Track image recognition device, track inspection device and track inspection vehicle

Abstract

To provide a track image recognition device which improves recognition accuracy of railway track components.SOLUTION: The present disclosure is a track image recognition device which includes an imaging unit configured to acquire three-dimensional images of railroad tracks including rails, sleepers and track materials from above the railroad tracks, and a recognition unit configured to recognize sleepers and track materials in three-dimensional images. The recognition unit is configured to execute a setting process for setting a sleeper detection area in a three-dimensional image, a sleeper detection process for detecting sleepers from a three-dimensional image based on height data in the sleeper detection area and a track material detection process for detecting track materials from a three-dimensional image based on the positions of the detected sleepers.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a track image recognition device, a track inspection device, and a track inspection vehicle. [Background technology]

[0002] The railway track on which railway vehicles run is made up of rails, sleepers, and other track materials such as fasteners, track pads, ballast, etc. To ensure the safe operation of railway vehicles, these elements that make up the railway track must be inspected periodically.

[0003] As such a railway track inspection means, a device that recognizes rails and sleepers using an image of the railway track captured from above is known (see Patent Document 1). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2006-176071 Summary of the Invention [Problem to be solved by the invention]

[0005] The above-mentioned device recognizes the positions of rails and sleepers using shadows created by lighting in the image. However, there is a risk that the recognition accuracy will be reduced due to the influence of light sources other than lighting, such as the sun.

[0006] One aspect of the present disclosure aims to provide a track image recognition device that can improve the recognition accuracy of components of a railway track. [Means for solving the problem]

[0007] One aspect of the present disclosure is a track image recognition device that includes an imaging unit configured to acquire a three-dimensional image of a railway track, including rails, sleepers, and track materials, from above the railway track, and a recognition unit configured to recognize the sleepers and track materials in the three-dimensional image.

[0008] The recognition unit is configured to perform a setting process for setting a sleeper detection area in a three-dimensional image, a sleeper detection process for detecting sleepers from the three-dimensional image based on height data in the sleeper detection area, and a track material detection process for detecting track materials from the three-dimensional image based on the positions of the detected sleepers.

[0009] With this configuration, the fact that the top surfaces of the sleepers are smooth can be utilized to detect the sleepers based on the height data contained in the 3D image, thereby improving the accuracy of recognizing the sleeper positions. Furthermore, by first setting a sleeper detection area that excludes areas where no smooth top surfaces appear, it is possible to improve the accuracy of recognizing sleepers and reduce the load of the detection process. Furthermore, since track materials other than sleepers are detected using the detected sleepers as the reference position, the accuracy of recognizing track materials is improved.

[0010] In one aspect of the present disclosure, the recognition unit may be configured to further perform an extraction process to extract a reference line along the extension direction of the rail from the 3D image, and to set a sleeper detection area based on the distance from the reference line in the setting process. This configuration allows the sleeper detection area to be set excluding the fastening portion between the sleeper and the rail. As a result, the accuracy of recognizing the position of the track material can be improved.

[0011] In one aspect of the present disclosure, the reference line may be a ridge that forms a corner of the bottom of the rail. This configuration allows for accurate detection based on height changes, and the corner of the bottom of the rail, which is less likely to deform due to wear, is used as the reference for setting the sleeper detection area, thereby improving the accuracy of recognizing the position of the track material.

[0012] In one aspect of the present disclosure, the recognition unit may be configured to detect track materials using the reference heights of multiple calculation sections, which are created by dividing the 3D image along a reference line, in the track material detection process. This configuration can improve the accuracy of recognizing track materials based on height data.

[0013] In one aspect of the present disclosure, the recognition unit may be configured to divide the sleeper detection area into a plurality of small areas in the sleeper detection process and detect the sleepers using a representative value of the height data contained in each of the plurality of small areas. With this configuration, it is possible to reduce the load of the detection process while maintaining the sleeper recognition accuracy.

[0014] In one embodiment of the present disclosure, the track material may include at least one of fasteners, track pads, and ballast. This configuration allows for position recognition of major components of the railway track other than the rails and sleepers.

[0015] Another aspect of the present disclosure is a track inspection device including a track image recognition device and an inspection unit configured to inspect track material recognized by the recognition unit. With this configuration, defects in the track material can be determined based on the position of the track material recognized with high accuracy.

[0016] Another aspect of the present disclosure is a track inspection vehicle including a track image recognition device and a mobile body on which the track image recognition device is installed and configured to travel on a railway track. According to this configuration, by running a mobile object on a railway track, it is possible to acquire a three-dimensional image of the railway track and recognize the components. [Brief explanation of the drawings]

[0017] [Figure 1] FIG. 1 is a schematic diagram of a track inspection vehicle according to an embodiment. [Figure 2] FIG. 2 is a schematic diagram of the imaging unit in the track inspection vehicle of FIG. [Figure 3]FIG. 3 is a block diagram showing the configuration of the track inspection device in the track inspection vehicle of FIG. [Figure 4] FIG. 4A is a schematic diagram showing a reference line, and FIG. 4B is a schematic cross-sectional view of a rail. [Figure 5] FIG. 5 is a schematic diagram showing a sleeper detection area. [Figure 6] FIG. 6 is a schematic diagram showing small areas in the sleeper detection area. [Figure 7] FIG. 7 is a schematic diagram showing a calculation interval. [Figure 8] FIG. 8 is a schematic diagram showing a joint portion of a rail. [Figure 9] FIG. 9 is a schematic diagram showing sleepers in a ballast-free section. [Figure 10] FIG. 10 is a flow chart schematically showing the processing executed by the recognition unit of the track inspection device of FIG. DETAILED DESCRIPTION OF THE INVENTION

[0018] Hereinafter, embodiments to which the present disclosure is applied will be described with reference to the drawings. [1. First embodiment] [1-1.Configuration] 1 travels on a railway track R1 to inspect components of the railway track R1. The track inspection vehicle 1 includes a moving body 2 and a track inspection device 3.

[0019] The components of the railway track R1 include rails R2, sleepers R3, and track materials other than the rails R2 and sleepers R3. The track materials include at least one of fasteners, track pads, and ballast.

[0020] <Mobile> The moving body 2 is configured to travel along a rail R2 on a railway track R1 to be inspected.

[0021] The moving body 2 has a plurality of wheels 21 that run along the rail R2, a carriage 22 to which the wheels 21 are attached and on which a track inspection device 3 is installed, a driving device 23 that controls the running and stopping of the moving body 2, and an electric motor (not shown) that drives the wheels 21. An operator operates the driving device 23 while riding on the carriage 22.

[0022] <Track inspection equipment> The track inspection device 3 includes a first imaging unit 31A, a second imaging unit 31B, and an information processing device 32.

[0023] <Image capture unit> The first imaging unit 31A and the second imaging unit 31B are each configured to continuously acquire three-dimensional images of the railway track R1 from above the railway track R1, from the moving object 2 while it is traveling.

[0024] 2, the first imaging unit 31A has a line laser 311 and a 3D camera 312. The line laser 311 irradiates the railway track R1 with an infrared laser in the vertical direction.

[0025] The 3D camera 312 is equipped with a horizontally long area sensor with multiple imaging lines in the scanning direction (i.e., the extension direction of the rail R2). The 3D camera 312 can acquire light-section images, which will be described later, at high speed and high resolution. The 3D camera 312 captures an image of the laser irradiation area of the line laser 311 from behind the moving object 2 in the traveling direction D. Note that the 3D camera 312 may also capture an image of the laser irradiation area of the line laser 311 from ahead of the moving object 2 in the traveling direction D.

[0026] The first imaging unit 31A acquires a plurality of light-section images along the extension direction of the railway track R1 using a line laser 311 and a 3D camera 312. The first imaging unit 31A synthesizes the light-section images to acquire a three-dimensional image including height data (i.e., Z values) in the XY plane. The configuration of the second imaging unit 31B is similar to that of the first imaging unit 31A.

[0027] The first imaging unit 31A acquires a 3D image of an area including the right rail R2 of the railway track R1 while the moving object 2 is traveling. The second imaging unit 31B acquires a 3D image of an area including the left rail R2 of the railway track R1 while the moving object 2 is traveling.

[0028] <Information processing device> The information processing device 32 shown in FIG. 3 is configured by a computer having a processor such as a CPU (Central Processing Unit), a recording medium such as a memory, and input / output devices such as a keyboard and a display.

[0029] The information processing device 32 has a recognition unit 33 and an inspection unit 34. The recognition unit 33 and the inspection unit 34 may each be configured as an independent computer, or the functions of the recognition unit 33 and the inspection unit 34 may each be executed by a single computer.

[0030] The computer constituting the information processing device 32 executes the function of the recognition unit 33 by an image recognition program recorded on a recording medium, and also executes the function of the inspection unit 34 by an inspection program recorded on a recording medium.

[0031] <Recognition part> The recognition unit 33 is configured to recognize rails, sleepers, and track materials in the three-dimensional images acquired by the first imaging unit 31A and the second imaging unit 31B.

[0032] The recognition unit 33, together with the first imaging unit 31A and the second imaging unit 31B, constitutes a track image recognition device 3A. The recognition unit 33 executes extraction processing, setting processing, sleeper detection processing, and track material detection processing.

[0033] (Extraction process) In the extraction process, the recognition unit 33 extracts a reference line along the extension direction (i.e., the longitudinal direction) of the rail from the three-dimensional image.

[0034] Specifically, as shown in Figures 4A and 4B, the recognition unit 33 extracts the ridge lines that form the corners of the bottom portions R21 of both the left and right rails R2 as the reference line S. The bottom portions R21 are the portions of the rails R2 that are placed on the ground and are wider than the head portions R23. Note that Figure 4B shows a cross section of the right rail R2 in Figure 4A.

[0035] In this embodiment, the recognition unit 33 extracts the ridge line forming the corner of the bottom portion R21 on the outer side O in the rail width direction as the reference line S. However, the recognition unit 33 may also extract the ridge line forming the corner of the bottom portion R21 on the inner side I in the rail width direction as the reference line S. Note that the inner side I in the rail width direction refers to the area where the two rails R2 face each other, and the outer side O in the rail width direction refers to the area opposite the inner side I in the rail width direction.

[0036] The ridgeline of the head portion R23 of the rail R2 is easily deformed or displaced due to wear caused by the wheels of the railway vehicle. In addition, the bottom portion R22 connecting the head portion R23 and the bottom portion R21 cannot be imaged from above. Therefore, the corner of the bottom portion R21 is optimal as the reference line S.

[0037] The recognition unit 33 first recognizes the head R23, and then recognizes the corners of the bottom R21 based on the height difference and widthwise distance from the head R23. That is, the recognition unit 33 extracts, as the corners of the bottom R21, edges whose height difference and widthwise distance from the corners of the head R23 are within predetermined ranges. The reference line S is set as a straight line.

[0038] (Setting process) In the setting process, a sleeper detection area is set in the 3D image. The sleeper detection area is an area where sleepers are recognized.

[0039] Specifically, as shown in FIG. 5, the recognition unit 33 sets a first sleeper detection area A11 and a second sleeper detection area A12 for each rail R2 based on the distance from the reference line S extracted in the extraction process.

[0040] The first sleeper detection area A11 is located outside the reference line S in the rail width direction. The first sleeper detection area A11 and the reference line S are separated by a first distance D1. The first distance D1 is a preset value.

[0041] The first distance D1 is set to be equal to or greater than the standard distance in the rail width direction between the outer end of the leaf spring R4 attached to the sleeper R3 and the bottom R21 of the rail R2. In other words, the first sleeper detection area A11 is set to a range that does not include the fastening device between the rail R2 and the sleeper R3.

[0042] The boundary line on the outer side in the rail width direction of the first sleeper detection area A11 is spaced a second distance D2 from the reference line S. The second distance D2 is a preset value. The second distance D2 is set to be equal to or less than the standard distance in the rail width direction between the end of the rail width direction of the sleeper R3 and the bottom R21 of the rail R2. In other words, the first sleeper detection area A11 is set to a range that does not include an area on the outer side in the rail width direction where the sleeper R3 does not exist.

[0043] The second sleeper detection area A12 is located on the inner side in the rail width direction of the reference line S. The second sleeper detection area A12 and the reference line S are separated by a third distance D3. The third distance D3 is a preset value.

[0044] The third distance D3 is the first distance D1 plus the width of the bottom R21 of the rail R2. Therefore, the second sleeper detection area A12, like the first sleeper detection area A11, is set to a range that does not include the fastening device between the rail R2 and the sleeper R3.

[0045] The inner boundary line of the second sleeper detection area A12 in the rail width direction is spaced a fourth distance D4 from the reference line S. The fourth distance D4 is a preset value. The fourth distance D4 is set to be the same as the second distance D2.

[0046] In each of the first sleeper detection area A11 and the second sleeper detection area A12, sleepers R3 and ballast areas are alternately present along the rail longitudinal direction.

[0047] (Sleeper detection process) In the sleeper detection process, the recognition unit 33 detects sleepers from the three-dimensional image based on the height data in the sleeper detection area set in the setting process.

[0048] Specifically, as shown in Figure 6, the recognition unit 33 divides the sleeper detection area (the second sleeper detection area A12 in Figure 6) into multiple small areas SA, and detects sleeper R3 using a representative value of the height data contained in each of the multiple small areas SA.

[0049] The small area SA is, for example, a rectangular area with a length of 30 mm in the X direction (i.e., the rail width direction) and a length of 8 mm in the Y direction (i.e., the rail longitudinal direction, or the direction along the reference line S). For example, if the intervals between height measurement points in the 3D image are 0.5 mm in the X direction and 4 mm in the Y direction, one small area SA will contain 120 pieces of height data.

[0050] The typical width (i.e., length in the Y direction) of the sleeper R3 is, for example, 180 mm to 200 mm. Therefore, if the length of the small area SA in the Y direction is 8 mm, the sleeper R3 includes at least 20 small areas SA in the Y direction.

[0051] The representative value of the height data of the small area SA may be, for example, the standard deviation of the population, which is the variation in the height data. Alternatively, the representative value of the height data may be the mean value, the median value, the mode value, or the like.

[0052] The recognition unit 33 determines that a small area SA in which the representative value of the height data is smaller than a predetermined threshold is a “smooth area.” The threshold is, for example, 1 when the standard deviation of the population is used as the representative value.

[0053] The recognition unit 33 determines that an area where the thus determined "smooth area" continues for a length of at least a lower limit in the X direction is an area where the sleeper R3 exists. The lower limit used for this determination can be, for example, 15. That is, in the above specific example, if the "smooth area" continues for at least 120 mm, the recognition unit 33 determines that this area is part of the sleeper R3.

[0054] In this way, the recognition unit 33 detects the position of the sleeper R3 in the Y direction of the sleeper detection area and the width W1 of the sleeper R3 in the rail longitudinal direction by scanning the "smooth area" in the X direction.

[0055] (Track material detection processing) In the track material detection process, the recognition unit 33 detects track materials from the three-dimensional image based on the positions of the detected sleepers.

[0056] 7, the recognition unit 33 first divides the 3D image into multiple calculation intervals A2 along the reference line S. The width W2 of the calculation interval A2 in the Y direction is approximately equal to the width W1 of the sleeper R3 detected in the sleeper detection process. For example, in the above specific example, the width W2 of the calculation interval A2 is set to 200 mm, which is the width of 50 small areas SA lined up in the Y direction.

[0057] Next, the recognition unit 33 calculates the width center line C of the sleeper R3 for each calculation section A2. The width center line C is a straight line that divides the calculation section A2 in half in the Y direction. Furthermore, the recognition unit 33 sets the representative value of the height data of the reference line S included in one calculation section A2 as the reference height. For example, the median value is used as the representative value of the height data.

[0058] The recognition unit 33 detects track materials (i.e., fasteners, track pads, and ballast) from the 3D image using the detected position of the sleeper R3 and the reference height set for each calculation section A2. The recognition procedure for each track material will be described below.

[0059] As shown in Fig. 8, the fasteners include a leaf spring R4, a fastening bolt R5 that fastens the leaf spring R4 to the sleeper R3, and a fish plate bolt R7 that fastens a fish plate R6 that connects the rails R2 to the rail R2. One fish plate R6 is arranged on each side of the rail R2.

[0060] The fastening bolt R5 is searched for in the calculation section A2 from the reference line S toward the outside and inside in the rail width direction. The recognition unit 33 recognizes the position of the fastening bolt R5 by counting measurement points that are estimated to include the top surface of the fastening bolt R5 by comparing the reference height with height data in the expected area where the fastening bolt R5 is located.

[0061] The fish plate bolt R7 is searched for in the area where the fish plate R6 exists, based on the intersection P between the reference line S and the width center line C. The recognition unit 33 first recognizes the position of the fish plate R6 by comparing the reference height with the height data and counting the measurement points that are estimated to include the fish plate R6.

[0062] Next, the recognition unit 33 recognizes the position of the joint plate bolt R7 by counting measurement points that are estimated to include the joint plate bolt R7 by comparing the reference height with the height data in an area ranging from one end to the other end of the joint plate R6 in the longitudinal direction in the Y direction and within a certain distance from the corner of the bottom R21 of the rail R2 in the X direction.

[0063] The leaf spring R4 is searched for in an estimated region of the leaf spring R4, based on the intersection P of the reference line S and the width center line C. The recognition unit 33 recognizes the position of the leaf spring R4 by counting measurement points estimated to include the leaf spring R4 by comparing the reference height with height data within a region within a certain distance from the fastening bolt R5.

[0064] As shown in Figure 9, when a tie plate R10 is laid under the rail R2, the track pad R8 is laid on the top surface of the sleeper R3. The tie plate R10 is fixed to the top surface of the sleeper R3 by a number of screws R9. The screws R9 are a type of fastener.

[0065] Under normal conditions, the entire rail pad R8 is located below the rail R2 and cannot be seen from above. However, if the rail pad R8 becomes misaligned for some reason, it may protrude from the rail R2 and become exposed.

[0066] Therefore, the recognition unit 33 sets a presence estimation area A3 for the displaced rail pad R8 and searches for the rail pad R8. The presence estimation area A3 is set to an area on both outer sides of the rail R2 in the X direction and on both outer sides of the sleepers R3 in the Y direction. The recognition unit 33 recognizes the displacement of the rail pad R8 laid under the rail R2 by counting measurement points in the presence estimation area A3 that are estimated to include the rail pad R8 by comparing the reference height with height data.

[0067] The screw R9 is searched for in an estimated presence area based on the intersection P of the reference line S and the width center line C. The recognition unit 33 recognizes the position of the screw R9 by counting measurement points estimated to include the screw R9 through a comparison between the reference height and the height data.

[0068] The ballast is laid between multiple sleepers R3 in the longitudinal direction of the rail R2. The ballast is searched for in an estimated presence area based on the intersection P of the reference line S and the width center line C. The recognition unit 33 counts measurement points estimated to contain ballast by comparing the reference height with the height data, thereby recognizing the ballast laying range and height variations.

[0069] <Inspection Department> The inspection unit 34 is configured to inspect the track material recognized by the recognition unit 33.

[0070] Specifically, the inspection unit 34 determines the presence or absence and positional deviation of track materials (i.e., fasteners, track pads, and ballast) using the recognition results of the recognition unit 33, and outputs or records the results. The contents of the determination by the inspection unit 34 include the presence or absence of track materials, loose fasteners, the condition of the ballast, and misalignment of sleepers.

[0071] In determining whether or not a track material is present, the inspection unit 34 determines that the track material is missing if the track material is not recognized at a predetermined assumed position.

[0072] When determining whether a fastener is loose, the inspection unit 34 checks whether the fastening bolt R5 and the screw R9 are loose based on the height of the top surface of the fastening bolt R5 and the screw R9. Specifically, if the height of the top surface of the fastening bolt R5 exceeds a threshold value, it is determined that there is looseness. The same applies to the screw R9. Using a similar procedure, the inspection unit 34 checks whether the fish plate bolt R7 is loose based on the position of the fish plate bolt R7 in the rail width direction.

[0073] In determining the state of the ballast, the inspection unit 34 uses the height distribution of the ballast on the railway track R1 to inspect for defects such as unevenness of the ballast due to ballast flow and the presence of obstacles within the railway track R1.

[0074] To determine whether a sleeper is out of alignment, the inspection unit 34 inspects the spacing and perpendicularity of the sleeper R3 from the position of the sleeper R3. Spacing is determined when the amount of variation from the reference value of the spacing between two adjacent sleepers R3 exceeds a threshold. Perpendicularity is determined when the deviation from the right angle of the angle formed by the reference line S and the width center line C exceeds a threshold.

[0075] [1-2. Processing] An example of the processing executed by the recognition unit 33 will be described below with reference to the flow chart of FIG.

[0076] In this process, the recognition unit 33 first acquires a three-dimensional image from the first imaging unit 31A and the second imaging unit 31B (step S110). After acquiring the three-dimensional image, the recognition unit 33 extracts a reference line in the three-dimensional image (step S120).

[0077] After extracting the reference line, the recognition unit 33 sets a sleeper detection area in the 3D image using the reference line (step S130). After setting the sleeper detection area, the recognition unit 33 detects sleepers in the set sleeper detection area (step S140).

[0078] After detecting the sleepers, the recognition unit 33 detects track materials other than the sleepers from the three-dimensional image based on the positions of the detected sleepers (step S150).

[0079] [1-3.Effects] According to the embodiment described above in detail, the following effects can be obtained. (1a) Taking advantage of the fact that the top surfaces of the sleepers are smooth, sleepers can be detected based on the height data contained in the 3D image, thereby improving the accuracy of recognizing the sleeper positions. Furthermore, by first setting a sleeper detection area that excludes areas where a smooth top surface does not appear, it is possible to improve the accuracy of recognizing sleepers and reduce the load of the detection process. Furthermore, since track materials other than sleepers are detected using the detected sleepers as the reference position, the accuracy of recognizing track materials is improved.

[0080] (1b) By setting the sleeper detection area based on the distance from the reference line along the rail extension direction, it is possible to set the sleeper detection area excluding the fastening part between the sleeper and the rail, thereby improving the accuracy of recognizing the position of the track material.

[0081] (1c) By using the corner of the bottom of the rail as the reference line, it is possible to detect accurately based on the change in height, and the corner of the bottom of the rail, which is less likely to deform due to wear, is used as the setting reference for the sleeper detection area, which promotes the improvement of the accuracy of recognizing the position of the track material.

[0082] (1d) By detecting track materials using the reference heights of multiple calculation sections divided into 3D images along the reference line, the accuracy of recognizing track materials based on height data can be improved.

[0083] (1e) By detecting sleepers using the representative value of the height data contained in each of the multiple small areas into which the sleeper detection area is divided, it is possible to reduce the load of the detection process while maintaining the recognition accuracy of the sleepers.

[0084] (1f) The track materials to be detected include at least one of fasteners, track pads, and ballast, so that the positions of major components of the railway track other than rails and sleepers can be recognized.

[0085] (1g) The inspection unit 34 can determine defects in the track material based on the position of the track material recognized with high accuracy. (1h) By running the moving body 2 on the railway track, it is possible to acquire a three-dimensional image of the railway track and recognize the components.

[0086] 2. Other Embodiments Although the embodiments of the present disclosure have been described above, it goes without saying that the present disclosure is not limited to the above-described embodiments and can take on various forms.

[0087] (2a) In the track image recognition device of the above embodiment, the recognition unit does not necessarily extract the corner of the bottom of the rail as the reference line. For example, the ridge line that forms the corner of the top of the rail may be used as the reference line.

[0088] (2b) In the track image recognition device of the above embodiment, the recognition unit does not necessarily need to set the sleeper detection area using a reference line along the extension direction of the rail. The recognition unit may set the sleeper detection area using, for example, a reference line or reference point other than those described above.

[0089] (2c) In the track image recognition device of the above embodiment, the recognition unit does not necessarily need to detect track materials using the reference height of each calculation section. For example, the recognition unit may detect track materials by referring to a reference value other than the reference height of the calculation section.

[0090] (2d) In the orbit image recognition device of the above embodiment, the recognition unit and the inspection unit do not necessarily have to be installed on the moving body. The moving body only needs to be equipped with at least an imaging unit and a device for recording the captured images. Furthermore, the recognition unit and the inspection unit may be configured to perform their respective processes at a location separate from the moving body using 3D images recorded during travel.

[0091] (2e) The function of one component in the above embodiments may be distributed among multiple components, or the functions of multiple components may be integrated into one component. Also, part of the configuration of the above embodiments may be omitted. Furthermore, at least part of the configuration of the above embodiments may be added to or substituted for the configuration of another of the above embodiments. All aspects included in the technical idea identified by the wording of the claims are embodiments of the present disclosure. [Explanation of symbols]

[0092] 1... Track inspection vehicle, 2... Moving body, 3... Track inspection device, 3A... Track image recognition device, 21...wheel, 22...cart, 23...driving device, 31A...first imaging unit, 31B... second imaging unit, 32... information processing device, 33... recognition unit, 34... inspection unit, 311...Line laser, 312...3D camera, R1...railroad track, R2...rail, R3...sleeper, R4...leaf spring, R5...fastening bolt, R6...Fish plate, R7...Fish plate bolt, R8...Railway pad, R9...Screw, R10...Tie plate, R21...Bottom, R22...Abdomen, R23...Head.

Claims

1. an imaging unit configured to acquire a three-dimensional image of a railway track, including rails, sleepers, and track materials, from above the railway track; a recognition unit configured to recognize the sleepers and the track material in the three-dimensional image; Equipped with The recognition unit an extraction process of extracting a reference line along the extension direction of the rail from the three-dimensional image; a setting process for setting a sleeper detection area in the three-dimensional image based on the distance from the reference line; a sleeper detection process for detecting the sleepers from the three-dimensional image based on height data in the sleeper detection area; a track material detection process for detecting the track material from the three-dimensional image based on the detected positions of the sleepers; configured to perform Orbital image recognition device.

2. The trajectory image recognition device according to claim 1, The reference line is a ridge line that forms a corner of the bottom of the rail.

3. The trajectory image recognition device according to claim 1 or 2, The recognition unit is configured to detect the track material using a reference height for each of a plurality of calculation sections obtained by dividing the three-dimensional image along the reference line in the track material detection process.

4. The orbit image recognition device according to any one of claims 1 to 3, A track image recognition device in which the recognition unit is configured to divide the sleeper detection area into a plurality of small areas in the sleeper detection process and detect the sleepers using a representative value of the height data contained in each of the plurality of small areas.

5. an imaging unit configured to acquire a three-dimensional image of a railway track, including rails, sleepers, and track materials, from above the railway track; a recognition unit configured to recognize the sleepers and the track material in the three-dimensional image; Equipped with The recognition unit a setting process for setting a sleeper detection area in the three-dimensional image; a sleeper detection process for dividing the sleeper detection area into a plurality of small areas and detecting the sleepers from the three-dimensional image using a representative value of height data included in each of the plurality of small areas; a track material detection process for detecting the track material from the three-dimensional image based on the detected positions of the sleepers; configured to perform Orbital image recognition device.

6. The trajectory image recognition device according to any one of claims 1 to 5, The track material includes at least one of fasteners, track pads, and ballast.

7. The trajectory image recognition device according to any one of claims 1 to 6, an inspection unit configured to inspect the track material recognized by the recognition unit; A track inspection device comprising:

8. The trajectory image recognition device according to any one of claims 1 to 6, a moving body on which the track image recognition device is installed and which is configured to travel on the railway track; A track inspection vehicle equipped with

Images