Seamless rail detection method for intelligent rail inspection system

Abstract

The present invention relates to the technical field of rail offset detection, and specifically to a seamless line rail detection method of an intelligent track inspection system. The method comprises: obtaining images of markers on seamless line rails at the current moment and at historical moments, obtaining the longitudinal displacement of the rails based on the position distribution of the target areas in the marker images to be analyzed at the historical moment and the marker images to be analyzed at the current moment; screening reference markers based on the shape similarity between the marker images to be analyzed and the target areas in other marker images and the degree of shape change of the marker target areas to be analyzed; determining the curvature of the rails based on the difference in the number of pixels at different gray levels in different target areas in each reference marker image at the current moment and the acceleration of the train when passing the reference marker; determining the deformation coefficient based on the grayscale confusion degree and the rail curvature, and correcting the longitudinal displacement of the rails based on the shape of the target area. The present invention improves the accuracy of the measurement results of the longitudinal displacement of the rails.

Description

Technical Field

[0001] The present invention relates to the technical field of rail offset detection, and in particular to a seamless line rail detection method of an intelligent track inspection system. Background Art

[0002] The longitudinal displacement of seamless line rails refers to the relative movement of the rails along the line direction, which is usually caused by factors such as temperature changes and train loads. The friction between the wheels and the rails and the longitudinal horizontal thrust cause longitudinal creep. If the longitudinal displacement is too large, it will cause the track geometry to change, affecting the safety and stability of train operation. Therefore, it is necessary to regularly detect the longitudinal displacement of seamless line rails, and if unqualified, they should be repaired immediately.

[0003] At present, the commonly used method for measuring the longitudinal displacement of rails is to set markers on the rails, then use image acquisition equipment to obtain rail images containing markers, and then determine the actual longitudinal displacement of the rails based on the offset of the markers in the image. This method can provide relatively accurate measurement results under ideal conditions. However, in practical applications, this method has some limitations. During the operation of the train, there are two types of contact forces between the wheelset and the rail contact surface: longitudinal contact force and lateral contact force. Under the long-term action of the train load, due to the reciprocating impact of the train, the rails will undergo local lateral displacement and torsion relative to the sleepers. This deformation will affect the markers attached at this position, causing them to deform, resulting in inaccurate longitudinal displacement of the rails originally measured. Summary of the invention

[0004] In order to solve the problem of inaccurate measurement results when measuring the longitudinal displacement of the rail in the existing method, the purpose of the present invention is to provide a seamless line rail detection method for an intelligent track inspection system, and the technical scheme adopted is as follows:

[0005] The present invention provides a seamless line rail detection method for an intelligent track inspection system, the method comprising the following steps:

[0006] Obtaining images of several markers on the seamless line rails at the current moment and at historical moments and the acceleration of each train passing through the markers in the current time period, wherein the markers include markers to be analyzed;

[0007] Based on the position distribution of the target area in the marker image to be analyzed at the historical moment and the marker image to be analyzed at the current moment, the longitudinal displacement of the rail is obtained; based on the shape similarity between the marker image to be analyzed and the target area in other marker images, as well as the degree of shape change between the image of the marker to be analyzed at the historical moment and the image at the current moment, the reference marker of the marker to be analyzed is screened;

[0008] Determine the grayscale confusion degree of each reference marker based on the difference in the number of pixel points at different grayscale levels in different target areas of each reference marker image at the current moment; determine the rail curvature at the location of each reference marker based on the acceleration of the train each time it passes the reference marker; determine the deformation coefficient of the marker to be analyzed by combining all the grayscale confusion degrees and all the rail curvatures;

[0009] The longitudinal displacement is corrected according to the shape distribution of the target area and the deformation coefficient to obtain a corrected longitudinal displacement of the rail.

[0010] Preferably, the obtaining of the longitudinal displacement of the rail based on the position distribution of the target area in the marker image to be analyzed at the historical moment and the marker image to be analyzed at the current moment comprises:

[0011] Calculate the Euclidean distance between the center point of each target area in the marker image to be analyzed at the current moment and the center point of the corresponding target area in the marker image to be analyzed at the historical moment, and record it as the offset distance of each target area;

[0012] Based on the position of the marking point on the seamless rail at the historical moment and the position of the marking point on the seamless rail at the current moment, the actual moving distance of the marking point is obtained; the first Euclidean distance between the coordinates of the marking point in the marker image to be analyzed at the historical moment and the coordinates of the marking point in the marker image to be analyzed at the current moment is calculated; the sum of the first Euclidean distance and the preset adjustment parameter is calculated; wherein the preset adjustment parameter is greater than 0;

[0013] Calculating a first ratio between the actual moving distance and the sum;

[0014] The longitudinal displacement of the rail is obtained according to the first ratio and the average value of the offset distances of all target areas, and the first ratio and the average value of the offset distance are both positively correlated with the longitudinal displacement.

[0015] Preferably, the method of screening reference markers for the marker to be analyzed by combining the shape similarity between the marker image to be analyzed and the target area in other marker images, and the degree of shape change between the image of the marker to be analyzed at a historical moment and the image at a current moment, comprises:

[0016] The shape similarity between each target region in the image of the marker to be analyzed at the current moment and the corresponding target region in the image of the candidate marker at the current moment is calculated respectively, and recorded as the first similarity corresponding to each target region of the marker to be analyzed; the shape similarity between each target region in the image of the marker at the historical moment to be analyzed and the corresponding target region in the image of the candidate marker at the historical moment is calculated respectively, and recorded as the second similarity corresponding to each target region of the marker to be analyzed;

[0017] Calculate the perimeter difference between each target area in the image at the current moment of each marker and the corresponding target area in the image at the historical moment, and record it as the perimeter difference corresponding to each marker;

[0018] Combining the first similarity, the second similarity, and the difference between the perimeter difference corresponding to the marker to be analyzed and the perimeter difference corresponding to the candidate marker, a similarity index between the candidate marker and the marker to be analyzed is obtained;

[0019] Based on the similarity index, it is judged whether the candidate marker is a reference marker for the marker to be analyzed; the candidate marker is any marker except the marker to be analyzed.

[0020] Preferably, judging whether the candidate marker is a reference marker for the marker to be analyzed based on the similarity index comprises:

[0021] If the similarity index is greater than a preset similarity threshold, the candidate marker is determined as a reference marker for the marker to be analyzed.

[0022] Preferably, the step of comprehensively analyzing the difference in the number distribution of pixels at different gray levels in different target areas of each reference marker image at the current moment to determine the grayscale confusion degree of each reference marker includes:

[0023] For any reference marker:

[0024] Divide the first target area into a preset number of equal-sized sub-areas; obtain a uniform factor of the first target area at each gray level according to the difference in the number of pixels at the same gray level between every two sub-areas and the variance of the number of pixels at the same gray level in all sub-areas; the first target area is any target area in any reference marker image at the current moment;

[0025] The grayscale disorder degree of any reference marker is obtained according to the overall distribution of uniformity factors of all target areas in any reference marker image at each grayscale level at the current moment.

[0026] Preferably, obtaining the grayscale disorder degree of any reference marker according to the overall distribution of uniformity factors of all target areas in any reference marker image at each grayscale level at the current moment comprises:

[0027] For any gray level: calculate the average value of the uniformity factors of all target areas in any reference marker image at the current moment at the any gray level, and record it as the uniform value at the any gray level;

[0028] The average value of the uniform values ​​at all gray levels is determined as the grayscale disorder degree of any one of the reference markers.

[0029] Preferably, determining the rail curvature at the location of each reference marker based on the acceleration of the train each time it passes the reference marker comprises:

[0030] For any reference marker:

[0031] The normalized result of the average value of the acceleration of all trains when passing through any reference marker is determined as the rail curvature at the location of any reference marker.

[0032] Preferably, the combining of all the grayscale disorder degrees and all the rail curvatures to determine the deformation coefficient of the marker to be analyzed includes:

[0033] Calculating a second ratio between the grayscale confusion degree of each reference marker and the corresponding rail curvature;

[0034] The normalized result of the average value of the second ratio corresponding to all reference markers is determined as the deformation coefficient of the marker to be analyzed.

[0035] Preferably, the correcting the longitudinal displacement according to the shape distribution of the target area and the deformation coefficient to obtain the corrected longitudinal displacement of the rail comprises:

[0036] A line segment is drawn through the center point of the target area to be evaluated, wherein the two end points of the line segment intersect with the edge line of the target area to be evaluated; the angle between the line segment and the travel direction of the train is recorded as the deformation angle corresponding to the target area to be evaluated; the target area to be evaluated is any target area in the marker image to be analyzed at the current moment;

[0037] Calculate a first average value of deformation angles corresponding to all target areas in the marker image to be analyzed at the current moment;

[0038] If the first average value is less than the preset angle threshold, the difference between the constant 1 and the deformation coefficient of the marker to be analyzed is used as the adjustment weight; if the first average value is greater than or equal to the preset angle threshold, the sum of the constant 1 and the deformation coefficient of the marker to be analyzed is used as the adjustment weight;

[0039] The longitudinal displacement is corrected using the adjustment weight to obtain a corrected rail longitudinal displacement.

[0040] Preferably, the step of correcting the longitudinal displacement by using the adjustment weight to obtain a corrected rail longitudinal displacement comprises:

[0041] The product of the adjustment weight and the longitudinal displacement is used as the corrected rail longitudinal displacement.

[0042] The present invention has at least the following beneficial effects:

[0043] The present invention firstly evaluates the longitudinal displacement of the rail according to the position distribution of the target area in the marker image to be analyzed at the historical moment and the marker image to be analyzed at the current moment, and then screens the reference marker of the marker to be analyzed based on the shape similarity between the marker image to be analyzed and the target area in other marker images and the degree of shape change of the target area in the image of the marker to be analyzed at the historical moment and the image at the current moment. The deformation coefficient of the marker to be analyzed is further determined through the difference in the number distribution of pixels at different gray levels in the target area and the acceleration of the train when passing the reference marker, and the longitudinal displacement of the rail is corrected based on the shape distribution of the target area. The present invention overcomes the error problem caused by marker deformation in the traditional method by introducing the screening of reference markers, the comprehensive analysis of the grayscale confusion degree and the curvature of the rail, thereby solving the measurement error problem caused by local deformation of the rail in the traditional method and improving the accuracy of the measurement result of the longitudinal displacement of the rail. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0045] Figure 1 A flow chart of a seamless line rail detection method of an intelligent track inspection system provided by an embodiment of the present invention;

[0046] Figure 2 A reference marker image provided by an embodiment of the present invention;

[0047] Figure 3 A structural block diagram of a seamless line rail detection system of an intelligent track inspection system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the seamless line rail detection method of the intelligent track inspection system proposed by the present invention is described in detail as follows in combination with the accompanying drawings and preferred embodiments.

[0049] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0050] The specific scheme of the seamless line rail detection method of the intelligent track inspection system provided by the present invention is described in detail below with reference to the accompanying drawings.

[0051] Implementation example of seamless rail detection method for intelligent rail inspection system:

[0052] The specific scenario targeted by this embodiment is: as the train travels on the seamless line rails for a longer time, the rails will shift. When the shift is serious, it will affect the smoothness of the train operation. Therefore, it is necessary to evaluate the shift situation regularly. This embodiment combines several marker images on the seamless line rails at the current moment and historical moments and the acceleration of each train passing the marker in the current time period to correct the longitudinal displacement of the rails, obtain accurate longitudinal displacement, and improve the accuracy of the measurement results of the longitudinal displacement of the rails.

[0053] This embodiment proposes a seamless rail detection method for an intelligent rail inspection system. Figure 1 As shown, the seamless line rail detection method of the intelligent track inspection system of this embodiment includes the following steps:

[0054] Step S1, obtaining images of several markers on the seamless line rails at the current moment and at historical moments and the acceleration of each train passing through the markers in the current time period, wherein the markers include markers to be analyzed.

[0055] First, the image acquisition device is used to collect images of each marker on the seam line rail at the current moment and the historical moment; in this embodiment, the number of markers is 20, so the images of 20 markers at the current moment and the historical moment are collected, that is, two images are collected for each marker, one collected at the historical moment and the other collected at the current moment. These 20 markers include 1 marker to be analyzed. In specific applications, the implementer can set the number of markers according to the specific situation, and set the historical moment according to the specific situation. In this embodiment, the time interval between the historical moment and the current moment is 5 days, and the time period composed of all moments between the historical moment and the current moment is recorded as the current time period. It should be noted that the current time period in this embodiment includes the historical moment and the current moment. And record the acceleration of each train passing each marker in the current time period. It should be noted that the acceleration can be calculated by the train's running speed and running time. The calculation method of the acceleration is a prior art and will not be repeated here.

[0056] So far, this embodiment has obtained images of multiple markers on the seamless line rails at current moments and images at historical moments, and has obtained the acceleration of each train when it passes through a marker in the current time period.

[0057] Step S2, based on the position distribution of the target area in the marker image to be analyzed at the historical moment and the marker image to be analyzed at the current moment, obtain the longitudinal displacement of the rail; combined with the shape similarity of the marker image to be analyzed and the target area in other marker images, as well as the degree of shape change of the target area in the image of the marker to be analyzed at the historical moment and the image at the current moment, screen the reference marker of the marker to be analyzed.

[0058] Considering that the position of the markers on the rail will change when the rail is displaced, this embodiment first makes a preliminary evaluation of the longitudinal displacement of the rail based on the position distribution of the target area in the marker images to be analyzed at the historical moment and the marker images to be analyzed at the current moment.

[0059] Specifically, first, the semantic segmentation network is trained. The semantic segmentation network is used to extract the target area in the image. The trained semantic segmentation network is used to process each image separately to extract multiple target areas in each image. The semantic segmentation network is an existing network and will not be described in detail here. Figure 2 As shown, the figure is a marker image, and each black circular area in the figure is a target area.

[0060] Then, the center point of each target area in each image is obtained respectively, and the Euclidean distance between the center point of each target area in the marker image to be analyzed at the current moment and the center point of the corresponding target area in the marker image to be analyzed at the historical moment is calculated, which is recorded as the offset distance of each target area in the marker image to be analyzed at the current moment; each target area in the marker image to be analyzed at the current moment corresponds to an offset distance.

[0061] A marking point is set near the marker to be analyzed on the seamless rail, and the position of the marking point on the seamless rail at the historical moment and the position of the marking point on the seamless rail at the current moment are collected; based on the position of the marking point on the seamless rail at the historical moment and the position of the marking point on the seamless rail at the current moment, the actual movement distance of the marking point is obtained; the Euclidean distance between the coordinates of the marking point in the image of the marker to be analyzed at the historical moment and the coordinates of the marking point in the image of the marker to be analyzed at the current moment is calculated, recorded as the first Euclidean distance; the sum of the first Euclidean distance and the preset adjustment parameter is calculated; wherein the preset adjustment parameter is greater than 0; in this embodiment, the preset adjustment parameter is 0.01, and in specific applications, the implementer can set it according to specific circumstances. Calculate the first ratio between the actual movement distance and the sum; according to the first ratio and the average value of the offset distance of all target areas, obtain the longitudinal displacement of the rail, and the first ratio and the average value of the offset distance are both positively correlated with the longitudinal displacement.

[0062] Among them, the positive correlation means that the dependent variable will increase as the independent variable increases, and the dependent variable will decrease as the independent variable decreases. It can be an additive relationship, a multiplicative relationship, etc., which is determined by practical applications.

[0063] In this embodiment, a specific calculation formula for the longitudinal displacement is given, and the longitudinal displacement of the rail can be expressed as:

[0064]

[0065] Where, L represents the longitudinal displacement of the rail, Indicates the actual moving distance of the marked point. represents the Euclidean distance between the coordinates of the marker point in the marker image to be analyzed at the historical moment and the coordinates of the marker point in the marker image to be analyzed at the current moment, that is, the first Euclidean distance; M represents the number of target areas in each marker image, represents the center point of the mth target area in the marker image to be analyzed at the current moment, represents the center point of the mth target area in the landmark image to be analyzed at the historical moment, represents the Euclidean distance between the center point of the mth target area in the marker image to be analyzed at the current moment and the center point of the mth target area in the marker image to be analyzed at the historical moment, that is, the offset distance of the mth target area in the marker image to be analyzed at the current moment; Indicates preset adjustment parameters.

[0066] In this embodiment, a preset adjustment parameter is introduced into the calculation formula of the longitudinal displacement in order to prevent the denominator from being 0. In this embodiment, the preset adjustment parameter is 0.01. In specific applications, the implementer can set it according to specific circumstances. The larger the first ratio is, the larger the Euclidean distance between the center point of the target area in the marker image to be analyzed at the current moment and the center point of the corresponding target area in the marker image to be analyzed at the historical moment is, and the larger the longitudinal displacement of the rail is.

[0067] In railway operation scenarios, the complexity of train loads causes rails to not only undergo longitudinal displacement but also lateral displacement. When the rail at the location of the marker undergoes lateral displacement, the marker will deform due to the stress transmitted by the rail, causing the original longitudinal displacement of the rail calculated based on the marker displacement to deviate from the true value and fail to accurately reflect the actual longitudinal displacement of the rail.

[0068] When a marker is deformed, its physical structure will change accordingly. For example, under the action of lateral displacement, the marker may bend, twist, and other deformations. These deformations will cause the flatness of the marker surface to change. The originally flat surface may become uneven after deformation, causing the light reflection to change. When the grayscale distribution on the marker surface is uneven, the deformation coefficient of the marker is large. Therefore, the deformation coefficient of the marker can be converted into the problem of grayscale uniformity on the marker surface. However, due to factors such as the environment, there may be noise points in the image at this time, making it inaccurate to calculate the grayscale uniformity of the marker surface to be analyzed only by the grayscale performance in the marker image to be analyzed. Therefore, it is necessary to first screen out reference markers at other rail positions that have similar deformation performance to the marker to be analyzed, and reflect the grayscale disorder of the marker to be analyzed, that is, the deformation coefficient, through the grayscale distribution of the reference marker. The deformation coefficient is the degree of deformation of the marker, and markers with the same degree of deformation have similar shapes. In this case, the parameter marker with the same deformation coefficient as the marker to be analyzed is the reference marker with the same shape as the marker to be analyzed.

[0069] The more similar the marker shapes are in the marker images of the marker to be analyzed taken before and after a period of time and the marker images of other location markers taken before and after a period of time, and the more similar the degree of change in the shapes of the markers photographed on the train before and after a period of time, it means that the markers at other locations have the same shape expression as the marker to be analyzed.

[0070] Next, this embodiment is described by taking any marker other than the marker to be analyzed as an example, and other markers can be processed by the method provided in this embodiment.

[0071] Specifically, any marker other than the marker to be analyzed is recorded as a candidate marker. The shape similarity between each target area in the image of the marker to be analyzed at the current moment and the corresponding target area in the image of the candidate marker at the current moment is calculated respectively, and recorded as the first similarity corresponding to each target area of ​​the marker to be analyzed; each target area of ​​the marker to be analyzed has a corresponding first similarity. The shape similarity is calculated by the Freeman chain code method. Specifically, the edge pixel point closest to the upper left vertex of the image in the target area is taken as the starting point. In a clockwise direction, the outline of the figure is represented by a series of line segments with a specific direction. The direction of each line segment is digitally encoded to form a Freeman chain code, and the edit distance of the two chain codes is calculated. The negative correlation mapping result of the edit distance is determined as the shape similarity. There are many methods for negative correlation mapping. In this embodiment, the value of an exponential function with a natural constant as the base and a negative edit distance as the exponent is determined as the corresponding shape similarity.

[0072] The shape similarity between each target region in the image of the historical moment of the marker to be analyzed and the corresponding target region in the image of the historical moment of the candidate marker is calculated respectively, and recorded as the second similarity corresponding to each target region of the marker to be analyzed.

[0073] For any marker: the absolute value of the difference between the perimeter of each target area in the image of the marker at the current moment and the corresponding target area in the image at the historical moment is recorded as the perimeter difference corresponding to the marker. Using this method, the perimeter difference corresponding to the marker to be analyzed and the perimeter difference corresponding to the candidate marker can be obtained.

[0074] Next, the similarity index between the candidate marker and the marker to be analyzed is determined by comprehensively considering the first similarity, the second similarity, and the difference between the perimeter difference corresponding to the marker to be analyzed and the perimeter difference corresponding to the candidate marker.

[0075] In this embodiment, a specific calculation formula for the similarity index is given, and the similarity index between the candidate marker and the marker to be analyzed can be expressed as:

[0076]

[0077] Among them, G is the similarity index between the candidate marker and the marker to be analyzed, represents the perimeter of the mth target area in the image of the marker to be analyzed at the current moment, represents the perimeter of the mth target area in the image at the historical moment of the landmark to be analyzed, represents the perimeter of the mth target area in the image of the candidate landmark at the current moment, represents the perimeter of the mth target region in the image at the historical moment of the candidate landmark, represents the average value of the first similarity corresponding to all target regions of the marker to be analyzed, represents the average value of the second similarities corresponding to all target regions of the marker to be analyzed, M represents the number of target regions in any reference marker image, Indicates the absolute value sign, represents the preset adjustment parameter, and norm() represents the normalization function.

[0078] Characterizes the difference in the perimeter of the mth target area in the image of the marker to be analyzed at the current moment and the image of the marker to be analyzed at the historical moment, Characterizes the difference in the perimeter of the mth target area in the image of the candidate marker at the current moment and the image of the candidate marker at the historical moment. It is used to characterize the difference between the degree of shape change of the marker to be analyzed and the degree of shape change of the candidate marker before and after a period of time. The preset adjustment parameter is introduced into the calculation formula of the similarity index in this embodiment to prevent the denominator from being 0. When the difference between the degree of shape change of the marker to be analyzed and the degree of shape change of the candidate marker before and after a period of time is smaller, and the sum of the average value of the first similarity corresponding to all target areas of the marker to be analyzed and the average value of the second similarity corresponding to all target areas of the marker to be analyzed is larger, it means that the shape change of the candidate marker and the marker to be analyzed is more similar, that is, the similarity index between the candidate marker and the marker to be analyzed is larger.

[0079] The larger the similarity index between the candidate marker and the marker to be analyzed, the more similar the shape changes of the candidate marker and the marker to be analyzed are. Therefore, if the similarity index between the candidate marker and the marker to be analyzed is greater than the preset similarity threshold, the candidate marker is determined as a reference marker for the marker to be analyzed. In this embodiment, the preset similarity threshold is 0.5. In specific applications, the implementer can set it according to specific circumstances.

[0080] The above method is used to judge all markers except the marker to be analyzed, and multiple reference markers of the marker to be analyzed are screened out.

[0081] Step S3, comprehensively analyzing the difference in the number distribution of pixels at different gray levels in different target areas in each reference marker image at the current moment, and determining the grayscale confusion degree of each reference marker; determining the curvature of the rail at the location of each reference marker based on the acceleration of the train each time it passes the reference marker; and determining the deformation coefficient of the marker to be analyzed by combining all the grayscale confusion degrees and all the rail curvatures.

[0082] For the grayscale of the pixels in any target area, when the difference in the number of pixels between any two blocks in the grayscale is smaller, and the variance of the number of pixels between blocks is smaller, it means that the grayscale distribution of the target area at the grayscale is more uniform, that is, the grayscale chaos is smaller. Therefore, this embodiment will evaluate the grayscale chaos of each reference marker in combination with the difference in the number of pixel distributions at different grayscales in different target areas of each reference marker image at the current moment.

[0083] Specifically, for any reference marker:

[0084] Any target area in any reference marker image at the current moment is recorded as the first target area, and the first target area is divided into a preset number of equal-sized sub-areas; according to the difference in the number of pixels at the same gray level between every two sub-areas and the variance of the number of pixels at the same gray level in all sub-areas in the first target area, the uniform factor of the first target area at each gray level is obtained. Using this method, the uniform factor of each target area at each gray level in any reference marker image at the current moment can be obtained. In this embodiment, the preset number is 8. In specific applications, the implementer can set the preset number according to the size and shape of the target area. The gray level division can be set according to the specific situation.

[0085] For any gray level: calculate the average value of the uniformity factors of all target areas in any reference marker image at the current moment at the gray level, and record it as the uniform value at the gray level; using this method, the uniform value at each gray level can be obtained. It should be noted that each gray level has a corresponding uniform value. The average value of the uniform values ​​at all gray levels is determined as the gray level disorder degree of any reference marker.

[0086] In this embodiment, a specific calculation formula for the grayscale disorder degree is given, and the grayscale disorder degree of the rth reference marker can be expressed as:

[0087]

[0088] in, represents the grayscale disorder degree of the r-th reference marker, K represents the number of grayscale levels, M represents the number of target areas in the r-th reference marker image at the current moment, and I represents the number of sub-areas into which each target area in the r-th reference marker image at the current moment is divided, that is, the preset number; represents the number of pixels at the kth gray level in the ith sub-region of the mth target region in the rth reference landmark image at the current moment, represents the number of pixels at the kth gray level in the jth sub-region of the mth target region in the rth reference landmark image at the current moment, represents the variance of the number of pixels at the kth gray level in all sub-regions of the mth target region in the rth reference landmark image at the current moment, Indicates the absolute value sign.

[0089] It indicates the difference in the number of pixels between the i-th sub-region and the j-th sub-region at the k-th gray level in the m-th target region. The larger the absolute value, the greater the difference in the number between the two. Represents the uniformity factor of the mth target area at each gray level. In this embodiment, the preset number is 8. In specific applications, the implementer can set it according to specific circumstances. When the difference in the number of pixels at the same gray level in different sub-areas of the target area is smaller and the variance of the number of pixels at the same gray level in all sub-areas of the target area is smaller, it means that the gray uniformity is greater, that is, the gray chaos of the rth reference marker is smaller.

[0090] By using the above method, the grayscale confusion degree of each reference marker can be obtained. The grayscale confusion degree of the reference marker can reflect the deformation coefficient of the current marker. The deformation of the marker is caused by lateral displacement, and the probability of lateral displacement that may occur at different rail positions is different. The displacement characteristics of the rails in the curved section are significantly different from those in the straight section. In addition to the longitudinal temperature force and the longitudinal displacement caused by the train load, the curved rails are also affected by the centrifugal force when the train passes, which may cause a large lateral displacement.

[0091] Therefore, the lateral displacement of reference markers at different positions is affected by train travel to different degrees. The reference markers at the bend are affected by the centrifugal force of the train, making it easier for the position to be lateral displaced. At this time, the grayscale confusion of the reference marker can better reflect the influence of the lateral displacement. Based on the above analysis, the grayscale confusion of each reference marker image is weighted according to the degree to which the lateral displacement of the reference marker is affected by the train travel. From the analysis of the causes of the lateral displacement, the weighted value can better represent the deformation coefficient of the marker to be analyzed.

[0092] When the reference marker is located at a straight section of the rail, the lateral displacement of the reference marker is less affected by the movement of the train. When the reference marker is located at a curve in the rail, the centrifugal force of the train passing through may cause a larger lateral displacement. At the same time, the sharper the curve, the more the lateral displacement of the marker is affected by the movement of the train. Therefore, the degree to which the lateral displacement of the reference marker is affected by the movement of the train can be converted into the curvature of the rail at the location of the reference marker. When the reference marker is located, the smaller the acceleration of the train before and after passing, the smaller the curvature of the rail at that location. The curvature of the rail at each reference marker is determined by the acceleration of the train each time it passes the reference marker.

[0093] Specifically, for any reference marker: based on the acceleration of all trains in the current time period when passing the reference marker, calculate the average value of the acceleration of all trains in the current time period when passing the reference marker, normalize the average value, and use the normalized result as the curvature of the rail at the location of the reference marker. There are many methods for normalizing the average value. This embodiment uses a linear normalization method so that the normalized result is at (0, 1). The linear normalization method is an existing method and will not be described in detail here. Using this method, the curvature of the rail at the location of each reference marker can be obtained.

[0094] Furthermore, the deformation coefficient of the marker to be analyzed is determined by combining all the grayscale confusion degrees and all the rail curvatures. Specifically, for any reference marker: calculate the ratio between the grayscale confusion degree of the reference marker and the rail curvature of the reference marker, and record the ratio as the second ratio. Using this method, multiple second ratios can be obtained, and each reference marker corresponds to a second ratio. The normalized result of the average value of the second ratios corresponding to all reference markers is determined as the deformation coefficient of the marker to be analyzed. This embodiment uses a linear normalization method to process the average value of the second ratio so that the normalized result is at (0, 1).

[0095] So far, the deformation coefficient of the marker to be analyzed is obtained by using the method provided in this embodiment.

[0096] Step S4, correcting the longitudinal displacement according to the shape distribution of the target area and the deformation coefficient to obtain a corrected longitudinal displacement of the rail.

[0097] In step S3, this embodiment obtains the deformation coefficient of the marker to be analyzed, but the direction of deformation is unknown. In this embodiment, the direction of deformation refers to the lateral displacement direction. The lateral displacement of the rail may be convex or concave. When the rail has a convexity, that is, the lateral displacement is positive, the marker is stretched, so that the calculated longitudinal displacement will be too large; when the rail has a concave, that is, the lateral displacement is negative, the marker is squeezed, so that the calculated longitudinal displacement will be too small. Based on the above analysis, this embodiment corrects the longitudinal displacement of the rail in combination with the above characteristics.

[0098] Specifically, any target area in the marker image to be analyzed at the current moment is recorded as the target area to be evaluated, and a line segment is made through the center point of the target area to be evaluated, wherein the two endpoints of the line segment intersect with the edge line of the target area to be evaluated; the angle between the line segment and the train driving direction is recorded as the deformation angle corresponding to the target area to be evaluated; it should be noted that: if the train is traveling to the right, the train driving direction is the horizontal right direction; if the train is traveling to the left, the train driving direction is the horizontal left direction. Using this method, the deformation angle corresponding to each target area in the marker image to be analyzed at the current moment is obtained, and the average value of the deformation angles corresponding to all target areas in the marker image to be analyzed at the current moment is calculated, and the average value is recorded as the first average value; if the first average value is less than the preset angle threshold, the difference between the constant 1 and the deformation coefficient of the marker to be analyzed is used as the adjustment weight; if the first average value is greater than or equal to the preset angle threshold, the sum of the constant 1 and the deformation coefficient of the marker to be analyzed is used as the adjustment weight.

[0099] Next, the longitudinal displacement is corrected using the adjustment weight to obtain a corrected rail longitudinal displacement. The corrected rail longitudinal displacement can be expressed as:

[0100]

[0101] in, represents the corrected longitudinal displacement of the rail, L represents the longitudinal displacement, represents the deformation coefficient of the marker to be analyzed, represents the average value of the deformation angles corresponding to all target areas in the marker image to be analyzed at the current moment, that is, the first average value; Indicates the preset angle threshold.

[0102] In this embodiment, the preset angle threshold is 45 degrees. In specific applications, the implementer can set it according to specific circumstances. Less than When , it means that the lateral displacement direction of the marker to be analyzed is positive, and the calculated longitudinal displacement will be too large. Therefore, the difference between the constant 1 and the deformation coefficient is used as the adjustment weight to correct the longitudinal displacement obtained by the initial calculation; on the contrary, when Greater than or equal to When , the sum of the constant 1 and the deformation coefficient is used as the adjustment weight to correct the longitudinal displacement obtained by the initial calculation.

[0103] So far, this embodiment has completed the correction of the longitudinal displacement of the rail and obtained the accurate longitudinal displacement of the rail. Based on the corrected longitudinal displacement of the rail obtained above, according to the rail displacement monitoring standard, such as the "Railway Track Engineering Construction Quality Acceptance Standard", the standard stipulates the allowable range of longitudinal displacement of the rail under different line conditions, and for railway lines of different speed levels, the standard gives the corresponding displacement limit. By judging whether the corrected longitudinal displacement of the rail exceeds the standard displacement limit, the displacement of the seamless line is monitored.

[0104] This embodiment firstly performs a preliminary evaluation on the longitudinal displacement of the rail based on the position distribution of the target area in the marker image to be analyzed at the historical moment and the marker image to be analyzed at the current moment, and then screens the reference marker of the marker to be analyzed based on the shape similarity between the marker image to be analyzed and the target area in other marker images and the degree of shape change between the image of the marker to be analyzed at the historical moment and the image at the current moment. The deformation coefficient of the marker to be analyzed is further determined based on the difference in the number distribution of pixels at different gray levels in the target area and the acceleration of the train when passing the reference marker, and the longitudinal displacement of the rail is corrected based on the shape distribution of the target area. This embodiment overcomes the error problem caused by marker deformation in traditional methods by introducing the screening of reference markers, the degree of grayscale confusion and the comprehensive analysis of the curvature of the rail, thereby solving the measurement error problem caused by local deformation of the rail in traditional methods and improving the accuracy of the measurement results of the longitudinal displacement of the rail.

[0105] Implementation example of seamless rail inspection system of intelligent rail inspection system:

[0106] like Figure 3 As shown, the figure shows a structural block diagram of a seamless line rail detection system of an intelligent track inspection system, which includes a data acquisition module, a reference marker screening module, a deformation coefficient determination module and a correction module.

[0107] The data acquisition module is used to obtain images of several markers on the seamless line rails at the current moment and at historical moments and the acceleration of each train passing through the markers in the current time period, wherein the markers include markers to be analyzed;

[0108] The reference marker screening module is used to obtain the longitudinal displacement of the rail based on the position distribution of the target area in the marker image to be analyzed at the historical moment and the marker image to be analyzed at the current moment; the reference marker of the marker to be analyzed is screened based on the shape similarity between the marker image to be analyzed and the target area in other marker images, and the degree of shape change between the image of the marker to be analyzed at the historical moment and the image at the current moment;

[0109] The deformation coefficient determination module is used to comprehensively determine the distribution difference of the number of pixels at different gray levels in different target areas of each reference marker image at the current moment, and determine the grayscale confusion degree of each reference marker; determine the rail curvature at the location of each reference marker based on the acceleration of the train each time it passes the reference marker; and determine the deformation coefficient of the marker to be analyzed by combining all the grayscale confusion degrees and all the rail curvatures;

[0110] The correction module is used to correct the longitudinal displacement according to the shape distribution of the target area and the deformation coefficient to obtain the corrected longitudinal displacement of the rail.

[0111] It should be understood that Figure 3 The structural block diagram of a seamless rail detection system of an intelligent track inspection system shown and its modules can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or a dedicated design hardware. Those skilled in the art can understand that the above-mentioned method and system can be implemented using computer executable instructions and / or included in a processor control code, for example, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The system and its modules of this specification can not only be implemented by hardware circuits such as ultra-large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., but can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software (for example, firmware).

[0112] For more details about the above modules, please refer to other places in this manual and will not be repeated here.

[0113] In other embodiments, a seamless track rail detection device of an intelligent track inspection system is also provided, including a memory and a processor. The memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, so that the device executes the seamless track rail detection method of the intelligent track inspection system. The device can be a chip, a component or a module, and the chip may include a connected processor and a memory; wherein the memory is used to store instructions, and when the processor calls and executes the instructions, the chip can execute the seamless track rail detection method of the intelligent track inspection system provided in the above embodiment.

[0114] In other embodiments, a computer program product is also provided. When the computer program product runs on a computer, the computer executes the above-mentioned related steps to implement the seamless line rail detection method of the intelligent track inspection system provided in the above-mentioned embodiment.

[0115] In other embodiments, a computer-readable storage medium is also provided, which stores a computer program code. When the computer program code runs on a computer, the computer executes the above-mentioned related method steps to implement the seamless line rail detection method of the intelligent rail inspection system provided in the above-mentioned embodiment.

[0116] Among them, the provided system, electronic device, computer program product, and computer-readable storage medium are all used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0117] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A seamless rail detection method for an intelligent rail inspection system, characterized in that: The method comprises the following steps: Obtaining images of several markers on the seamless line rails at the current moment and at historical moments and the acceleration of each train passing through the markers in the current time period, wherein the markers include markers to be analyzed; Based on the position distribution of the target area in the marker image to be analyzed at the historical moment and the marker image to be analyzed at the current moment, the longitudinal displacement of the rail is obtained; combined with the shape similarity of the target area in the marker image to be analyzed and other marker images, and the degree of shape change of the target area in the image of the marker to be analyzed at the historical moment and the image at the current moment, the reference marker of the marker to be analyzed is screened, including: The shape similarity between each target region in the image of the marker to be analyzed at the current moment and the corresponding target region in the image of the candidate marker at the current moment is calculated respectively, and recorded as the first similarity corresponding to each target region of the marker to be analyzed; the shape similarity between each target region in the image of the marker at the historical moment to be analyzed and the corresponding target region in the image of the candidate marker at the historical moment is calculated respectively, and recorded as the second similarity corresponding to each target region of the marker to be analyzed; Calculate the perimeter difference between each target area in the image at the current moment of each marker and the corresponding target area in the image at the historical moment, and record it as the perimeter difference corresponding to each marker; The similarity index between the candidate marker and the marker to be analyzed is obtained by comprehensively considering the first similarity, the second similarity, and the difference between the perimeter difference corresponding to the marker to be analyzed and the perimeter difference corresponding to the candidate marker; if the similarity index is greater than a preset similarity threshold, the candidate marker is determined as a reference marker for the marker to be analyzed; the candidate marker is any marker other than the marker to be analyzed; Determine the grayscale confusion degree of each reference marker based on the difference in the number of pixel points at different grayscale levels in different target areas of each reference marker image at the current moment; determine the rail curvature at the location of each reference marker based on the acceleration of the train each time it passes the reference marker; determine the deformation coefficient of the marker to be analyzed by combining all the grayscale confusion degrees and all the rail curvatures; The longitudinal displacement is corrected according to the shape distribution of the target area and the deformation coefficient to obtain a corrected longitudinal displacement of the rail.

2. The seamless track rail detection method of the intelligent track inspection system according to claim 1 is characterized in that: The method of obtaining the longitudinal displacement of the rail based on the position distribution of the target area in the marker image to be analyzed at the historical moment and the marker image to be analyzed at the current moment includes: Calculate the Euclidean distance between the center point of each target area in the marker image to be analyzed at the current moment and the center point of the corresponding target area in the marker image to be analyzed at the historical moment, and record it as the offset distance of each target area; Based on the position of the marking point on the seamless rail at the historical moment and the position of the marking point on the seamless rail at the current moment, the actual moving distance of the marking point is obtained; the first Euclidean distance between the coordinates of the marking point in the marker image to be analyzed at the historical moment and the coordinates of the marking point in the marker image to be analyzed at the current moment is calculated; the sum of the first Euclidean distance and the preset adjustment parameter is calculated; wherein the preset adjustment parameter is greater than 0; Calculating a first ratio between the actual moving distance and the sum; The longitudinal displacement of the rail is obtained according to the first ratio and the average value of the offset distances of all target areas, and the first ratio and the average value of the offset distance are both positively correlated with the longitudinal displacement.

3. The seamless track rail detection method of the intelligent track inspection system according to claim 1, characterized in that: The step of comprehensively analyzing the difference in the number distribution of pixels at different gray levels in different target areas of each reference marker image at the current moment to determine the grayscale confusion degree of each reference marker includes: For any reference marker: Divide the first target area into a preset number of equal-sized sub-areas; obtain a uniform factor of the first target area at each gray level according to the difference in the number of pixels at the same gray level between every two sub-areas and the variance of the number of pixels at the same gray level in all sub-areas; the first target area is any target area in any reference marker image at the current moment; The grayscale disorder degree of any reference marker is obtained according to the overall distribution of uniformity factors of all target areas in any reference marker image at each grayscale level at the current moment.

4. The seamless track rail detection method of the intelligent track inspection system according to claim 3 is characterized in that: The grayscale disorder degree of any reference marker is obtained according to the overall distribution of uniformity factors of all target areas in any reference marker image at each grayscale level at the current moment, including: For any gray level: calculate the average value of the uniformity factors of all target areas in any reference marker image at the current moment at the any gray level, and record it as the uniform value at the any gray level; The average value of the uniform values ​​at all gray levels is determined as the grayscale disorder degree of any one of the reference markers.

5. The seamless track rail detection method of the intelligent track inspection system according to claim 1, characterized in that: The method of determining the rail curvature at the location of each reference marker based on the acceleration of the train each time the train passes the reference marker comprises: For any reference marker: The normalized result of the average value of the acceleration of all trains when passing through any reference marker is determined as the rail curvature at the location of any reference marker.

6. The seamless track rail detection method of the intelligent track inspection system according to claim 1, characterized in that: The step of combining all the grayscale disorder degrees and all the rail curvatures to determine the deformation coefficient of the marker to be analyzed includes: Calculating a second ratio between the grayscale confusion degree of each reference marker and the corresponding rail curvature; The normalized result of the average value of the second ratio corresponding to all reference markers is determined as the deformation coefficient of the marker to be analyzed.

7. The seamless track rail detection method of the intelligent track inspection system according to claim 1, characterized in that: The step of correcting the longitudinal displacement according to the shape distribution of the target area and the deformation coefficient to obtain the corrected longitudinal displacement of the rail includes: A line segment is drawn through the center point of the target area to be evaluated, wherein the two end points of the line segment intersect with the edge line of the target area to be evaluated; the angle between the line segment and the travel direction of the train is recorded as the deformation angle corresponding to the target area to be evaluated; the target area to be evaluated is any target area in the marker image to be analyzed at the current moment; Calculate a first average value of deformation angles corresponding to all target areas in the marker image to be analyzed at the current moment; If the first average value is less than the preset angle threshold, the difference between the constant 1 and the deformation coefficient of the marker to be analyzed is used as the adjustment weight; if the first average value is greater than or equal to the preset angle threshold, the sum of the constant 1 and the deformation coefficient of the marker to be analyzed is used as the adjustment weight; The longitudinal displacement is corrected using the adjustment weight to obtain a corrected rail longitudinal displacement.

8. The seamless track rail detection method of the intelligent track inspection system according to claim 7, characterized in that: The method of correcting the longitudinal displacement by using the adjustment weight to obtain a corrected rail longitudinal displacement comprises: The product of the adjustment weight and the longitudinal displacement is used as the corrected rail longitudinal displacement.

Images