shown in formula (17):

[0236] (17);

[0237] where, represents the crack length, and are the coordinates of adjacent points in the main axis direction of the crack.

[0238] The crack width is calculated according to the distance between the left and right boundaries, as shown in formula (18):

[0239] (18);

[0240] where, represents the crack width, and are the points of the left and right boundaries of the crack respectively, is the number of sampling points in the transverse direction of the crack.

[0241] (3)Output of crack morphology information, where the crack morphology features are used to provide detailed crack description information, including the extension direction and geometric features of the crack;

[0242] In this embodiment, the crack morphology information is mainly described by the extension direction and geometric features of the crack. The system analyzes the change of the normal vector in the crack area to generate the crack morphology feature information, which further helps the tunnel maintenance personnel to understand the trend and shape of the crack.

[0243] (4)Three-dimensional visualization display of the crack identification results, including: the cracks are displayed in the tunnel point cloud model with marked colors or lines. Through this visualization display, users can clearly understand the position, size of the cracks and their distribution in the tunnel structure.

[0244] S52. After the crack identification result is output, classify the cracks by type and analyze the development trend, evaluate the impact of the cracks on the tunnel structure, and generate a complete crack detection report.

[0245] In this embodiment, S52 is specifically as follows:

[0246] (1) Crack type analysis: Classify the cracks according to the length, width and morphological characteristics of the cracks, and mark the high-risk cracks.

[0247] The system classifies the cracks according to the length, width and morphological characteristics of the cracks. Long cracks or cracks with a larger width are considered high-risk cracks, and the system will mark them as key attention objects. By analyzing the crack type, maintenance personnel can take corresponding treatment measures according to the actual situation.

[0248] (2) Crack development trend analysis: Combine multiple detection data to analyze the change trend of the cracks and evaluate the risk level of the cracks.

[0249] The system can combine multiple detection data to analyze the change trend of the cracks. If it is detected that the cracks have a significant increase in length and width, the system will identify the area where the cracks develop rapidly. For such cracks with an obvious trend, the system will judge the risk level of the cracks according to their change speed.

[0250] (3) Crack impact assessment: Evaluate the impact of the cracks on the tunnel structure according to the location and size of the cracks, and provide early warning information.

[0251] The crack impact assessment is based on the location and size of the cracks. The system will judge its impact on the overall safety of the tunnel according to the different structural parts where the cracks are located. If the cracks are located in the area of tunnel stress concentration, the system will give corresponding warnings to remind relevant personnel to take repair measures in time.

[0252] Embodiment 2

[0253] This embodiment provides a tunnel crack identification system based on refined point cloud data for implementing the method of Embodiment 1, including:

[0254] A data acquisition module for acquiring three-dimensional point cloud data and preprocessing the three-dimensional point cloud data.

[0255] A feature enhancement module for enhancing the features of the preprocessed three-dimensional point cloud data.

[0256] A tunnel wall extraction module for automatically identifying the tunnel wall based on the feature-enhanced three-dimensional point cloud data.

[0257] A crack identification module for identifying cracks based on the point cloud data of the tunnel wall surface.

[0258] A result output module, configured to output and analyze the results of the crack area after crack identification is completed.

[0259] The present invention also provides a memory storing multiple instructions for implementing the method as in Embodiment 1.

[0260] As Figure 7 As shown, the present invention also provides an electronic device, including a processor 301 and a memory 302 connected to the processor 301. The memory 302 stores multiple instructions that can be loaded and executed by the processor, enabling the processor to execute the method as in Embodiment 1.

[0261] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A tunnel crack identification method based on refined point cloud data, characterized in that: include: S1, collecting three-dimensional point cloud data and preprocessing the three-dimensional point cloud data; S2, performing feature enhancement on the preprocessed three-dimensional point cloud data; S3, automatically identifying the tunnel wall based on the three-dimensional point cloud data after the feature enhancement; S4, crack identification based on tunnel wall point cloud data; S5, after completing the crack identification, the result output and analysis of the crack area; The S1 includes: S11, collecting three-dimensional point cloud data of the internal structure of the tunnel based on a high-precision laser radar system, wherein the three-dimensional point cloud data is used to characterize the three-dimensional coordinates and reflection intensity information of each sampling point; S12, preprocessing the three-dimensional point cloud data; The S2 includes: S21, performing normal vector calculation, including: calculating the normal vector of each point, and enhancing the normal vector feature of the crack area by analyzing the normal vector change rate; S22, performing local curvature calculation and curvature enhancement, including: calculating the local curvature of each point, and enhancing the geometric difference between the crack area and other areas; S23, performing surface roughness calculation and surface roughness enhancement, including: enhancing the surface roughness characteristics of the crack area by calculating the height difference between the point cloud surface and the neighboring points; S24, performing color intensity or reflection intensity enhancement, including: analyzing reflection intensity information of the laser radar, and enhancing reflection intensity characteristics of the crack area by using multi-scale analysis and adaptive fusion methods; The S4 includes: S41, using a region growing algorithm based on feature enhancement, automatically selects seed points of the crack region from points with large normal vector change rate, high curvature, and large surface roughness; S42, construct a multi-dimensional similarity measurement model, combining the features of normal vector, curvature, surface roughness and reflection intensity to calculate the similarity between adjacent points and seed points; S43, based on the similarity metric, the fracture area is expanded along the fracture main axis direction to maintain the linear characteristics of the fracture; S44, setting the termination condition of crack growth, when the similarity metric value is lower than the preset threshold or the crack width reaches the maximum threshold, stop the regional growth, and prevent misdetection caused by excessive expansion by limiting the crack length; S45, smoothing and correcting the crack boundary based on the post-processing algorithm; S46, removing redundant small regions identified during the growth process based on morphological analysis.

2. The tunnel crack identification method based on refined point cloud data according to claim 1 is characterized in that: The pre-processing comprises: (1) A noise reduction method based on statistical analysis is used to remove noise from the collected three-dimensional point cloud data to eliminate abnormal points caused by equipment noise and environmental interference; wherein the noise reduction method based on statistical analysis calculates the average neighborhood distance by calculating the average distance between each point corresponding to each data in the three-dimensional point cloud data and the point corresponding to its neighboring data, and determines whether the neighborhood average distance exceeds a preset threshold to identify abnormal points; (2) filtering the noise-removed three-dimensional point cloud data based on a Gaussian filter to remove local fluctuations to enhance continuity and maintain the geometric characteristics of the tunnel cracks; (3) The three-dimensional point cloud data is collected multiple times through the iterative closest point algorithm to form a data set, and the data set is uniformly and accurately registered to ensure that all three-dimensional point cloud data are aligned in a unified coordinate system; the ICP algorithm adjusts the rotation matrix of the point cloud through continuous iteration and translation vectors , so that the distance between the source point cloud and the target point cloud is minimized; (4) Smoothing the three-dimensional point cloud data after the precise registration based on a bilateral filtering algorithm; the calculation formula of the bilateral filtering is formula (4): （4）； in, represents a Gaussian function based on spatial distance, represents the weight function based on the difference in reflection intensity, and Represent the reflection intensity of points in the source point cloud and the target point cloud respectively; represents the three-dimensional coordinate vector of the point to be smoothed, Indicate point The three-dimensional coordinates of the jth neighboring point in the neighborhood, Represents the Euclidean space distance between point i and its neighboring point j; (5) Based on the known geometric information of the tunnel structure, the three-dimensional point cloud data is globally corrected to ensure the consistency of the point cloud data during the crack detection process.

3. The tunnel crack identification method based on refined point cloud data according to claim 2 is characterized in that: The covariance matrix corresponding to the normal vector calculation is shown in formula (5): （5）； in, represents the centroid of the neighborhood points, Indicates that the neighborhood For all k points except The eigenvector corresponding to the minimum eigenvalue of the point The normal vector , represents the offset vector of the jth neighborhood point relative to the centroid, Represents the transpose operation; Normal vector enhancement is performed, including: after calculating the normal vector of each point, the crack area is enhanced by analyzing the change in the angle between the normal vectors of the points in the neighborhood. The calculation formula of the normal vector is shown in formula (6): （6）； In the formula, represents the change in the angle between the normal vectors of points in the neighborhood, Indicate point The normal vector of Represents neighborhood points The normal vector of The S22 includes: (1) Local curvature calculation For point , calculate the rate of change of the normal vector in its neighborhood to obtain the local curvature κ. The calculation formula of the local curvature κ is formula (7): （7）； in, and Points and neighboring points The normal vector of , k is the number of neighborhood points; (2) Curvature enhancement The local curvature is enhanced by nonlinear transformation, and the enhanced curvature formula is formula (8): （8）； In the formula, represents the minimum local curvature, represents the maximum local curvature; The S23 includes: (1) Surface roughness calculation: For point , describing the surface roughness by calculating the height variance of points in its neighborhood , as shown in formula (9): （9）； in, Indicate point The height of a point in the neighborhood of Indicate point The mean height of all points in the neighborhood of ; (2) Surface roughness enhancement: The surface roughness formula after enhancement is as follows: （10）； in, Indicates the minimum surface roughness, Indicates the maximum surface roughness; The specific implementation of S24 includes the following process: (1) Perform multi-scale color intensity analysis; Based on the multi-scale analysis method, each point is calculated at different scales. The difference in reflection intensity between points in its neighborhood , the formula is shown in formula (11): （11）； in, is the neighborhood point set at scale s, and Points and neighboring points The reflection intensity; (2) Perform adaptive enhancement coefficient fusion; Calculate the enhancement coefficient at each scale , and adaptive fusion; enhancement coefficient The standard deviation of the reflected intensity Determine the enhancement factor The calculation formula is shown in formula (12): （12）； In the formula, represents the enhancement coefficient; represents the standard deviation of the reflection intensity; Indicates the maximum standard deviation of the reflection intensity; (3) Calculating the enhanced color intensity or reflection intensity; Enhanced color intensity or reflection intensity Calculated by the following formula (13): （13）； in, Represents the point before enhancement The intensity of the color or reflection; Indicates the minimum color intensity or reflection intensity of all points; Indicates the minimum color intensity value or reflection intensity value in the current point cloud dataset; The S3 includes: S31, based on the point cloud density and normal vector consistency, preliminarily identifying the tunnel wall and various parts of the infrastructure, and distinguishing the tunnel wall from the interference area based on the spatial density of the tunnel wall point cloud and the surface normal vector change characteristics, wherein the interference area includes one or more of the track, equipment and ground; (1) The spatial density of the tunnel wall point cloud is obtained based on point cloud density analysis, and the point cloud density analysis includes: screening out the tunnel wall area by analyzing the density difference of the tunnel point cloud; (2) The surface normal vector change feature is obtained based on the normal vector consistency analysis, and the normal vector consistency analysis includes: further distinguishing the tunnel wall area by calculating the change of the normal vector. The calculation formula is shown in formula (14): （14）； In the formula, Represents the absolute value of the cosine of the angle between the normal vectors of points in the neighborhood. Indicate point The normal vector of Represents neighborhood points The normal vector of S32, based on the characteristics of tunnel point cloud, combined with the spatial position constraints of tunnel point cloud data, using height constraints, curvature constraints and symmetry analysis, accurately extracts the tunnel wall area; S33, extracting and outputting accurate tunnel wall point cloud data for subsequent crack detection.

4. The tunnel crack identification method based on refined point cloud data according to claim 3 is characterized in that: The multi-dimensional similarity measurement formula of S41 is shown in the following formula (16): （16）； in, , Respectively indicate points , The unit normal vector of the point, whose direction describes the geometric orientation of the local surface where the point is located; , Indicate point , The local curvature value reflects the degree of surface concavity and convexity changes around the point; , Indicate point , Surface roughness index; , Points , The reflection intensity value of , , , is the weight factor of each feature item; each division term is Normalization ensures that the feature components of different dimensions are weighted and calculated at a unified scale; similarity measurement , the larger the value, the greater the difference between the two points in comprehensive characteristics.

5. The tunnel crack identification method based on refined point cloud data according to claim 4 is characterized in that The S5 includes: S51, result output, includes generating the spatial position, size and morphological characteristics of the crack through the analysis of the three-dimensional point cloud data. The specific steps are as follows: (1) Crack spatial position output, including: recording the three-dimensional spatial coordinates of the crack and generating a crack distribution map to intuitively display the precise location of the crack on the tunnel wall; (2) Calculation of crack size, including: based on the crack identification area, calculating the length of the crack by accumulating the distances between adjacent points in the crack main axis direction; and calculating the width of the crack by the distances between the left and right boundary points of the crack; (3) outputting crack morphology information, wherein the crack morphology information is used to provide detailed crack description information, including the extension direction and geometric characteristics of the crack; (4) 3D visualization of the crack identification results, including: displaying the cracks in the tunnel point cloud model with marked colors or lines; S52, after the crack identification results are output, the crack types are classified and the development trend is analyzed, the impact of the cracks on the tunnel structure is evaluated, and a complete crack detection report is generated; including: (1) Crack type analysis: Cracks are classified according to their length, width, and morphological characteristics, and high-risk cracks are marked; (2) Crack development trend analysis: Combine multiple detection data to analyze the change trend of cracks and assess the risk level of cracks; (3) Crack impact assessment: Based on the location and size of the cracks, assess their impact on the tunnel structure and provide early warning information.

6. A tunnel crack identification system based on refined point cloud data, used to implement any of the methods of claims 1-5, characterized in that: include: A data acquisition module, used for acquiring three-dimensional point cloud data and preprocessing the three-dimensional point cloud data; A feature enhancement module, used for performing feature enhancement on the preprocessed three-dimensional point cloud data; A tunnel wall extraction module, used for automatically identifying tunnel walls based on the three-dimensional point cloud data after the feature enhancement; A crack identification module, used to identify cracks based on tunnel wall point cloud data; The result output module is used to output and analyze the results of the crack area after the crack identification is completed.

7. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a plurality of instructions, and the processor is used to read the instructions and execute the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the plurality of instructions can be read by a processor to execute the method according to any one of claims 1 to 5.

Images