A method for segmenting and extracting defective areas of corridor concrete point cloud

By using Gaussian fusion of local curvature change and normal curvature information in concrete point cloud data, combined with European clustering and growth area methods, the problems of unversatile data and low accuracy in concrete surface defect identification are solved, and concrete is achieved Accurate segmentation and quantification of defect areas, adapt to different types of defects, and reduce artificial intervention.

CN118887239BActive Publication Date: 2024-12-03THREE GORGES JINSHAJIANG CHUANYUN HYDROPOWER DEV CO LTD
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411336566.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-25
Publication Date
2024-12-03
Estimated Expiration
2044-09-25

AI Technical Summary

Technical Problem

The prior art has problems such as uncommon data and low accuracy in the identification of concrete surface defects, especially the poor identification of concrete surface flat defects, and traditional methods are not applicable to nonlinear data, making it difficult to adjust parameters of RANSAC algorithms.

Method used

By collecting concrete point cloud data, using local curvature changes and normal curvature information for Gaussian fusion, combining European clustering and growth area methods, defect areas are sliced ​​and separated, which is suitable for different types of defect forms.

Benefits of technology

Accurate segmentation and quantification of concrete defect areas is achieved, artificial intervention is reduced, different types of defect forms are adapted, and the accuracy and efficiency of detection are improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118887239B_ABST
    Figure CN118887239B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for segmenting and extracting defective areas of corridor concrete point clouds, which relates to the field of point cloud processing technology, and comprises: collecting point cloud data of concrete and using it as initial data, judging the plane part in the point cloud data by local curvature change, and then using the normal curvature information of each point and the distance information from the point to the fitting plane for Gaussian fusion, and finally obtaining the defective area; then slicing the defective area, using the method combining Euclidean clustering with growth area, and combining the boundary neighbor relationship of complex defect point cloud slices, to separate the slice contours. The present invention can accurately segment and extract concrete defects, which is beneficial to the subsequent volume quantification of concrete defects.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of point cloud processing, and in particular to a method for segmenting and extracting defective areas of corridor concrete point clouds. Background Art

[0002] Concrete plays a vital role in the fields of infrastructure and construction. Over time, various factors can cause problems such as cracks, voids, spalling, and uneven internal structures on the concrete surface. Such concrete structures damaged by aging will affect the performance and life of infrastructure and buildings, and become a potential danger. Therefore, it is crucial to quantitatively analyze the concrete surface structure and take timely repair measures for apparent defects such as cracks, voids, and spalling on the concrete surface to ensure the safety and reliability of concrete structures. Among them, accurate segmentation and extraction of concrete apparent defect areas is a key step in concrete defect monitoring. The increase in the volume of concrete apparent defects will cause the steel bars inside the concrete to be exposed, thereby oxidizing and corroding, resulting in a decrease in the strength and durability of the steel bars, and reducing the overall safety of the concrete structure.

[0003] Defect area extraction is mainly divided into two aspects: traditional methods and deep neural network-based methods.

[0004] Defect area identification based on deep neural networks. In view of the problem that traditional defect detection algorithms are not suitable for defects of standard workpieces of various shapes, there is currently a method for constructing a neighboring graph based on fusion distance, which improves the average prediction accuracy of the model. For the problem of surface defect detection of processed parts, the recognition accuracy of surface defects can be improved by combining unsupervised and supervised machine learning technologies. Regarding point cloud processing methods, it can be used for semantic segmentation of concrete surface defects in infrastructure such as bridges, improving the performance of detecting cracks and defects. At present, effective segmentation of the area to be measured and the area not to be measured can be achieved through point cloud segmentation technology based on deep learning, which facilitates further quantification of the area to be measured. At present, for end-to-end steel surface defect detection, specific categories and precise positioning of defects can be achieved by fusing features at multiple levels. The above-mentioned deep neural network methods have problems in concrete defect recognition, such as data is not universal, and the accuracy of the training results using concrete defect data is not as good as expected.

[0005] In terms of traditional defect area identification and extraction. Regarding the surface defect detection of cabbage based on the curvature features of 3D point clouds, the curvature features can be used to achieve high-precision and high-adaptability defect detection, which is suitable for defect areas with large curvature changes, but not suitable for relatively flat concrete defects. It is also possible to generate the surface point cloud of the pipeline by registration, use voxel sampling and Gaussian filtering to improve data processing efficiency and reduce noise, and use the RANSAC algorithm to remove the surface information of the pipeline, which is conducive to subsequent volume quantification. However, RANSAC has parameters that are difficult to adjust for different data and is not very suitable for nonlinear data such as concrete appearance. Summary of the invention

[0006] In view of this, the present invention provides a method for segmenting and extracting defective areas of corridor concrete point cloud.

[0007] The present invention discloses a method for segmenting and extracting defective areas of corridor concrete point clouds, which comprises:

[0008] The point cloud data of concrete is collected and used as the initial data. The plane part in the point cloud data is judged by the local curvature change. Then, the normal curvature information of each point and the distance information from the point to the fitting plane are Gaussian fused to finally get the defect area. Then, the defect area is sliced, and the slice contour is separated by combining the Euclidean clustering and growth area method with the boundary neighbor relationship of the complex defect point cloud slices.

[0009] Furthermore, the process of obtaining the normal curvature of each point is:

[0010] The local point cloud of concrete is processed, the defect point cloud data is tree-structured, and the normal curvature of the point cloud is calculated.

[0011] Furthermore, the calculating of the normal curvature of the point cloud includes:

[0012] The least squares method is used to calculate the curvature of each point in the point cloud. Suppose any point in the point cloud , the normal vector of the point is , using its point coordinates and normal vector to estimate the point The normal curvature of .

[0013] Furthermore, assuming that Nearby Points, For point No. neighboring points, point cloud midpoint The normal vector is ,set up Orthogonal coordinate system for points The local coordinate system of Indicate point The normal vector of , X and Y are orthogonal unit vectors; In the coordinate system, through a point Osculating circle estimated point The normal curvature .

[0014] Further, , , The coordinates are:

[0015]

[0016]

[0017]

[0018] in, , , They are X-axis coordinate, Y-axis coordinate, N-axis coordinate; , , They are The X-axis coordinate, Y-axis coordinate, and N-axis coordinate.

[0019] Further, the point Relative to point The estimated value of the normal curvature is :

[0020]

[0021]

[0022] in, For vector With vector The angle between For vector With vector The angle between for and The Euclidean distance between is an intermediate variable.

[0023] Furthermore, the determining of the plane part in the point cloud data by the local curvature change includes:

[0024] Select points in the concrete point cloud whose curvature is less than the preset curvature and whose rate of change is less than the preset rate of change, and use RANSAC to fit the non-defective area plane to all selected points. The equation of the fitting plane is: ; Among them, A, B, C, D are fitting coefficients; , , They are X-axis coordinate, Y-axis coordinate, and N-axis coordinate in the coordinate system.

[0025] Furthermore, the Gaussian fusion using the normal curvature information of each point and the distance information from the point to the fitting plane includes:

[0026] For the points in the concrete point cloud that are not involved in the fitting plane, calculate their relative distance to the fitting plane; record the relative distance of each point to the fitting plane , is the point number, For the The three-dimensional coordinates of a point;

[0027]

[0028] Curvature and relative distance weighting: Selecting points Relative distance and point to the fitted plane The curvature value of is used as the weighted term, and the weighted function uses a two-dimensional Gaussian function:

[0029] in, is the weighting function, For point The weighted value of , is the curvature of the point and its distance to the fitting plane; , Separate control , The weighted influence range of .

[0030] Further, obtaining the defective area includes:

[0031] according to The defect area and the non-defect area are determined, the bottom of the defect area is filled, and the defect area is converted into a pseudo-closed area.

[0032] Furthermore, the initial data is subjected to point cloud filtering.

[0033] Due to the adoption of the above technical solution, the present invention has the following advantages:

[0034] 1. It can accurately segment and extract concrete defects, which is conducive to the subsequent volume quantification of concrete defects.

[0035] 2. Greatly reduce human intervention and interference from human factors.

[0036] 3. It can adapt to the segmentation of different types of defects such as deformation, peeling, holes and other defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0038] Figure 1 A schematic diagram of a flow chart of a method for segmenting and extracting defective areas of corridor concrete point cloud according to an embodiment of the present invention;

[0039] Figure 2 This is a schematic diagram of an ideal defect area extraction result according to an embodiment of the present invention;

[0040] FIG3( a ) is a schematic diagram of a point cloud normal curvature solution process according to an embodiment of the present invention;

[0041] FIG3( b ) is a schematic diagram of another point cloud normal curvature solution process according to an embodiment of the present invention;

[0042] FIG4( a ) is a left side view of a defect region extraction result according to an embodiment of the present invention;

[0043] FIG4( b ) is a right side view of the defect area extraction result of the embodiment of the present invention;

[0044] FIG. 4( c ) is a top view of the defect region extraction result according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The present invention is further described in conjunction with the accompanying drawings and embodiments, and the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field should fall within the scope of protection of the embodiments of the present invention.

[0046] See also Figure 1 The present invention provides an embodiment of a method for segmenting and extracting defective areas of corridor concrete point clouds, which comprises:

[0047] The point cloud data of concrete is collected and used as the initial data. The plane part in the point cloud data is determined by the local curvature change. Then, the normal curvature information of each point and the distance information from the point to the fitting plane are used for Gaussian fusion to finally obtain the defect area. Then, the defect area is sliced ​​and the slice contour is separated by combining the method of Euclidean clustering with the growth area and the boundary neighbor relationship of the complex defect point cloud slices. The point cloud data of concrete can be collected by a 3D camera.

[0048] In one embodiment, point cloud filtering is performed on the initial data.

[0049] In one embodiment, tree structured defect point cloud data is used to calculate the normal curvature of the point cloud. The least square method is used to calculate the curvature of each point in the point cloud. In FIG3(a), assume that the point in the point cloud ,point The normal vector is , using its point coordinates and normal vector to estimate the point The normal curvature of . Assume that point Nearby are Points, For point No. neighboring points, The normal vector is ,set up Orthogonal coordinate system for points The local coordinate system, where Indicate point The normal vector of and Orthogonal unit vectors. In this orthogonal coordinate system, , , The coordinates of are shown below, through a point The osculating circle is used to estimate the point The normal curvature , see Figure 3(b). for arrive The vertical projection point of the plane, for and The virtual circle formed by the two points is The tangent plane at the point.

[0050]

[0051]

[0052]

[0053] in, , , They are X-axis coordinate, Y-axis coordinate, N-axis coordinate; , , They are The X-axis coordinate, Y-axis coordinate, and N-axis coordinate.

[0054] Points Relative to point The estimated value of the normal curvature is :

[0055]

[0056] in, For vector With vector The angle between For vector With vector The angle between for and The Euclidean distance between is an intermediate variable.

[0057] Select points in the concrete point cloud whose curvature is less than the preset curvature and whose rate of change is less than the preset rate of change, and use RANSAC (random sampling consensus algorithm) to fit the non-defective area plane to the selected points. The equation is ; Among them, A, B, C, D are fitting coefficients; , , They are X-axis coordinate, Y-axis coordinate, and N-axis coordinate in the coordinate system.

[0058] Relative distance calculation: For points that are not involved in the fitted plane, calculate their relative distance to the fitted plane. Record the relative distance of each point to the fitted plane. , is the point number, For the The three-dimensional coordinates of a point.

[0059]

[0060] The weighting of curvature and relative distance. The characteristics of the defect area point are: large curvature change, large and small curvature values, and far and near distances from the fitting plane. Select point Relative distance and point to the fitted plane The curvature value of is used as the weighting term. The characteristic is that the greater the relative distance and the greater the curvature value, the greater the probability of being judged as a defect area. Therefore, the weighting function uses a two-dimensional Gaussian function:

[0061]

[0062] in, is the weighting function, For point The weighted value of , is the curvature of the point and its distance to the fitting plane; , Separate control , The weighted influence range of .

[0063] Larger parameter values ​​indicate and The change of has a greater impact on the weighting range, and vice versa. According to multiple experimental verifications, setting 0.25, is 0.75, the ideal region extraction result can be obtained as follows Figure 2 , the red part in the figure is the extracted defect area, and the green part is the non-defect area. The bottom surface of the defect area is filled to convert the defect area into a pseudo-closed area. Figure 4 (a) is the left view of the defect area extraction result, Figure 4 (b) is the right view of the defect area extraction result, and Figure 4 (c) is the top view of the defect area extraction result.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for segmenting and extracting defective areas of corridor concrete point cloud, characterized in that: include: The point cloud data of concrete is collected and used as the initial data. The plane part in the point cloud data is determined by the local curvature change. Then, the normal curvature information of each point and the distance information from the point to the fitting plane are used for Gaussian fusion to finally obtain the defect area. Then, the defect area is sliced, and the slice contour is separated by combining the method of Euclidean clustering with the growth area and the boundary neighbor relationship of the complex defect point cloud slices. The process of obtaining the normal curvature of each point is: Process the local point cloud of concrete, structure the defect point cloud data, and calculate the normal curvature of the point cloud; The calculating of the normal curvature of the point cloud comprises: The least squares method is used to calculate the curvature of each point in the point cloud. Suppose any point in the point cloud ,point The normal vector is , using its point coordinates and normal vector to estimate the point The normal curvature of Assumptions Nearby Points, For point No. neighboring points, point cloud midpoint The normal vector is ,set up Orthogonal coordinate system for points The local coordinate system of Indicate point The normal vector of , X and Y are orthogonal unit vectors; In the coordinate system, through a point Osculating circle estimated point The normal curvature ; The method of determining the plane part in the point cloud data by the local curvature change includes: Select points in the concrete point cloud whose curvature is less than the preset curvature and whose rate of change is less than the preset rate of change, and use RANSAC to fit the non-defective area plane to all selected points. The equation of the fitting plane is: ; Among them, A, B, C, D are fitting coefficients; , , They are X-axis coordinate, Y-axis coordinate, and N-axis coordinate in the coordinate system; The Gaussian fusion using the normal curvature information of each point and the distance information from the point to the fitting plane includes: For the points in the concrete point cloud that are not involved in the fitting plane, calculate their relative distance to the fitting plane; record the relative distance of each point to the fitting plane , is the point number, For the The three-dimensional coordinates of a point; Curvature and relative distance weighting: Selecting points Relative distance and point to the fitted plane The curvature value of is used as the weighted term, and the weighted function uses a two-dimensional Gaussian function: in, is the weighting function, For point The weighted value of , is the curvature of the point and its distance to the fitting plane; , Separate control , The weighted influence range of .

2. The method according to claim 1, characterized in that: , , The coordinates are: in, , , They are X-axis coordinate, Y-axis coordinate, N-axis coordinate; , , They are The X-axis coordinate, Y-axis coordinate, and N-axis coordinate.

3. The method according to claim 2, characterized in that Points Relative to point The estimated value of the normal curvature is : in, For vector With vector The angle between For vector With vector The angle between for and The Euclidean distance between is an intermediate variable.

4. The method according to claim 1, characterized in that: The step of obtaining the defective area comprises: according to The defect area and the non-defect area are determined, the bottom of the defect area is filled, and the defect area is converted into a pseudo-closed area.

5. The method according to claim 1, characterized in that Perform point cloud filtering on the initial data.

Citation Information

Patent Citations

  • Method for detecting underwater surface crack of hydraulic concrete structure by using underwater robot system

    CN103364408A

  • A bridge crack detection method based on tensor voting

    CN109886939A

  • Device and method for collecting dust generated by friction

    CN116441272A

  • Dam displacement missing data completion method and system based on clustering and deep learning

    CN117951124A

  • Apparatus and methods for detecting optical components and their misalignment in optical coherence tomographic systems

    US20150085294A1