Method for determining mileage deviation of track inspection data based on DTW and Hough transform

Abstract

The present invention provides a method for determining the mileage deviation of track inspection data based on DTW and Hough transform, comprising: selecting a mileage range [t0, t end ] is used as the query sequence Q, and the same mileage range [t0,t end ] as the matching sequence C; use the DTW algorithm to calculate the regular path P of the query sequence Q and the matching sequence C; perform Hough transform on the regular path P and find the maximum value H in the Hough space max (ρ,θ); by H max (ρ,θ) finds the most collinear matching path #imgabs0# to determine the mileage offset Δ between the query sequence Q and the matching sequence C. The present invention utilizes the good anti-noise performance of Hough transform and the insensitivity to partial covering phenomenon to achieve the determination of the mileage deviation between multiple track inspection data under the condition of moving track disturbance or data loss, has good robustness, and can be used to improve the accuracy of track inspection data analysis.

Description

Technical Field

[0001] The present invention relates to the field of track detection technology, and in particular to a method for determining mileage deviation of track detection data based on DTW and Hough transform. Background Art

[0002] The principle of track inspection for my country's high-speed railways is "dynamic inspection as the main method, combined with dynamic and static inspections". Dynamic inspections (hereinafter referred to as dynamic inspections) represented by track inspection vehicles (hereinafter referred to as track inspection vehicles) and static inspections (hereinafter referred to as static inspections) represented by track inspection instruments (hereinafter referred to as track inspection instruments) periodically collect a huge amount of track geometry data (hereinafter referred to as track inspection data). However, the application of track inspection data still has problems such as low data management level and insufficient data mining depth. To break through the information islands between track inspection data, mine the implicit patterns in the data, and detect potential risks early, all rely on the realization of track inspection data matching. The premise of track inspection data matching is to be able to quickly and accurately determine the mileage deviation between data in complex scenarios.

[0003] In order to correct the mileage deviation of track inspection data, mileage benchmarks can be established through external features such as DGPS, RFID, curve control points, and turnout equipment, or mileage deviations can be estimated by time-delay cross-correlation, gray correlation, or overall information matrix of track inspection data. However, mileage correction based on external features is often not accurate and requires the installation of external equipment. Methods such as time-delay cross-correlation are sensitive to noise and signal periodicity, and assume linear mileage deviations, so they fail to handle the mileage correction problem between online data and historical data well. To this end, Wei Hui (ZL 202010514649.0; ZL 202110174127.5) and He Qing (He Qing, Ma Yusong, Li Chenzhong, et al. Mileage alignment and error correction of dynamic and static track inspection data for high-speed railways [J]. Journal of the China Railway Society, 2024, 46(1): 129-136.) used the DTW (Dynamic Time Wraping) algorithm to match track inspection data. Among them, Wei Hui achieved the matching of dynamic and static height unevenness through "robust estimation + secondary regularization", while He Qing used the method of "time-delay cross-correlation + dynamic regularization" to match dynamic and static gauge unevenness. The above methods have solved the problem of track inspection data matching under undisturbed track conditions. However, when there are disturbance factors such as comprehensive maintenance and large machine operations on the track, the similarity between online data and historical data will be destroyed, and both robust estimation and least squares estimation are difficult to converge to the actual mileage deviation. In order to improve the robustness of mileage deviation estimation, a method for determining the mileage deviation of track inspection data under conditions of dynamic track disturbance or data loss should be established. Summary of the invention

[0004] In view of the above situation, it is necessary to provide a method for determining the mileage deviation of track inspection data based on DTW and Hough transform to address the problem that the accuracy of determining the mileage deviation of track inspection data in the prior art is easily affected by track disturbance or data loss.

[0005] A method for determining mileage deviation of track inspection data based on DTW and Hough transform, comprising:

[0006] Select mileage range [t 0 ,t end ] as the query sequence Q, and the same mileage range [t 0 ,t end ] as the matching sequence C;

[0007] Use the DTW algorithm to calculate the regular path P of the query sequence Q and the matching sequence C;

[0008] Perform Hough transform on the regular path P and find the maximum value H in the Hough space max (ρ,θ);

[0009] By H max (ρ,θ) finds the matching path with the most collinearity Determine the mileage offset Δ between the query sequence Q and the matching sequence C.

[0010] The above method for determining the mileage deviation of track inspection data based on DTW and Hough transform, wherein the same mileage range [t 0 ,t end ] as the matching sequence C, specifically including:

[0011] Retrieving a historical track inspection database corresponding to the line name and line type according to the line name and line type detected by the online track inspection data;

[0012] According to the mileage range [t 0 ,t end ], extract one or more groups of historical track inspection data with the same name from the historical track inspection database as a matching sequence C.

[0013] The above-mentioned method for determining the mileage deviation of track inspection data based on DTW and Hough transform, wherein the regular path P of the query sequence Q and the matching sequence C is calculated by using the DTW algorithm, specifically includes:

[0014] Calculate the distance matrix D between the query sequence Q and the matching sequence C;

[0015] A dynamic programming method is used to find a regular path P in the distance matrix D.

[0016] In the above-mentioned method for determining the mileage deviation of track inspection data based on DTW and Hough transform, Hough transform is performed on the regular path P, and the maximum value H is found in the Hough space. max (ρ,θ), specifically including:

[0017] Discretize the parameters ρ and θ and construct the Hough space H;

[0018] Map each point of the regular path P to the Hough space H through Hough transform, and find the maximum value H in the Hough space H max (ρ,θ).

[0019] The above method for determining the mileage deviation of track inspection data based on DTW and Hough transform, wherein H max (ρ,θ) finds the matching path with the most collinearity Determine the mileage offset Δ between the query sequence Q and the matching sequence C, specifically including:

[0020] Using the point-line duality relationship, H max (ρ,θ) finds the matching path with the most collinearity

[0021] Calculate matching paths The intercept with the index axis of the query sequence Q or the matched sequence C;

[0022] The mileage offset Δ between the query sequence Q and the matching sequence C is determined by the intercept and the sampling step.

[0023] In the above-mentioned method for determining mileage deviation of track inspection data based on DTW and Hough transform, the online track inspection data is collected by a level 0 track inspection instrument or by a track inspection vehicle.

[0024] According to the method for determining the mileage deviation of track inspection data based on DTW and Hough transform provided by the present invention, the most collinear matching path is extracted by Hough transform of the regular path, and then the mileage deviation between the online track inspection data and the historical track inspection data is determined based on the most collinear matching path. Since the mileage deviation is estimated by the most collinear points, the influence of mismatching caused by moving track disturbance or data loss on the estimation accuracy is effectively avoided, and the mileage deviation between multiple track inspection data under the condition of moving track disturbance or data loss can be accurately estimated, which has good robustness and can be used to improve the accuracy of track inspection data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a flow chart of a method for determining mileage deviation of track inspection data based on DTW and Hough transform in a first embodiment of the present invention;

[0026] Figure 2 are the query sequence Q and matching sequence C data samples in the first embodiment of the present invention;

[0027] Figure 3 A schematic diagram of the DTW algorithm calculating the regularized path P of Q and C in the first embodiment of the present invention;

[0028] Figure 4 It is a schematic diagram of mapping the regularized path P to the Hough space through Hough transform in the first embodiment of the present invention;

[0029] Figure 5 In the first embodiment of the present invention, H max (ρ,θ) finds the matching path with the most collinearity Schematic diagram of

[0030] Figure 6 In the second embodiment of the present invention, H max (ρ,θ) finds the matching path with the most collinearity Schematic diagram of . DETAILED DESCRIPTION

[0031] In order to facilitate understanding of the present invention, the present invention will be described more fully below with reference to various embodiments, but the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0032] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0033] See also Figure 1 The first embodiment of the present invention provides a method for determining the mileage deviation of track inspection data based on DTW and Hough transform. The detection method is applied to track inspection data processing of railway tracks to determine the mileage deviation between online track inspection data and historical track inspection data under the condition of track disturbance or data loss. The method for determining the mileage deviation of track inspection data based on DTW and Hough transform includes steps S11 to S14.

[0034] Step S11, select mileage range [t 0 ,t end ] is used as the query sequence Q, and the same mileage range [t 0 ,t end ]’s static inspection data with the same name is taken as the matching sequence C.

[0035] In this embodiment, the data of a high-speed railway up train was collected at the end of April 2021. 0 ,t end ] as the query sequence Q. Specifically, in this embodiment, t 0 =K1224.1km,t end =K1225.1km, see Figure 2 a). Among them, the dynamic inspection data is selected as the left high and low data, which can be collected through existing inspection equipment, such as track inspection vehicles, comprehensive inspection trains, etc.

[0036] According to the query sequence Q related information, retrieve the upline static inspection data collected in early April 2021 from the static inspection historical track inspection database, and intercept the left high and low data of the track inspection instrument static inspection with the same name in the mileage range of K1224.1km~K1225.1km as the matching sequence C, see Figure 2 b). Among them, other technical means have been used to estimate the mileage deviation between Q and C to be 66.315m, and the data from K1224.638km to K 1224.588km in sequence C is lost.

[0037] Step S12, using the DTW algorithm to calculate the regularized path P of Q and C.

[0038] The size of the measured distance determines the degree of similarity between the data, and the similarity measurement method determines the effect of the measurement. DTW is a flexible measurement based on dynamic programming. It calculates the similarity between two time series by regularizing the time series. Therefore, it has significant advantages in matching time series with problems such as unequal length, offset or amplitude, and time axis expansion.

[0039] The steps of calculating the regularized path P of Q and C using the DTW algorithm include:

[0040] Step S121: Calculate the distance matrix D between two sequences Q and C.

[0041] Let the i-th element in the query sequence Q be q i , the jth element in the matching sequence C is c j , then the expression of the distance between elements d(i,j) is:

[0042] d(i,j)=||q i -c j || w (1)

[0043] Where i is the index of the query sequence Q, i = 1, 2, ..., I; j is the index of the matching sequence C, j = 1, 2, ..., J. I is the maximum value of the query sequence Q index, and J is the maximum value of the matching sequence C index; when w = 1, d(i, j) is the Manhattan distance; when w = 2, it is the Euclidean distance.

[0044] Get the distance matrix D:

[0045]

[0046] Step S122: Use dynamic programming method to find a regular path P in D.

[0047] P = {p 1 ,p 2 ,...,p k ,...,p K}

[0048] where p k Represents the regular path position, that is, p k =(i,j) k Indicates q i With c j Match, where K is the length of the regular path P, p 1 、p 2 、p k 、p K They are the 1st, 2nd, kth, and Kth points in the matching path P, respectively, and satisfy: k = 1, 2, ..., K, min (I, J) ≤ K ≤ I + J;

[0049] For details, see Figure 3 , using the idea of ​​dynamic programming to determine the regular path position p k , when calculating the DTW distance, a cost matrix needs to be constructed, and the matrix element γ(i,j) is defined as:

[0050]

[0051] The dynamic time warping distance DTW(Q,C) between Q and C should be calculated to minimize the cumulative distance value. The dynamic time warping distance is calculated as follows:

[0052]

[0053] Step S13, perform Hough transform on the regular path P, and find the maximum value H in the Hough space. max (ρ,θ).

[0054] In 1962, Paul Hough proposed the Hough transform, which realizes the mapping of a straight line from image space to parameter space. Its basic idea is the point-line duality, as shown in equation (4), that is, using the mapping relationship from image space to parameter space, by counting the calculation results of all contour points, the peak coordinates in the parameter space are determined to be a straight line. The Hough transform has the advantages of being insensitive to noise, good robustness, and easy to handle occlusion problems. It has been widely used in the fields of image processing and pattern recognition.

[0055] H:(i,j)→ρ=i·cosθ+j·sinθ (-90°≤θ≤90°) (3)

[0056] Among them, ρ is the distance from the origin to the straight line, and θ is the angle between the straight line and the i-axis.

[0057] Perform Hough transform on the regular path P and find the maximum value H in the Hough space max The (ρ,θ) step includes:

[0058] Step S131, discretize the parameters ρ and θ to construct the Hough space H.

[0059] Specifically, the parameter ρ is divided into N equal parts, and θ is divided into M equal parts, and discretized into N×M units to form the Hough space H. The elements of H are accumulator cells, and the initial value of the accumulator cells is set to 0.

[0060] Step S132: Map each point of the regular path P to the Hough space through Hough transform, and find the maximum value H in the Hough space. max (ρ,θ).

[0061] Specifically, take out the point (i, j) in the regular path along P and substitute it into equation (3) and use the quantized θ m ρ n , where m = 1, 2, ..., M; n = 1, 2, ..., N; ; in Hough space, ρ n and θ m The corresponding accumulator cell is added by 1, that is, H(n,m)=H(n,m)+1; after traversing the regular path P, check each accumulator cell in the Hough space to find the maximum value H max (ρ,θ). See Figure 4 , H max (237,-26.5°)=571.

[0062] Step S14, by H max (ρ,θ) finds the matching path with the most collinearity Determine the mileage offset Δ between the query sequence Q and the matching sequence C.

[0063] Step S141, using the point-line duality relationship, H max (ρ,θ) finds the matching path with the most collinearity

[0064] For details, see Figure 5 , by H max (ρ,θ) determined Find the matching paths with the most collinearity like

[0065]

[0066] Step S142, calculate matching path The intercept with the sequence Q or C index axis.

[0067] Specifically, from formula (4), we can get the matching path The intercept δ with the index axis of sequence Q is as shown in formula (6):

[0068]

[0069] That is, in this embodiment, δ 1 =265.015;

[0070] Matching Path The intercept with the index axis of sequence C is as shown in formula (7)

[0071]

[0072] That is, in this embodiment, δ 2 =-530.030;

[0073] Step S143: The mileage offset Δ between the query sequence Q and the matching sequence C can be determined based on the intercept and the sampling step.

[0074] Specifically, in this embodiment, the sampling step s of the query sequence Q is 0.25m, and the sampling step s of the matching sequence C is 0.125m.

[0075] Δ=δ 1 ·s 1 =-δ 2 ·s 2 (7)

[0076] That is, in this embodiment, |Δ|=66.254m.

[0077] This embodiment uses DTW to establish a regular path for matching online dynamic inspection data with historical static inspection data, and uses Hough transform of the regular path to extract the most collinear matching path, and then determines the mileage deviation between the online track inspection data and the historical track inspection data based on the most collinear matching path. Since it estimates the mileage deviation with the most collinear points, it fully utilizes the similarity between the track inspection data in the unmoved track section, effectively avoids the influence of mismatching caused by moving track disturbance or data loss on the estimation accuracy, and can achieve accurate estimation of the mileage deviation between multiple track inspection data under the conditions of moving track disturbance or data loss. It has good robustness and can be used to improve the accuracy of track inspection data analysis.

[0078] The method for determining mileage deviation of track inspection data based on DTW and Hough transform in the second embodiment of the present invention includes steps S21 to S24.

[0079] Step S21, select mileage range [t 0 ,t end ] as the query sequence Q, and the same mileage range [t 0 ,t end ]’s static inspection data with the same name is taken as the matching sequence C.

[0080] In this embodiment, the data of a high-speed railway downlink [t 0 ,t end ] as the query sequence Q. Specifically, in this embodiment, t 0 =K443.1km,t end = K443.6km. Among them, the static inspection data is selected as the left high and low data, which can be collected through existing inspection equipment, such as level 0 track inspection instrument, etc.

[0081] According to the query sequence Q related information, the static inspection historical track inspection data is constructed as the matching sequence C. Among them, through other technical means, a mileage deviation of 20.0m is set between Q and C, and white noise with a standard deviation of 0.2mm is superimposed in sequence C.

[0082] Step S22, using the DTW algorithm to calculate the regularized path P of Q and C.

[0083] The steps include:

[0084] Step S221: Calculate the distance matrix D between the two sequences Q and C.

[0085] Step S222: Using dynamic programming method to find a regular path P in D

[0086] Step S23, perform Hough transform on the regular path P, and find the maximum value H in the Hough space. max (ρ,θ).

[0087] The steps include:

[0088] Step S231, discretize the parameters ρ and θ to construct the Hough space H

[0089] Step S232: Map each point of the regular path P to the Hough space through Hough transform, and find the maximum value H in the Hough space. max (ρ,θ).

[0090] Step S24, by H max (ρ,θ) finds the matching path with the most collinearity Determine the mileage offset Δ between the query sequence Q and the matching sequence C.

[0091] Step S241, using the point-line duality relationship, from H max (ρ,θ) finds the matching path with the most collinearity

[0092] For details, see Figure 6 .

[0093] Step S242, calculate matching path The intercept with the sequence Q or C index axis

[0094] In this embodiment, δ = -158;

[0095] Step S243: The mileage offset Δ between the query sequence Q and the matching sequence C can be determined based on the intercept and the sampling step.

[0096] Specifically, in this embodiment, the sampling step s of the query sequence Q and the matching sequence C are both 0.125 m, and Δ=-19.75 m.

[0097] In the above steps, the regularized path, Hough transform and the calculation method of the mileage offset Δ are basically the same as the regularized path, Hough transform and the calculation method of the mileage offset Δ in the first embodiment, and are not described in detail here.

[0098] In summary, according to the method for determining the mileage deviation of track inspection data based on DTW and Hough transform provided by the present invention, the most collinear matching path is extracted by Hough transform of the regular path, and then the mileage deviation between the online track inspection data and the historical track inspection data is determined based on the most collinear matching path. Since the mileage deviation is estimated by the most collinear points, the influence of mismatching caused by moving track disturbance or data loss on the estimation accuracy is effectively avoided, and the mileage deviation between multiple track inspection data under the conditions of moving track disturbance or data loss can be accurately estimated, which has good robustness and can be used to improve the accuracy of track inspection data analysis.

[0099] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0100] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for determining mileage deviation of track inspection data based on DTW and Hough transform, characterized in that: include: Select the mileage range [t0,t end ] is used as the query sequence Q, and the same mileage range [t0,t end ] as the matching sequence C; Use the DTW algorithm to calculate the regular path P of the query sequence Q and the matching sequence C; Perform Hough transform on the regular path P and find the maximum value H in the Hough space max (ρ,θ); By H max (ρ,θ) finds the matching path with the most collinearity Determine the mileage offset Δ between the query sequence Q and the matching sequence C; The same mileage range [t0,t end ] as the matching sequence C, specifically including: Retrieving a historical track inspection database corresponding to the line name and line type according to the line name and line type detected by the online track inspection data; According to the mileage range [t0,t end ], intercepting one or more groups of historical track inspection data with the same name from the historical track inspection database as a matching sequence C; The DTW algorithm is used to calculate the regular path P of the query sequence Q and the matching sequence C, which includes: Calculate the distance matrix D between the query sequence Q and the matching sequence C; Use dynamic programming method to find regular path P in distance matrix D; Perform Hough transform on the regular path P and find the maximum value H in the Hough space max (ρ,θ), specifically including: Discretize the parameters ρ and θ and construct the Hough space H; Map each point of the regular path P to the Hough space H through Hough transform, and find the maximum value H in the Hough space H max (ρ,θ); By H max (ρ,θ) finds the matching path with the most collinearity Determine the mileage offset Δ between the query sequence Q and the matching sequence C, specifically including: Using the point-line duality relationship, H max (ρ,θ) finds the matching path with the most collinearity Calculate matching paths The intercept with the index axis of the query sequence Q or the matched sequence C; The mileage offset Δ between the query sequence Q and the matching sequence C is determined by the intercept and the sampling step.

2. The method for determining mileage deviation of track inspection data based on DTW and Hough transform according to claim 1, characterized in that: The online track inspection data is collected by a level 0 track inspection instrument or a track inspection vehicle.

Images