Ballast-condition monitoring system, ballast-condition monitoring device, and ballast-condition monitoring method

Abstract

The purpose of the present invention is to monitor the condition of ballast with reference to the fine particle size of a mass placed as the ballast. This ballast condition monitoring system comprises: an input unit into which imaging data obtained by imaging the surface condition of ballast is input; and a processing unit that executes processing based on the imaging data. The processing unit calculates a statistic on the basis of the imaging data, and determines whether maintenance is necessary on the basis of the statistic.

Description

Ballast condition monitoring system, ballast condition monitoring device, and ballast condition monitoring method

[0001] This disclosure relates to techniques for monitoring the condition of ballast on railroad tracks over which railroad vehicles travel.

[0002] Patent document 1 discloses a system including a laser, a camera, and a processor. The laser emits a beam of light across a railroad track, and the camera records an image of the railroad track where the beam of light is emitted. Patent document 1 describes that an area of ​​interest containing ballast can be analyzed in the image obtained by scanning the railroad track, and the size of the ballast stones can be estimated using the curves in the ballast outline.

[0003] Special Publication No. 2008-505260

[0004] As the ballast becomes finer due to age-related deterioration, the crushed stone laid as ballast is replaced. Therefore, it is desirable to monitor the condition of the ballast based on the fineness of the crushed stone.

[0005] However, the technology disclosed in Patent Document 1 cannot monitor the state of the ballast based on the fineness of the crushed stone.

[0006] Therefore, the present disclosure aims to monitor the condition of ballast based on the fineness of the lumps laid as ballast.

[0007] This ballast condition monitoring system comprises an input unit to which imaging data of the surface condition of the ballast is input, and a processing unit that performs processing based on the imaging data, and the processing unit calculates statistics based on the imaging data and determines whether maintenance is required based on the statistics.

[0008] The ballast condition monitoring device also includes an input unit to which imaging data of the surface condition of the ballast is input, and a processing unit that performs processing based on the imaging data, and the processing unit calculates statistical quantities based on the imaging data.

[0009] In addition, the ballast condition monitoring method is a ballast condition monitoring method that images the surface condition of the ballast, calculates statistics based on the imaging data of the surface condition of the ballast, and determines whether maintenance is required based on the statistics.

[0010] According to the present disclosure, the condition of the ballast can be monitored based on the fineness of the lumps laid as ballast.

[0011] FIG. 1 is a block diagram showing the overall configuration of a ballast state monitoring system according to a first embodiment. FIG. 2 is a block diagram showing a ballast state monitoring device. FIG. 3 is an explanatory diagram showing an example of the state of ballast in an area adjacent to the first rail. FIG. 4 is an explanatory diagram showing another example of the state of ballast in an area adjacent to the first rail. FIG. 5 is a flowchart showing an example of processing by a statistical calculation device. FIG. 6 is an image in which the state of the ballast shown in FIG. 3 has been enhanced in contrast. FIG. 7 is an image in which the state of the ballast shown in FIG. 4 has been enhanced in contrast. FIG. 8 is a flowchart showing an example of processing by a determination processing device. FIG. 9 is a diagram showing an example of a ballast state image. FIG. 10 is an explanatory diagram showing an example of a trained model according to a second embodiment. FIG. 11 is an explanatory diagram showing a state during training of a learning model. FIG. 12 is a block diagram showing a machine learning device according to a third embodiment. FIG. 13 is a diagram showing an example of a trained decision tree model. FIG. 14 is a block diagram showing a determination processing device according to a fourth embodiment. FIG. 15 is a flowchart showing an example of processing by a determination processing device.

[0012] 1 is a block diagram showing the overall configuration of a ballast state monitoring system 30. The ballast state monitoring system 30 includes a ballast state monitoring device 30a, a ballast state monitoring apparatus 30b, and a ballast state monitoring method 30c.

[0013] An example of a railway track 10 to be monitored by the system 30 will now be described. The railway track 10 is a road that guides a railway vehicle 20 along a predetermined route. In this example, the railway track 10 includes a first rail 12a and a second rail 12b. The two rails 12a, 12b are fixed on ballast 16 via sleepers 13.

[0014] The ballast 16 is a trackbed that supports the rails 12a, 12b. The ballast 16 includes a plurality of blocks 17 laid on the roadbed. The blocks 17 are, for example, crushed rocks or gravel. The roadbed may be the surface of the land, the lower surface of a tunnel, or the upper surface of a bridge or viaduct. Sleepers 13 are placed on the ballast 16. The sleepers 13 are rectangular parallelepiped members that are interposed between the ballast 16 and the two rails 12a, 12b and support the rails 12a, 12b. That is, a plurality of sleepers 13 are arranged on the ballast 16 in a parallel orientation with a gap between them in the extension direction of the rails 12a, 12b. The two rails 12a, 12b are arranged on the sleepers 13 with a gap between them in the extension direction of the sleepers 13 and in an orientation perpendicular to the sleepers 13. The rails 12a and 12b are fixed to the sleepers 13 by fasteners such as bucklers.

[0015] The railway vehicle 20 includes a carbody 22 and a bogie 24. The bogie 24 includes a bogie frame 25 and a plurality of wheels 25W. The plurality of wheels 25W are rotatably supported on left and right portions of the bogie frame 25 via axles. The portions supporting the axles are sometimes called axle boxes. In this embodiment, the forward direction of the railway vehicle 20 is sometimes referred to as the front side, and the backward direction is sometimes referred to as the rear side. Also, left and right may be referenced based on the forward direction as viewed from the railway vehicle 20. The side to which gravity is applied in the direction of gravity may be sometimes referred to as the lower side, and the opposite side may be sometimes referred to as the upper side. The left and right wheels 25W run on two rails 12a, 12b while being guided by the rails 12a, 12b, respectively. The bogie 24 supports the carbody 22 from below. As the bogie 24 runs on the railway track 10, the railway vehicle 20, including the carbody 22, runs along the railway track 10. The railway vehicle 20 may be any vehicle that travels on the railway track 10, including an electric train, a freight train locomotive, a freight car, a passenger train locomotive, or a passenger car. The freight car or passenger car may be a trailer pulled by a locomotive, or may be a powered vehicle that has its own power. The locomotive may be an electric locomotive or an internal combustion engine such as a diesel locomotive. The railway vehicle 20 may be a commercial vehicle for transporting people or cargo, or may be a business vehicle for monitoring the condition of the railway track. The railway vehicle 20 may also be a road-rail vehicle that can travel on both railway tracks and roads.

[0016] The ballast 16 has the function of dispersing vibrations generated when the railway vehicle 20 passes over the rails 12 a, 12 b to the roadbed, and also has the function of improving drainage and preventing the growth of weeds.

[0017] It is conceivable that the lumps 17 come into contact with each other as the railway vehicle 20 travels, causing the lumps 17 contained in the ballast 16 to break down into smaller particles. As the lumps 17 become finer, they lose the functions of dispersing vibrations, draining water, and preventing weed growth. As the lumps 17 become finer, the laid lumps 17 are replaced. To determine whether replacement is necessary, it is conceivable to visually observe the fineness of the lumps 17. In this case, an inspector must visit the site where the ballast 16 is laid to check the fineness of the lumps 17, which increases the workload for inspection. Furthermore, visual inspections tend to be qualitative, which may lead to variations in the judgment of whether replacement is necessary.

[0018] This embodiment relates to a technique for easily and quantitatively determining the fineness of the lumps 17 laid as ballast 16.

[0019] As shown in FIG. 1 , the ballast condition monitoring system 30 is a system for monitoring the condition of a mass 17 laid on ballast 16 in a railway track 10 , and includes a ballast condition monitoring device 40 and a judgment processing device 70 .

[0020] The ballast condition monitoring device 40 is a device that calculates statistics for monitoring the condition of the ballast 16. In this embodiment, the ballast condition monitoring device 40 includes an imaging device 42, a running position detection unit 44, a statistics calculation device 50, and a communication device 46. The ballast condition monitoring device 40 is incorporated into the railway vehicle 20.

[0021] The imaging device 42 is a device that images the surface condition of the ballast while the railway track 10 is in motion and acquires imaging data. The running position detection unit 44 detects the running position of the railway track 10 on the railway track 10. The outputs of the imaging device 42 and the running position detection unit 44 are input to the statistical quantity calculation device 50. The statistical quantity calculation device 50 calculates statistics based on the imaging data obtained by imaging the surface condition of the ballast 16 on the railway track 10. The calculated statistical quantities are output to the communication device 46 as data associated with the position on the railway track 10. The communication device 46 includes a communication circuit and transmits data that associates the statistical quantities with the position on the railway track 10.

[0022] In this embodiment, an example will be described in which the imaging data is two-dimensional image data obtained by a two-dimensional imaging device. If the imaging data is two-dimensional data, the amount of calculation can be reduced compared to when processing is performed based on three-dimensional data.

[0023] The imaging data may be one-dimensional image data obtained by a line sensor camera, or may be three-dimensional data obtained by a stereoscopic camera.

[0024] The ballast state monitoring device 40 and the determination processing device 70 are communicatively connected to the data server 90. For example, the ballast state monitoring device 40 is communicatively connected to the data server 90 via the communication network 38 and the communication device 76. Furthermore, the determination processing device 70 is communicatively connected to the data server 90. The determination processing device 70 and the data server 90 may each have a communication device capable of communicating via the communication network 38. In this case, communication between the ballast state monitoring device 40, the determination processing device 70, and the data server 90 may be performed via the communication network 38. The determination processing device 70 may also be communicatively connected directly to the data server 90 via a wired system. The data server 90 is a computer equipped with a storage device 92. The data server 90 may also be a cloud server. The statistical quantities, which are the calculation results, are transmitted to the data server 90 via the communication network 38 as data associated with the traveling position of the railway vehicle 20. As a result, collected data 92a, in which the statistical quantities are associated with positions on the railway track 10, is stored in the storage device 92 of the data server 90.

[0025] The data server 90 may store data transmitted from a plurality of railway vehicles 20. The data server 90 can comprehensively collect statistics for the railway track 10 by collecting data transmitted from a plurality of railway vehicles 20. When a plurality of statistics for a specific railway track 10 are collected, the collected statistics may be successively updated with the latest data.

[0026] The determination processing device 70 downloads data in which statistics are associated with positions on the railway track 10 from the data server 90, and determines whether maintenance is required based on the statistics.

[0027] In this embodiment, the determination processing device 70 is provided at any location other than the railway track 10. For example, the determination processing device 70 is provided at a management base 28 installed on the ground to monitor the railway track 10. The determination processing device 70 may also be provided at a base for performing maintenance work. The determination processing device 70 is implemented in a portable terminal device and may be carried by a worker. The determination processing of the determination processing device may be implemented on a data server 90. If the data server 90 is a cloud server, the determination processing function of the determination processing device may be implemented on the cloud server.

[0028] The communication network 38 may be wired or wireless, or may be a combination of these. The communication network 38 may also be a public communication network or a communication network using dedicated lines. Data associating statistics with positions on the railway track 10 may also be transmitted directly from the ballast condition monitoring device 40 to the determination processing device 70. In this case, the data server 90 may be omitted.

[0029] FIG. 2 is a block diagram showing the ballast condition monitoring device 40.

[0030] The imaging device 42 is supported on the railway vehicle 20 and captures an image of the surface condition of the ballast 16 at the traveling position of the railway vehicle 20 while the railway vehicle 20 is traveling. The imaging device 42 is supported on the railway track 10 at a position where it can capture an image of the surface condition of at least a portion of the ballast 16 between the first rail 12a and the second rail 12b. For example, the imaging device 42 is supported on the lower part of the car body 22. The imaging device 42 is located between the left and right wheels 25W in the vehicle width direction. The imaging area captured by the imaging device 42 includes at least a portion of the area between the first rail 12a and the second rail 12b. For example, the imaging device 42 captures an image of the first rail 12a and the area of ​​the ballast 16 extending on both sides of the first rail 12a. The area extending adjacent to the second rail 12b with respect to the first rail 12a is the area between the rails 12a and 12b. The imaging device 42 may be supported on the bogie 24.

[0031] 3 and 4 are explanatory diagrams showing examples of the state of the ballast 16 in the area adjacent to the first rail 12a.

[0032] As shown in Figure 3, ballast 16 is laid on the roadbed, and multiple lumps 17 are present in the area next to the first rail 12a. Immediately after laying, the lumps 17 are large enough to disperse vibrations caused by the passage of railway vehicles 20, improve drainage, and prevent the growth of weeds. As railway vehicles 20 repeatedly travel, the lumps 17 become finer due to contact with each other, and the size of the lumps 17 decreases as shown in Figure 4.

[0033] The imaging device 42 captures, for example, images of the state of the ballast 16 in an area adjacent to the first rail 12a. The differences in the imaging data before and after the granulation of the ballast 16 are as follows: The imaging data before granulation captures lumps 17 that are larger than the lumps 17 after granulation. The imaging data after granulation captures lumps 17 that are smaller than the lumps 17 before granulation. Therefore, the imaging data after granulation captures more of the boundaries of the lumps 17 than the imaging data before granulation. It is expected that the boundaries often appear as dark lines. Furthermore, recessed valleys are observed at the boundaries of the lumps 17. It is expected that these valleys appear as shadows. Therefore, the imaging data after granulation captures more dark areas than the imaging data before granulation. Therefore, by calculating statistics that are affected by the total area of ​​the darkened areas for the imaging data, the granularity of the ballast 16 can be estimated based on the statistics, and thereby it is possible to determine whether or not maintenance of the ballast 16 is necessary based on the statistics.

[0034] As shown in FIG. 2 , the running position detection unit 44 detects a state for identifying the running position of the railcar 20 while the railcar 20 is traveling. The running position of the railcar 20 is the position of the railcar 20 in the longitudinal direction of the railroad track 10. The running position of the railcar 20 may be a position (e.g., kilometers) based on a fixed position in the longitudinal direction of the railroad track 10 (e.g., the starting point of the track or any station), or may be a position based on an arbitrary position in the longitudinal direction of the railroad track 10. For example, the running position detection unit 44 may include a rotation speed detection sensor that detects the number of rotations of the wheels and output a running distance from any position based on the detection result of the rotation speed detection sensor. A sensor that detects the vehicle speed based on the number of rotations of the railcar 20 is sometimes called a tachometer generator. Because the running distance can be determined by integrating the speed, the running position detection unit 44, which includes a rotation speed detection sensor, may output the speed at regular intervals.

[0035] Furthermore, for example, the running position detection unit 44 may include a GPS (Global Positioning System) receiving unit in a GNSS (Global Navigation Satellite System), and output latitude and longitude information determined by a signal received by the GPS receiving unit, or a position in the longitudinal direction of the railway track 10 based on the latitude and longitude information.

[0036] The statistical calculation device 50 is configured by a computer including a processor 52 such as a CPU, a storage device 54, an input / output interface 56, etc. The outputs from the imaging device 42 and the traveling position detection unit 44 are input to the input / output interface 56 which includes an input / output circuit.

[0037] The processor 52 includes an arithmetic circuit. The processor 52 is an example of a processing unit that calculates statistics based on image data of the surface condition of the ballast 16. The processor 52 is also an example of a processing unit that generates data that associates the statistics with position data on the railway track 10. The storage device 54 is configured with a non-volatile storage device such as an HDD (hard disk drive) or an SSD (solid-state drive). The storage device 54 stores, for example, a program 54a and data 54b that associates statistical data with position data.

[0038] The program 54a describes the processing that causes the processor 52 to function as a processing unit. Therefore, the processor 52 executes the processing described in the program 54a stored in the storage device 54 or the like, thereby performing the processing as a processing unit that calculates statistics. For example, the processor 52 performs the function of a statistical quantity calculation unit 52a that calculates statistics. There may be one processor 52 or multiple processors 52. The multiple processors 52 may be incorporated into one computer. The multiple processors 52 may be incorporated into multiple computers, and the multiple computers may perform the processing as processing units that calculate statistics in a distributed manner.

[0039] The data 54b stored in the storage device 54 is data that associates a statistical quantity calculated based on the imaging data with the position of the railway vehicle 20 that captured an image of the state of the ballast 16 that corresponds to the statistical quantity.

[0040] An example of processing by the processing unit in the statistical quantity calculation device 50 will be described with reference to the flowchart shown in FIG.

[0041] In step S1, it is determined whether the railway vehicle 20 has traveled a determination distance based on the output from the travel position detection unit 44. The determination distance is a preset value as the interval for calculating statistics. For example, assume that the determination distance is set to d (m). In this case, if the travel distance obtained by subtracting the initial position or the travel position at which the previous statistics were calculated from the current travel position of the railway vehicle 20 is less than d (m), the determination is NO. If the travel distance exceeds d (m), the determination is YES. When the travel distance of the railway vehicle 20 is d (m), the determination may be either YES or NO. The determination distance d (m) may be, for example, 0.5 (m), 1 (m), 2 (m), 5 (m), or a longer value such as 10 (m) or 50 (m). If the determination is NO in step S1, the process of step S1 is repeated. If the determination is YES, the process proceeds to the next step S2.

[0042] In step S2, image data of the surface shape of the ballast 16 is acquired from the image capturing device 42.

[0043] In steps S1 and S2, the imaging device 42 may operate to acquire imaging data for each determined distance traveled, based on the output of the traveling position detection unit 44. In this case, the operations from step S3 onward may be performed when the imaging data of the imaging device 42 is input to the statistics calculation device 50 for each determined distance traveled.

[0044] In the next step S3, a target region R is cut out from the image data. The target region R is preferably a region of the image data excluding the non-ballast region Rx.

[0045] Fig. 6 is a high-contrast image of the state of the ballast shown in Fig. 3. Fig. 7 is a high-contrast image of the state of the ballast shown in Fig. 4.

[0046] The non-ballast region Rx is a region where the ballast 16 is not visible, for example, a region where the sleepers 13 are observed. It is preferable that the target region R be set to a region excluding the non-ballast region Rx in the image data.

[0047] The non-ballast region Rx may be extracted, for example, as follows: That is, an edge detection process is performed on the image data. The edge detection process may be performed, for example, by the Canny method.

[0048] A detection line L(i) is set along the width direction of the image data that has undergone edge detection processing. The detection line L(i) is in the direction along the extension direction of the sleepers 13, and is also the width direction in which the rails 12a, 12b are aligned. A plurality of detection lines L(i) are set so as to be aligned along the extension direction of the rails 12a, 12b.

[0049] The pixel values ​​are summed for each detection line L(i). The amount of change in the total pixel values ​​of adjacent detection lines L(i) in the extension direction of the rails 12a, 12b is calculated, and if the amount of change exceeds or is equal to or greater than a predetermined reference amount of change, it is determined that the boundary of the sleeper 13 exists.

[0050] The moving average of the total pixel values ​​of each detection line L(i) for a predetermined number may be calculated, and the presence or absence of the boundary of the sleeper 13 may be determined by comparing the change in the moving average with a reference change.

[0051] When the existence of multiple boundaries of the sleepers 13 is determined, the distance between the two boundaries is calculated. When it is determined that the boundary distance is outside a predetermined reference range corresponding to the width of the actually existing sleeper 13, it is determined that the two boundaries correspond to boundaries of different sleepers 13. When it is determined that the boundary distance is within a predetermined reference range corresponding to the width of the actually existing sleeper 13, it is determined that the two boundaries correspond to both side boundaries of one sleeper 13.

[0052] The center position of the two boundaries corresponding to the boundaries on both sides of one sleeper 13 can be identified as the center position of that one sleeper 13 in the width direction.

[0053] By repeating the above process, the widthwise central positions of a plurality of sleepers 13 are identified in the image data. By setting the target region R based on the widthwise central position of the sleeper 13, it is possible to set the region of the image data excluding the non-ballast region Rx as the target region R. For example, the center of the target region R may be set at the center between the widthwise central positions of two adjacent sleepers 13.

[0054] The imaging range of the imaging device 42 from the railway track 10 is roughly determined, and the position of the rail 12a in the imaging data is also roughly determined. Therefore, the target area R is set in the imaging data to an area that avoids the area in which the rail 12a is captured.

[0055] The process for setting the target region R is not limited to the above example. For example, the sleeper 13 may be extracted by pattern matching or the like. The target region R may also be set as a region that is a predetermined distance away from the edge of the sleeper 13. The rail 12a may also be extracted by image processing, and the target region R may be set at a position away from the extracted position of the rail 12a.

[0056] In the next step S4, statistics are calculated based on the acquired imaging data. Statistics are values ​​obtained by applying a function to the imaging data. For example, the imaging data includes the amount of change in the total value of multiple pixels having gradation values. The statistics may be calculated using the gradation values ​​of multiple pixel data as random variables. Here, the imaging data may be the imaging data itself output from the imaging device 42, or may be data that has been subjected to image processing. The image processing may be high-contrast processing, noise removal processing, edge detection processing, etc.

[0057] 3 and 4, and their high-contrast versions, 6 and 7, as the ballast 16 becomes finer and the number of edge region pixels that form the boundaries of the lumps 17 increases, the area of ​​dark pixels in the target region R increases, and the entire target region R also becomes darker. Therefore, the degree of fineness can be estimated based on statistics calculated using the gradation values ​​of multiple pixel data as random variables.

[0058] The statistical amount may be a value calculated based on the pixel values ​​of edge area pixels extracted from the imaging data.

[0059] Extraction of edge area pixels from the imaging data may be performed based on gradation values ​​from multiple pixel data. That is, because the boundary of the lump 17 is darker than other areas, it is conceivable that the edge area pixels corresponding to the boundary of the lump 17 have smaller gradation values ​​than other pixels. Therefore, the gradation value of each pixel data is compared with a gradation reference value, and if the gradation value is smaller than or equal to the gradation value, the edge area pixel can be extracted. The gradation reference value may be a predetermined value or may be set based on pixel data in the entire region of the imaging data. For example, the average pixel value of the gradation values ​​of pixel data in the entire region of the imaging data may be calculated, and the gradation reference value may be set based on the average pixel value. In this case, the gradation reference value may be set according to the brightness at the time of shooting.

[0060] According to this extraction process, the gradation values ​​of a plurality of pixel data items can be compared with the gradation reference value, and therefore the extraction process can be easily performed.

[0061] The extraction of edge area pixels from the image data may be performed by other processing methods, such as by comparing the pixel values ​​with those of adjacent pixels.

[0062] The statistic may be the total number of pixels in a plurality of edge regions. The statistic may also be a percentage value of the number of pixels in an edge region relative to the total number of pixels in the target region R. When the statistic is the total number or percentage value of the number of pixels in a plurality of edge regions, the larger the statistic, the more dark pixels there are. Therefore, in this case, it is estimated that the larger the statistic, the more advanced the granularity is, and the finer the granularity is.

[0063] The statistical quantity may be the average value of the gradation values ​​of multiple edge region pixels. The statistical quantity may also be a percentage value of the total gradation values ​​of multiple edge region pixels relative to the total gradation values ​​of all pixels in the target region R. When the statistical quantity is the average or percentage value of the gradation values ​​of multiple edge region pixels, the smaller the statistical quantity, the darker the image is considered to be overall. Therefore, in this case, the smaller the statistical quantity, the more advanced the granularity is, and the finer the granularity is presumed to be.

[0064] The statistic may be the sum of the gradation values ​​of a plurality of edge region pixels. In this case, the correlation between the sum and the fine granularity is considered to depend on the gradation and gradation reference value that each pixel in the target region R can express.

[0065] The statistic may be the sum or average of the gradation values ​​of multiple pixel data in the target region R. In this case, the smaller the statistic, the darker the image is considered to be overall. Therefore, in this case, the smaller the statistic, the more advanced the granularity is, and the finer the granularity is presumed to be.

[0066] In this case, statistics can be easily obtained without performing an extraction process for edge region pixels on the target region R. When the edge region of the lump 17 is large, the target region R becomes dark overall, and therefore it can be assumed that the smaller the total or average value of the gradation values ​​of the multiple pixel data in the target region R, the more refined the grains become.

[0067] Since each of the above statistical quantities is calculated by addition or subtraction processing, the calculation processing can be easily performed.

[0068] The statistical amount is not limited to the above example, and may be other statistical amounts such as the amount of image information or an entropy value.

[0069] The number of statistics to be calculated may be one or more.

[0070] In the next step S5, the statistical amount calculated in step S4 is associated with the traveling position of the railway vehicle 20 used in the determination in step S1, and the associated data is stored in the storage device 54 as data 54b.

[0071] In the next step S6, data 54b specifying the traveling position of the railway vehicle 20 as the statistical quantity is transmitted via the communication device 46. The data 54b is stored in the data server 90. The data 54b may be transmitted each time the statistical quantity is calculated, or may be transmitted every time a predetermined distance is traveled or a predetermined number of times the statistical quantity is calculated, or may be transmitted each time the railway vehicle 20 stops or at the end of operation. After transmission, the data 54b in the storage device 54 may be erased. The communication device 46 may be omitted. In this case, the data 54b stored in the storage device 54 may be collected via a portable storage medium.

[0072] In the next step S7, it is determined whether or not the railway vehicle 20 has finished traveling. Whether or not the railway vehicle 20 has finished traveling is determined, for example, by whether or not the railway vehicle 20 has reached a terminal station or whether or not the power has been turned off. If it is determined that the railway vehicle 20 has finished traveling, the process ends, and if it is determined that the railway vehicle 20 has not finished traveling, the process returns to step S1 and the above process is repeated.

[0073] As a result, data is obtained in which the statistical amount is associated with a position on the railway track 10 for each of the determined distances traveled by the railway vehicle 20 as it travels.

[0074] The calculated statistics are associated with the position of the railway track 10 and stored in the storage device 54 of the data server 90. The determination processing device 70 downloads data from the data server 90 and executes processing.

[0075] 1, the determination processing device 70 is configured by a computer including a processor 72 such as a CPU, a storage device 74, a communication device 76, etc. The determination processing device 70 is connected to a data server 90 via wired or wireless communication so as to be able to communicate with the data server 90.

[0076] The determination processing device 70 receives collected data 92a stored in the data server 90 and stores it in the storage device 74. The collected data 74b downloaded to the storage device 74 may be a portion of the collected data 92a in the data server 90 that belongs to the railway track 10 to be evaluated. The processor 72, as a maintenance necessity determination processing unit, executes processing in accordance with a program 74a stored in the storage device 74 to determine whether maintenance is necessary based on statistics. For example, the processor 72 determines whether maintenance is necessary by comparing the statistics with reference values ​​included in reference value data 74d stored in the storage device 74. The necessity of maintenance may be determined as a degree of maintenance request (maintenance level). A determination result 74c is stored in the storage device 74 in association with the position of the railway track 10.

[0077] A display device 78 and an input unit 79 are also connected to the determination processing device 70. The display device 78 may be a liquid crystal display device, an organic EL (Electro-Luminescence) display device, or the like. A display device provided on a smartphone, a tablet terminal, or the like may also be used as the display device 78. The input unit 79 receives various instructions from a user for the determination processing device 70. The input unit 79 may be a keyboard including multiple switches, a mouse, a touch panel, or the like. The determination result as to whether maintenance is required for the ballast 16 of the railway track 10 may be displayed on the display device 78.

[0078] An example of processing in the determination processing device 70 will be described with reference to the flowchart shown in FIG.

[0079] In step S11, statistics of the evaluation target position are acquired. For example, a section of the railway track 10 is designated as the evaluation target section by the user via the input unit 79. Statistics of one position in the evaluation target section are acquired from the collected data 74b in the storage device 74.

[0080] In the next step S12, the statistical quantity is compared with a reference value to determine a maintenance level indicating whether maintenance is required. For example, the more the ballast 16 is refined, the greater the degree to which maintenance of the ballast 16 is required. For this reason, multiple maintenance levels are pre-set according to the degree of maintenance required (attention level). The maintenance level corresponding to the statistical quantity can be empirically set according to the refinement of the ballast 16 corresponding to the statistical quantity. The maintenance level at each position on the railway track 10 is determined by comparing the statistical quantity with the reference value. The maintenance level indicates the degree to which the ballast 16 has been refined, and it can be said that the higher the maintenance level, the more refined the ballast 16 is. There may be one reference value, and the maintenance level may simply be two levels indicating whether attention is required. There may be multiple reference values, and the maintenance level may be three or more levels. The maintenance levels may be distinguished into levels using the reference value as a threshold. If the statistical quantity is the same as the reference value, it may be determined to belong to either a level before or after the reference value.

[0081] The maintenance level may be determined based on a plurality of statistical quantities. For example, each of the plurality of statistical quantities may be compared with a separately set criterion value, and the maintenance level may be determined according to a combination of the comparison results. For example, if all or at least one of the plurality of statistical quantities meets a criterion that granularity is increasing, it may be determined that maintenance is required. Furthermore, one of a plurality of maintenance levels may be determined according to a combination of the determination results for each of the plurality of statistical quantities.

[0082] In the next step S13, the determination result 74c is written in the storage device 74 in association with the position on the railway track 10.

[0083] In the next step S14, it is determined whether the evaluation of the target section has been completed. If it is determined that the evaluation of all positions included in the target section has been completed, the process proceeds to step S15. If it is determined that the evaluation of all positions included in the target section has not been completed, the process returns to step S11 and the above process is repeated. In this way, a maintenance level is determined for the statistics corresponding to the positions for each judgment distance included in the target section.

[0084] In the next step S15, based on the result of the determination as to whether maintenance is required, a ballast condition image showing the condition of the ballast 16 on the railway track 10 is displayed on the display device 78. After this, the process ends.

[0085] 9 is a diagram showing an example of a ballast state image 100. The ballast state image 100 is an image in which the state of the ballast 16 is associated with the position of the railway track 10, for example.

[0086] In FIG. 9 , a maintenance level is associated with each section of the railway track 10. The image includes a track image 102 that represents the actual railway track 10. The track image 102 includes a maintenance level image 103 that displays the maintenance level. The maintenance level image 103 may be distinguished by color, shade, pattern, or the like. For example, the maintenance level may be distinguished so that the degree of maintenance increases as the maintenance level transitions from green to yellow to red. By looking at this image, it is easy to understand at which position on the railway track 10 attention should be paid to the condition of the ballast 16.

[0087] Depending on the display scale of the railway track 10, it is conceivable that each section of the railway track 10 includes a plurality of positions that are the subject of statistical evaluation. In this case, the maintenance level image 103 corresponding to the highest maintenance level among the evaluation results of any of the plurality of positions may be displayed.

[0088] Separately from the track image 102, a detailed image 104 showing statistics may be displayed in an enlarged area of ​​a portion of the track image 102. The detailed image 104 is a graph with the horizontal axis representing the longitudinal position (e.g., kilometers) on the railway track 10 and the horizontal axis representing the statistics. The detailed image 104 may be displayed, for example, by selecting a portion of the track image 102 by clicking, touching, or the like. This detailed image 104 allows the state of a portion of the railway track 10 to be grasped in more detail.

[0089] The ballast status image may be an image that displays the areas requiring maintenance in a table format according to the maintenance level, or may be an image that includes a message specifying the areas requiring maintenance and the maintenance level.

[0090] According to the ballast condition monitoring device 40, ballast condition monitoring system 30, and ballast condition monitoring method configured in this manner, statistical quantities indicating the fineness of the lumps 17 laid as ballast 16 are calculated based on image data obtained by capturing images of the surface condition of the ballast 16 on the railway track 10.

[0091] The inventors of the present invention have found that differences in the fineness of the ballast result in differences in the statistical quantities of the image data of the ballast surface condition. Therefore, it is possible to determine the degree to which the ballast has become finer and to determine whether maintenance is required based on these statistical quantities.

[0092] The statistics are calculated based on image data of the surface condition of the ballast 16, so the condition of the ballast 16 is monitored quantitatively and without individual differences.

[0093] Furthermore, the boundaries of the lumps 17 in the ballast 16 affect the gradation values ​​of the plurality of pixel data. Therefore, by calculating statistics in which the gradation values ​​of the plurality of image data are used as random variables, it is possible to more appropriately determine the degree to which the granulation of the ballast 16 has progressed.

[0094] Edge area pixels of the ballast 16 may be extracted from the plurality of pixel data based on the gradation values, and the statistical value may be calculated by calculating the total value or percentage value of the edge area pixels, or the total value, average value, or percentage value of the gradation values ​​of the edge area pixels. In this case, edge area pixels can be extracted with simpler calculations than when edge area pixels are extracted based on a comparison of adjacent pixels. Calculating the statistical value based on the edge area pixels allows the statistical value to be calculated relatively easily.

[0095] Alternatively, the statistical quantity may be calculated as the sum or average of the gradation values ​​of a plurality of pixel data. In this case, it is not necessary to extract edge pixel regions, and therefore the statistical quantity can be calculated with simpler calculations than when edge region pixels are extracted.

[0096] Furthermore, since the target area R is set to an area excluding the non-ballast area Rx corresponding to the sleepers 13 and the rails 12a, 12b, the influence of the sleepers 13 can be eliminated and the need for maintenance can be appropriately determined.

[0097] Furthermore, since the necessity of maintenance is determined by comparing the statistical amount with a predetermined reference value, variations in the criteria for determining the necessity of maintenance can be suppressed.

[0098] Furthermore, data is generated that associates statistics indicating fine granularity with position data on the railway track 10, and is saved as collected data 92a and 74b. Position data on the railway track 10 is generated for the maintenance level, which is the result of the determination of whether or not maintenance is required, and is saved as determination result 74c. This makes it possible to grasp the relationship between the position on the railway track and the determination result of whether or not maintenance is required or the fine granularity.

[0099] The value indicating the fine-granularity associated with the position data on the railway track 10 does not have to be a statistical quantity. For example, the value indicating the fine-granularity may be a value that is scaled based on a statistical quantity so as to appropriately indicate the fine-granularity.

[0100] In addition, a ballast status image 100 serving as a monitoring image showing the determination result of whether maintenance is required is displayed on the display device 78. In addition, a detailed image 104 serving as a monitoring image showing fine granularity based on statistics is displayed on the display device 78.

[0101] Therefore, by looking at the display device 78, the user can easily recognize the determination result of whether maintenance is necessary or the fine granularity based on the statistical amount.

[0102] Moreover, in the images 100 and 104, the determination result or granularity of the necessity of maintenance is displayed in association with the position data on the railway track 10.

[0103] Therefore, the user can easily check the relationship between the position on the railway track 10 and the determination result or granularity of whether maintenance is required by looking at the display device 78.

[0104] The imaging device 42 is supported on the railway track 10 and captures images of the surface condition of the ballast 16 while the railway track 10 is in motion. This allows surface shape data to be acquired sequentially while the railway vehicle 20 is in motion. This makes it possible to easily monitor the condition of the ballast 16 on the railway track 10 without an inspector having to visit each location on the railway track 10.

[0105] If the railway track 10 on which the imaging device 42 is supported is a commercial vehicle for transporting people or cargo, the condition of the ballast 16 on the railway track 10 on which the commercial vehicle runs can be obtained over a wide area and at high frequency.

[0106] Furthermore, if the imaging device 42 is supported on the railway vehicle 20 at a position where it can detect the surface condition of at least a portion of the ballast 16 between the rails 12a and 12b, the imaging device 42 can easily continuously detect the surface condition of the ballast 16. For example, assume that the imaging device 42 is detecting an outer area outside the area between the rails 12a. In this case, if the railway vehicle 20 tilts around a curve, for example, the target area R is likely to move outward from the ballast 16. If the imaging device 42 is detecting the area between the rails 12a, 12a, the target area R is unlikely to move outward from the ballast 16 even if the railway vehicle 20 tilts around a curve, for example.

[0107] In the above embodiment, an example has been described in which the statistical quantity calculation device 50 is mounted on the railway vehicle 20. The statistical quantity calculation device 50 may be implemented in the data server 90 or the determination processing device 70, and the statistical quantity calculation processing may be performed in the data server 90 or the determination processing device 70. In this case, the railway vehicle 20 may be equipped with an imaging device 42 and a running position detection unit 44, and output data from the imaging device 42 and the running position detection unit 44 may be transmitted to the data server 90 or the determination processing device 70 via the communication device 46.

[0108] Furthermore, in the above embodiment, an example has been described in which the determination processing device 70 is installed in a location separate from the railway vehicle 20. The determination processing device 70 may be mounted on the railway vehicle 20, and the determination processing may be performed on the railway vehicle 20. In this case, the determination result may be transmitted to the data server 90 or a computer operated by a maintenance manager or maintenance worker, and a ballast state image based on the determination result may be displayed on the computer operated by the maintenance manager or maintenance worker.

[0109] Second Embodiment A ballast state monitoring system, a ballast state monitoring device, and a ballast state monitoring method according to a second embodiment will be described below, focusing on the differences from the first embodiment.

[0110] The calculation of statistics by the processor 52 of the statistical calculation device 50 may be performed by a trained model 210 that has undergone machine learning, as shown in Fig. 10. The trained model 210 is configured, for example, by a multilayer neural network or an SVM (support vector machine) and is stored in the storage device 54. The processor 52 reads out the program and parameters written in the trained model and executes an identification process, thereby performing a process of calculating statistics. For example, imaging data is input into the trained model 210, and the statistics are calculated.

[0111] The trained model 210 is generated by a machine learning device 230 configured by a computer including a storage device 220 storing a training model 222, a processor 231 having a model generation unit 232, and the like, as shown in FIG. 11 . The model generation unit 232 learns how to calculate statistics based on training data. The training data may be multiple data sets combining imaging data and statistics calculated from the imaging data. Then, the trained model 210 that infers statistics based on the imaging data is generated from the training data.

[0112] Here, the imaging data used as the training data is data that is actually captured while the railway track 10 is in operation. The statistical quantities are values ​​calculated by a computer or an operator.

[0113] The learning model 222 and the learning model 222 may be a regression model that infers the relationship between at least one parameter obtained based on the imaging data and a statistical quantity, or may be a classification model that classifies the imaging data into multi-stage statistical quantities.

[0114] The machine learning device 230 may be built into the railway vehicle, or may be a device separate from the railway vehicle and connected to the railway vehicle via a network.

[0115] According to this embodiment, statistics can be calculated using a trained model 210 that has been trained by machine learning.

[0116] {Third Embodiment} A ballast state monitoring system, a ballast state monitoring device, and a ballast state monitoring method according to a third embodiment will be described. In this embodiment, differences from the first embodiment will be mainly described.

[0117] In the above-mentioned determination processing device 70, the reference value used to determine whether maintenance is required is set based on a trained model 310 that has undergone machine learning to estimate the reference value using statistics based on the imaging data and the results of the determination of whether maintenance is required based on the imaging data as learning data.

[0118] 12, the trained model 310 is generated by a machine learning device 330 configured by a computer including a storage device 320 storing a training model 322 and a processor 331 having a model generation unit 332. The trained model is generated by training the training model 222.

[0119] The model generation unit 332 learns to infer the maintenance level determination result from the statistics Sa and Sb based on the training data. The training data may be a plurality of data sets in which the statistics Sa and Sb based on the imaging data are associated with the determination result of whether or not maintenance is required for the imaging data. The statistics Sa and Sb may be one or more of the statistics described in the first embodiment. The maintenance level is determined, for example, as one of three levels: low, medium, and high. The determination result of whether or not maintenance is required for the imaging data may be determined by a person based on observation of the imaging data or the ballast 16 on site.

[0120] The learning model 322 is configured, for example, by a decision tree model, and is stored in the storage device 320. The processor 331 generates, from the training data, a learned model 310 that infers a determination result of whether maintenance is required based on the imaging data.

[0121] The learning may be supervised learning or semi-supervised learning. In the case of semi-supervised learning, it is advisable to adopt a method in which the highly confident judgment results of the classifier are added to the supervised data and re-learned. For example, the bootstrap method, label propagation, self-training, etc. may be adopted.

[0122] 13 is a diagram showing an example of a trained decision tree model. As shown in the figure, if the statistic Sa is smaller than the value a, it is inferred that the maintenance request level is low. If the statistic Sa is equal to or greater than the value a and the statistic Sb is smaller than the value b, it is inferred that the maintenance request level is medium. If the statistic Sa is equal to or greater than the value a and the statistic Sb is equal to or greater than the value b, it is inferred that the maintenance request level is high.

[0123] The processor 331 includes a function as a parameter analysis reference value calculation unit 331a that outputs a reference value from the trained model 310 by executing a program stored in the storage device 320.

[0124] The parameter analysis reference value calculation unit 331a outputs a reference value based on, for example, parameters a and b set as branching conditions for each node.

[0125] For example, the parameter a may be output as a reference value for distinguishing between a low maintenance level and a medium or high maintenance level based on the statistical quantity Sa. Also, for example, the parameter b may be output as a reference value for distinguishing between a medium maintenance level and a high maintenance level based on the statistical quantity Sb.

[0126] Depending on the setting example of the decision tree model, the number of statistics used, and the number of maintenance level classifications, fewer or more nodes may be set, or more branches may branch from any node, and therefore the output reference value may be fewer or more.

[0127] For this reason, it is conceivable that the maintenance level determination process based on the output reference value may also be required to be divided into more frequent categories depending on the output reference value.

[0128] According to this embodiment, the reference value can be easily set by the machine-learned trained model 310. Therefore, even a person who does not have the knowledge or experience to set a reference value from statistics can easily set the reference value.

[0129] Furthermore, even when the statistical quantity is multidimensional or the maintenance level determination result has three or more stages, the reference value can be easily set.

[0130] {Fourth Embodiment} A ballast state monitoring system, a ballast state monitoring device, and a ballast state monitoring method according to a fourth embodiment will be described. In this embodiment, differences from the first embodiment will be mainly described.

[0131] 14, in the determination processing device 70, a weighting factor 474d is stored in the storage device 74. The weighting factor 474d is one or more coefficients set according to the position of the railway track 10.

[0132] For example, the railway track 10 does not necessarily continue in a uniform state in the direction of its extension. Furthermore, the running behavior of the railway vehicles 20 on the railway track 10 is also not necessarily uniform. For this reason, it may be better to change the maintenance priority of the ballast 16 in the direction of extension of the railway track 10 depending on the position of the railway track 10. The weighting element 474d is set as a value indicating the maintenance priority depending on the position of the railway track 10.

[0133] The weighting factor 474d may be set, for example, according to the radius of the curve of the railway track 10. For example, the weighting factor 474d may be set so that the smaller the radius of the curve, the higher the priority of maintenance.

[0134] The weighting factor 474d may be set, for example, according to the speed of the railcar 20 traveling on the railroad track 10. The speed may be the maximum operating speed. For example, the weighting factor 474d may be set so that the higher the speed, the higher the priority of maintenance.

[0135] The weighting factor 474d may be set to give a higher priority to maintenance at locations where switches are present.

[0136] The weighting factor 474d may be set separately depending on the curve radius, the speed, and the position of the switch. In this case, at each position on the railway track 10, the weighting factor 474d at that position may be calculated by multiplying or accumulating a plurality of weighting factors 474d depending on the curve radius, the speed, and the position of the switch.

[0137] Furthermore, the curve radius, speed, or presence or absence of a turnout may be stored in association with the position of the railway track 10, and the value of the weighting factor 474d may be stored separately as a table in association with the curve radius, speed, or presence or absence of a turnout. In this case, the weighting factor 474d may be read from the table using the curve radius, speed, or presence or absence of a turnout corresponding to the position of the railway track 10 as a key. This case also corresponds to an example in which the weighting factor 474d is set according to the position of the railway track 10.

[0138] The weighting factor 474d may be set as one coefficient for each position on the railway track 10, taking into consideration the curve radius, speed, and switch position.

[0139] 15 is a flowchart showing an example of a process for determining the maintenance level in consideration of the weighting factor 474d. This flowchart differs from the flowchart shown in FIG. 8 in that after step S11, the process proceeds to step S13 via steps S21 and S22.

[0140] In step S21, an evaluation value is calculated by taking the weighting element 474d into account for the statistical quantity. For example, if a larger statistical quantity indicates a greater degree of refinement, it is assumed that the evaluation value is calculated by adding the weighting element 474d to the statistical quantity. Also, for example, if the weighting element 474d is set to a value of 1 or greater and a larger statistical quantity indicates a greater degree of refinement, it is assumed that the evaluation value is calculated by multiplying the statistical quantity by the weighting element 474d. Also, for example, if a smaller statistical quantity indicates a greater degree of refinement, it is assumed that the evaluation value is calculated by subtracting the weighting element 474d from the statistical quantity. Also, for example, if the weighting element 474d is set to a value of 1 or greater and a smaller statistical quantity indicates a greater degree of refinement, it is assumed that the evaluation value is calculated by dividing the statistical quantity by the weighting element 474d.

[0141] In step S22, similarly to step S12, the evaluation value is compared with the reference value to determine the maintenance level indicating whether maintenance is required.

[0142] According to this embodiment, maintenance priorities are determined based on statistics indicative of the condition of the ballast 16 and a weighting factor 474 d that takes into account maintenance priorities depending on the location on the railway track 10 .

[0143] For example, even if the granularity of the ballast 16 is the same based on the statistics, it is likely that the maintenance level is high at locations with small curve radii, high speeds, or locations with switches. This provides information for making decisions so that locations where maintenance is considered to be highly required according to the location of the railway track 10 are given priority.

[0144] {Additional Note} The present disclosure discloses the following aspects.

[0145] The first aspect is a ballast condition monitoring system that includes an input unit to which imaging data of the surface condition of the ballast is input, and a processing unit that performs processing based on the imaging data, wherein the processing unit calculates statistics based on the imaging data and determines whether maintenance is required based on the statistics.

[0146] This ballast condition monitoring system calculates statistics of the image data of the ballast surface condition. The inventors discovered that differences in the granularity of the ballast lead to differences in the statistics of the image data of the ballast surface condition. Therefore, based on these statistics, it is possible to determine the degree to which the ballast has become finer, and to determine whether maintenance is required.

[0147] A second aspect is a ballast status monitoring system according to the first aspect, wherein the imaging data includes a plurality of pixel data each having a gradation value, and the processing unit calculates the statistical quantity using the gradation values ​​of the plurality of pixel data as random variables.

[0148] The boundaries of the ballast affect the gradation values ​​of multiple pixel data. Therefore, by calculating statistics using the gradation values ​​of multiple image data as random variables, it is possible to more appropriately determine the degree to which the ballast has been refined.

[0149] A third aspect is a ballast status monitoring system according to the second aspect, wherein the processing unit extracts edge area pixels of the ballast from the plurality of pixel data based on the gradation values, and calculates the total value or percentage value of the edge area pixels, or the total value, average value or percentage value of the gradation values ​​of the edge area pixels, as the statistical quantity.

[0150] This allows the statistical quantities to be calculated with relatively simple calculations.

[0151] A fourth aspect is the ballast state monitoring system according to the second aspect, wherein the processing unit calculates a sum or average value of the gradation values ​​of the plurality of pixel data as the statistical quantity.

[0152] This eliminates the need to extract edge regions, and allows the statistical amounts to be calculated with relatively simple calculations.

[0153] A fifth aspect is a ballast condition monitoring system according to any one of the first to fourth aspects, wherein the processing unit includes a trained model that has undergone machine learning to estimate the statistical quantity based on learning data that corresponds the imaging data having a plurality of pixel data to the statistical quantity in which the gradation values ​​of each of the plurality of pixel data are random variables, and the statistical quantity is calculated by inputting the imaging data into the trained model.

[0154] This allows statistics to be calculated using a trained model that has been machine-learned.

[0155] A sixth aspect is a ballast condition monitoring system relating to any one of the first to fifth aspects, wherein the processing unit extracts non-ballast areas corresponding to sleepers in the image data, sets the area excluding the non-ballast areas from the image data as a target area, and calculates the statistical quantity based on the data of the target area.

[0156] In this way, by calculating statistics for the target area excluding the non-ballast area corresponding to the sleepers, it is possible to appropriately determine whether maintenance is required.

[0157] A seventh aspect is a ballast condition monitoring system according to any one of the first to sixth aspects, wherein the processing unit determines whether or not maintenance is required by comparing the statistical quantity with a predetermined reference value.

[0158] This makes it possible to reduce variations in the criteria for determining whether or not maintenance is necessary.

[0159] An eighth aspect is a ballast condition monitoring system according to the seventh aspect, in which the reference value is a value set based on a learned model in which machine learning has been performed using the statistical quantity based on the imaging data and the determination result of the need for maintenance based on the imaging data as learning data.

[0160] This makes it easy to set the reference value using a machine-learned model.

[0161] A ninth aspect is a ballast condition monitoring system according to any one of the first to eighth aspects, wherein the processing unit generates data that associates the determination result of the need for maintenance or fine granularity based on the statistical quantity with position data on the railway track.

[0162] This makes it possible to grasp the relationship between the position on the railway track and the determination result or fine granularity of whether maintenance is required.

[0163] A tenth aspect is a ballast condition monitoring system according to the ninth aspect, wherein the processing unit determines the maintenance priority based on a weighting factor set according to the position of the railway track and the statistical quantity.

[0164] This allows maintenance priorities to be determined according to the location of the railway track.

[0165] An eleventh aspect is a ballast condition monitoring system according to any one of the first to tenth aspects, further comprising a display device, wherein the processing unit displays on the display device a monitoring image showing the determination result of whether or not maintenance is required, or a fine-grained image based on the statistical quantity.

[0166] This allows the user to see the display device and recognize the result of the determination as to whether maintenance is required or the fine granularity based on the statistics.

[0167] A twelfth aspect is a ballast condition monitoring system according to the eleventh aspect, wherein the monitoring image may be an image in which the result of the determination of the need for maintenance or a fine granularity based on the statistical quantity is associated with position data on the railway track.

[0168] This allows the user to check the relationship between the position on the railway track and the determination result or fine granularity of whether maintenance is required by looking at the display device.

[0169] A thirteenth aspect is a ballast condition monitoring system according to any one of the first to twelfth aspects, further comprising an imaging device that is supported by a railway vehicle running on the railway track at a position where it can image the surface condition of at least a portion of the ballast between the first rail and the second rail on the railway track, and that images the surface condition of the ballast while the railway vehicle is running.

[0170] This allows statistical calculations to be performed based on image data captured while the railcar is running. Furthermore, even if the railcar leans on a curve, the image capture device can easily continue to detect the condition of the ballast surface between the rails.

[0171] The ballast condition monitoring device of the fourteenth aspect is a ballast condition monitoring device that includes an input unit to which imaging data of the surface condition of the ballast is input, and a processing unit that performs processing based on the imaging data, and the processing unit calculates statistical quantities based on the imaging data.

[0172] This ballast condition monitoring device calculates statistics of the image data of the ballast surface condition, and can provide these statistics as values ​​appropriate for determining the degree of ballast refinement, making it possible to monitor the condition of the ballast based on the fineness of the lumps laid as ballast.

[0173] The ballast condition monitoring method according to the fifteenth aspect is a ballast condition monitoring method that images the surface condition of the ballast, calculates statistics based on the imaging data of the surface condition of the ballast, and determines whether maintenance is required based on the statistics.

[0174] According to this ballast condition monitoring method, statistics of image data of the ballast surface condition are calculated. Based on the statistics, it is possible to determine the degree to which the ballast has been refined and to determine whether maintenance is required.

[0175] It should be noted that the functions of the elements disclosed herein can be performed using circuits or processing circuits, including general-purpose processors, special-purpose processors, integrated circuits, application-specific integrated circuits (ASICs), conventional circuits, and / or combinations thereof, configured or programmed to perform the disclosed functions. A processor is considered a processing circuit or circuit because it includes transistors and other circuitry. In this disclosure, a circuit, unit, or means is hardware that performs the recited functions or hardware that is programmed to perform the recited functions. The hardware may be hardware disclosed herein or other known hardware that is programmed or configured to perform the recited functions. Where hardware is a processor, which is considered a type of circuit, the circuit, means, or unit is a combination of hardware and software, and the software is used to configure the hardware and / or processor.

[0176] The above description is illustrative in all respects and is not intended to limit the disclosure to the above. It is understood that numerous variations not illustrated can be envisioned without departing from the scope of the disclosure.

Claims

1. A ballast condition monitoring system comprising: an input unit to which imaging data capturing the surface condition of the ballast is input; and a processing unit that executes processing based on the imaging data, wherein the processing unit calculates statistics based on the imaging data, and determines whether or not maintenance is required based on the statistics.

2. A ballast status monitoring system as described in claim 1, wherein the imaging data includes a plurality of pixel data each having a gradation value, and the processing unit calculates the statistics using the gradation values ​​of the plurality of pixel data as random variables.

3. A ballast status monitoring system as described in claim 2, wherein the processing unit extracts edge region pixels of the ballast based on the gradation values ​​from the plurality of pixel data, and calculates, as the statistical quantity, a total value or a percentage value of the edge region pixels, or a total value, an average value or a percentage value of the gradation values ​​of the edge region pixels.

4. A ballast status monitoring system according to claim 2, wherein the processing unit calculates a total or average value of the gradation values ​​of the plurality of pixel data as the statistical quantity.

5. A ballast condition monitoring system as claimed in any one of claims 1 to 4, wherein the processing unit includes a trained model in which machine learning has been performed to estimate the statistical quantity based on learning data in which the imaging data having a plurality of pixel data is associated with the statistical quantity in which the gradation values ​​of each of the plurality of pixel data are random variables, and the ballast condition monitoring system calculates the statistical quantity by inputting the imaging data into the trained model.

6. A ballast condition monitoring system as described in any one of claims 1 to 4, wherein the processing unit extracts non-ballast areas corresponding to sleepers in the imaging data, sets an area of ​​the imaging data excluding the non-ballast areas as a target area, and calculates the statistics based on the data of the target area.

7. A ballast condition monitoring system as claimed in any one of claims 1 to 4, wherein the processing unit determines whether or not maintenance is required by comparing the statistical amount with a predetermined reference value.

8. A ballast condition monitoring system as described in claim 7, wherein the reference value is a value set based on a learned model in which machine learning has been performed using the statistics based on the imaging data and the determination result of the need for maintenance based on the imaging data as learning data.

9. A ballast condition monitoring system as claimed in any one of claims 1 to 4, wherein the processing unit generates data in which the determination result of the necessity of maintenance or a fine granularity based on the statistical quantity is associated with position data on the railway track.

10. A ballast condition monitoring system as claimed in claim 9, wherein the processing unit determines the maintenance priority based on a weighting factor set according to the position of the railway track and the statistical amount.

11. A ballast status monitoring system as described in any one of claims 1 to 4, further comprising a display device, wherein the processing unit displays on the display device a monitoring image indicating the result of the determination as to whether or not maintenance is required, or a fine-grained image based on the statistical quantity.

12. A ballast condition monitoring system as described in claim 11, wherein the monitoring image is an image in which the result of the determination of the necessity of maintenance or a fine granularity based on the statistical quantity is associated with position data on the railway track.

13. A ballast condition monitoring system as described in any one of claims 1 to 4, further comprising an imaging device supported by a railway vehicle running on the railway track at a position capable of imaging the surface condition of at least a portion of the ballast between a first rail and a second rail on the railway track, and which images the surface condition of the ballast while the railway vehicle is running.

14. A ballast condition monitoring device comprising: an input unit to which imaging data capturing an image of the surface condition of the ballast is input; and a processing unit that executes processing based on the imaging data, wherein the processing unit calculates statistics based on the imaging data.

15. A ballast condition monitoring method comprising: imaging the surface condition of the ballast; calculating statistics based on imaging data of the surface condition of the ballast; and determining whether or not maintenance is required based on the statistics.

Images