Tunnel quality safety detection system and method based on laser radar data

Abstract

The invention provides a tunnel quality safety detection system and method based on laser radar data, and the method achieves the fusion and complementation of crack detection based on point cloud density change and crack detection based on curvature change, and achieves the quick recognition and precise detection of a crack region. The potential crack area is rapidly positioned through point cloud density change detection, and the boundary of the crack is further accurately positioned in combination with curvature change detection, so that the accuracy and integrity of crack detection are improved. For tiny deformation which is very tiny in change in point cloud density and curvature and is difficult to directly detect, a projection plane is further determined on the basis of the identified crack, and after point cloud data is projected to a two-dimensional plane, the characteristics of the tiny crack in the image are highlighted through image enhancement, so that the crack is identified. And then an edge detection algorithm is used to identify the micro crack so as to make up for the defects of a previous point cloud density and curvature detection method in the aspect of micro crack detection, and the accuracy and integrity of crack detection are further improved.

Description

Technical Field

[0001] The present invention relates to the field of tunnel detection technology, and more particularly to a tunnel quality and safety detection system and method based on laser radar data. Background Art

[0002] Tunnels (including highway tunnels, railway tunnels, etc.) as basic transportation facilities play an important role in efficient transportation and fast passage. With the years of tunnel construction and investment, traffic volume, and changes in environmental and geological conditions, the later operation and management of tunnels is the key to ensuring safe passage in tunnels, especially the quality, risk detection and assessment of tunnels, which are the focus of ensuring safety and stability.

[0003] At present, the main methods for tunnel inspection include ultrasonic inspection, infrared inspection, micro-broadcasting inspection, geological radar inspection, laser radar scanning inspection, etc., mainly non-destructive inspection. In particular, high-precision laser radar scanning technology has the advantages of high-resolution imaging, comprehensive data collection, fast and accurate measurement capabilities in tunnel internal structure inspection and evaluation. Based on laser radar scanning data (point cloud data), a three-dimensional model inside the tunnel is constructed to detect defects such as cracks, staggered layers, shedding, water leakage, etc., and to evaluate tunnel quality risks, which is of great significance to improving the safety and stability of the tunnel.

[0004] At present, tunnel disease detection based on laser radar scanning point cloud data is mainly based on point cloud data, and image processing is used to identify and mark tunnel surface diseases, such as lining shedding, cracks, etc. The typical technical path in the prior art is to use the geometric characteristics and spatial distribution information of point cloud data, such as identifying and detecting cracks based on the point cloud density change and curvature change characteristics. This method is more obvious for cracks with large areas or crack clusters (i.e., crack features with obvious geometric features), and can achieve effective crack identification and screening. However, the recognition ability for tiny cracks is insufficient, and the changes in point cloud density and curvature are subtle, making it difficult to directly detect them through the above method. Summary of the invention

[0005] In view of the defects and deficiencies of the prior art, the purpose of the present invention is to provide a detection method that integrates and complements crack detection based on point cloud density changes and crack detection based on curvature changes, so as to achieve rapid identification and accurate detection of crack areas, quickly locate potential crack areas through point cloud density change detection, and further accurately locate crack boundaries in combination with curvature change detection, thereby improving the accuracy and integrity of crack detection. As for small deformations, the changes in point cloud density and curvature are so subtle that they are difficult to detect directly. The present invention further determines the projection plane based on the identified cracks, and proposes to project the point cloud data onto a two-dimensional plane, highlight the features of small cracks in the image through image enhancement, and then use edge detection algorithms to identify small cracks, so as to make up for the defects of the previous point cloud density and curvature detection methods in small crack detection, and further improve the accuracy and integrity of crack detection from a microscopic level.

[0006] According to a first aspect of the present invention, a tunnel quality safety detection method based on laser radar data is proposed, comprising the following steps:

[0007] Step 1: Input the tunnel point cloud data obtained by laser radar scanning;

[0008] Step 2: preprocessing the tunnel point cloud data and performing coordinate changes to obtain point cloud data in the tunnel global coordinate system;

[0009] Step 3: For the point cloud data in the global coordinate system of the tunnel, local point cloud density change detection is used to preliminarily screen crack defects and mark the point cloud data of possible crack areas;

[0010] Step 4: For the point cloud data marked as possible crack areas, accurate detection based on curvature changes is used to accurately determine and mark crack boundary points to obtain the crack area and its boundary;

[0011] Step 5: Select a projection plane according to the position and direction information of the identified crack area, project the point cloud data in the tunnel global coordinate system according to the selected projection plane, and analyze the micro cracks based on the projected two-dimensional image to identify the micro cracks and their edge information;

[0012] Step 6: Visually represent the crack areas and micro-cracks in the three-dimensional tunnel model.

[0013] In a further embodiment, in step 2, preprocessing the tunnel point cloud data includes:

[0014] Step 2.1, based on statistical filtering, remove isolated points that deviate from the spatial distribution of the surrounding point cloud;

[0015] Step 2.2: According to the spatial range and height information of the point cloud, the point cloud of objects outside the tunnel and the point cloud of non-target structures at the top or bottom of the tunnel are removed.

[0016] In a further embodiment, in step 2, the coordinate change refers to converting the point cloud data acquired by the laser radar into a unified tunnel global coordinate system, including:

[0017] Based on the posture and position information of the mobile platform equipped with the laser radar scanning device, the point cloud data in the local coordinate system of the laser radar is converted to the global coordinate system of the tunnel using the homogeneous transformation matrix to obtain the point cloud data in the global coordinate system of the tunnel.

[0018] In a further embodiment, in step 3, for the point cloud data in the global coordinate system of the tunnel, local point cloud density change detection is used to preliminarily screen crack defects and mark the point cloud data of possible crack areas, including:

[0019] Step 3.1, dividing the point cloud data in the tunnel global coordinate system into several sub-areas, so that each point is divided into a sub-area;

[0020] Step 3.2: For any sub-area A, calculate its point cloud density ρ A , represents the point cloud density of sub-area A;

[0021] Step 3.3: Calculate the average density of sub-regions within a certain neighborhood around sub-region A

[0022] Step 3.4: According to the point cloud density ρ of sub-area A A The average density of the sub-regions within a certain neighborhood around it if Then the sub-area A is determined as a possible crack area and marked, where k represents the set proportional coefficient;

[0023] Step 3.5: For each sub-area, perform traversal processing according to the above steps 3.2 to 3.4, preliminarily screen all possible crack areas in the point cloud data, and quickly locate the range where cracks may exist.

[0024] In a further embodiment, the sub-areas are set as evenly divided square areas, the side length of which is l;

[0025] Traverse each point in the point cloud data and determine the square area where it is located based on the coordinates of the point.

[0026] In a further embodiment, in step 3.3, the average density of sub-regions in a certain neighborhood around sub-region A is calculated. include:

[0027]

[0028] Where m represents the number of sub-regions in the neighborhood around sub-region A, ρ i represents the point cloud density of sub-region i in the neighborhood around sub-region A;

[0029] in,

[0030] Where n i represents the number of points in sub-region i, S i Represents the area of ​​sub-region i, and counts the number n by traversing the points in the sub-region i , and calculate the area S based on the geometry of the sub-region i .

[0031] In a further embodiment, in step 4, for the point cloud data marked as a possible crack area, accurate detection based on curvature change is used to accurately determine and mark the crack boundary points to obtain the crack area and its boundary, including:

[0032] Step 4.1, according to the point cloud data in each possible crack area, the local point cloud is fitted into a surface using the least squares method;

[0033] Step 4.2: For the fitted surface, calculate the curvature at each point;

[0034] Step 4.3, according to the preset curvature threshold, traverse each point in the fitting surface area, compare and judge the curvature at each point, mark the point whose curvature is greater than the preset curvature threshold as a crack boundary point, and use the marking data to record the marking status of each point;

[0035] Step 4.4: Based on the labeled data, a neighborhood search and connection algorithm is used to form continuous crack boundaries and construct crack areas.

[0036] In a further embodiment, in step 5, a projection plane is selected according to the position and direction information of the identified crack area, the point cloud data in the tunnel global coordinate system is projected according to the selected projection plane, and based on the projected two-dimensional image, micro-cracks are analyzed to identify micro-cracks and their edge information, including:

[0037] Step 5.1, selecting a projection plane according to the position and direction information of the identified crack area, projecting the point cloud data in the tunnel global coordinate system according to the selected projection plane, and obtaining a projected two-dimensional image; wherein the method for selecting the projection plane is: according to the position and direction of the crack area, selecting a plane perpendicular to the crack development direction as the projection plane;

[0038] Step 5.2, traverse each pixel point in the projected two-dimensional image, count the frequency of each gray level, and obtain a gray level histogram;

[0039] Step 5.3, calculating the cumulative distribution function according to the grayscale histogram, and performing grayscale mapping according to the cumulative distribution function to obtain an enhanced image;

[0040] Step 5.4, based on the enhanced image, calculate the gradient of the image in the horizontal and vertical directions respectively;

[0041] Step 5.5: Traverse each pixel in the enhanced image to perform non-maximum suppression, and compare the gradient amplitude of the pixel with the gradient amplitude of the adjacent pixel in the gradient direction: if the gradient amplitude of the pixel is not the local maximum, mark it as a non-edge point.

[0042] Step 5.6, traverse the image after non-maximum suppression, and determine the non-edge points, strong edge points and weak edge points through high and low double threshold detection;

[0043] Step 5.7: Connect the weak edge points to form a continuous edge, and screen and identify the tiny cracks to obtain the tiny cracks and their edge information.

[0044] In a further embodiment, in step 5.7, the weak edge points are connected to form a continuous edge, and small cracks are screened and identified to obtain small cracks and their edge information, including:

[0045] Step 5.7.1: For continuous edges, obtain the pixel point set marked as the edge to construct the edge point sequence, calculate the distance between adjacent points and accumulate them to obtain the edge length;

[0046] Step 5.7.2, based on the pixel point set, perform local curve fitting and calculate the curvature and curvature distribution of the fitted local curve;

[0047] Step 5.7.3, for adjacent points in the edge point sequence, calculate the adjacent point vectors and direction angles;

[0048] Step 5.7.4, for each point in the edge point sequence, according to the area of ​​predetermined pixel width on both sides of the edge, obtain the grayscale values ​​on both sides of the edge and calculate the grayscale difference and grayscale change rate;

[0049] Step 5.7.5: Screen out tiny cracks based on edge length, curvature of local curve, change of edge direction angle, and grayscale change rate.

[0050] According to a second aspect of the present invention, a computer system is also provided, comprising:

[0051] one or more processors;

[0052] A memory storing operable instructions, wherein when the instructions are executed by the one or more processors, the one or more processors perform operations, wherein the operations include the process of executing the aforementioned tunnel quality safety detection method based on lidar data.

[0053] According to the tunnel quality and safety detection method based on lidar data of the above embodiment of the present invention, based on the tunnel point cloud data, local point cloud density change detection is first used to preliminarily screen crack defects, so as to realize rapid positioning of possible crack areas, and then combined with the point cloud data of possible crack areas, precise detection based on curvature change is adopted to accurately judge and mark crack boundary points; finally, based on the accurately identified and marked crack areas, a projection plane is selected according to their position and direction information, and the point cloud data in the tunnel global coordinate system is projected according to the selected projection plane, and based on the projected two-dimensional image, tiny cracks are analyzed and identified, thereby realizing rapid detection and identification of crack features with macroscopic and relatively obvious geometric feature changes and identification of tiny cracks with subtle changes in point cloud density and curvature, thereby improving crack detection capabilities.

[0054] The tunnel quality safety detection method based on laser radar data according to the embodiment has the following significant advantages:

[0055] On the one hand, the detection based on the change of point cloud density focuses on the proportional relationship between the number of points and the area of ​​the region in the local area. Since cracks usually lead to a decrease in the point cloud density, the area where cracks may exist can be quickly screened out from a macro perspective; the detection based on curvature change focuses on the geometric shape changes of the tunnel wall surface. In the possible crack area initially screened out, based on the curvature change detection, the geometric characteristics of the point cloud are further analyzed, and the surface mutation points are found by calculating the curvature of the surface. These mutation points are usually the boundaries of the cracks. Not only for narrow but deep cracks, the curvature changes will be more significant, even if the point cloud density changes are not so obvious, the curvature change detection can also play a role; thus, the boundary position of the crack is determined more accurately, which plays a role of refinement and precision, and accurately locates the crack; therefore, the two methods of point cloud density and curvature change detection complement each other, the point cloud density change detection quickly locates the potential crack area, and the curvature change detection further accurately locates the boundary of the crack, thereby improving the accuracy and completeness of crack detection;

[0056] On the other hand, for tiny cracks that are still difficult to identify using the first two methods, image enhancement and edge detection methods for tiny cracks are used. In terms of the selection of the projection plane, the optimized projection plane is selected based on the position and direction of the crack area determined by the point cloud density and curvature change detection, thereby improving the efficiency and accuracy of tiny crack detection. The point cloud data is converted into a two-dimensional image for processing based on the selected projection plane, and the information of tiny cracks is deeply mined to ensure that tiny cracks are detected as comprehensively as possible, thereby improving the quality and level of tunnel safety detection and providing an accurate, comprehensive and scientific basis for tunnel maintenance.

[0057] It should be understood that all combinations of the aforementioned concepts and the additional concepts described in more detail below can be considered as part of the inventive subject matter of the present disclosure as long as such concepts are not mutually inconsistent. In addition, all combinations of the claimed subject matter are considered as part of the inventive subject matter of the present disclosure.

[0058] The foregoing and other aspects, embodiments and features of the present invention can be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the present invention, such as the features and / or beneficial effects of the exemplary embodiments, will be apparent from the following description or learned from the practice of the specific embodiments according to the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The accompanying drawings are not intended to be drawn to scale. In the accompanying drawings, each identical or nearly identical component shown in each figure may be represented by the same reference numeral. For clarity, not every component is labeled in each figure. Embodiments of various aspects of the present invention will now be described by way of example and with reference to the accompanying drawings.

[0060] Figure 1 4 is a flow chart of a tunnel quality and safety detection method based on laser radar data according to an embodiment of the present invention.

[0061] Figure 2 This is a flow chart of using local point cloud density change detection to preliminarily screen crack defects according to an embodiment of the present invention.

[0062] Figure 3 This is a flow chart of accurately detecting, accurately judging and marking crack boundary points based on curvature changes according to an embodiment of the present invention.

[0063] Figure 4 This is a flow chart of a method for image enhancement and edge detection of micro cracks according to an embodiment of the present invention.

[0064] Figure 5 This is a flow chart of screening micro cracks according to an embodiment of the present invention. DETAILED DESCRIPTION

[0065] In order to better understand the technical content of the present invention, specific embodiments are given and described as follows in conjunction with the accompanying drawings.

[0066] Various aspects of the present invention are described in this disclosure with reference to the accompanying drawings, in which many illustrative embodiments are shown. The embodiments of the present disclosure are not necessarily intended to include all aspects of the present invention. It should be understood that the various concepts and embodiments introduced above, as well as those described in more detail below, can be implemented in any of many ways, because the concepts and embodiments disclosed by the present invention are not limited to any implementation. In addition, some aspects disclosed by the present invention can be used alone or in any appropriate combination with other aspects disclosed by the present invention.

[0067] {Example 1}

[0068] Combination Figure 1 As shown, the tunnel quality safety detection method based on laser radar data according to an embodiment of the present invention includes the following steps:

[0069] Step 1: Input the tunnel point cloud data obtained by laser radar scanning;

[0070] Step 2: preprocessing the tunnel point cloud data and performing coordinate changes to obtain point cloud data in the tunnel global coordinate system;

[0071] Step 3: For the point cloud data in the global coordinate system of the tunnel, local point cloud density change detection is used to preliminarily screen crack defects and mark the point cloud data of possible crack areas;

[0072] Step 4: For the point cloud data marked as possible crack areas, accurate detection based on curvature changes is used to accurately determine and mark crack boundary points to obtain the crack area and its boundary;

[0073] Step 5: Select a projection plane according to the position and direction information of the identified crack area, project the point cloud data in the tunnel global coordinate system according to the selected projection plane, and analyze the micro cracks based on the projected two-dimensional image to identify the micro cracks and their edge information;

[0074] Step 6: Visualize and characterize the crack areas and micro-cracks in the three-dimensional model of the tunnel.

[0075] It should be understood that in the process of tunnel inspection, cracks, misalignment, shedding, water leakage, etc. on the tunnel surface can be comprehensively evaluated and analyzed based on the laser radar scanning data and combined with the data of other sensors. In the tunnel quality and safety inspection of the embodiment of the present invention, attention is paid to the crack detection on the tunnel surface, such as the method of the above embodiment and the following specific implementation process, through the processing of the laser radar scanning point cloud data for analysis and comprehensive evaluation.

[0076] During the inspection process, the laser radar scanning equipment is carried out on mobile platforms such as track-mounted robots and trolleys to collect comprehensive tunnel data. The mobile platform is controlled to move along the preset track in the tunnel, and the moving speed and position accuracy are accurately controlled to ensure the density of data collection and scanning efficiency (completed within a reasonable time). At the same time, the position accuracy control ensures that the area scanned by the laser radar each time can be accurately spliced, and there will be no data blanks or excessive repeated scanning.

[0077] In an optional embodiment, in step 2, preprocessing the tunnel point cloud data includes:

[0078] Step 2.1, based on statistical filtering, remove isolated points that deviate from the spatial distribution of the surrounding point cloud;

[0079] Step 2.2: According to the spatial range and height information of the point cloud, the point cloud of objects outside the tunnel and the point cloud of non-target structures at the top or bottom of the tunnel are removed.

[0080] Specifically, the collected raw point cloud data will contain noise points caused by factors such as the laser radar's own electronic noise and stray light in the environment, which appear as isolated points that are significantly deviated from the spatial distribution of the surrounding point clouds. The noise points can be removed through the pre-processing denoising algorithm to improve the quality of the point cloud data.

[0081] For example, based on the statistical analysis method, a distance threshold is set to identify and remove isolated points that are significantly deviated from the spatial distribution of the surrounding point cloud as noise points. At the same time, abnormal fluctuation points that appear in multiple scans can also be smoothed by filtering algorithms.

[0082] As an optional example, in step 2.1, isolated points that deviate from the spatial distribution of surrounding point clouds are removed based on statistical filtering, including:

[0083] For each point, calculate the average distance to other points in a certain neighborhood;

[0084] For the set distance threshold, if the average distance from a point to other points in its neighborhood is greater than the threshold, it is determined to be a noise point and removed.

[0085] In order to highlight the surface features of the tunnel, it is necessary to filter out irrelevant background information. For example, the point clouds of objects outside the tunnel (such as passing vehicles, personnel, etc.) and some non-target structure point clouds at the top or bottom of the tunnel can be removed based on the spatial range and height information of the point cloud. In an embodiment of the present invention, the upper and lower limits of the spatial coordinates are set to remove unnecessary background information and highlight the features of the tunnel surface.

[0086] As an optional example, in step 2.2, based on the spatial range and height information of the point cloud, the point cloud of objects outside the tunnel and the point cloud of non-target structures at the top or bottom of the tunnel are removed, including:

[0087] The upper and lower limits of the tunnel space coordinates can be set according to the actual size of the tunnel and the installation location of the LiDAR equipment;

[0088] Traverse the point cloud data, and if the coordinates of a point exceed the set spatial range, remove it.

[0089] In step 2, since the coordinate system of the point cloud data collected by the laser radar on the mobile platform is relative to the local coordinate system of the sensor itself, it needs to be converted to a unified tunnel global coordinate system for subsequent analysis and processing. In an embodiment of the present invention, this is achieved through a coordinate transformation algorithm (such as a homogeneous transformation matrix) through the posture and position information of the mobile platform combined with the scanning data of the laser radar.

[0090] In a further embodiment, the step of converting the point cloud data acquired by the laser radar into a unified tunnel global coordinate system includes:

[0091] Based on the posture and position information of the mobile platform equipped with the laser radar scanning device, the point cloud data in the local coordinate system of the laser radar is converted to the global coordinate system of the tunnel using the homogeneous transformation matrix to obtain the point cloud data in the global coordinate system of the tunnel.

[0092] For example, the attitude and position information of the mobile platform can be obtained by installing high-precision sensors such as inertial measurement units (IMUs) and odometers on the mobile platform.

[0093] For the coordinates of a point P in the local coordinate system of the laser radar, (x l ,y l , z l ), the position of the mobile platform in the tunnel global coordinate system is (x g ,y g , z g ), the attitude matrix is ​​R (which can be obtained by sensors such as IMU); then the coordinates of point P in the tunnel global coordinate system are (x g ,y g , z g )+R(x l,y l , z l ).

[0094] The homogeneous transformation matrix is ​​expressed as:

[0095]

[0096] Where t=(x g ,y g , z g ), then the homogeneous coordinates of point P in the tunnel global coordinate system are:

[0097]

[0098] In a further embodiment, in combination Figure 2 As shown, in step 3, for the point cloud data in the global coordinate system of the tunnel, local point cloud density change detection is used to preliminarily screen crack defects and mark the point cloud data of possible crack areas, including:

[0099] Step 3.1, dividing the point cloud data in the tunnel global coordinate system into several sub-areas, so that each point is divided into a sub-area;

[0100] Step 3.2: For any sub-area A, calculate its point cloud density ρ A , represents the point cloud density of sub-area A;

[0101] Among them, the point cloud density of sub-area A is ρ A The calculation process is as follows:

[0102]

[0103] Among them, S A represents the area of ​​sub-region A, n A Represents the number of points in sub-region A, and counts the number n by traversing the points in the sub-region A , and calculate the area S based on the geometry of the sub-region A ;

[0104] Step 3.3: Calculate the average density of sub-regions within a certain neighborhood around sub-region A

[0105] Step 3.4: According to the point cloud density ρ of sub-area A A The average density of the sub-regions within a certain neighborhood around it if Then the sub-area A is determined as a possible crack area and marked, where k represents the set proportional coefficient;

[0106] Step 3.5: For each sub-area, perform traversal processing according to the above steps 3.2 to 3.4, preliminarily screen all possible crack areas in the point cloud data, and quickly locate the range where cracks may exist.

[0107] Therefore, by calculating the density of the point cloud in the local area and comparing it with the average density of the surrounding area, the possible crack area can be preliminarily and quickly located.

[0108] In a further embodiment, the sub-areas are set to be evenly divided square areas, whose side length is l;

[0109] Traverse each point in the point cloud data and determine the square area where it is located based on the coordinates of the point. For example, for any point Q(x,y,z), based on its coordinate position on the xy plane, the index i in the x direction is The index j in the y direction is in Indicates rounding down.

[0110] Each point is thus divided into a sub-region of a square represented by a specific (i, j) index.

[0111] Furthermore, for a uniformly divided square area, its area S = l 2 , where l is the side length of the subregion.

[0112] The size of the sub-region division can be determined according to the size of the tunnel and the resolution of the lidar, so as to ensure that the point cloud data in each sub-region is representative and the amount of calculation is controllable.

[0113] In a further embodiment, in step 3.3, the average density of sub-regions in a certain neighborhood around sub-region A is calculated. include:

[0114]

[0115] Where m represents the number of sub-regions in the neighborhood around sub-region A, ρ i represents the point cloud density of sub-region i in the neighborhood around sub-region A;

[0116] in,

[0117] Where n i represents the number of points in sub-region i, S i Represents the area of ​​sub-region i, and counts the number n by traversing the points in the sub-region i , and calculate the area S based on the geometry of the sub-region i .

[0118] Therefore, through the above point cloud density change detection, the sub-region whose point cloud density is obviously lower than the surrounding average density is marked as a possible crack region.

[0119] In a further embodiment, in combination Figure 4 As shown, in step 3, for the point cloud data marked as possible crack areas, accurate detection based on curvature changes is used to accurately determine and mark crack boundary points to obtain the crack area and its boundary, including:

[0120] Step 4.1, according to the point cloud data in each possible crack area, the local point cloud is fitted into a surface using the least squares method;

[0121] Step 4.2: For the fitted surface, calculate the curvature at each point;

[0122] Step 4.3, according to the preset curvature threshold, traverse each point in the fitting surface area, compare and judge the curvature at each point, mark the point whose curvature is greater than the preset curvature threshold as a crack boundary point, and use the marking data to record the marking status of each point;

[0123] Step 4.4: Based on the labeled data, a neighborhood search and connection algorithm is used to form continuous crack boundaries and construct crack areas.

[0124] As an optional embodiment, in step 4.1, based on the point cloud data marked as possible crack areas, which contains three-dimensional coordinate information, the least squares method is used for surface fitting. The goal of the least squares method is to find a surface function so that the sum of squared errors between the function and the actual point cloud data is minimized.

[0125] In this embodiment, different types of surfaces can be selected for fitting according to the shape of the tunnel wall and the point cloud data, including quadratic surfaces (such as elliptical paraboloids, hyperbolic paraboloids) or polynomial surfaces. If the tunnel wall is roughly cylindrical, a parametric form of a cylindrical surface equation is used for fitting; if the shape is more complex and there is no prior model, a high-order polynomial surface is used.

[0126] In this example, taking the typical cylindrical wall shape of the tunnel as an example, it is more appropriate to use cylindrical surface fitting for possible crack areas.

[0127] The cylindrical surface equation (for a cylinder with its axis parallel to the axis) can be expressed in rectangular coordinates as:

[0128] (x-x0) 2 +(y-y0) 2 =r 2

[0129] Where (x0, y0) is the coordinate of the center of the cylinder base, and r is the radius. In practical applications, it can be converted into a parametric equation form according to the local area where the possible crack area is located:

[0130] x=x0+rcosθ, y=y0+rsinθ, z=z;

[0131] For each point in the point cloud data of the possible crack area, the least squares method is used to fit the point cloud data to minimize the error between the point cloud data and the fitted cylindrical surface, and the coefficient of the fitted surface is obtained.

[0132] In an optional embodiment, in step 4.2, for the fitted surface, the curvature at each point is calculated, including:

[0133] For the fitted surface equation, calculate its first-order partial derivative and second-order partial derivative;

[0134] Then, according to the curvature calculation formula and the first-order partial derivatives and second-order partial derivatives obtained above, the calculated partial derivatives are substituted into the formula to calculate the curvature value at each point.

[0135] For example, for the fitted surface equation z = f(x, y), the numerical differentiation method is used to calculate its first-order partial derivative f x 、f y With the second-order partial derivative f xx 、f yy 、f xy .

[0136] For the first-order partial derivative f x , can be calculated using the central difference formula: Where h represents a small step value (selected according to the accuracy and density of the point cloud data, such as h = 0.01). Similarly, f can be calculated y 、f xx 、f yy 、f xy .

[0137] Then, according to the curvature formula Substitute the calculated partial derivative into the formula to calculate the curvature value K at each point.

[0138] Then, in step 4.3, based on the curvature value K of each point in the point cloud data of the possible crack area, each point in the fitting surface area is traversed and the points with curvature values ​​greater than the curvature threshold K are selected. t The points are marked as crack boundary points, and a data structure (such as a marking array) can be used to record the marking status of each point.

[0139] The curvature threshold Kt is determined by analyzing sample data of known cracks and normal tunnel wall areas. For example, point cloud data of multiple cracks of different types and crack-free areas are collected, their curvature distribution is calculated, and then a suitable value is selected based on the statistical results.

[0140] Finally, in step 4.4, based on the labeled data, a neighborhood search and connection algorithm is used to form continuous crack boundaries and construct the crack area.

[0141] As an optional embodiment, in step 4.4, a neighborhood search and connection algorithm is used to form a continuous crack boundary, including: starting from a marked crack boundary point, searching for its adjacent marked points, and connecting them according to certain rules (such as the shortest distance or a specific geometric direction), gradually forming a continuous crack boundary and constructing a crack area.

[0142] In a further embodiment, in combination Figure 5 As shown, in step 4, a projection plane is selected according to the position and direction information of the identified crack area, the point cloud data in the tunnel global coordinate system is projected according to the selected projection plane, and based on the projected two-dimensional image, small cracks are analyzed to identify small cracks and their edge information, including:

[0143] Step 5.1, selecting a projection plane according to the position and direction information of the identified crack area, projecting the point cloud data in the tunnel global coordinate system according to the selected projection plane, and obtaining a projected two-dimensional image; wherein the method for selecting the projection plane is: according to the position and direction of the crack area, selecting a plane perpendicular to the crack development direction as the projection plane;

[0144] Step 5.2, traverse each pixel point in the projected two-dimensional image, count the frequency of each gray level, and obtain a gray level histogram;

[0145] Step 5.3, calculating the cumulative distribution function according to the grayscale histogram, and performing grayscale mapping according to the cumulative distribution function to obtain an enhanced image;

[0146] Step 5.4, based on the enhanced image, calculate the gradient of the image in the horizontal and vertical directions respectively;

[0147] Step 5.5: Traverse each pixel in the enhanced image to perform non-maximum suppression, and compare the gradient amplitude of the pixel with the gradient amplitude of the adjacent pixel in the gradient direction: if the gradient amplitude of the pixel is not the local maximum, mark it as a non-edge point.

[0148] Step 5.6, traverse the image after non-maximum suppression, and determine the non-edge points, strong edge points and weak edge points through high and low double threshold detection;

[0149] Step 5.7: Connect the weak edge points to form a continuous edge, and screen and identify the tiny cracks to obtain the tiny cracks and their edge information.

[0150] As an optional embodiment, in step 5.1, the projection plane is selected according to the position and direction information of the crack area obtained in the previous crack detection step. For example, if the crack mainly develops along the longitudinal direction of the tunnel wall, the point cloud data can be projected onto a plane perpendicular to the longitudinal direction; if the crack direction is complex, multiple projection directions can be used, such as projecting onto three planes parallel to the tunnel coordinate axis (i.e., the xy plane, the yz plane, and the xz plane).

[0151] For each point cloud data point, assuming that the projection is to the xy plane, the z coordinate is ignored and the coordinates of the projection point on the xy projection plane are obtained. For projection to other planes, perform the corresponding coordinate selection operation

[0152] As an optional embodiment, in step 5.2, each pixel in the projected two-dimensional image is traversed, and the frequency of occurrence of each gray level is counted to obtain a gray histogram.

[0153] Assume that the grayscale range of the image is [0, L-1] (for example, for an 8-bit image, L=256), create an array H of length L, where H[i] represents the number of pixels with grayscale level i.

[0154] As an optional embodiment, in step 5.3, the cumulative distribution function is calculated according to the grayscale histogram, assuming that the cumulative distribution function is CDF(i), and the calculation formula is:

[0155]

[0156] Where N is the total number of pixels in the two-dimensional image.

[0157] On this basis, grayscale mapping is performed according to the cumulative distribution function to obtain the enhanced image. Specifically, assuming that the grayscale value of the pixel in the original image is p and the enhanced grayscale value is s, then:

[0158] s = (L-1) × CDF(p).

[0159] Therefore, through the grayscale mapping relationship, the grayscale value of each pixel in the two-dimensional image is updated, thereby achieving histogram equalization and enhancing the contrast of the image.

[0160] As an optional embodiment, in step 5.4, based on the enhanced image, the Sobel operator is used to calculate the gradients of the two-dimensional image in the horizontal and vertical directions respectively, including:

[0161] For each pixel (x, y) in a two-dimensional image, the horizontal gradient G x (x, y) is calculated by convolution of the horizontal template of the Sobel operator with the two-dimensional image, and the vertical gradient G y (x, y) is similarly obtained through convolution calculation.

[0162] Then according to the formula Calculate the gradient magnitude, and Compute the gradient direction.

[0163] As an optional embodiment, in step 5.5, each pixel in the enhanced image is traversed to perform non-maximum suppression, and its gradient amplitude is compared with the gradient amplitude of the adjacent pixel in the gradient direction: if the gradient amplitude of the pixel is not a local maximum, it is marked as a non-edge point to suppress its influence in subsequent processing. For example, for a pixel with a gradient direction of 0° (horizontal direction), its gradient amplitude is compared with the gradient amplitude of the adjacent pixel on the left and right; for a pixel with a gradient direction of 45°, its gradient amplitude is compared with the gradient amplitude of the adjacent pixel in the diagonal direction, etc.

[0164] As an optional embodiment, in step 5.6, the image after non-maximum suppression is traversed, and non-edge points, strong edge points and weak edge points are determined by high and low double threshold detection, including:

[0165] Set the high threshold T h and low threshold T l (T h >T l );

[0166] Traverse the image after non-maximum suppression and convert the gradient amplitude greater than T h The pixel points with gradient amplitude less than T are marked as strong edge points; l The pixel points with gradient amplitude between T l and T h If the pixel between is adjacent to a strong edge point, it is marked as a weak edge point.

[0167] Finally, the weak edge points are connected by checking the neighborhood around the weak edge points. If there are strong edge points, the weak edge points are connected to the edge where the strong edge points are located, thus forming a continuous edge.

[0168] Therefore, the distribution and further analysis of geometric features such as edge length, curvature, and direction of continuous edges can be further conducted to distinguish crack edges from other types of edges, and to screen out edges that do not meet the crack characteristics. That is, only when the edge length, curvature, direction angle, and grayscale characteristics all meet the crack characteristics can it be determined as a crack edge, thereby accurately determining the location of tiny cracks.

[0169] In a further embodiment, in combination Figure 5 As shown, in step 5.7, the weak edge points are connected to form a continuous edge, and the small cracks are screened and identified, so as to finally determine the small cracks and their edge information, including:

[0170] Step 5.7.1, for continuous edges, obtain the pixel point set marked as the edge to construct an edge point sequence, calculate the distance between adjacent points and accumulate to obtain the edge length; for example, for the pixel point set of the continuous edge, the points therein are arranged into a point sequence in a certain order (such as along the direction of the edge) as the edge point sequence; further calculate the distance between adjacent points based on the Euclidean distance formula, and then obtain the edge length by adding the distances between all adjacent points;

[0171] Step 5.7.2: Based on the pixel point set, perform local curve fitting and calculate the curvature and curvature distribution of the fitted local curve; specifically, based on the edge point sequence, for each point in it, select a number of adjacent points before and after it (preferably 5-8 points) and use the least squares method to fit the quadratic curve, and solve the curve coefficient by the least squares method to obtain the curve equation, based on which the curvature of each point in the edge point sequence can be calculated; after calculating the curvature of all points on the edge, a curvature sequence is obtained, which is used for subsequent curvature change and distribution judgment;

[0172] Step 5.7.3: For the adjacent points in the edge point sequence, calculate the adjacent point vector and direction angle; specifically, based on the edge point sequence, the adjacent point pair p formed by any two adjacent points i With p i+1 , calculate the adjacent point vector: Then use the inverse tangent function to calculate the direction angle of the vector, then the vector of the adjacent point The direction angle is:

[0173]

[0174] Based on this, the direction angle of each adjacent point pair in the edge point sequence can be calculated to obtain the direction angle sequence, which is used to observe the change of the direction angle sequence in the future and judge whether the direction of the edge has a certain continuity and regularity, so as to distinguish cracks from other messy edges.

[0175] Step 5.7.4: For each point in the edge point sequence, based on the area of ​​predetermined pixel width on both sides of the edge, obtain the grayscale values ​​on both sides of the edge and calculate the grayscale difference and grayscale change rate; specifically, based on the edge point sequence, for each edge point, select an area of ​​a certain pixel width (e.g., 1-3 pixels) perpendicular to the edge direction (the vertical direction can be determined by the edge direction angle), obtain the grayscale values ​​on both sides of the edge, and then calculate the difference ΔG of the grayscale values ​​on both sides of the edge. i Grayscale change rate where d i，i+1 is the distance between adjacent points, such as the Euclidean distance;

[0176] Step 5.7.5: Based on the edge length, the curvature of the local curve, the change in the edge direction angle, and the grayscale change rate, tiny cracks are screened out. That is, when the edge length, curvature, direction angle, and grayscale characteristics all meet the crack characteristics, it is determined to be a crack edge, thereby accurately determining the location of the tiny crack.

[0177] Specifically, in step 5.7.5, the edge length is judged by presetting the maximum and minimum thresholds of the crack length, and all continuous edges are traversed. If the length exceeds the maximum and minimum threshold range, it is deleted from the crack edge candidate set.

[0178] Similarly, by presetting a reasonable range of curvature changes, setting the maximum and minimum thresholds of curvature to judge the curvature changes, traversing all continuous edges, and calculating the curvature value for each point on the edge, if there are more than a certain number of curvatures less than the minimum curvature threshold (indicating that the edge is too smooth) or greater than the maximum curvature threshold (indicating that the edge is too curved and does not meet the crack characteristics), then the edge is excluded from the crack edge candidate set.

[0179] Similarly, by setting the lower threshold of grayscale difference (such as 10, grayscale range is 0-255) and the lower threshold of grayscale change rate (such as 0.1) through the reasonable range of grayscale feature change, it can be determined according to the projected 2D image of point cloud data and the material of tunnel wall. On this basis, for continuous edges, if the mean of their grayscale difference and grayscale change rate is less than the corresponding threshold, they will be removed from the crack edge candidate set.

[0180] For direction angle screening, the variation range of edge direction angle is calculated. The direction of the crack edge usually has a certain continuity. By setting the direction angle variation threshold, for example, 33°, the edge direction angle sequence is checked. If the number of points where the direction angle variation between adjacent points exceeds the direction angle variation threshold exceeds a certain number, the edge is considered to not meet the directional continuity feature of the crack and is deleted from the crack edge candidate set.

[0181] The above-mentioned crack edge feature screening and judgment based on edge length, curvature, direction angle and grayscale characteristics can more accurately determine the true crack edge from the edge detection results, reduce misjudgment and improve the accuracy of tiny crack detection.

[0182] Combined with the implementation of steps 3, 4, and 5 above, the complete process of crack detection is completed from rapid screening of point cloud density to precise boundary determination based on curvature and image detection of tiny cracks. Mutually complementary and dependent, point cloud and crack feature data are gradually analyzed more finely during the processing process, thereby improving the accuracy and comprehensiveness of crack detection.

[0183] {Example 2}

[0184] In combination with the tunnel quality and safety detection method based on laser radar data of the above embodiment, according to the present invention, a computer system is also proposed, including: one or more processors, and a memory.

[0185] The aforementioned memory is used to store operable instructions, which, when executed by the one or more processors, enable the one or more processors to perform operations, including the process of executing the tunnel quality and safety detection method based on lidar data of the aforementioned embodiment.

[0186] Although the present invention has been disclosed as above with preferred embodiments, it is not intended to limit the present invention. A person with ordinary knowledge in the technical field to which the present invention belongs may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be determined by the definition of the claims.

Claims

1. A tunnel quality safety detection method based on laser radar data, characterized in that: The following steps are involved: Step 1: Input the tunnel point cloud data obtained by laser radar scanning; Step 2: preprocessing the tunnel point cloud data and performing coordinate changes to obtain point cloud data in the tunnel global coordinate system; Step 3: For the point cloud data in the global coordinate system of the tunnel, local point cloud density change detection is used to preliminarily screen crack defects and mark the point cloud data of possible crack areas; Step 4: For the point cloud data marked as possible crack areas, accurate detection based on curvature changes is used to accurately determine and mark crack boundary points to obtain the crack area and its boundary; Step 5: Select a projection plane according to the position and direction information of the identified crack area, project the point cloud data in the tunnel global coordinate system according to the selected projection plane, and analyze the micro cracks based on the projected two-dimensional image to identify the micro cracks and their edge information; Step 6: Visually represent the crack areas and micro-cracks in the three-dimensional tunnel model.

2. The tunnel quality and safety detection method based on laser radar data according to claim 1 is characterized in that: In the step 2, the tunnel point cloud data is preprocessed, including: Step 2.1, based on statistical filtering, remove isolated points that deviate from the spatial distribution of the surrounding point cloud; Step 2.2: According to the spatial range and height information of the point cloud, the point cloud of objects outside the tunnel and the point cloud of non-target structures at the top or bottom of the tunnel are removed.

3. The tunnel quality and safety detection method based on laser radar data according to claim 1 is characterized in that: In step 2, the coordinate change refers to converting the point cloud data acquired by the laser radar into a unified tunnel global coordinate system, including: Based on the posture and position information of the mobile platform equipped with the laser radar scanning device, the point cloud data in the local coordinate system of the laser radar is converted to the global coordinate system of the tunnel using the homogeneous transformation matrix to obtain the point cloud data in the global coordinate system of the tunnel.

4. The tunnel quality and safety detection method based on laser radar data according to claim 1 is characterized in that: In step 3, for the point cloud data in the global coordinate system of the tunnel, local point cloud density change detection is used to preliminarily screen crack defects and mark the point cloud data of possible crack areas, including: Step 3.1, dividing the point cloud data in the tunnel global coordinate system into several sub-areas, so that each point is divided into a sub-area; Step 3.2: For any sub-area A, calculate its point cloud density ρ A , represents the point cloud density of sub-area A; Step 3.3: Calculate the average density of sub-regions within a certain neighborhood around sub-region A Step 3.4: According to the point cloud density ρ of sub-area A A The average density of the sub-regions within a certain neighborhood around it if Then the sub-area A is determined as a possible crack area and marked, where k represents the set proportional coefficient; Step 3.5: For each sub-area, perform traversal processing according to the above steps 3.2 to 3.4, preliminarily screen all possible crack areas in the point cloud data, and quickly locate the range where cracks may exist.

5. The tunnel quality and safety detection method based on laser radar data according to claim 4 is characterized in that: The sub-area is set to be a uniformly divided square area, the side length of which is l; Traverse each point in the point cloud data and determine the square area where it is located based on the coordinates of the point.

6. The tunnel quality and safety detection method based on laser radar data according to claim 4 is characterized in that: In step 3.3, the average density of sub-regions in a certain neighborhood around sub-region A is calculated. include: Where m represents the number of sub-regions in the neighborhood around sub-region A, ρ i represents the point cloud density of sub-region i in the neighborhood around sub-region A; in, Where n i represents the number of points in sub-region i, S i Represents the area of ​​sub-region i, and counts the number n by traversing the points in the sub-region i , and calculate the area S based on the geometry of the sub-region i .

7. The tunnel quality and safety detection method based on laser radar data according to claim 1 is characterized in that: In step 4, for the point cloud data marked as possible crack areas, accurate detection based on curvature changes is used to accurately determine and mark crack boundary points to obtain the crack area and its boundary, including: Step 4.1, according to the point cloud data in each possible crack area, the local point cloud is fitted into a surface using the least squares method; Step 4.2: For the fitted surface, calculate the curvature at each point; Step 4.3, according to the preset curvature threshold, traverse each point in the fitting surface area, compare and judge the curvature at each point, mark the point whose curvature is greater than the preset curvature threshold as the crack boundary point, and use the marking data to record the marking status of each point; Step 4.4: Based on the labeled data, a neighborhood search and connection algorithm is used to form continuous crack boundaries and construct crack areas.

8. The tunnel quality and safety detection method based on laser radar data according to claim 1 is characterized in that: In step 5, a projection plane is selected according to the position and direction information of the identified crack area, the point cloud data in the tunnel global coordinate system is projected according to the selected projection plane, and based on the projected two-dimensional image, small cracks are analyzed to identify small cracks and their edge information, including: Step 5.1, selecting a projection plane according to the position and direction information of the identified crack area, projecting the point cloud data in the tunnel global coordinate system according to the selected projection plane, and obtaining a projected two-dimensional image; wherein the method for selecting the projection plane is: according to the position and direction of the crack area, selecting a plane perpendicular to the crack development direction as the projection plane; Step 5.2, traverse each pixel point in the projected two-dimensional image, count the frequency of each gray level, and obtain a gray level histogram; Step 5.3, calculating the cumulative distribution function according to the grayscale histogram, and performing grayscale mapping according to the cumulative distribution function to obtain an enhanced image; Step 5.4, based on the enhanced image, calculate the gradient of the image in the horizontal and vertical directions respectively; Step 5.5: Traverse each pixel in the enhanced image to perform non-maximum suppression, and compare the gradient amplitude of the pixel with the gradient amplitude of the adjacent pixel in the gradient direction: if the gradient amplitude of the pixel is not the local maximum, mark it as a non-edge point. Step 5.6, traverse the image after non-maximum suppression, and determine the non-edge points, strong edge points and weak edge points through high and low double threshold detection; Step 5.7: Connect the weak edge points to form a continuous edge, and screen and identify the tiny cracks to obtain the tiny cracks and their edge information.

9. The tunnel quality and safety detection method based on laser radar data according to claim 8 is characterized in that: In step 5.7, the weak edge points are connected to form a continuous edge, and small cracks are screened and identified to obtain small cracks and their edge information, including: Step 5.7.1: For continuous edges, obtain the pixel point set marked as the edge to construct the edge point sequence, calculate the distance between adjacent points and accumulate them to obtain the edge length; Step 5.7.2, based on the pixel point set, perform local curve fitting and calculate the curvature and curvature distribution of the fitted local curve; Step 5.7.3, for adjacent points in the edge point sequence, calculate the adjacent point vectors and direction angles; Step 5.7.4, for each point in the edge point sequence, according to the area of ​​predetermined pixel width on both sides of the edge, obtain the grayscale values ​​on both sides of the edge and calculate the grayscale difference and grayscale change rate; Step 5.7.5: Screen out tiny cracks based on edge length, curvature of local curve, change of edge direction angle, and grayscale change rate.

10. A computer system, characterized in that: include: one or more processors; A memory storing operable instructions, wherein when the instructions are executed by the one or more processors, the one or more processors are caused to perform operations, wherein the operations include the process of executing the tunnel quality and safety detection method based on lidar data as described in any one of claims 1 to 9.

Images