PC sleeper soundness evaluation device and PC sleeper soundness evaluation method

Abstract

To provide a PC sleeper soundness evaluation device capable of clearly showing a determination criterion regarding necessity or priority of the replacement of the PC sleeper with cracks.SOLUTION: A PC sleeper soundness evaluation device comprises a first creation unit 11 that creates a second image showing extracted cracks, a second creation unit 12 that creates a third image with the second image binarized, a third creation unit 13 that creates a fourth image with noise removed for the third image, a fourth creation unit 14 that creates a fifth image with a fourth region thinned into a fifth region for the fourth image, a first calculation unit 15 that calculates the crack length on the basis of the number of pixels of multiple second pixels included in the fifth region in the fifth image, a first recording unit 16 that records first data indicating the relation between the crack length and loading capacity of the PC sleeper, and an evaluation unit 17 that evaluates soundness of the PC sleeper on the basis of the first calculation value of the crack length and the first data.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates to a PC sleeper soundness evaluation device and a PC sleeper soundness evaluation method for evaluating the soundness of a PC sleeper. [Background technology]

[0002] PC sleepers, which are made of prestressed concrete (PC), are an important component of track that is essential for increasing railway speeds and improving safety. PC sleepers have higher durability than wooden sleepers. For this reason, the introduction of PC sleepers has been promoted from the perspective of reducing maintenance costs and suppressing track displacement. Some existing PC sleepers still have sufficient strength for their design, even though they were installed more than 50 years ago.

[0003] On the other hand, for example, in track environments where the total weight of each vehicle passing on the track, i.e. the passing tonnage, is relatively large, there are many cases where wear has progressed and where the load-bearing capacity has significantly decreased due to damage that appears to be frost damage. For this reason, there has been an increasing demand in recent years for replacing PC sleepers with PC sleepers, and investigations into the actual conditions of existing PC sleepers and studies on replacement standards are underway. In other words, it is becoming important to efficiently evaluate the soundness of PC sleepers.

[0004] JP 2022-12627 A (Patent Document 1) discloses a technology for evaluating the soundness of PC sleepers, which includes an evaluation process for evaluating the soundness of PC sleepers based on the scaling area of ​​the PC sleepers.

[0005] Japanese Patent Publication No. 2016-89339 (Patent Document 2) discloses a technology in a PC sleeper deterioration determination system that includes a microphone that picks up sound generated by excitation of the PC sleeper, a natural frequency detection unit that detects the natural frequency of the PC sleeper's third-order flexure mode based on the sound picked up by the microphone, and a deterioration determination unit that determines that the PC sleeper has deteriorated if it determines that the natural frequency of the third-order flexure mode detected by the natural frequency detection unit is a lower frequency than a predetermined condition.

[0006] JP 2021-189124 A (Patent Document 3) discloses a technology that includes a modeling process for modeling each of a railway vehicle and a track on which the railway vehicle runs, a time history waveform calculation process for calculating the time history waveform of the bending moment of the sleeper loaded on the track sleeper when the railway vehicle is running by dynamic interaction analysis between the railway vehicle and the track, and a static detailed analysis process for calculating the bending moment of the PC sleeper and the PC steel stress relationship for any PC sleeper by static detailed analysis. The technology described in Patent Document 3 discloses a technology that includes a PC steel stress time history waveform calculation process for calculating the time history waveform of the PC steel stress of the PC sleeper when the railway vehicle is running from the results of the time history waveform calculation process and the static detailed analysis process, and a tensile fatigue strength calculation process for calculating the tensile fatigue strength of the PC steel into the future using the time history waveform of the PC steel stress. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] Patent Publication No. 2022-12627 [Patent Document 2] JP 2016-89339 A [Patent Document 3] JP 2021-189124 A [Non-patent literature]

[0008] [Non-Patent Document 1] V. Badrinarayanan et al., “Segnet: A deep convolutional encoder-decoder architecture for image segmentation”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 39, No. 12, pp. 2481-2495, (2017). [Non-Patent Document 2] OpenCV, “Structural Analysis and Shape Descriptors”, [online], [Retrieved March 15, 2023], Internet〈URL:https: / / docs.opencv.org / 3.4 / d3 / dc0 / group__imgproc__shape.html〉 [Non-Patent Document 3] OpenCV, “Extended Image Processing”, [online], [Retrieved March 15, 2023], Internet〈URL:https: / / docs.opencv.org / 4.x / df / d2d / group__ximgproc.html〉 [Non-Patent Document 4] Shintaro Minoura and 3 others, "Evaluation of the Soundness of PC Sleepers with Vertical Cracks," Proceedings of the Japan Concrete Institute, Vol. 38, No. 2, 2016, pp. 1585-1590 [Non-Patent Document 5] Scikit-image, “Module: registration”, [online], [Retrieved March 15, 2023], Internet〈URL:https: / / scikit-image.org / docs / stable / api / skimage.registration.html#r065f4156fcfd-1〉 Summary of the Invention [Problem to be solved by the invention]

[0009] On the top surface of PC sleepers, cracks were occasionally found to have occurred in the longitudinal direction of the sleepers, but the relationship between the occurrence of these cracks and the performance of the PC sleepers (load-bearing capacity, etc.) was not clear, and the criteria for determining the necessity of replacement or prioritization were not clear. In addition, replacement decisions were made on a qualitative basis through visual inspection at each site, and there were issues with differences in standards depending on the site or inspector, as well as the rationalization of maintenance and work efficiency by smoothing out the replacement process.

[0010] The present invention has been made to solve the problems of the conventional technology as described above, and has an object to provide a PC sleeper soundness evaluation device that, when a crack has occurred in the longitudinal direction of a PC sleeper, can accurately calculate the length of the crack and clearly indicate the judgment criteria for the necessity of replacing PC sleepers or the prioritization of PC sleepers. [Means for solving the problem]

[0011] Among the inventions disclosed in this application, a brief summary of representative inventions is as follows.

[0012] One embodiment of the PC sleeper soundness evaluation device of the present invention is a PC sleeper soundness evaluation device that evaluates the soundness of a PC sleeper that includes an upper surface and has a crack on the upper surface. The PC sleeper soundness evaluation device has a first creation unit that creates a second image in which the cracks are extracted by performing semantic segmentation on a first image in which the upper surface of the PC sleeper is captured, and a second creation unit that creates a third image in which the second image is binarized by performing a binarization process using a first threshold on the second image created by the first creation unit. The PC sleeper soundness evaluation device also has a third creation unit that creates a fourth image from which noise has been removed by dividing a first region including a plurality of first pixels each having a luminance value equal to or greater than a first threshold value into a plurality of second regions each including a plurality of second pixels which are adjacent first pixels among the plurality of first pixels, and performing a noise removal process to set to zero the luminance value of each of the plurality of second pixels included in a third region which is a second region having an area equal to or less than the second threshold value among the divided plurality of second regions; and a fourth creation unit that creates a fifth image from which the fourth region has been thinned into a fifth region by performing a thinning process on a fourth region which is a second region having an area exceeding the second threshold value among the plurality of second regions, in the fourth image created by the third creation unit. In addition, the PC sleeper soundness evaluation device has a first calculation unit that calculates the crack length based on the pixel count of multiple second pixels included in the fifth region in the fifth image created by the fourth creation unit, a first recording unit that records first data indicating the relationship between the crack length and the load-bearing capacity of the PC sleeper, and an evaluation unit that evaluates the soundness of the PC sleeper based on the first calculated value of the crack length calculated by the first calculation unit and the first data recorded by the first recording unit.

[0013] In another aspect, the evaluation unit may evaluate the soundness of the PC sleeper by classifying the soundness of the PC sleeper into one of a plurality of first ranks based on the first calculated value and the first data.

[0014] In another aspect, the PC sleeper soundness evaluation device may have a first acquisition unit that acquires multiple third thresholds for crack length for classifying the soundness of the PC sleeper into multiple first ranks based on the first data. The evaluation unit may evaluate the soundness of the PC sleeper based on the second rank by classifying the soundness of the PC sleeper into a second rank, which is one of the multiple first ranks, based on the first calculated value and the multiple third thresholds acquired by the first acquisition unit.

[0015] In another aspect, the PC sleeper soundness evaluation device may evaluate the soundness of a first PC sleeper, which is a PC sleeper having a first rail fastened to its first top surface by a first rail fastening device. The PC sleeper soundness evaluation device may further include a fifth creation unit that calculates an amount of deviation between a first position of the first rail fastening device in a sixth image in which the first top surface of the first PC sleeper is captured and a second position of the second rail fastening device in a seventh image in which the second top surface of the second PC sleeper is captured, which includes a second top surface and has a second rail fastened to its second top surface by a second rail fastening device, and shifts the sixth image by the calculated amount of deviation to create a first image in which the sixth image is shifted, and a second calculation unit that calculates a first distance from the first rail fastening device of a crack in the width direction of the first rail based on a third position of the fifth region in the fifth image. The first data indicates the relationship between the crack length, the load-bearing capacity, and the first distance, and the evaluation unit may evaluate the soundness of the first PC sleeper based on the first calculated value, the first data, and the second calculated value of the first distance calculated by the second calculation unit.

[0016] In another aspect, the PC sleeper soundness evaluation device may evaluate the soundness of a third PC sleeper, which is a PC sleeper having a third rail fastened to its third upper surface by a third rail fastening device. The PC sleeper soundness evaluation device may further have a first imaging unit that captures a first image of the third top surface from a fourth position that is a first height above the third rail, a second imaging unit that captures an eighth image of a fourth PC sleeper that includes a first part and a second part that are arranged at a distance from each other in a planar view, a first actual distance between the first part and the second part being known, and to which the fourth rail is fastened, captured from a fifth position that is a second height above the fourth rail that is equal to the first height, and a third calculation unit that calculates a first pixel distance, which is the distance in pixels between the first part and the second part, based on the eighth image captured by the second imaging unit, and calculates a first ratio between the first pixel distance and the first actual distance based on the calculated first pixel distance and first actual distance. The first calculation unit may calculate the length of the crack based on the number of the multiple second pixels included in the fifth region in the fifth image and the third calculation value of the first ratio calculated by the third calculation unit.

[0017] In another aspect, the PC sleeper soundness evaluation device may evaluate the soundness of a fifth PC sleeper, which is a PC sleeper having a crack on its top surface among a plurality of PC sleepers arranged along the track. The PC sleeper soundness evaluation device may further include a second recording unit that records second data indicating first information representing a second distance along the track from a reference position on the track to the fifth PC sleeper, and a determination unit that determines whether to replace the fifth PC sleeper based on the soundness of the fifth PC sleeper evaluated by the evaluation unit and the second data recorded by the second recording unit.

[0018] In another aspect, the PC sleeper soundness evaluation device may have a third recording unit that records third data indicating a first parameter indicating the number of years since the PC sleeper was installed or the cumulative tonnage, a fourth recording unit that records fourth data indicating the relationship between the first parameter and the load-bearing capacity, and a second acquisition unit that acquires a plurality of fourth thresholds for the first parameter for classifying the soundness of the PC sleeper into a plurality of third ranks based on the fourth data recorded by the fourth recording unit. The evaluation unit may classify the soundness of the PC sleeper into the second rank based on the first calculated value and the plurality of third thresholds, and may evaluate the soundness of the PC sleeper based on the second rank and the fourth rank by classifying the soundness of the PC sleeper into one of the plurality of third ranks, based on the third data recorded in the third recording unit and the plurality of fourth thresholds acquired by the second acquisition unit.

[0019] One aspect of the present invention is a PC sleeper soundness evaluation method for evaluating the soundness of a PC sleeper that includes an upper surface and has a crack on the upper surface. The PC sleeper soundness evaluation method includes the steps of (a) having a first creation unit perform semantic segmentation on a first image capturing the upper surface of the PC sleeper to create a second image from which the cracks are extracted, and (b) having the second creation unit perform binarization processing using a first threshold on the second image created in step (a) to create a third image in which the second image is binarized. The method for evaluating the soundness of PC sleepers also includes a step (c) of dividing the third image created in step (b) into a plurality of second regions each including a plurality of second pixels which are adjacent first pixels among the plurality of first pixels, and performing a noise removal process by the third creation unit to set to zero the luminance values ​​of the plurality of second pixels included in the third region which is a second region having an area equal to or less than the second threshold among the divided plurality of second regions, thereby creating a fourth image from which noise has been removed; and a step (d) of creating a fifth image by the fourth creation unit by performing a thinning process on the fourth region which is a second region having an area exceeding the second threshold among the plurality of second regions, using the fourth creation unit. In addition, the method for evaluating the soundness of a PC sleeper includes the steps of (e) calculating the length of the crack by a first calculation unit based on the number of second pixels included in the fifth region in the fifth image created in (d), (f) recording first data indicating the relationship between the crack length and the load-bearing capacity of the PC sleeper by a first recording unit, and (g) evaluating the soundness of the PC sleeper by an evaluation unit based on the first calculated value of the crack length calculated in (e) and the first data recorded in (f). Effect of the Invention

[0020] By applying one aspect of the present invention, in a PC sleeper soundness evaluation device that evaluates the soundness of PC sleepers, when a crack occurs in the longitudinal direction of a PC sleeper, it is possible to accurately calculate the length of the crack and clearly indicate the criteria for determining the need to replace or prioritization of PC sleepers. [Brief description of the drawings]

[0021] [Figure 1] FIG. 1 is a perspective view showing a ballasted track in which the soundness of PC sleepers is evaluated by a PC sleeper soundness evaluation device according to an embodiment. [Diagram 2] This is an image of the area around the rail fastening device of the PC sleeper shown in Figure 1. [Diagram 3] 1 is a block diagram showing a configuration of a PC sleeper soundness evaluation device according to an embodiment. FIG. [Figure 4] FIG. 2 is a flow diagram showing some steps of a method for evaluating the soundness of a PC sleeper according to an embodiment. [Diagram 5] FIG. 2 is a flow diagram showing some steps of a method for evaluating the soundness of a PC sleeper according to an embodiment. [Figure 6] 1A to 1C are diagrams showing examples of images created in each step included in the method for evaluating the soundness of PC sleepers according to an embodiment. [Figure 7] 1 is a graph showing the relationship between crack length and load-bearing capacity. [Figure 8] 7A to 7C are diagrams showing other examples of images created in each step included in the method for evaluating the soundness of PC sleepers according to an embodiment. [Figure 9] FIG. 13 is a diagram showing an example of an image in the vicinity of a rail fastening device extracted from a captured image. [Figure 10] FIG. 1 is a diagram showing the location and frequency of crack occurrence. [Figure 11] FIG. 13 shows the positions at which lengths were measured for pixel to actual length conversion. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0022] Hereinafter, each embodiment of the present invention will be described with reference to the drawings.

[0023] Note that the disclosure is merely an example, and for those that can be easily conceived by those skilled in the art with appropriate modifications while maintaining the gist of the invention, they are naturally included in the scope of the present invention. Also, for the purpose of making the description clearer, the drawings may schematically represent the width, thickness, shape, etc. of each part compared to the embodiments, but this is merely an example and does not limit the interpretation of the present invention.

[0024] In addition, in this specification and each figure, the same elements as those described above with respect to the previously shown figures may be given the same reference numerals, and detailed descriptions may be omitted as appropriate.

[0025] (Embodiment) <Method for Evaluating the Soundness of a PC Sleeper by a PC Sleeper Soundness Evaluation Device> First, the method for evaluating the soundness of a PC sleeper by the PC sleeper soundness evaluation device of the embodiment will be described. The PC sleeper soundness evaluation device of this embodiment is a PC sleeper soundness evaluation device that includes an upper surface and evaluates the soundness of a PC sleeper with cracks occurring on the upper surface. Also, the method for evaluating the soundness of a PC sleeper of this embodiment is a method for evaluating the soundness of a PC sleeper by the PC sleeper soundness evaluation device of this embodiment, and is a method for evaluating the soundness of a PC sleeper that includes an upper surface and has cracks occurring on the upper surface.

[0026] FIG. 1 is a perspective view schematically showing a ballast track in which the soundness of a PC sleeper is evaluated by the PC sleeper soundness evaluation device of the embodiment. FIG. 2 is an image of the periphery of the rail fastening device among the PC sleepers shown in FIG. 1. FIG. 2(a) is an image of the periphery of the rail fastening device 6 that presses the bottom 3a of the rail 3 shown in FIG. 1, and FIG. 2(b) is an image of the periphery of the rail fastening device 6 that presses the bottom 3b of the rail 3 shown in FIG. 1.

[0027] As shown in Fig. 1, the PC sleeper soundness evaluation device of this embodiment evaluates the soundness of the PC sleepers 2 included in a ballasted track 5, which includes a ballast track bed 1 installed on a roadbed (not shown), a plurality of PC sleepers 2 arranged at intervals from one another on the ballast track bed 1, and a rail 3 and a rail 4 installed straddling the plurality of PC sleepers 2 and arranged at intervals from one another. As shown in Figs. 1 and 2(a), a bottom 3a of the rail 3 on the rail 4 side is pressed and fastened to the PC sleeper 2 by a pressing part 7 included in a rail fastening device 6 arranged on the rail 4 side from the rail 3, and a bottom 3b of the rail 3 on the opposite side to the rail 4 side is pressed and fastened to the PC sleeper 2 by a pressing part 7 included in a rail fastening device 6 arranged on the opposite side to the rail 4 side from the rail 3.

[0028] In the example shown in Figures 2(a) and 2(b), cracks have occurred in the width direction of the rail 3 (see Figure 1), i.e., in the longitudinal direction of the PC sleeper 2. This type of crack is called a vertical crack, and is one of the typical damage forms of PC sleepers.

[0029] Fig. 3 is a block diagram showing the configuration of a PC sleeper soundness evaluation device according to an embodiment. Fig. 4 and Fig. 5 are flow diagrams showing some steps of a PC sleeper soundness evaluation method according to an embodiment. Fig. 4 mainly shows the flow included in an area RG surrounded by a dashed line in Fig. 5.

[0030] Figure 6 shows examples of images created in each step of the method for evaluating the soundness of PC sleepers according to the embodiment. Figure 6(a) shows the original image, Figure 6(b) shows a crack image, Figure 6(c) shows a binarized image, Figure 6(d) shows an image divided into continuous regions, Figure 6(e) shows an image from which minute regions have been deleted, and Figure 6(f) shows a thinned image.

[0031] Figure 7 is a graph showing the relationship between crack length and load-bearing capacity. Figure 7(a) shows the results of a positive bending test, and Figure 7(b) shows the results of a negative bending test. Figures 7(c) and 7(d) show the results of a plug pull-out test.

[0032] As shown in Fig. 3, the PC sleeper soundness evaluation device 10 of this embodiment has a first preparation unit 11, a second preparation unit 12, a third preparation unit 13, a fourth preparation unit 14, a first calculation unit 15, a first recording unit 16, and an evaluation unit 17. As described above, the PC sleeper soundness evaluation device 10 of this embodiment can use a computer 18, and in this case, the computer 18 can be used as each of the first preparation unit 11, the second preparation unit 12, the third preparation unit 13, the fourth preparation unit 14, the first calculation unit 15, the first recording unit 16, and the evaluation unit 17.

[0033] The first creation unit 11 creates an image (second image) IM2 (see FIG. 6(b)) from which cracks have been extracted by performing semantic segmentation on an image (first image, captured image) IM1 (see FIG. 6(a)) of the top surface of the PC sleeper 2. The image IM1 of the top surface of the PC sleeper 2 can be acquired by capturing an image using an imaging unit such as a camera fixed to a maintenance vehicle or the like. A specific method for performing semantic segmentation will be described later.

[0034] The second creation unit 12 performs a binarization process using a first threshold on the image IM2 created by the first creation unit 11, thereby creating an image (third image) IM3 (see Figure 6 (c)) in which the image IM2 is binarized.

[0035] The third creation unit 13 divides an area (first region) RG1 including a plurality of first pixels each having a brightness value equal to or greater than a first threshold value for the image IM3 created by the second creation unit 12 into a plurality of areas (second regions) RG2 (see FIG. 6(d)) including a plurality of second pixels which are adjacent first pixels among the plurality of first pixels, and performs a noise removal process to set the brightness value of each of the plurality of second pixels included in an area (third region) RG3 which is an area of ​​the divided plurality of areas RG2 that is equal to or less than the second threshold value to 0 (or equal to or less than a predetermined upper limit value), thereby creating a noise-removed image (fourth image) IM4 (see FIG. 6(e)).

[0036] The fourth creation unit 14 performs thinning processing on a region (fourth region) RG4, which is a region RG2 among the multiple regions RG2 having an area exceeding a second threshold, of the image IM4 created by the third creation unit 13, to create an image (fifth image) IM5 (see Figure 6(f)) in which the region RG4 is thinned into a region (fifth region) RG5.

[0037] The first calculation unit 15 calculates the length of the crack based on the number of second pixels included in the region RG5 in the image IM5 created by the fourth creation unit .

[0038] The first recording unit 16 records data (first data) DT1 (see FIG. 5) indicating the relationship between the crack length and the load-bearing capacity of the PC sleeper.

[0039] The evaluation unit 17 evaluates the soundness of the PC sleeper based on the crack length calculation value (first calculation value) CV1 (see Figure 5) calculated by the first calculation unit 15 and the data DT1 recorded by the first recording unit 16.

[0040] The computer 18 used as the PC sleeper soundness evaluation device in this embodiment is composed of a central processing unit (CPU), a RAM (Random Access Memory), a memory unit, a data / command input unit, an image display unit, an output unit, etc.

[0041] Although not shown, the CPU is a part that executes arithmetic operations (addition, subtraction, multiplication and division), logical operations (logical AND, logical OR, negation and exclusive OR, etc.), data comparison, data shifting, etc., on various data. Although not shown, the storage unit has a hard disk device (Hard Disk Drive: HDD) or a ROM (Read Only Memory), etc., and is a part that stores a control program for controlling the CPU and various data used by the CPU. The ROM is generally composed of a semiconductor chip, etc.

[0042] Further, the computer 18 used as the PC sleeper soundness evaluation device of this embodiment executes a program for performing the operation and processing of each step of the PC sleeper soundness evaluation method described below.

[0043] In the method for evaluating the soundness of a PC sleeper using the PC sleeper soundness evaluation device 10 of this embodiment, first, semantic segmentation is performed by the first creation unit 11 on an image (first image) IM1 (see FIG. 6(a)) of the top surface of the PC sleeper 2, so that an image (second image, crack image) IM2 from which cracks have been extracted is created by the first creation unit 11 (step S11 in FIG. 4), as shown in FIG. 6(b). Extracting cracks means performing a process of setting the brightness values ​​of pixels included in areas of image IM1 other than those corresponding to cracks to 0 (or a predetermined upper limit value or less).

[0044] A specific method for performing semantic segmentation can be as follows:

[0045] As described in Figure 1 of the above Non-Patent Document 1, semantic segmentation is one of the deep learning algorithms, which classifies the target image on a pixel-by-pixel basis. Deep learning is one of the machine learning techniques, and is a learning technique that uses a model in which neural networks, which are pattern recognition algorithms modeled on the nerve cells of the human brain, are connected in multiple layers, and has produced great results in various fields such as voice recognition and image recognition. Semantic segmentation makes it possible to extract only pixels that correspond to cracks in an image.

[0046] As described in FIG. 2 of Non-Patent Document 1, semantic segmentation uses a technique called a full-layer convolutional neural network (FCN), which is an improved version of a convolutional neural network (CNN). CNN outputs "what the object in the target image is," whereas FCN outputs "where the object is located in the target image." Semantic segmentation is composed of an encoder that extracts features while downsampling the target image, and a decoder that receives the extracted features and displays the classification results at the original image size. The encoder extracts features while downsampling using FCN, and the decoder performs the reverse process to upsample to the original pixel size and present the classification results in pixel units.

[0047] That is, in step S11 of FIG. 4, by utilizing semantic segmentation and learning an image showing the crack positions, the crack positions can be extracted pixel by pixel.

[0048] Next, the second creation unit 12 performs binarization processing using a first threshold on the image IM2 created in step S11, thereby creating an image (third image) IM3 in which the image IM2 is binarized, as shown in FIG. 6(c) (step S12 in FIG. 4).

[0049] In step S12, a binarization process is performed to extract only the cracked parts, leaving only parts with pixel values ​​equal to or greater than a certain value, as a process necessary for subsequent noise removal and thinning processes. Note that binarization means converting the luminance value of a pixel whose luminance value is between 0 and 1 (gray) into only two colors, 0 (black) or 1 (white), based on a certain threshold value between 0 and 1, or into a first luminance value equal to or greater than 0 and less than a threshold value (first threshold value), or into a second luminance value equal to or greater than a threshold value (first threshold value) and less than or equal to 1.

[0050] Next, the third creation unit 13 executes a division process for dividing the image IM3 created in step S12 into a plurality of regions (second regions) RG2 each including a pixel group made up of a plurality of second pixels, which are a plurality of first pixels adjacent to each other among the plurality of first pixels, as shown in Fig. 6(d) by dividing the region (first region) RG1 including a plurality of first pixels each having a luminance value equal to or greater than the first threshold, preferably 1 (white) (step S13 in Fig. 4). At this time, the region RG1 is divided into a plurality of regions RG2 so that a pixel group made up of a plurality of second pixels adjacent to each other does not straddle two different regions RG2. Furthermore, in this step S13, the third creation unit 13 executes a noise removal process to set the luminance value of each of the multiple second pixels contained in region (third region) RG3, which is region RG2 among the multiple divided regions RG2 having an area equal to or less than the second threshold, to 0 (or a predetermined upper limit value or less), and thereby creates a noise-removed image (fourth image) IM4 by the third creation unit 13, as shown in Figure 6 (e).

[0051] In step S13, the binarized image is divided into connected regions, and minute regions whose areas are equal to or smaller than a certain value are removed to remove erroneously detected minute cracks. The operation of dividing the binarized image into continuous regions can be performed using "connectedComponentsWithStats", a module of "OpenCV", a Python image processing library, as described in Non-Patent Document 2.

[0052] As shown in FIG. 6(b), a method using semantic segmentation can create an image (crack image) in which only the crack position is extracted. However, since the area shown in the crack image has a width, and the width varies depending on the image, the length of the crack cannot be calculated by simply calculating the area. In addition, in the crack image, a part of the ballast or aggregate may be erroneously detected as a microcrack. Therefore, in step S12, the crack image created in step S11 is first subjected to a binarization process that leaves only parts with pixel values ​​above a certain value, and only the crack parts are extracted as shown in FIG. 6(c). Then, step S13 is performed on the image binarized in step S12, where the image is divided into connected areas as shown in FIG. 6(d), and micro areas whose areas are below a certain value are removed, thereby removing the erroneously detected microcracks as shown in FIG. 6(e).

[0053] By performing the above-mentioned noise removal process in step S13, it is possible to remove small cracks that are erroneously detected, and to reliably extract long cracks. However, as a specific embodiment of the noise removal process, a general noise removal process other than the above-mentioned noise removal process in step S13 can also be used, although the effect is less than that of the noise removal process in step S13.

[0054] Next, for the image IM4 created in step S13, the fourth creation unit 14 performs thinning processing on region (fourth region) RG4, which is a region RG2 among the multiple regions RG2 having an area exceeding the second threshold, to create an image (fifth image) IM5 in which region RG4 is thinned into region (fifth region) IM5, as shown in Figure 6(f) (step S14 in Figure 4).

[0055] In step S14, a process is performed to convert the given binary image into a line image with a width of 1 pixel. Here, the crack image extracted by semantic segmentation has non-uniform crack widths, so the length of the crack cannot be calculated by calculating the area as is. Therefore, by performing a thinning process, the effect of the variation in crack width can be removed, making it possible to calculate the crack length.

[0056] As described above, thinning is a process of converting a given binary image into a line image with a width of one pixel. Thinning can be performed using “ximgproc.thinning” of “OpenCV” as described in Non-Patent Document 3, for example.

[0057] Next, the first calculation unit 15 calculates the length of the crack based on the number of second pixels included in the region RG5 in the image IM5 created in step S14 (step S15 in FIG. 4).

[0058] In step S15, by performing thinning in step S14, it is possible to convert the crack position into a line with a width of 1 pixel, and the length of the crack can be found by counting the number of pixels in this state.

[0059] Next, the soundness of the PC sleeper is evaluated and ranked based on the length of the crack (Step S16 in FIG. 4).

[0060] In step S16, first, data DT1 showing the relationship between crack length and the load-bearing capacity of the PC sleeper is recorded by the first recording unit 16 as a quantitative evaluation index showing the relationship between cracks and load-bearing capacity based on various tests (step S17 in Fig. 4). In step S17, as shown in Figs. 7(a), 7(b), 7(c) and 7(d), for example, data DT1 showing the relationship between crack length and the load-bearing capacity of the PC sleeper is recorded by the first recording unit 16 based on the results of a positive bending test, a negative bending test or a plug pull-out test performed on the PC sleeper. That is, as the data DT1, data shown in a graph in which the horizontal axis represents crack length and the vertical axis represents load-bearing capacity (load) can be used, as shown in Figs. 7(a) to 7(d), for example.

[0061] Bending tests of PC sleepers can be carried out in accordance with a method conforming to JIS E 1201. For example, as shown in Figure 3(a) of Non-Patent Document 4, a positive bending test can be carried out, which is a three-point bending test with the rail position as the loading point. Also, as shown in Figure 3(b) of Non-Patent Document 4, a negative bending test can be carried out, which is a three-point bending test with the sleeper turned upside down and the center in the longitudinal direction of the sleeper as the loading point. The loading span can be 700 mm. The bending proof load value is set at a tensile edge stress of 3.0 N / mm2 in concrete with an effective rate of 90% of the initial prestress. 2 The bending failure load is set as three times the stress that the effective prestress generates in the sleeper.

[0062] The results of the positive bending test are shown in Figure 7(a), and the results of the negative bending test are shown in Figure 7(b). Cracks that occur during the bending test are referred to as "bending cracks" or "cracks that occur in the rail direction on the surface of the PC sleeper opposite the loading point." In this case, in Figures 7(a) and 7(b), for each of several PC sleepers that already have longitudinal cracks, the load at which a new "bending crack" is visually determined to have occurred when a load is applied to the loading point is shown as "crack occurrence," and the maximum load at which the PC sleeper breaks in bending and can no longer support the load is shown as "ultimate." Also, although the horizontal axis of the graphs in Figures 7(a) and 7(b) is abbreviated as "crack length in the longitudinal direction of the sleeper" in Figures 7(a) and 7(b), it actually indicates "the length of the longitudinal crack on the top surface of the PC sleeper", and the approximation line is an approximation line that shows the relationship between the length of the longitudinal crack on the top surface of the PC sleeper and the ultimate state (the same is true for Figures 7(c) and 7(d)). Also, in Figures 7(a) and 7(b), the bending proof load specified in JIS is listed as "JIS guaranteed" and the bending failure load specified in JIS is listed as "JIS failure".

[0063] A pull-out test of embedded plugs for PC sleepers can also be carried out in accordance with a method conforming to JIS E 1201. For example, a pull-out test of embedded plugs can be carried out as shown in Figure 4 of Non-Patent Document 4. The allowable pull-out fracture load specified by JIS is 50 kN. Resins such as polyester or nylon are generally used as materials for embedded plugs.

[0064] The results of the embedded plug pull-out test are shown in Figures 7(c) and 7(d). Note that the crack length in Figure 7(d) shows only the crack length on the side where the target embedded plug is located. In Figures 7(c) and 7(d), when the embedded plug pull-out test was conducted on each of several PC sleepers that already had vertical cracks, the maximum load at which the embedded plug breaks and can no longer support the load is shown as "ultimate". Additionally, in Figures 7(c) and 7(d), the allowable pull-out failure load mentioned above is listed as "JIS failure".

[0065] In step S16, the soundness of the PC sleeper is then evaluated by the evaluation unit 17 based on the crack length calculation value (first calculation value) CV1 (see Figure 5) calculated in step S15 and the data DT1 recorded in step S17 (step S18 in Figure 4).

[0066] In this step S16, the length of the crack can be found by counting the number of pixels in the crack area in the crack image after thinning. Also, as described later, if the PC sleeper image has been aligned in advance, it is possible to know the position of the PC sleeper (for example, near the buried plug or near the end) by checking the position (coordinates) in the image where the crack area is located.

[0067] In this way, in the method for evaluating the soundness of a PC sleeper using the PC sleeper soundness evaluation device of this embodiment, an image IM2 is created in which cracks that have occurred on the top surface of the PC sleeper 2 are extracted, image IM3 is created in which image IM2 is binarized, image IM4 is created in which noise is removed from image IM3, and image IM5 is created in which region RG4, which is region RG2 having an area exceeding the second threshold among multiple regions RG2, is thinned into region RG5 in image IM4, the length of the crack is calculated based on the number of multiple second pixels included in region RG5 in image IM5, data DT1 showing the relationship between the crack length and the load-bearing capacity of the PC sleeper is recorded, and the soundness of the PC sleeper is evaluated based on the calculated crack length value CV1 and the data DT1.

[0068] Here, we will explain the problems that arise when evaluating the soundness of PC sleepers.

[0069] On the top surface of PC sleepers, there are occasional cases where cracks have developed in the longitudinal direction of the PC sleeper, i.e. vertical cracks. These are caused by, for example, alkali aggregate reaction or sulfate expansion, and have developed in the longitudinal direction of the PC sleeper.

[0070] If these vertical cracks develop, there is concern that the bending load-bearing capacity of the PC sleepers or the pull-out load-bearing capacity of the embedded plugs of the rail fastening bolts will decrease, making it necessary to replace the sleepers. However, the relationship between the occurrence of these cracks and the performance (load-bearing capacity, etc.) of the PC sleepers was not clear, and the criteria for determining the necessity of replacement or prioritization were not clear. In addition, the decision to replace was made on a qualitative basis through visual inspection at each site, and there were issues with regard to differences in standards depending on the site or inspector, as well as the rationalization of maintenance and work efficiency by smoothing the replacement process.

[0071] The technology described in Patent Document 1 above can evaluate the soundness of PC sleepers that have experienced a phenomenon known as scaling, in which surface concrete peels off due to frost damage. However, although the technology described in Patent Document 1 can evaluate the soundness of PC sleepers based on the area of ​​scaling, when cracks have developed in the longitudinal direction of the PC sleeper, it is difficult to accurately calculate the length of the cracks that have developed and to clearly indicate criteria for determining the necessity of replacing PC sleepers or for prioritizing them.

[0072] The technology described in Patent Document 2 above can evaluate the presence or absence of cracks in PC sleepers and their soundness by measuring their natural frequency. However, although the technology described in Patent Document 2 can detect the natural frequency of the third-mode deflection of PC sleepers based on the sound picked up by a microphone, it is difficult to accurately calculate the length of any cracks that have occurred and clearly indicate the criteria for determining the necessity of replacing PC sleepers or for prioritizing them.

[0073] The technology described in Patent Document 3 above can estimate the fatigue life of PC sleepers. However, although the technology described in Patent Document 3 can calculate the time history waveform of the PC steel stress of a PC sleeper while a railway vehicle is running from the results of the time history waveform calculation process and the static detailed analysis process, and can calculate the future tensile fatigue strength of the PC steel using the time history waveform of the PC steel stress, it is difficult to accurately calculate the length of a crack that has occurred in the longitudinal direction of a PC sleeper, and to clearly indicate the criteria for determining the necessity of replacing PC sleepers or prioritizing them.

[0074] On the other hand, according to this embodiment, an image IM2 is created in which cracks that have occurred on the top surface of the PC sleeper 2 are extracted, image IM3 is created in which image IM2 is binarized, image IM4 is created in which noise is removed from image IM3, and image IM5 is created in which region RG4, which is region RG2 having an area exceeding a second threshold among multiple regions RG2, is thinned into region RG5 in image IM4, the length of the crack is calculated based on the number of multiple second pixels included in region RG5 in image IM5, data DT1 showing the relationship between the crack length and the load-bearing capacity of the PC sleeper is recorded, and the soundness of the PC sleeper is evaluated based on the calculated crack length value CV1 and the data DT1.

[0075] In such cases, the length of the cracks can be accurately calculated by extracting the cracks, binarizing them, removing noise, and thinning them.The load-bearing capacity of the PC sleepers can then be estimated based on the calculated crack length and data DT1 showing the relationship between the crack length and the load-bearing capacity of the PC sleepers, making it possible to evaluate the soundness of the PC sleepers and clearly indicate the criteria for determining the need to replace or prioritizing PC sleepers.

[0076] That is, according to the PC sleeper soundness evaluation method using the PC sleeper soundness evaluation device of this embodiment, the inspection of PC sleepers can be made more efficient by using a device that automatically detects and quantifies longitudinal cracks that have occurred in PC sleepers by image analysis. It is also possible to clarify the relationship between longitudinal cracks in PC sleepers and the load-bearing capacity of PC sleepers and to propose evaluation criteria based on that relationship. Furthermore, a decision on replacement can be automatically made based on the position or length of cracks on PC sleepers detected by the PC sleeper soundness evaluation device of this embodiment and the quantitative evaluation criteria.

[0077] Figure 8 shows other examples of images created in each step of the method for evaluating the soundness of PC sleepers according to the embodiment. Figure 8(a) shows the original image, Figure 8(b) shows a crack image, Figure 8(c) shows an image that has been binarized, divided into continuous regions, and with minute regions deleted, and Figure 8(d) shows an image that has been thinned and with the crack lengths calculated.

[0078] As shown in Fig. 8(b), a small area (area RG3) that was erroneously detected near the ballast can be seen in image IM2, which is a crack image. However, by performing binarization and deleting the small area (area RG3), as shown in Fig. 8(b), it can be seen that the small area (area RG3) that was erroneously detected in image IM2, as shown in Fig. 8(b), has been properly deleted in image IM4, as shown in Fig. 8(c).

[0079] Preferably, in step S17 in Fig. 4, the evaluation unit 17 evaluates the soundness of the PC sleepers by classifying the soundness of the PC sleepers into one of a plurality of first ranks based on the calculated value CV1 and the data DT1. This makes it possible to more clearly indicate the judgment criteria regarding the necessity of replacing PC sleepers or prioritization. A specific method of rank judgment will be described below.

[0080] <Specific method of ranking> Preferably, the soundness evaluation apparatus for the PC mat has a first acquisition unit 19 (see FIG. 3). Further, in step S17 of FIG. 4, the first acquisition unit 19 acquires a plurality of third thresholds regarding the crack length for classifying the soundness of the PC mat into a plurality of first ranks based on the data DT1. Further, in step S17, the evaluation unit 17 classifies the soundness of the PC mat into a second rank, which is any one of the plurality of first ranks, based on the calculated value CV1 (see FIG. 5) and the plurality of third thresholds acquired by the first acquisition unit 19, and evaluates the soundness of the PC mat based on the second rank. Thereby, the criteria for determining the necessity or priority of replacing the PC mat can be shown more clearly.

[0081] That is, in step S17 of FIG. 4, a threshold for rank determination is determined from the investigation results of the PC mat laid at the site, and rank determination is performed by comparing the threshold with the measured value.

[0082] As an example of rank determination, the following examples can be exemplified. Sound: No damage Rank A: The cracks that have occurred are minor damages, and progress observation is to be carried out. Rank B: The cracks that have occurred are damages that may cause problems in the future, and it is necessary to replace the PC mat in a planned manner. Rank C: The cracks that have occurred are damages that will cause problems soon or already do not meet the performance, and immediate replacement is necessary.

[0083] <Alignment of PC mat images> Next, the alignment of the PC mat images will be described. FIG. 9 is a diagram showing an example of an image near the rail fastening device extracted from the captured image. FIG. 9(a) shows the image IM6 of the rail fastening device as a reference image, FIG. 9(b) shows the image IM7 of the rail fastening device as a target image, and FIG. 9(c) shows the image IM1 with the misalignment corrected.

[0084] Preferably, the PC sleeper soundness evaluation device of this embodiment evaluates the soundness of a first PC sleeper 2a (see Fig. 1), which is a PC sleeper 2 having a rail (first rail) 3c (see Fig. 1) fastened to its first upper surface, which is its upper surface, by a first rail fastening device 6a (see Fig. 1).The PC sleeper soundness evaluation device of this embodiment further has a fifth creation unit 21 (see Fig. 3) and a second calculation unit 22 (see Fig. 3).

[0085] The fifth creation unit 21 calculates the amount of deviation between the position (first position) PS1 of the first rail fastening device 6a in the image (sixth image) IM6 in which the first top surface of the first PC sleeper 2a is captured, and the position (second position) PS2 of the second rail fastening device 6b in the image (seventh image) IM7 in which the second top surface of the second PC sleeper 2b (see Figure 1) is captured, which includes a second top surface and has a rail (second rail) 3d (see Figure 1) fastened to the second top surface by the second rail fastening device 6b (see Figure 1), shifts the image IM6 by the calculated amount of deviation, and creates an image IM1 in which the image IM6 has been shifted, i.e., the image IM6 has been aligned with the image IM7 (step S21 in Figure 4).

[0086] 4, step S21 can be performed after step S11 and before step S12, or step S21 can be performed before step S11. Also, when step S21 is performed after step S11 and before step S12, step S11 can be performed again after performing step S21.

[0087] For example, among multiple images taken by an imaging unit consisting of a camera fixed to a maintenance vehicle, in some images the PC sleepers may not be positioned at the center of the image, but may be photographed shifted in the longitudinal direction of the rail. On the other hand, to more accurately estimate the location of cracks, it is desirable for all PC sleepers to be captured in the same position in the image.

[0088] Therefore, preferably, in step S21 in Figure 4, first one image (reference image) is selected to serve as a positional reference, and the amount of deviation between the position of the PC sleeper shown in the other images (target images) and the position of the PC sleeper in the reference image is calculated, and the target image is shifted by that amount of deviation, so that the position of the PC sleeper is made approximately the same in all target images.

[0089] To calculate the amount of deviation in the target image, images corresponding to the vicinity of the rail fastening devices, which have many features, are extracted from each of the reference image and the target image, the cross-correlation function of these images is calculated, and the position where the cross-correlation function is maximum is found. The reason for extracting images corresponding to the vicinity of the fastening devices in advance is that although the surface condition of the PC sleepers or the condition of the ballast varies greatly from image to image, the fastening devices themselves have the same shape and there is little difference in condition, and their shapes (edges) are clear, making it relatively easy to calculate the amount of deviation. Next, the phase is found for the reference image and the target image using FFT, and the cross-correlation function for that phase is found to calculate the amount of deviation. Generally, cross-correlation using phase enables more accurate estimation of the amount of deviation than cross-correlation using brightness values.

[0090] The amount of deviation can be calculated using the “registration.phase_cross_correlation” module included in the “Scikit-image” Python image processing library, as described in Non-Patent Document 5.

[0091] This method makes it possible to accurately determine the location of the crack (whether it is near a bolt, etc.).

[0092] Preferably, in step S16 of FIG. 4, the second calculation unit 22 calculates the distance (first distance) DS1 (see FIG. 6(a)) of the crack from the first rail fastening device 6a in the width direction of the rail 3c based on the position (third position) PS3 (see FIG. 6(f)) of the region RG5 in the image IM5. The data DT1 indicates the relationship between the crack length, the load-bearing capacity, and the distance DS1. The evaluation unit 17 evaluates the soundness of the first PC sleeper 2a based on the calculation value CV1, the data DT1, and the calculation value (second calculation value) CV2 (see FIG. 5) of the distance DS1 calculated by the second calculation unit 22. As the data DT1, for example, data described in multiple graphs corresponding to any one of FIG. 7(a) to FIG. 7(d) at different distances DS1 can be used. Alternatively, instead of the graphs corresponding to Figures 7(a) to 7(d), data DT1 can be used that is data depicted in multiple graphs in which the horizontal axis represents distance DS1 and the vertical axis represents load-bearing capacity (load), for different crack lengths.

[0093] Figure 10 shows the location and frequency of cracks. Figure 10(a) shows images of all PC sleepers averaged after position correction. Figure 10(b) shows the original image (400 x 360 pixels) divided into 5 x 5 pixel segments, with the frequency of cracks in each segment displayed on a 255-point scale, with darker areas indicating a higher frequency of cracks. Figure 10(c) shows an overlay of Figure 10(a) and Figure 10(b). Note that in the example shown in Figure 10, the PC sleepers have been aligned, i.e., their positions corrected, as explained above with reference to Figure 9.

[0094] As shown in Figure 10(a), the shapes of the PC sleeper 2 and the rail fastening device 6 can be clearly seen even in the averaged image, confirming that position correction was performed appropriately. As shown in Figures 10(b) and 10(c), it can be seen that cracks in the PC sleeper 2 occur frequently near the rail fastening device 6, and also near the center of the PC sleeper 2 in the longitudinal direction of the rail 3 (see Figure 1). Also, near the center of the PC sleeper 2 in the longitudinal direction of the rail 3, cracks occur more frequently the closer to the rail fastening device 6 in the width direction of the rail 3.

[0095] Thus, cracks are more likely to occur at positions closer to the rail fastening device in the width direction of the rail on the top surface of the PC sleeper. For this reason, for example, a PC sleeper in which a crack first occurs at a position farther from the rail fastening device is more likely to propagate than a PC sleeper in which a crack first occurs at a position closer to the rail fastening device, and so on. Therefore, even if the crack lengths are equal, differences in the ease of crack propagation may occur. The second calculation unit 22 calculates the distance DS1 (see FIG. 6(a)) of the crack from the first rail fastening device 6a in the width direction of the rail 3c, and the evaluation unit 17 evaluates the soundness of the first PC sleeper 2a based on the calculated value CV2 of the distance DS1 calculated by the second calculation unit 22, thereby making it possible to predict the ease of future crack propagation, and thus making it possible to more clearly indicate the criteria for determining the necessity of replacing PC sleepers or prioritizing them.

[0096] <Scale assignment> Next, the scaling will be described. Fig. 11 is a diagram showing the positions at which the length is measured for conversion between pixels and actual length.

[0097] Preferably, the PC sleeper soundness evaluation device of this embodiment evaluates the soundness of the third PC sleeper 2c (see FIG. 1), which is the PC sleeper 2 having a rail (third rail) 3e (see FIG. 1) fastened to its third upper surface, which is the upper surface, by a third rail fastening device 6c (see FIG. 1). The PC sleeper soundness evaluation device further has a first imaging unit 31 (see FIG. 3), a second imaging unit 32 (see FIG. 3), and a third calculation unit 33 (see FIG. 3). Also, preferably, in the PC sleeper soundness evaluation method using the PC sleeper soundness evaluation device of this embodiment, a scale is applied to estimate the length of a crack (step S31 in FIG. 4).

[0098] As the first imaging unit 31 and the second imaging unit 32, for example, the same camera fixed to a maintenance vehicle or the like or two different cameras can be used.

[0099] First, before performing step S11, the first imaging unit 31 captures an image IM1 of the third upper surface from a position (fourth position) PS4 (see FIG. 1) that is a height (first height) HT1 (see FIG. 1) above the rail 3e (step S32 in FIG. 4).

[0100] Also, in step S31, the second imaging unit 32 captures an image (8th image) IM8 of a fourth PC sleeper 2d (see Figure 1) which includes a part (first part) PR1 and a part (second part) PR2 which are arranged at a distance from each other in a planar view, where the actual distance (first actual distance) DS2 between parts PR1 and PR2 is known and to which a rail (fourth rail) 3f (see Figure 1) is fastened, from a position (fifth position) PS5 (see Figure 1) which is a height (second height) HT2 (see Figure 1) equal to the height HT1 above the rail 3f (step S33 in Figure 4).

[0101] In this step S31, the third calculation unit 33 then calculates a pixel distance (first pixel distance) DS3, which is the distance in pixels between parts PR1 and PR2, based on the image IM8 captured by the second imaging unit 32, and calculates a first ratio between the pixel distance DS3 and the actual distance DS2 based on the calculated pixel distance DS3 and the actual distance DS2 (step S34 in Figure 4).

[0102] After step S31 is performed in this manner, in step S15, the first calculation unit 15 calculates the length of the crack based on the number of second pixels included in area RG5 in image IM5 and the third calculation value of the first ratio calculated by the third calculation unit 33.

[0103] The crack length obtained by the thinning process is in pixels, and needs to be converted to actual length (in mm). Therefore, in step S31 in Fig. 4, the distance between two points in the part of the image whose length is known from a drawing or the like is found in pixels, and the pixel value is converted to actual scale (mm, etc.) from this value and the known length. In other words, since the image itself does not have a scale, when calculating the crack length in actual scale such as mm, it is necessary to separately provide how many mm 1 pixel corresponds to (mm / pix), that is, the scale.

[0104] On the other hand, as mentioned above, for PC sleeper images taken with a camera fixed to a maintenance vehicle, etc., the distance between the camera and the PC sleeper hardly changes, so this (mm / pix) value is almost constant. Therefore, by selecting one representative image and selecting a portion of the image whose length is known (for example, two points: the end of the PC sleeper and the center of the bolt), it is possible to convert the distance in pixels (pix) between those two points and the distance (mm) on the drawing to (mm / pix). In other words, it is possible to provide a scale for estimating the length of the crack.

[0105] Specifically, a large number of, for example, a total of 14 PC pillow blocks are selected, and on the selected images, the length in pixel units at the position shown in FIG. 11 is measured. Then, from the measured values of the lengths in those pixel units and the design values of the PC pillow blocks, the length per pixel (mm / pix) is obtained respectively and averaged. Using the value (the first ratio) thus obtained, the length of the crack obtained in pixel units can be accurately converted into the actual length (mm). Also, in FIG. 11, as an example of the pixel distance DS3 between the partial PR1 and the partial PR2, the width of the shoulder portion of the PC pillow block, the distance from the bolt center to the end of the PC pillow block, and the maximum value of the width of the spacer are illustrated.

[0106] <Position information of the PC pillow block> Next, a method for evaluating the soundness of the PC pillow block based on the position information of the PC pillow block will be described.

[0107] Preferably, the soundness evaluation device for the PC pillow block of the present embodiment evaluates the soundness of the fifth PC pillow block 2e (see FIG. 1), which is a PC pillow block with a crack on its upper surface, among the plurality of PC pillow blocks 2 arranged along the track. The soundness evaluation device for the PC pillow block has a second recording unit 41 (see FIG. 3) and a determination unit 42 (see FIG. 3). A computer 18 can be used as each of the second recording unit 41 and the determination unit 42.

[0108] In step S16 of FIG. 4, the second recording unit 41 records data (second data) DT2 (see FIG. 5) indicating information (first information) representing the distance (second distance, in kilometers) along the track from the reference position on the track to the fifth PC pillow block 2e (see FIG. 1). Note that the data DT2 is recorded in the second recording unit 41 based on, for example, the pillow block ledger 43 (see FIG. 5).

[0109] Also, in step S16 in Fig. 4, the judgment unit 42 judges (screens) whether or not to replace the fifth PC sleeper 2e based on the soundness of the fifth PC sleeper 2e evaluated by the evaluation unit 17 and the data DT2 recorded by the second recording unit 41 (step S19 in Fig. 5). In this step S19, the PC sleeper list and rank, crack information, and crack images can be output as judgment results 44 (see Fig. 5).

[0110] In such a case, the evaluation results obtained by the PC sleeper soundness evaluation device of this embodiment can provide information for formulating a replacement plan or prioritizing by showing statistics for each 100m lot, for example, in addition to the PC sleeper's position (kilometers) information and a list of judgment results. Also, the soundness or necessity of replacement of PC sleepers that have developed vertical cracks can be evaluated automatically and quantitatively, which can contribute to the streamlining of PC sleeper maintenance. Specifically, for example, by waiting to perform replacement work until the number of PC sleepers that need replacement per unit distance exceeds a threshold value, PC sleepers can be replaced collectively, which can contribute to the streamlining of PC sleeper maintenance.

[0111] <Age information or cumulative tonnage information> Next, a method for evaluating the soundness of PC sleepers based on the elapsed age information or cumulative tonnage information of the PC sleepers will be described.

[0112] Preferably, the PC sleeper soundness evaluation device of the present embodiment has a third recording unit 51 (see FIG. 3), a fourth recording unit 52 (see FIG. 3), and a second acquisition unit 53 (see FIG. 3). The computer 18 can be used as each of the third recording unit 51, the fourth recording unit 52, and the second acquisition unit 53.

[0113] In step S16 in Fig. 4, the third recording unit 51 records data (third data) DT3 (see Fig. 5) indicating a parameter (first parameter) representing the number of years elapsed since the installation of the PC sleepers or the cumulative tonnage. The data DT3 is recorded in the third recording unit 51 based on the sleeper register 43, for example.

[0114] In step S16, the fourth recording unit 52 records data (fourth data) DT4 (see Figure 5) indicating the relationship between the parameter (first parameter) and the load-bearing capacity as a quantitative evaluation index showing the relationship between age / tonnage and load-bearing capacity based on various tests and on-site investigations.

[0115] In step S16 of Fig. 4, the second acquiring unit 53 acquires a plurality of fourth thresholds for the parameter (first parameter) for classifying the soundness of the PC sleeper into a plurality of ranks (third ranks) based on the data DT4 recorded by the fourth recording unit 52. For example, data shown in a plurality of graphs corresponding to any of Figs. 7(a) to 7(d), where the number of years elapsed or the cumulative tonnage is different from each other, can be used as the data DT4. Alternatively, data shown in a plurality of graphs, where the horizontal axis represents the number of years elapsed or the cumulative tonnage and the vertical axis represents the load-bearing capacity (load), where the number of crack lengths is different from each other, can be used as the data DT4, instead of the graphs corresponding to Figs. 7(a) to 7(d).

[0116] In addition, in step S16 of Figure 4, the evaluation unit 17 classifies the soundness of the PC sleeper into a second rank based on the calculated value CV1 and a plurality of third thresholds, and classifies the soundness of the PC sleeper into a fourth rank, which is one of the multiple third ranks, based on the data DT3 recorded in the third recording unit 51 and a plurality of fourth thresholds acquired by the second acquisition unit 53, thereby evaluating the soundness of the PC sleeper based on the second rank and the fourth rank.

[0117] In such cases, it is possible to present judgment criteria that take into account data on PC sleepers owned by railway operators, i.e., sleeper register 43, and also consider data on age or cumulative passing tonnage DT3, etc. Also, it is possible to automatically and quantitatively evaluate the soundness or necessity of replacement of PC sleepers that have developed vertical cracks, which can contribute to the streamlining of maintenance and management of PC sleepers.

[0118] The invention made by the inventor has been specifically described above based on the embodiment thereof. However, it goes without saying that the present invention is not limited to the above embodiment, and various modifications can be made without departing from the spirit of the invention.

[0119] Within the scope of the concept of the present invention, a person skilled in the art may conceive of various modifications and alterations, and it is understood that these modifications and alterations also fall within the scope of the present invention.

[0120] For example, any embodiment in which a person skilled in the art has appropriately added or deleted components or modified the design, or added or omitted steps or modified conditions, is included within the scope of the present invention as long as it satisfies the essence of the present invention. [Industrial Applicability]

[0121] INDUSTRIAL APPLICABILITY The present invention is effective when applied to a PC sleeper soundness evaluation device and a PC sleeper soundness evaluation method for evaluating the soundness of PC sleepers. [Explanation of symbols]

[0122] 1 Ballast bed 2 PC sleeper 2a 1st PC sleeper 2b No. 2 PC sleeper 2c 3rd PC sleeper 2d 4th PC sleeper 2e No. 5 PC sleeper 3, 3c~3f, 4 rails 3a, 3b bottom 5. Ballast track 6 Rail fastening device 6a First rail fastening device 6b Second rail fastening device 6c 3rd rail fastening device 7 Pressing part 10. PC sleeper soundness evaluation device 11 First Creation Section 12. Second Creation Section 13 Third Creation Division 14 Fourth Creation Division 15 First Calculation Section 16 First Recording Section 17 Evaluation Section 18 Computers 19 First acquisition part 21 5th Creation Section 22 Second Calculation Section 31 First imaging unit 32 Second imaging unit 33 Third Calculation Section 41 Second Recording Section 42 Judgment section 43 Sleeper Register 44 Judgment result 51 Third Recording Section 52 4th Record Section 53 Second acquisition part CV1, CV2 calculated values DS1 distance DS2 Actual Distance DS3 pixel distance DT1~DT4 data HT1, HT2 Height IM1~IM8 Images PR1, PR2 part PS1~PS5 position RG, RG1~RG5 area

Claims

1. A PC sleeper soundness evaluation device for evaluating the soundness of a PC sleeper that includes an upper surface and has a crack on the upper surface, a first creation unit that creates a second image in which the cracks are extracted by performing semantic segmentation on a first image in which the top surface of the PC sleeper is captured; a second generation unit that performs a binarization process using a first threshold value on the second image generated by the first generation unit to generate a third image obtained by binarizing the second image; a third creation unit that creates a fourth image from which noise has been removed by performing a noise removal process on the third image created by the second creation unit, by dividing a first region including a plurality of first pixels each having a luminance value equal to or greater than the first threshold into a plurality of second regions each including a plurality of second pixels which are a plurality of first pixels adjacent to each other among the plurality of first pixels, and setting the luminance value of each of the plurality of second pixels included in a third region which is a second region having an area equal to or less than the second threshold among the plurality of divided second regions to zero; a fourth creation unit that creates a fifth image in which the fourth region is thinned into a fifth region by performing a thinning process on a fourth region that is a second region having an area exceeding the second threshold value among the plurality of second regions in the fourth image created by the third creation unit; A first calculation unit that calculates the length of the crack based on the number of the second pixels included in the fifth area in the fifth image created by the fourth creation unit; a first recording unit that records first data indicating a relationship between the length of the crack and the load-bearing capacity of the PC sleeper; an evaluation unit that evaluates soundness of the PC sleeper based on a first calculated value of the crack length calculated by the first calculation unit and the first data recorded by the first recording unit; and A PC sleeper soundness evaluation device having the above structure.

2. In the PC sleeper soundness evaluation device according to claim 1, The evaluation unit evaluates the soundness of the PC sleeper by classifying the soundness of the PC sleeper into one of a plurality of first ranks based on the first calculated value and the first data.

3. In the PC sleeper soundness evaluation device according to claim 2, a first acquisition unit that acquires a plurality of third threshold values ​​for the length of the crack for classifying the soundness of the PC sleeper into the plurality of first ranks based on the first data; The evaluation unit classifies the soundness of the PC sleeper into a second rank, which is one of the plurality of first ranks, based on the first calculated value and the plurality of third threshold values ​​acquired by the first acquisition unit, and evaluates the soundness of the PC sleeper based on the second rank.

4. In the PC sleeper soundness evaluation device according to any one of claims 1 to 3, the PC sleeper soundness evaluation device evaluates the soundness of a first PC sleeper, which is the PC sleeper having a first rail fastened to a first upper surface, which is the upper surface, by a first rail fastening device; The PC sleeper soundness evaluation device further includes: a fifth creation unit that calculates an amount of deviation between a first position of the first rail fastening device in a sixth image in which the first top surface of the first PC sleeper is captured and a second position of the second rail fastening device in a seventh image in which the second top surface of a second PC sleeper is captured, the second top surface of the second PC sleeper including a second top surface and to which a second rail is fastened by a second rail fastening device, shifts the sixth image by the calculated amount of deviation, and creates the first image in which the sixth image has been shifted; a second calculation unit that calculates a first distance of the crack from the first rail fastening device in a width direction of the first rail based on a third position of the fifth area in the fifth image; having The first data indicates a relationship between the length of the crack, the load-bearing capacity, and the first distance, The evaluation unit evaluates the soundness of the first PC sleeper based on the first calculated value, the first data, and the second calculated value of the first distance calculated by the second calculation unit.

5. In the PC sleeper soundness evaluation device according to any one of claims 1 to 3, the PC sleeper soundness evaluation device evaluates the soundness of a third PC sleeper, which is the PC sleeper having a third rail fastened to a third upper surface, which is the upper surface, by a third rail fastening device; The PC sleeper soundness evaluation device further includes: a first imaging unit that captures the first image of the third upper surface from a fourth position that is a first height above the third rail; a second imaging unit that captures an eighth image of a fourth PC sleeper including a first portion and a second portion that are spaced apart from each other in a plan view, a first actual distance between the first portion and the second portion being known, and a fourth rail fastened thereto, captured from a fifth position that is a second height above the fourth rail that is equal to the first height; a third calculation unit that calculates a first pixel distance, which is a distance in pixels between the first portion and the second portion, based on the eighth image captured by the second imaging unit, and calculates a first ratio between the first pixel distance and the first actual distance based on the calculated first pixel distance and the first actual distance; having The first calculation unit calculates the length of the crack based on the number of the second pixels included in the fifth area in the fifth image and the third calculation value of the first ratio calculated by the third calculation unit.

6. In the PC sleeper soundness evaluation device according to any one of claims 1 to 3, the PC sleeper soundness evaluation device evaluates the soundness of a fifth PC sleeper, which is the PC sleeper having a crack on the upper surface among the plurality of PC sleepers arranged along a track; The PC sleeper soundness evaluation device further includes: a second recording unit configured to record second data indicating first information that represents a second distance along the track from a reference position on the track to the fifth PC sleeper; a determination unit that determines whether or not to replace the fifth PC sleeper based on the soundness of the fifth PC sleeper evaluated by the evaluation unit and the second data recorded by the second recording unit; and A PC sleeper soundness evaluation device having the above structure.

7. In the PC sleeper soundness evaluation device according to claim 3, A third recording unit that records third data indicating a first parameter that represents the number of years elapsed since the installation of the PC sleeper or the cumulative tonnage; a fourth recording unit configured to record fourth data indicating a relationship between the first parameter and the load-bearing capacity; a second acquisition unit that acquires a plurality of fourth threshold values ​​for the first parameter for classifying the soundness of the PC sleeper into a plurality of third ranks based on the fourth data recorded by the fourth recording unit; having The evaluation unit classifies the soundness of the PC sleeper into the second rank based on the first calculated value and the plurality of third threshold values, and classifies the soundness of the PC sleeper into a fourth rank, which is one of the plurality of third ranks, based on the third data recorded in the third recording unit and the plurality of fourth threshold values ​​acquired by the second acquisition unit, thereby evaluating the soundness of the PC sleeper based on the second rank and the fourth rank.

8. A method for evaluating the soundness of a PC sleeper, the method comprising: (a) performing semantic segmentation on a first image in which the top surface of the PC sleeper is captured, by a first creation unit, to create a second image in which the cracks are extracted, by the first creation unit; (b) performing a binarization process on the second image generated in the step (a) using a first threshold value by a second generation unit, thereby generating a third image obtained by binarizing the second image by the second generation unit; (c) dividing the third image created in step (b) into a plurality of second regions each including a plurality of second pixels that are a plurality of first pixels adjacent to each other among the plurality of first pixels, and performing a noise removal process by a third creation unit to set the luminance value of each of the plurality of second pixels included in a third region that is a second region having an area equal to or less than the second threshold among the plurality of divided second regions to 0, thereby creating a fourth image from which noise has been removed by the third creation unit; (d) performing a thinning process on a fourth region, which is a second region among the plurality of second regions having an area exceeding the second threshold value, in the fourth image created in the (c) step by a fourth creation unit, thereby creating a fifth image in which the fourth region is thinned into a fifth region by the fourth creation unit; (e) calculating the length of the crack by a first calculation unit based on the number of the second pixels included in the fifth region in the fifth image created in the (d) step; (f) recording first data indicating a relationship between the length of the crack and the load-bearing capacity of the PC sleeper by a first recording unit; (g) evaluating the soundness of the PC sleeper by an evaluation unit based on the first calculated value of the crack length calculated in the step (e) and the first data recorded in the step (f); The method for evaluating the soundness of PC sleepers comprises:

Images