A method for rapid detection and evaluation of fastener service status

Through a rapid detection method based on point cloud data, the problem of single detection dimension and low automation level in track fastener detection is solved, and the service status description and prevention functions of fastener are realized in all-round and multi-angle fastener, which improves the accuracy of detection and efficiency replace manual inspections and are highly economical and popularizable.

CN118761994BActive Publication Date: 2024-11-26CHINA RAILWAY DESIGN GRP CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411223115.7
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-03
Publication Date
2024-11-26
Estimated Expiration
2044-09-03

AI Technical Summary

Technical Problem

In the detection of track fasteners, the prior art has problems such as single detection dimensions, weak description ability, lack of prevention assistance, cumbersome detection process, low time efficiency, and low automation level, which makes it easy to introduce errors.

Method used

Using a fast detection method based on point cloud data, through noise filtering, inter-frame registration and noise ablation, segmentation and registration solution, the geometric parameters and pose parameters of the fastener are extracted, and the service status of the fastener is judged in combination with the logic tree. Real-time updates using adversarial learning models and reinforced weight transfers.

Benefits of technology

It realizes a comprehensive and multi-angle fastener service status description, improves the accuracy and automation of detection, improves time efficiency, can prevent accidents in advance, replaces manual inspections, is economical and popularizable.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118761994B_ABST
    Figure CN118761994B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for rapid detection and evaluation of the service status of fasteners, comprising the following steps: S1, raw data collection; S2, generating a digital point cloud of the fasteners to be detected; S3, digital point cloud status evaluation. The present invention relies on the dimensional features extracted from the point cloud to set relevant criteria, perform data registration and modification, and eliminate jitter and burrs affected by noise, thereby further improving the descriptiveness and accuracy of the point cloud; the present invention combines the key geometric parameters extracted from the point cloud and the state matrix parameters solved by the nonlinear registration algorithm, and uses a logic tree to evaluate the service status of the fasteners. The detection dimension is comprehensive, the description ability is strong, and accidents can be prevented in advance; compared with traditional detection methods, the present invention has the advantages of high detection efficiency, large situation evaluation coverage, and comprehensive service status awareness.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of rapid comprehensive detection of rail transit, and in particular relates to a method for rapid detection and evaluation of the service status of fasteners. Background Art

[0002] The supporting fasteners of the track are an important part of the entire rail transit project, and their status will directly affect the operational safety of the train. After long-term operation and maintenance of rail transit lines, the service status of the fasteners will change, thus affecting the geographical location and smoothness of the line. Fastener anomalies include but are not limited to local fastener fractures, shear failure, spring bar failure, posture errors, relative position distortion, etc. With the development of high-speed railways and urban rail transit, the demand for detecting abnormal fastener status is increasing, and it is urgent to find a comprehensive and multi-dimensional detection method for fasteners in service.

[0003] In addition to long-standing manual inspection, in recent years, with the advancement of digital imaging technology and the improvement of hardware level, many researchers at home and abroad have begun to try to use data sources such as images, videos and point clouds to automatically inspect rail fasteners. According to the attributes of the source data, rail fastener inspection methods can be divided into three categories:

[0004] (1) Rail fastener detection method based on two-dimensional data source;

[0005] (2) Track fastener detection method based on three-dimensional data source;

[0006] (3) Vibration-based rail fastener detection method.

[0007] The above method has the following problems:

[0008] a) Fastener detection has a single dimension and weak description ability;

[0009] b) It is limited to the detection of defects to a greater extent and has no effective prevention assistance capability;

[0010] c) The testing process is repetitive and cumbersome, which greatly increases the workload and reduces time efficiency;

[0011] d) The whole process has low digitization and automation levels, which can easily introduce gross errors.

[0012] In response to the above-mentioned problems and weaknesses, there is an urgent need to develop a multi-angle detection and evaluation technology for fasteners, which can efficiently, accurately and comprehensively describe the service status and life range of the fasteners, and provide reasonable suggestions for subsequent processing methods during operation. Summary of the invention

[0013] To solve the problems existing in the prior art, the present invention provides an accurate, rapid and comprehensive fastener detection method, which can describe the service status and life span of the fasteners in an all-round and multi-angle manner, and provide reasonable suggestions for subsequent processing methods during operation.

[0014] To this end, the present invention adopts the following technical solutions:

[0015] A method for quickly detecting and judging the service status of a fastener comprises the following steps:

[0016] S1, raw data acquisition: collecting point cloud data of the fastener to be inspected and the supporting area, wherein the point cloud data is frame data divided according to timestamps;

[0017] S2, generating a digitized point cloud of the fastener to be inspected, including the following steps:

[0018] S21, noise filtering: remove noise and reduce errors of the point cloud data collected by S1;

[0019] S22, inter-frame registration and noise ablation: extracting feature dimension data from the data processed in S21, the feature dimension data including affine transformation length, Euclidean curvature and data reprojection cost, performing point cloud fine registration on the feature dimension data, and performing noise ablation on the result of the point cloud fine registration to obtain high-quality point cloud data of the fastener to be inspected;

[0020] S3, digitized point cloud status evaluation, includes the following steps:

[0021] S31, point cloud data segmentation:

[0022] Pre-segmenting the fastener into different partitions, wherein the partitions include a spring bar region and a central bolt region; using the partitions to segment the high-quality point cloud data obtained in S22 to obtain a fastener sub-point cloud set consisting of a plurality of mutually uncoupled sub-point clouds;

[0023] S32, point cloud registration solution:

[0024] The sub-point clouds in the fastener sub-point cloud set are respectively registered and solved to obtain the posture parameters of each fastener sub-point cloud, wherein the posture parameters include a PSNR value (Peak signal-to-noise ratio), the number of registration point pairs, and a registration posture conversion matrix;

[0025] S33, using the sub-point cloud in the fastener sub-point cloud set to solve and obtain geometric parameters of each fastener, wherein the geometric parameters include the spring clip envelope space, the fixing bolt center point, the fastener gasket thickness, the spring clip inclination angle, and the spring clip gap value;

[0026] S34, judging the service status of the center bolt and the spring bar on the fastener to be inspected by using the fastener service threshold, the pose parameters of each fastener sub-point cloud obtained in S32, and the geometric parameters of each fastener obtained in S33.

[0027] In the above step S22, the affine transformation length is calculated by the following steps:

[0028] First, in the frame data processed by S21, Get any two feature points in the frame and ,

[0029] Then, build with and The abscissa equation of the independent variable ; Constructed with and The ordinate equation of the independent variable ; and Have the same functional structure; is the total number of frames of the point cloud data;

[0030] The feature points and The affine length between for:

[0031] ;

[0032] in, for The first derivative of for The second-order derivative of for The first derivative of for The second derivative of and is the number of the feature point;

[0033] Finally, calculate the affine transformation length :

[0034] ;

[0035] in: For along The scaling factor of the axis, For along The scaling factor of the axis, For along The shear factor of the axis, For along The scaling factor and shear factor are set according to the actual situation and are both between 0 and 1.

[0036] The Euclidean curvature in step S22 is calculated by the following steps:

[0037] Calculate the Frame affine mapping Euclidean curvature of the affine length parameterized curve in the frame data corresponding to the frame , ,have:

[0038] ;

[0039] in:

[0040] ;

[0041] ;

[0042] ;

[0043] .

[0044] The data reprojection cost in step S22 is calculated by the following steps:

[0045] First, for the Frame data corresponding to the frame , and the adjacent Frame and Frame data corresponding to the frame and , calculate the light source projection consistency ,have:

[0046] ;

[0047] in, and Frame data and The two corresponding feature points in express and The Euclidean distance of is the correction factor, Corresponding to Frame, Frame and Frame, correction factor is determined according to the resolution of the acquisition device and the actual working parameters;

[0048] Then take the Any point on the frame , calculate the surface consistency between adjacent frame data ,have:

[0049] ;

[0050] ;

[0051] ;

[0052] in, is the weight of the acquisition device in the x-axis direction; is the weight of the acquisition device in the y-axis direction; The coordinates are The depth value of the feature point; The coordinates are The depth value of the feature point; The coordinates are The depth value of the feature point; The coordinates are The depth value of the feature point; The coordinates are The depth value of the feature point; , , , Indicate point The coordinates of the adjacent points of ;

[0053] Then, combining the above light source projection consistency and surface consistency, calculate the Overall consistency of frames ,have:

[0054] ;

[0055] ;

[0056] in, It is the optimization threshold of the adaptive smoothing index, and its value is determined according to the required noise ablation degree, and its value range is (0, 1); For the The correction factor fitting result of the frame light source projection consistency;

[0057] Finally, calculate the Frame projection mapping Data reprojection cost corresponding to frame data , ,have:

[0058] ;

[0059] in, It is frame data The matching cost corresponding to the depth map; Frame data The matching cost corresponding to the depth map; Frame data The matching cost corresponding to the depth map; For the Comprehensive consistency of frames; For the Comprehensive consistency of frames; For the Comprehensive consistency of frames; are constants, and their value range is .

[0060] In step S31, the fastener is pre-divided into 6 partitions A, B, C, D, E, and F; wherein A, C, D, and F are spring bar areas, B is a locking spring bar area, and E is a center bolt area.

[0061] Step S34 includes the following steps:

[0062] S341, judging the sub-point cloud of the central bolt area, when the registration posture conversion matrix and the center point of the fixing bolt both meet the service threshold, proceeding to S342; otherwise, proceeding to S345;

[0063] S342, judging the sub-point cloud of the elastic strip area, when the PSNR value and the number of the registration point pairs both meet the service threshold, proceeding to S343; otherwise, proceeding to S346;

[0064] S343, for the D, B, and F regions in the spring bar region, when the spring bar envelope space, the spring bar inclination angle, and the spring bar gap value all meet the service threshold, proceed to S344; otherwise, proceed to S347;

[0065] S344, for the D and F regions in the spring bar region, when the registration posture conversion matrix and the thickness of the fastener gasket both meet the service threshold, output "the state of the fastener is normal"; otherwise, output "the spring bar of the fastener is installed reversely";

[0066] S345, when the registration posture conversion matrix does not meet the service threshold, output "the center bolt of the fastener is loose"; when the center point of the fixing bolt does not meet the service threshold, output "the center bolt of the fastener is missing";

[0067] S346, when the number of coordinate points of the sub-point cloud filtered by S342 is greater than the service threshold, output that the spring clip of the fastener is partially broken; otherwise, output “the spring clip of the fastener is lost”;

[0068] S347, when the spring clip envelope space and the spring clip inclination angle of the sub-point cloud filtered by S343 do not meet the service threshold, the "spring clip plane of the fastener is skewed" is output; when the spring clip gap value does not meet the service threshold, the "spring clip elevation of the fastener is skewed" is output.

[0069] Preferably, in S1, point cloud data is collected by a self-mobile rail transit mobile three-dimensional scanning system.

[0070] Preferably, in S21 , a Gaussian filtering algorithm is used to perform noise filtering.

[0071] Compared with the prior art, the present invention has the following beneficial effects:

[0072] 1. The present invention relies on the dimensional features extracted from the point cloud to set relevant criteria, perform data registration and modification, remove jitter and burrs affected by noise, and further improve the descriptiveness and accuracy of the point cloud;

[0073] 2. The present invention combines the key geometric parameters extracted from the point cloud and the state matrix parameters solved by the nonlinear registration algorithm to judge the service status of the fastener with a logic tree. The detection dimension is comprehensive, the description ability is strong, and accidents can be prevented in advance.

[0074] 3. The detection process of the present invention is highly digitalized and automated, and the time efficiency and accuracy are greatly improved. It can replace manual inspections and has a high degree of economy and applicability;

[0075] 4. The detection results of the present invention integrate the adversarial learning model and enhanced weight transfer, assign corresponding weights to each parameter in the matrix describing the service status and update it in real time, so as to achieve the effect of adaptive matching and improve the ability to describe the service status, and have a high degree of confidence in the assistance and guidance suggestions for the operation and maintenance process. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 is a flow chart of the present invention;

[0077] Figure 2 This is a schematic diagram of the point cloud obtained during the filtering process;

[0078] Figure 3 This is a top view of the fastener point cloud, showing the six pre-segmented partitions of the fastener upper surface point cloud. DETAILED DESCRIPTION

[0079] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings and embodiments.

[0080] like Figure 1 As shown, the method for quickly detecting and judging the service status of fasteners of the present invention comprises the following steps:

[0081] S1, raw data acquisition: using a self-mobile rail transit mobile 3D scanning system, or using other acquisition systems to collect and obtain point cloud data in the fastener to be inspected and the supporting area, the point cloud data is frame data divided according to timestamps;

[0082] S2, generating a digitized point cloud of the fastener to be inspected, including the following steps:

[0083] S21, noise filtering: Use Gaussian filtering algorithm to remove noise and reduce errors from the point cloud data collected in S1. In one embodiment of the present invention, the filtering result is as follows: Figure 2 As shown in the figure, the darker ones are outliers and noise points, which are point clouds to be removed; the lighter ones are regular point cloud data that need to be retained.

[0084] S22, inter-frame registration and noise ablation: extract feature dimension data from the data processed by S21, the feature dimension data includes affine transformation length, Euclidean curvature and data reprojection cost, input the above feature dimension data into the input of the conventional registration algorithm, and perform noise ablation on the obtained result, and finally obtain high-quality point cloud data of the fastener to be detected, which specifically includes the following steps:

[0085] S221, in the frame data Get two feature points at random in the frame and , constructed with and The abscissa equation of the independent variable ; Constructed with and The ordinate equation of the independent variable ; and Have the same functional structure; is the total number of frames of point cloud data.

[0086] Feature Points and The affine length between for:

[0087] ;

[0088] in, for The first derivative of , for The first derivative of ; and is the number of the feature point;

[0089] Calculate the affine transformation length :

[0090] ;

[0091] in: For along The scaling factor of the axis, For along The scaling factor of the axis, For along The shear factor of the axis, For along The shear factor of the axis. Each factor should be determined based on the actual mapping between the data and the real target, and all are between 0 and 1.

[0092] S222, calculate the Frame affine mapping Frame data corresponding to the frame Euclidean curvature of affine length parameterized curves , ,have:

[0093] ;

[0094] in:

[0095] ;

[0096] ;

[0097] ;

[0098] .

[0099] S223, calculate the data reprojection cost:

[0100] First, for the Frame data corresponding to the frame , and the two adjacent frames Frame and Frame data corresponding to the frame and , calculate the light source projection consistency ,have:

[0101] ;

[0102] in, and Frame data and The two corresponding feature points in express and The Euclidean distance of is the correction factor, Corresponding to Frame, Frame and The correction factor is determined according to the resolution of the acquisition device and the actual working parameters.

[0103] Then take the Any point on the frame , calculate the surface consistency between adjacent frame data ,have:

[0104] ;

[0105] ;

[0106] ;

[0107] in, is the weight of the acquisition device in the x-axis direction; is the weight of the acquisition device in the y-axis direction; The coordinates are The depth value of the feature point; The coordinates are The depth value of the feature point; The coordinates are The depth value of the feature point; The coordinates are The depth value of the feature point; The coordinates are The depth value of the feature point; , , , Indicate point The coordinates of the neighboring points.

[0108] Then, combining the above light source projection consistency and surface consistency, calculate the Overall consistency of frames ,have:

[0109] ;

[0110] ;

[0111] in, It is the optimization threshold of the adaptive smoothing index, and its value is determined according to the required noise ablation degree, and its value range is (0, 1); For the The correction factor fitting result of the frame light source projection consistency;

[0112] Finally, calculate the Frame projection mapping Data reprojection cost corresponding to frame data , ,have:

[0113] ;

[0114] in, It is frame data The matching cost corresponding to the depth map; Frame data The matching cost corresponding to the depth map; Frame data The matching cost corresponding to the depth map; For the The overall consistency of the frames; For the The overall consistency of the frames; For the The overall consistency of the frames; are constants, and their value range is .

[0115] S224, inter-frame registration and noise ablation: use the methods of step S221, step S222 and step S223 to calculate the affine transformation length, Euclidean curvature and data reprojection cost of all frame data, and then use them as features to perform point cloud fine registration (preferably ICP algorithm) on the data processed by S21; and perform noise ablation on the result of point cloud fine registration to obtain high-quality point cloud data of the fastener to be inspected;

[0116] S3, digitized point cloud status evaluation, includes the following steps:

[0117] S31, point cloud data segmentation: pre-segment the fastener into 6 partitions A to F, A, C, D and F are spring bar areas, B is the locking spring bar area, and E is the center bolt area; use the A to F partitions to segment the high-quality point cloud obtained in S23 to obtain a fastener sub-point cloud set consisting of several uncoupled sub-point clouds. The fastener sub-point cloud set includes the spring bar area sub-point cloud set and the center bolt area sub-point cloud set. The division of the A to F partitions of a single fastener is as follows: Figure 3 shown.

[0118] S32, point cloud registration solution: perform registration solutions on the sub-point clouds in the fastener sub-point cloud set respectively to obtain the pose parameters of each fastener sub-point cloud, wherein the pose parameters include a PSNR value, a number of registration point pairs, and a registration pose conversion matrix;

[0119] S33, using the sub-point cloud in the fastener sub-point cloud set to solve the spring clip envelope space and the center point of the fixing bolt of the fastener, and then solving the fastener gasket thickness, spring clip inclination angle and spring clip gap value by Euclidean distance;

[0120] S34, performing a fastener service evaluation, including the following steps:

[0121] S341, judging the sub-point cloud of the central bolt area, and when the registration attitude conversion matrix and the center point of the fixed bolt both meet the service threshold, proceed to S342, otherwise proceed to S345;

[0122] S342, judging the sub-point cloud of the elastic strip area, when the PSNR value and the number of registration point pairs both meet the service threshold, proceed to S343, otherwise proceed to S346;

[0123] S343, for the D, B, and F regions, when the spring clip envelope space, the spring clip inclination angle, and the spring clip gap value all meet the service threshold, enter S344, otherwise enter S347;

[0124] S344, for regions D and F, when the registration attitude conversion matrix and the thickness of the fastener gasket both meet the service threshold, output "the fastener is in normal condition", otherwise output "the spring clip of the fastener is installed reversely";

[0125] S345, when the registration posture conversion matrix does not meet the service threshold, output "the center bolt of the fastener is loose"; when the center point of the fixing bolt does not meet the service threshold, output "the center bolt of the fastener is missing";

[0126] S346, when the number of coordinate points of the sub-point cloud is greater than the service threshold, output "the spring clip of the fastener is partially broken", otherwise output "the spring clip of the fastener is lost";

[0127] S347, when the spring clip envelope space and the spring clip inclination angle of the sub-point cloud do not meet the service threshold, output "the spring clip plane of the fastener is skewed", and when the spring clip gap value does not meet the service threshold, output "the spring clip elevation of the fastener is skewed". The service threshold is a standard threshold that can maintain the normal use of the fastener.

[0128] The present invention can also combine the probability matrix to predict the life span of different fasteners, providing data support and reasonable suggestions for subsequent operation and maintenance processing methods. In subsequent practical applications, relying on adversarial learning models and enhanced weight migration, each parameter in the matrix describing the service status is given a corresponding weight and updated in real time, achieving the effect of adaptive matching and improving the ability to describe the service status.

Claims

1. A method for quickly detecting and judging the service status of a fastener, characterized in that: The following steps are involved: S1, raw data acquisition: collecting point cloud data of the fastener to be inspected and the supporting area, wherein the point cloud data is frame data divided according to timestamps; S2, generating a digitized point cloud of the fastener to be inspected, including the following steps: S21, noise filtering: remove noise and reduce errors of the point cloud data collected by S1; S22, inter-frame registration and noise ablation: extracting feature dimension data from the data processed in S21, the feature dimension data including affine transformation length, Euclidean curvature and data reprojection cost, performing point cloud fine registration on the feature dimension data, and performing noise ablation on the result of the point cloud fine registration to obtain high-quality point cloud data of the fastener to be inspected; S3, digitized point cloud status evaluation, includes the following steps: S31, point cloud data segmentation: The fastener is pre-divided into 6 partitions A, B, C, D, E, and F; wherein A, C, D, and F are spring clip areas, B is a locking spring clip area, and E is a center bolt area; the high-quality point cloud data obtained by S22 is segmented using the partitions to obtain a fastener sub-point cloud set composed of a number of mutually uncoupled sub-point clouds; S32, point cloud registration solution: The sub-point clouds in the fastener sub-point cloud set are respectively registered and solved to obtain the pose parameters of each fastener sub-point cloud, wherein the pose parameters include a PSNR value, a number of registration point pairs, and a registration pose conversion matrix; S33, using the sub-point cloud in the fastener sub-point cloud set to solve and obtain geometric parameters of each fastener, wherein the geometric parameters include the spring clip envelope space, the fixing bolt center point, the fastener gasket thickness, the spring clip inclination angle, and the spring clip gap value; S34, judging the service status of the center bolt and the spring clip on the fastener to be detected by using the fastener service threshold, the pose parameters of each fastener sub-point cloud obtained in S32, and the geometric parameters of each fastener obtained in S33, specifically comprising the following steps: S341, judging the sub-point cloud of the central bolt area, when the registration posture conversion matrix and the center point of the fixing bolt both meet the service threshold, proceeding to S342; otherwise, proceeding to S345; S342, judging the sub-point cloud of the elastic strip area, when the PSNR value and the number of the registration point pairs both meet the service threshold, proceeding to S343; otherwise, proceeding to S346; S343, for the D, B, and F regions in the spring bar region, when the spring bar envelope space, the spring bar inclination angle, and the spring bar gap value all meet the service threshold, proceed to S344; otherwise, proceed to S347; S344, for the D and F regions in the spring bar region, when the registration posture conversion matrix and the thickness of the fastener gasket both meet the service threshold, output "the state of the fastener is normal"; otherwise, output "the spring bar of the fastener is installed reversely"; S345, when the registration posture conversion matrix does not meet the service threshold, output "the center bolt of the fastener is loose"; when the center point of the fixing bolt does not meet the service threshold, output "the center bolt of the fastener is missing"; S346, when the number of coordinate points of the sub-point cloud filtered by S342 is greater than the service threshold, output that the spring clip of the fastener is partially broken; otherwise, output "the spring clip of the fastener is lost"; S347: When the spring clip envelope space and spring clip inclination angle of the sub-point cloud filtered by S343 do not meet the service threshold, the "spring clip plane of the fastener is skewed" is output; when the spring clip gap value does not meet the service threshold, the "spring clip elevation of the fastener is skewed" is output.

2. The method for rapid detection and evaluation of the service status of fasteners according to claim 1, characterized in that: The affine transformation length in step S22 is calculated by the following steps: First, in the frame data processed by S21, Get any two feature points in the frame and , Then, build with and The abscissa equation of the independent variable ; Constructed with and The ordinate equation of the independent variable ; and Have the same functional structure; , is the total number of frames of the point cloud data; The feature points and The affine length between for: ; in, for The first derivative of for The second-order derivative of for The first derivative of for The second derivative of and is the number of the feature point; Finally, calculate the affine transformation length : ; in: For along The scaling factor of the axis, For along The scaling factor of the axis, For along The shear factor of the axis, For along The scaling factor and shear factor are set according to the actual situation and are both between 0 and 1.

3. The method for rapid detection and evaluation of the service status of fasteners according to claim 2, characterized in that: The Euclidean curvature in step S22 is calculated by the following steps: Calculate the The first frame affine mapping Euclidean curvature of an affine length parameterized curve in the frame data of the frame , ,have: ; in: ; ; ; 。 4. The method for rapid detection and evaluation of the service status of fasteners according to claim 3, characterized in that: The data reprojection cost in step S22 is calculated by the following steps: First, for the Frame data corresponding to the frame , and the adjacent Frame and Frame data corresponding to the frame and , calculate the light source projection consistency ,have: ; in, and Frame data and The two corresponding feature points in express and The Euclidean distance of is the correction factor, Corresponding to Frame, Frame and Frame, correction factor is determined according to the resolution of the acquisition device and the actual working parameters; Then take the Any point on the frame , calculate the surface consistency between adjacent frame data ,have: ; ; ; in, is the weight of the acquisition device in the x-axis direction; is the weight of the acquisition device in the y-axis direction; The coordinates are The depth value of the feature point; The coordinates are The depth value of the feature point; The coordinates are The depth value of the feature point; The coordinates are The depth value of the feature point; The coordinates are The depth value of the feature point; , , , Indicate point The coordinates of the adjacent points of ; Then, combining the above light source projection consistency and surface consistency, the comprehensive consistency of the frame is calculated ,have: ; ; in, It is the optimization threshold of the adaptive smoothing index, and its value is determined according to the required noise ablation degree, and its value range is (0, 1); For the The correction factor fitting result of the frame light source projection consistency; Finally, calculate the Frame projection mapping Data reprojection cost of frame data , ,have: ; in, It is frame data The matching cost corresponding to the depth map; Frame data The matching cost corresponding to the depth map; Frame data The matching cost corresponding to the depth map; For the The overall consistency of the frames; For the The overall consistency of the frames; For the The overall consistency of the frames; are constants, and their value range is .

5. The method for rapid detection and evaluation of the service status of fasteners according to claim 1, characterized in that: In S1, point cloud data is collected by a self-mobile rail transit mobile 3D scanning system.

6. The method for rapid detection and evaluation of fastener service status according to claim 1, characterized in that: Gaussian filtering algorithm is used in S21 for noise filtering.

Citation Information

Patent Citations

  • Tightness state detection method for track elastic strip fastener

    CN115100109A

  • Space target reconstruction and pose estimation method and system based on multi-view vision

    CN117197333A

  • Three-dimensional geometric detection system for steel rail fasteners

    CN117433447A

  • Method and System for Remote Clothing Selection

    US20210049811A1