A platform gauge detection method

Abstract

The application discloses a kind of platform limit detection methods, comprising: S1, data acquisition;S2, platform data is carried out time synchronization and space synchronization;S3, point cloud fusion;S4, point cloud denoising;S5, based on the matrix transformation platform rough matching;S6, based on the improved particle swarm optimization algorithm platform fine matching;S7, platform limit detection;S8, real-time display, result upload.The method of the application can realize the matching of platform and design profile, obtain the change of platform, and complete the extraction of platform height and side limit value in platform limit detection.The method of the application can realize the matching of platform and design profile, obtain the change of platform, and complete the extraction of platform height and side limit value in platform limit detection.

Description

Technical Field

[0001] This invention relates to the field of railway and urban rail transit platform inspection technology, specifically to a platform clearance inspection method. Background Technology

[0002] By 2024, my country's railway and urban rail transit operating mileage will reach nearly 200,000 kilometers, and the demand for operation and maintenance testing will continue to grow. Several railway bureaus, including Guangzhou Railway Group and Beijing Railway Bureau, require platform clearance testing to be conducted no less than once per quarter, which has created a huge market demand for testing.

[0003] Currently, manual measurement methods for platform clearance inspection mostly employ total stations or platform clearance measuring rulers. These single-point measurement methods lack continuity and efficiency, and may also result in missed points. Among intelligent inspection methods, the Taiyuan Institute of Science and Technology has developed a portable railway structure clearance detector that uses laser ranging for measurement. Distance data is calculated based on calibrated measurement angles, controllable optical path switches, and the number of steps taken by a stepper motor. However, this system has low measurement accuracy and a long inspection time. Large-scale vehicle-mounted structure clearance inspection systems, suitable for continuous measurement, offer higher single-operation efficiency and satisfactory accuracy. These systems typically integrate high-precision scanners from manufacturers such as Z+F and Leica, using mobile measurement to collect line point cloud data, which is then processed and analyzed to complete clearance inspection. They are effective for rapid inspection on long lines. However, multi-platform inspection faces the problem of repeated up-and-down work. In this type of intelligent platform clearance inspection, the integrated inspection systems from various manufacturers are generally heavy and relatively expensive, resulting in low efficiency for multi-platform clearance inspection.

[0004] Employing low-cost, portable integrated inspection systems can improve operational efficiency while reducing inspection costs. However, using low-cost LiDAR results in lower accuracy compared to high-precision LiDAR, necessitating algorithm improvements and optimizations to meet the tolerance requirements of inspection specifications. Currently, there is limited research and development of similar optimization algorithms for inspection results both domestically and internationally. Summary of the Invention

[0005] To address the problems in the background art, this invention proposes a platform clearance detection method with high matching accuracy and strong timeliness.

[0006] Therefore, the present invention adopts the following technical solution:

[0007] A platform clearance detection method includes the following steps:

[0008] S1, Data Collection:

[0009] Platform data at different times is collected on the rails of the platform in the area to be tested. The platform data includes tilt angle data, platform point cloud data, and mileage data.

[0010] S2, perform time synchronization and spatial synchronization on the platform data;

[0011] S3, point cloud fusion:

[0012] The synchronized station data obtained in S2 are fused together to generate cross-sectional point cloud data with tilt correction at the corresponding mileage; the cross-sectional point cloud data includes cross-sectional point cloud frames at different times.

[0013] S4, Point Cloud Denoising:

[0014] S41, Calculate the spatial weights G s (p i ,p j ), where i and j are point numbers, p j For p i The neighborhood N(p) i For any point in ), the neighborhood N(p) i ) is p i Within the point cloud frame of the section, p i The neighborhood N(p) is a region centered at R and with radius R. i Point p is not included in the range. i The value of R is set according to the actual situation;

[0015] S42, Calculate the intensity weight G r ;

[0016] S43, calculate the midpoint p in the cross-sectional point cloud data obtained in S3. i The normal vector is shown in the following formula:

[0017]

[0018] Among them, C i For point p i The covariance matrix, This represents all points within the neighborhood to point p. i The average distance, where T is the transpose operator;

[0019] Then, for the C i Eigenvalue decomposition is performed to obtain several eigenvalues ​​λ and corresponding eigenvectors v. The smallest eigenvector is selected as the point p. i normal vector n i ;

[0020] S44, Calculate the normal vector weight G n (n i ,n j As shown in the following formula:

[0021]

[0022] Among them, point p j normal vector n j The solution obtained using the S43 method is n. i *n j σ is the inner product of the normal vectors, which represents the similarity between the normal vectors. n For n i and n j The standard deviation of the difference;

[0023] S45, Calculate the curvature weighting function G c :

[0024] S451, calculate the curvature of points in the cross-sectional point cloud data, as shown in the following formula:

[0025]

[0026] Where, λ max,i λ is the maximum value among several eigenvalues ​​λ in S43. sum,i c is the sum of all eigenvalues ​​calculated in S43. i For point p i The curvature;

[0027] Then use the solution c i The same method is used to calculate point p j curvature c j :

[0028] S452, Calculate the curvature weighting function G c As shown in the following formula:

[0029]

[0030] Where, σ c For c i and c j The standard deviation of the difference;

[0031] S46, use the following formula to denoise the cross-sectional point cloud data obtained in S3 to obtain denoised point cloud data:

[0032]

[0033] Where, p i ′ represents a point in the denoised point cloud data;

[0034] S5, Platform coarse matching based on matrix transformation:

[0035] Based on the design data of the platform in the area to be detected, a standard dimension map of the platform outline is obtained; then, the RANSAC algorithm is used to obtain the optimal transformation matrix of the denoised point cloud data based on the standard dimension map of the platform outline.

[0036] S6, platform fine matching based on improved particle swarm optimization algorithm, includes the following steps:

[0037] S61 initializes all particles in the particle swarm optimization algorithm;

[0038] S62, the iterative method for determining the position and velocity of particles includes the following sub-steps:

[0039] S621, the basic velocity of the particle at each iteration is determined by the following formula:

[0040] v k ′(t+1)=ω(t)*v k (t)+a1r1(P k (t)-x k (t))+a2r2(g(t)-x k (t));

[0041] Among them, v k (t) represents the velocity of particle k after t iterations, v k ′(t+1) represents the base velocity of particle k at iteration t+1; ω(t) is the inertia weight at iteration t; a1 and a2 are acceleration coefficients set according to actual working conditions; r1 and r2 are random numbers in the range [0,1]; local optimal solution P k (t) represents the best position traversed by particle k up to iteration t; the global optimal solution g(t) is the best position traversed by all particles up to iteration t, x k (t) is the position of particle k after t iterations;

[0042] S622, determine the velocity of the particle after each iteration:

[0043] The base velocity obtained from S621 is updated to obtain the velocity v of particle k after the (t+1)th iteration. k (t+1), we have the following equation:

[0044]

[0045] S623, determines the position of the particle after each update:

[0046] The position x of particle k after iteration t+1 is determined by the following formula. k (t+1):

[0047] x k(t+1)=x k (t)+v k (t+1);

[0048] S63, evaluate the position of each particle using the following formula, calculate the objective function value of the position of each particle after t iterations, and record the particle position with the optimal objective function value;

[0049] MinF(x k (t))=(f1(x k (t)), f2(x) k (t)), f3(x) k (t)));

[0050] In the formula, MinF is the objective function, f1 is the root mean square error evaluation function, f2 is the maximum error evaluation function, and f3 is the matching time;

[0051] S64, terminate the iteration, obtain the optimal solution:

[0052] When the optimal particle position in S63 no longer changes, it is used as the final result and applied to the denoised point cloud data obtained in S46, so that the denoised point cloud data is precisely aligned with the standard dimension drawing of the platform outline, and the final cross-sectional point cloud data is obtained, thus completing the fine matching.

[0053] S7, Inspection Station Clearance:

[0054] The test results for platform clearance include the platform height value and the minimum clearance value in the horizontal direction;

[0055] In step S6, the vertical coordinates of the platform corner points in the final cross-sectional point cloud data are the platform height values.

[0056] The minimum absolute value of the x-coordinate of the points on the side of the platform in the final cross-section point cloud data in S6 is taken as the minimum horizontal limit value for platform clearance detection.

[0057] Step S2 specifically involves: synchronizing the platform data obtained in S1 at the same time; and spatially synchronizing the tilt angle data and mileage data in the vehicle coordinate system and the platform point cloud data in the scanner coordinate system to the corresponding time in the track coordinate system with the center of the track surface as the origin, thus completing the spatial synchronization.

[0058] Step S41 specifically involves calculating the spatial weight G using the following formula. s :

[0059]

[0060] Among them, ||p i -p j ||For point pi and point p j The Euclidean distance between them, σ s For p i and p j The standard deviation of the Euclidean distance.

[0061] Step S42 specifically involves calculating the intensity weight G using the following formula. r :

[0062]

[0063] Among them, I i and I j For point p i and p j The strength, σ r The standard deviation of the intensity difference; the intensity of the points in the cross-sectional point cloud data is obtained from the corresponding points in the station data in S1.

[0064] Step S5 includes the following sub-steps:

[0065] S51, randomly select a certain number of points from the denoised point cloud data obtained from S4 as a coarse matching point set;

[0066] S52, estimate the transformation matrix using the coarse matching point set;

[0067] S53, rotate the denoised point cloud data using the transformation matrix calculated in S52, and calculate the error between the points in the rotated denoised point cloud data and the points in the standard dimension drawing of the platform outline.

[0068] S54. Select a threshold based on the actual working conditions. If the error obtained in S53 is within the threshold range, then the point in the denoised point cloud data corresponding to this error is an interior point, and count the number of all interior points obtained by the corresponding transformation matrix.

[0069] S55, Repeat steps S51 to S54, and select the transformation matrix with the most interior points as the optimal transformation matrix.

[0070] The initialization of all particles in S61 includes: the total number of particles n, the initial position of each particle, the initial velocity of each particle, and the optimal transformation matrix obtained in S5 as the initial individual value of each particle.

[0071] Step S63 also includes the following steps:

[0072] S631, calculate convergence:

[0073] The convergence N(t) of the particle swarm after the t-th iteration is calculated as follows:

[0074]

[0075] S632 adjusts the inertia weight in S621 as shown in the following formula:

[0076]

[0077] Where ω(0) and ω min Let N(0) be the initial and minimum values ​​among all inertial weights, and let N(0) be the initial convergence.

[0078] S633 guides the search for the global optimal solution:

[0079] In each iteration, the local optimum is solved using the following formula:

[0080] P k (t+1)=P k (t)+v k (t+1)+γ(g(t)-P k (t)) (14)

[0081] Here, γ is the guiding factor, which is determined based on the actual working conditions.

[0082] Preferably, the initial value of the guiding factor γ is 2. When the convergence N(t) is low, the value of γ is increased by 0.1 each time to balance the particle's local and global search capabilities.

[0083] Preferably, the platform clearance detection method of the present invention further includes:

[0084] S8, real-time display, results upload:

[0085] During actual testing, the test results of the platform clearance at the current location are displayed in real time according to steps S1-S7 and uploaded to the host computer.

[0086] Compared with the prior art, the present invention has the following beneficial effects:

[0087] 1. The method of the present invention uses data output from multiple sensors integrated on the same track inspection vehicle; compared with manual inspection using total stations, gauge rulers and other methods, it has strong continuity and high efficiency.

[0088] 2. The method of this invention can automatically complete data processing and quickly generate detection results. By using coarse matching of stations based on matrix transformation and fine matching of stations based on particle swarm optimization, and introducing a dynamic convergence control mechanism, the algorithm's matching accuracy and robustness are significantly improved by adaptively adjusting the ratio of global and local searches, inertia weights, etc.; it can also effectively avoid premature convergence and improve the algorithm's global optimization capability, thereby quickly completing station extraction and boundary detection.

[0089] 3. The method of this invention can output the clearance detection results of the platform in real time and automatically upload the results. When abnormalities are found in the detection results, on-site verification can be carried out in a timely manner, which greatly improves the timeliness of the detection operation and the validity of the detection results.

[0090] 4. The method of this invention introduces normal vector weights and curvature weights, and can adaptively adjust the standard deviation parameter. In contrast, traditional bilateral filtering generally only considers the spatial and intensity information of point clouds. This method enhances the applicability and robustness of station detection in different scenarios and ensures the accuracy of boundary detection.

[0091] 5. This invention is applicable to clearance detection of all railway and urban rail transit platforms, providing data support for operation and maintenance departments and ensuring the safety of platform operation. Simultaneously, the platform matching results can also be used to determine changes in platform dimensions from design dimensions, whether there is damage, etc., and can also provide a reference for platform repair. Attached Figure Description

[0092] Figure 1 This is a flowchart of a platform clearance detection method according to the present invention;

[0093] Figure 2 A schematic diagram of a track inspection system operating the method of the present invention;

[0094] Figure 3 This is a schematic diagram of the transformation between two spatial rectangular coordinate systems in step S2 of the present invention;

[0095] Figure 4 This is a schematic diagram of the point cloud of the platform cross-section after point cloud fusion in an embodiment of the present invention;

[0096] Figure 5 This is a standard dimension drawing of the platform outline in an embodiment of the present invention;

[0097] Figure 6 This is a schematic diagram of station fine matching based on particle swarm optimization in an embodiment of the present invention.

[0098] In the diagram: 1. Scanner; 2. Encoder; 3. Inclinometer; 4. Tablet PC; 5. Scanner stand; 6. Hand lever; 7. Battery compartment; 8. Mounting bracket; 9. Top caster;

[0099] 10. Platform surface; 11. Platform side; 12. Platform bottom edge. Detailed Implementation

[0100] The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0101] In order to realize the platform boundary detection method of the present invention, the platform boundary detection system should be able to simultaneously acquire the tilt angle data, platform point cloud data and mileage data at the current moment.

[0102] The platform clearance detection system integrates scanner 1, encoder 2, inclinometer 3, tablet computer 4, and hand-push vehicle platform. The hand-push vehicle platform includes scanner bracket 5, push rod 6, battery compartment 7, mounting bracket 8, top wheel 9, and vehicle body. The vehicle platform carries various sensors for motion measurement, using a high-precision scanner to collect platform point cloud data, and simultaneously acquiring inclinometer and mileage data at the current moment. A schematic diagram of the platform clearance detection system is shown below. Figure 2 The specific components are as follows:

[0103] (1) Driving structure:

[0104] 1. The vehicle body support uses hollow square tubes, which reduces the wall thickness of the square tubes and lowers the weight while ensuring sufficient structural strength; the vehicle body adopts modular assembly, and quick-release screws are used at all T-shaped connection points for easy installation; the vehicle body platform is equipped with handles for easy handling across lines.

[0105] 2. The push rod adopts a rotatable and adjustable buckle mechanism, which can be pushed at different heights and angles, making it suitable for different operators.

[0106] 3. The scanner is fixed to the vehicle platform using a scanner bracket and a triangular fixing bracket to ensure the stability of the scanner's data acquisition.

[0107] 4. The encoder is embedded in the wheel, and the inclinometer is installed inside the vehicle body. Each sensor is equipped with a connector plug, which can be plugged in and connected accordingly, making installation quick and reducing the exposure of data cables.

[0108] 5. The wheels of the vehicle are made of nylon, and the top wheel is made of ceramic to ensure the insulation of the vehicle body.

[0109] (2) Tightening structure:

[0110] The hand-push type vehicle platform adopts a three-point clamping structure to reduce swaying in the plane. The right end clamping wheel is equipped with a spring, which adaptively adjusts according to changes in track gauge during the pushing process to ensure the fit between the vehicle body and the rail.

[0111] (3) System communication and storage:

[0112] The system powers each sensor via a lithium battery, allowing for continuous operation of over 3 hours on a single battery. The battery is located in a battery compartment at the bottom of the scanner holder. The battery powers the scanner, encoder, tiltmeter, and tablet computer; all sensor data is stored on the tablet computer.

[0113] The platform clearance detection system used in this invention employs lightweight materials and a modular design. It utilizes a hand-pushed vehicle platform to carry various sensors for mobile scanning and data collection. During the process, the scanner collects platform point cloud data; the encoder collects mileage data; the inclinometer collects inclinometer data; and a tablet computer is then used to complete the method of this invention. Figure 1 As shown, the method of the present invention specifically includes the following steps:

[0114] S1, Data Collection:

[0115] Platform data at different times is acquired on the tracks of the platform in the area to be inspected using a platform clearance detection system. The platform data includes tilt angle data, platform point cloud data, and mileage data. The platform point cloud data includes the coordinates and intensity of the platform point cloud.

[0116] S2 performs time and spatial synchronization of platform data:

[0117] The platform data obtained at the same time from S1 is synchronized in time; the tilt angle data and mileage data in the time-synchronized vehicle coordinate system, and the platform point cloud data in the scanner coordinate system are all spatially synchronized to the corresponding time in the track coordinate system with the center of the track surface as the origin, thus completing spatial synchronization. Specifically:

[0118] Time synchronization: The encoder and inclinometer data are stored on the tablet computer and the same system time is recorded. The encoder and scanner are directly connected, and the collected scanner data carries the corresponding mileage information, thus completing the time synchronization of each sensor.

[0119] Spatial synchronization: A calibration method is used to establish the transformation relationship between different coordinate systems, namely two spatial rectangular coordinate systems O. a X a Y a Z a and O b X b Y b Z b Transformation such as Figure 3 The two coordinate axes are not parallel and their scales are inconsistent. First, the two coordinate systems are translated so that the origins coincide. Then, the three coordinate axes are made parallel by three rotations. The relationship between the two coordinate systems is as shown in equation (1):

[0120]

[0121] In the formula, X a Y a Z a The coordinates before transformation, X b Y b Z bThe coordinates after transformation; X0, Y0, Z0 are translation parameters, δμ is the scale factor, and R X R Y R z The equations for the rotation matrix are as follows. The translation parameters, scale factor, and rotation matrix equations are obtained based on the actual working conditions.

[0122] S3, point cloud fusion:

[0123] The platform point cloud data, tilt data, and mileage data obtained in the track coordinate system from S2 are fused to generate tilt-corrected cross-sectional point cloud data at the corresponding mileage. The cross-sectional point cloud data includes cross-sectional point cloud frames at different times; the intensity of the point cloud in the cross-sectional point cloud data is the same as the intensity of the corresponding point in the platform point cloud.

[0124] S4, Point Cloud Denoising:

[0125] Cross-sectional point cloud data fused from S3 point clouds, incorporating mileage information and tilt correction, often contains noise. These outliers can significantly impact the accuracy of station detection. Since the surface and sides of a station are generally straight lines, an improved bilateral filtering method is used for initial processing of the point cloud data. Compared to traditional filtering methods, this method introduces normal vectors and curvature information to further improve denoising. The specific steps include:

[0126] S41, Calculate the spatial weights G s As shown in equation (2):

[0127]

[0128] Among them, ||p i -p j ||For point p i and point p j The Euclidean distance between them, p j For p i The neighborhood N(p) i For any point in ), σ s For p i and p j The standard deviation of the Euclidean distance. Neighborhood N(p) i ) is p i Within the point cloud frame of the section, p i The region centered at R and with radius R, excluding the point p in its neighborhood. i R is set according to the actual situation.

[0129] S42, Calculate the intensity weight G r As shown in equation (3):

[0130]

[0131] Among them, Ii and I j For point p i and p j The strength, σ r This represents the standard deviation of the intensity difference.

[0132] S43, Calculate the normal vector of a point in the cross-sectional point cloud data:

[0133] For point p i It can be calculated by its covariance matrix C i And calculate point p by determining the magnitude of the characteristic vector. i normal vector n i The specific calculation method is as follows:

[0134]

[0135] In the formula, This represents all points within the neighborhood to point p. i The average distance.

[0136] For the covariance matrix C i Eigenvalue decomposition is performed to obtain several eigenvalues ​​λ and corresponding eigenvectors v. The smallest eigenvector v among these eigenvectors is selected. min As point p i normal vector n i v min For point p i The midpoint p of the cross-section point cloud frame i The direction with the least change.

[0137] S44, Calculate the normal vector weight G n As in equation (5):

[0138]

[0139] Among them, point p j normal vector n j The solution obtained using the S43 method is n. i *n j σ is the inner product of the normal vectors, used to represent the similarity between normal vectors. n For n i and n j The standard deviation of the difference.

[0140] S45, Calculate the curvature weighting function G c :

[0141] S451, Calculate the curvature of the midpoint of the point cloud:

[0142] Point p i curvature c iCalculated using equation (6):

[0143]

[0144] Where, λ max,i λ is the maximum value among several eigenvalues ​​λ in S43. sum,i This is the sum of all eigenvalues ​​calculated in S43.

[0145] Then calculate point p using the same method. j curvature c j :

[0146] S452, Calculate the curvature weighting function G c As shown in equation (7):

[0147]

[0148] Where, σ c For c i and c j The standard deviation of the difference;

[0149] S46, Denoise the cross-sectional point cloud data:

[0150] Based on steps S41 to S45, calculate the spatial weight, intensity weight, normal vector weight, and curvature weight of all points in the cross-sectional point cloud data, and then use equation (8) to denoise the cross-sectional point cloud data to obtain denoised point cloud data:

[0151]

[0152] Where, p i ′ represents a point in the denoised point cloud data.

[0153] S5, Platform coarse matching based on matrix transformation:

[0154] Based on the design data of the platform in the area to be inspected, obtain the standard dimension drawing of the platform outline.

[0155] The RANSAC algorithm is used to obtain the optimal transformation matrix through random sampling and model validation. The specific process is as follows:

[0156] S51, randomly select a certain number of points from the denoised point cloud data as a coarse matching point set;

[0157] S52, estimate the transformation matrix using a coarsely matched point set;

[0158] S53, rotate the denoised point cloud data using the transformation matrix calculated in S52, and calculate the error between the points in the rotated denoised point cloud data and the points in the standard dimension drawing of the platform outline.

[0159] S54. Select a suitable threshold based on the actual working conditions. If the error obtained in S53 is within the threshold range, then the point is an interior point, and count the number of all interior points obtained by the corresponding transformation matrix.

[0160] S55, Repeat steps S51 to S54, and select the transformation matrix with the most interior points as the best transformation matrix.

[0161] S56 uses the optimal transformation matrix to transform the denoised point cloud data, completing a coarse match between it and the standard dimension drawing of the platform outline.

[0162] S6, platform matching based on an improved particle swarm optimization algorithm:

[0163] The platform matching results obtained after the S5 coarse matching step still have some errors and cannot meet the accuracy requirements of platform detection. Fine matching is needed to further improve the matching accuracy between the platform cross-section and the design contour. Fine matching employs Particle Swarm Optimization (PSO) algorithm, which searches for the global optimum through individual cooperation and information sharing within the swarm. A particle with velocity and position attributes is designed, where velocity represents the speed of movement and position represents the direction of movement. Each particle individually searches for the optimal solution of the transformation matrix in the search space and records it as its current individual extreme value. This individual extreme value is shared with other particles in the swarm, and the optimal individual extreme value is taken as the current global optimum for the entire swarm. All particles in the swarm adjust their velocity and position based on their current individual extreme value and the shared global optimum, iteratively seeking the optimal solution. The fine matching process based on PSO is as follows: Figure 6 As shown, the specific process is as follows:

[0164] S61 initializes all particles, including:

[0165] (1) Initialize the number of particles: Determine the total number of particles n;

[0166] (2) Initialize the particle positions: Randomly initialize the particle swarm and determine the initial position of each particle. The position of each particle represents a possible solution.

[0167] (3) Initialize the particle velocity: Initialize the particle velocity and determine the initial velocity of each particle. The particle velocity is used to determine the update direction and step size of the particle position in each iteration.

[0168] (4) Use the optimal transformation matrix obtained in S5 as the initial individual value for each particle.

[0169] S62, the iterative method for determining the position and velocity of particles:

[0170] S621, determine the basic velocity of the particle in each iteration, which is obtained through equation (9):

[0171] v k ′(t+1)=ω(t)*v k (t)+a1r1(P k (t)-x k (t))+a2r2(g(t)-x k (t)) (9)

[0172] Among them, v k (t) represents the velocity of particle k after t iterations, v k '(t+1) represents the base velocity of particle k in the (t+1)th iteration; ω(t) is the inertia weight in the t-th iteration, the initial value of which is generally set to 0.9, and a large inertia weight helps to enhance the global search capability; a1 and a2 are acceleration coefficients, set according to actual working conditions; r1 and r2 are random numbers, and their values ​​range from [0,1]; local optimal solution P k (t) represents the best position traversed by particle k up to iteration t; the global optimal solution g(t) is the best position traversed by all particles up to iteration t, x k (t) is the position of particle k after t iterations. When t = 0, it is the initial value set in S61.

[0173] S622, determine the velocity of the particle after each iteration:

[0174] The base velocity determined using S621 may have local optima. Therefore, a nonlinear adjustment factor is introduced into the velocity update formula to improve the velocity update mechanism and increase the accuracy of velocity calculation, thus obtaining the velocity v of particle k after t+1 iterations. k (t+1), we have equation (10):

[0175]

[0176] S623, determines the position of the particle after each update:

[0177] The position x of particle k after the (t+1)th iteration is determined by equation (11). k (t+1):

[0178] x k (t+1)=x k (t)+v k (t+1) (11)

[0179] S63, Improved dynamic convergence calculation of particle swarm optimization algorithm:

[0180] After the particle swarm initialization, optimal matching is required. We introduce an improved adaptive dynamic convergence control mechanism, which dynamically adjusts the weights of global and local searches by calculating the convergence state of the particle swarm in real time, so as to avoid premature convergence and improve matching accuracy, thereby optimizing the convergence speed and accuracy of the algorithm.

[0181] S631, calculate convergence:

[0182] The convergence state of the current particle swarm is calculated as shown in equation (12):

[0183]

[0184] Where N(t) is the convergence degree of the t-th iteration.

[0185] S632, adjusts the inertia weight:

[0186] At the same time, adaptive weights are introduced to dynamically adjust the number of iterations, and the calculation of adaptive weights is improved as shown in equation (13).

[0187]

[0188] Where, ω max and ω min N(0) represents the initial and minimum values ​​among all inertia weights, and N(0) represents the initial convergence. The inertia weights can be dynamically adjusted based on the global and local optima of the particle swarm, thereby balancing the capabilities of global and local searches. The initial convergence N(0) is calculated using the initial particle values ​​set in S61.

[0189] When the convergence N(t) is large, ω(t) approaches ω max To enhance global search capability; when N(t) is small, ω(t) decreases to enhance local search capability.

[0190] S633 guides the search for the global optimal solution:

[0191] In each iteration, the guiding role of g(t) is enhanced to prevent the algorithm from getting trapped in local optima, as shown in equation (14):

[0192] P k (t+1)=P k (t)+v k (t+1)+γ(g(t)-P k (t)) (14)

[0193] Among them, v k(t) represents the velocity of particle i at iteration t. γ is the guiding factor, which is generally set to 2 initially. When the convergence metric is low (i.e., close to convergence), the value of γ is increased by about 0.1 each time to balance the particle's local and global search capabilities.

[0194] S634, Objective function evaluation incorporating multiple matching errors:

[0195] An improved fusion multi-matching error evaluation method is adopted, which optimizes the matching accuracy through a multi-objective optimization method and can also avoid the range caused by a single evaluation index, as shown in equation (15):

[0196] MinF(x k (t))=(f1(x k (t)), f2(x) k (t)), f3(x) k (t))) (15)

[0197] In the formula, MinF is the objective function, f1 is the root mean square error evaluation function, f2 is the maximum error evaluation function, and f3 is the matching time.

[0198] Equation (15) is used to evaluate the position of each particle obtained in each iteration, calculate the objective function value of each particle at the position after the iteration, and record the particle position with the optimal objective function value.

[0199] S64, terminate the iteration, obtain the optimal solution:

[0200] When the optimal particle position in S634 no longer changes, it is used as the final result and applied to the denoised point cloud data to accurately align it with the standard dimension drawing of the platform outline to obtain the final cross-sectional point cloud data, thus completing the fine matching.

[0201] S7, Inspection Station Clearance:

[0202] The test results for platform clearance include the platform height value and the minimum clearance value in the horizontal direction;

[0203] The final cross-sectional point cloud data obtained after fine matching in step S6 can be directly used for platform clearance detection and analysis. The ordinate of the platform corner points of the matched standard platform outline is the platform height value for platform clearance detection. Then, the side point cloud is processed and analyzed to obtain the minimum absolute value of the abscissa of the points on the side of the platform in the final cross-sectional point cloud data as the minimum clearance value in the horizontal direction for platform clearance detection.

[0204] S8, real-time display, results upload:

[0205] During actual testing, the system can display the detection results of the platform clearance at the current location of each point cloud frame in real time and upload them to the host computer. When abnormalities occur in the detection results, they can also be reviewed promptly, greatly improving testing efficiency.

[0206] Example

[0207] In one embodiment of the present invention, the cross-sectional point cloud data of the station to be tested is obtained through steps S1-S3; such as Figure 4 The image shown is a frame of point cloud data from the cross-sectional point cloud data. After performing point cloud denoising in step S4 on this frame of point cloud data, step S5 is performed to compare the denoised point cloud frame with... Figure 5 The platform outline standard dimension diagram at the corresponding location is used for coarse matching; finally, the coarse matching result is input into S6 for fine matching. The fine matching result of the cloud frame data at this point is as follows: Figure 6 As shown.

[0208] From Figure 6 As can be seen, this point cloud frame data is basically completely consistent with the standard dimension drawing of the platform outline at the corresponding location. In other words, this point cloud frame can be used as the platform outline at this location to detect the platform boundary.

Claims

1. A method for detecting clearance at a platform, characterized in that, Includes the following steps: S1, Data Collection: Platform data at different times is collected on the rails of the platform in the area to be tested. The platform data includes tilt angle data, platform point cloud data, and mileage data. S2, perform time synchronization and spatial synchronization on the platform data; S3, point cloud fusion: The synchronized station data obtained in S2 are fused together to generate cross-sectional point cloud data with tilt correction at the corresponding mileage; the cross-sectional point cloud data includes cross-sectional point cloud frames at different times. S4, Point Cloud Denoising: S41, Calculate the spatial weights G s (p i ,p j ), where i and j are point numbers, p j For p i The neighborhood N(p) i For any point in ), the neighborhood N(p) i ) is p i Within the point cloud frame of the section, p i The neighborhood N(p) is a region centered at R and with radius R. i Point p is not included in the range. i The value of R is set according to the actual situation; S42, Calculate the intensity weight G r ; S43, calculate the midpoint p in the cross-sectional point cloud data obtained in S3. i The normal vector is shown in the following formula: Among them, C i For point p i The covariance matrix, This represents all points within the neighborhood to point p. i The average distance, where T is the transpose operator; Then, for the C i Eigenvalue decomposition is performed to obtain several eigenvalues ​​λ and corresponding eigenvectors v. The smallest eigenvector is selected as the point p. i normal vector n i ; S44, Calculate the normal vector weight G n (n i ,n j As shown in the following formula: Among them, point p j normal vector n j The solution obtained using the S43 method is n. i *n j σ is the inner product of the normal vectors, which represents the similarity between the normal vectors. n For n i and n j The standard deviation of the difference; S45, Calculate the curvature weighting function G c : S451, calculate the curvature of points in the cross-sectional point cloud data, as shown in the following formula: Where, λ max,i λ is the maximum value among several eigenvalues ​​λ in S43. sum,i c is the sum of all eigenvalues ​​calculated in S43. i For point p i The curvature; Then use the solution c i The same method is used to calculate point p j curvature c j : S452, Calculate the curvature weighting function G c As shown in the following formula: Where, σ c For c o and c j The standard deviation of the difference; S46, use the following formula to denoise the cross-sectional point cloud data obtained in S3 to obtain denoised point cloud data: Where, p i ′ The points in the denoised point cloud data; S5, Platform coarse matching based on matrix transformation: Based on the design data of the platform in the area to be detected, a standard dimension map of the platform outline is obtained; then, the RANSAC algorithm is used to obtain the optimal transformation matrix of the denoised point cloud data based on the standard dimension map of the platform outline. S6, platform fine matching based on improved particle swarm optimization algorithm, includes the following steps: S61 initializes all particles in the particle swarm optimization algorithm; S62, the iterative method for determining the position and velocity of particles includes the following sub-steps: S621, the basic velocity of the particle at each iteration is determined by the following formula: v k ′(t+1)=ω(t)*v k (t)+a1r1(P k (t)-x k (t))+a2r2(g(t)-x k (t)); Among them, v k (t) represents the velocity of particle k after t iterations, v k ′ (t+1) represents the base velocity of particle k at iteration t+1; ω(t) is the inertia weight at iteration t; a1 and a2 are acceleration coefficients set according to actual working conditions; r1 and r2 are random numbers in the range [0,1]; local optimal solution P k (t) represents the best position traversed by particle k up to iteration t; the global optimal solution g(t) is the best position traversed by all particles up to iteration t, x k (t) is the position of particle k after t iterations; S622, determine the velocity of the particle after each iteration: The base velocity obtained from S621 is updated to obtain the velocity v of particle k after the (t+1)th iteration. k (t+1), we have the following equation: S623, determines the position of the particle after each update: The position x of particle k after iteration t+1 is determined by the following formula. k (t+1): x k (t+1)=x k (t)+v k (t+1); S63, evaluate the position of each particle using the following formula, calculate the objective function value of the position of each particle after t iterations, and record the particle position with the optimal objective function value; MinF(x k (t))=(f1(x k (t)),f2(x k (t)),f3(x k (t))); In the formula, MinF is the objective function, f1 is the root mean square error evaluation function, f2 is the maximum error evaluation function, and f3 is the matching time; S64, terminate the iteration, obtain the optimal solution: When the optimal particle position in S63 no longer changes, it is used as the final result and applied to the denoised point cloud data obtained in S46, so that the denoised point cloud data is precisely aligned with the standard dimension drawing of the platform outline, and the final cross-sectional point cloud data is obtained, thus completing the fine matching. S7, Inspection Station Clearance: The test results for platform clearance include the platform height value and the minimum clearance value in the horizontal direction; In step S6, the vertical coordinates of the platform corner points in the final cross-sectional point cloud data are the platform height values. The minimum absolute value of the x-coordinate of the points on the side of the platform in the final cross-section point cloud data in S6 is taken as the minimum horizontal limit value for platform clearance detection.

2. The platform clearance detection method according to claim 1, characterized in that, Step S2 specifically involves: synchronizing the platform data obtained in S1 at the same time; and spatially synchronizing the tilt angle data and mileage data in the vehicle coordinate system and the platform point cloud data in the scanner coordinate system to the corresponding time in the track coordinate system with the center of the track surface as the origin, thus completing the spatial synchronization.

3. The platform clearance detection method according to claim 1, characterized in that, Step S41 specifically involves calculating the spatial weight G using the following formula. s : Among them, ||p i -p j ||For point p i and point p j The Euclidean distance between them, σ s For p i and p j The standard deviation of the Euclidean distance.

4. The platform clearance detection method according to claim 1, characterized in that, Step S42 specifically involves calculating the intensity weight G using the following formula. r : Among them, I i and I j For point p i and p j The strength, σ r The standard deviation of the intensity difference; the intensity of the points in the cross-sectional point cloud data is obtained from the corresponding points in the station data in S1.

5. The platform clearance detection method according to claim 1, characterized in that, Step S5 It includes the following steps: S51, randomly select a certain number of points from the denoised point cloud data obtained from S4 as a coarse matching point set; S52, estimate the transformation matrix using the coarse matching point set; S53, rotate the denoised point cloud data using the transformation matrix calculated in S52, and calculate the error between the points in the rotated denoised point cloud data and the points in the standard dimension drawing of the platform outline. S54. Select a threshold based on the actual working conditions. If the error obtained in S53 is within the threshold range, then the point in the denoised point cloud data corresponding to this error is an interior point, and count the number of all interior points obtained by the corresponding transformation matrix. S55, Repeat steps S51 to S54, and select the transformation matrix with the most interior points as the optimal transformation matrix.

6. The platform clearance detection method according to claim 1, characterized in that, The initialization of all particles in S61 includes: the total number of particles n, the initial position of each particle, the initial velocity of each particle, and the optimal transformation matrix obtained in S5 as the initial individual value of each particle.

7. The platform clearance detection method according to claim 6, characterized in that, Step S63 also includes the following steps: S631, calculate convergence: The convergence N(t) of the particle swarm after the t-th iteration is calculated as follows: S632 adjusts the inertia weight in S621 as shown in the following formula: Where ω(0) and ω min Let N(0) be the initial and minimum values ​​among all inertial weights, and let N(0) be the initial convergence. S633 guides the search for the global optimal solution: In each iteration, the local optimum is solved using the following formula: P k (t+1)=P k (t)+v k (t+1)+γ(g(t)-P k (t)) (14) Here, γ is the guiding factor, which is determined based on the actual working conditions.

8. The platform clearance detection method according to claim 7, characterized in that, The initial value of the guiding factor γ is 2. When the convergence N(t) is low, the value of γ is increased by 0.1 each time to balance the particle's local and global search capabilities.

9. The platform clearance detection method according to claim 1, characterized in that, Also includes: S8, real-time display, results upload: During actual testing, the test results of the platform clearance at the current location are displayed in real time according to steps S1-S7 and uploaded to the host computer.

Images