Fixed-position stop detection device, fixed-position stop detection system, and fixed-position stop detection method

Abstract

The present invention provides a fixed-position stopping detection device that can accurately determine the position of a railway vehicle attempting to stop at a station. [Solution] The fixed-position stop detection device according to the present disclosure includes: a wheel image acquisition unit that acquires a wheel image including the wheels of a railway vehicle; a target point extraction unit that extracts a specific position of the wheel as a target point and determines the coordinates of the target point; a stop tolerance range setting unit that sets the range between the overrun tolerance coordinates and the short run tolerance coordinates as the stop tolerance range; a stop position determination unit that determines that the railway vehicle has stopped at a fixed position if the target point coordinates are within the stop tolerance range from the time the target point coordinates exceed the short run tolerance coordinates in the direction of travel of the railway vehicle until a set time has elapsed; and a determination result output unit that outputs a determination result of whether or not the railway vehicle has stopped at a fixed position.

Description

[Technical Field]

[0001] This disclosure relates to a fixed-position stop detection device, a fixed-position stop detection system, and a fixed-position stop detection method. [Background technology]

[0002] In recent years, the installation of platform screen doors has increased at train stations to prevent passengers from falling onto the tracks. When using platform screen doors, it is necessary to stop trains at designated positions. Therefore, there is a growing demand for systems that can determine whether or not a train has stopped at a designated position corresponding to the platform screen doors. To determine whether or not a train has stopped at a designated position, it is necessary to accurately know the stopping position of the train. The designated stopping detection system disclosed in Patent Document 1 is equipped with a distance sensor at the station, which emits a laser beam from the front of the train towards the train to measure the distance from the distance sensor to the train. Based on the measured distance, it is then determined whether or not the train has stopped at the stopping target. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2015-105081 [Overview of the Initiative] [Problems that the invention aims to solve]

[0004] However, with conventional technology, distance measurements can be unstable due to disturbances such as vibrations of the train itself, making it difficult to accurately determine the stopping position of the train when it is stopped.

[0005] This disclosure was made to solve the aforementioned problems and aims to provide a fixed-position stopping detection device that can accurately determine the position of a railway vehicle that is about to stop at a station. It also aims to provide a fixed-position stopping detection system and a fixed-position stopping detection method. [Means for solving the problem]

[0006] The fixed-position stop detection device according to this disclosure includes: a wheel image acquisition unit that acquires a wheel image including the wheels of a railway vehicle; a target point extraction unit that extracts a specific position of the wheel from the wheel image as a target point and determines the coordinates of the target point; a stop tolerance range setting unit that determines the overrun tolerance coordinates from the overrun tolerance position of the railway vehicle on the rails on which the wheel in the wheel image is running, or from the short run tolerance position set in advance on the wheel image, and sets the range between the overrun tolerance coordinates and the short run tolerance coordinates as the stop tolerance range; a stop position determination unit that determines that the railway vehicle has stopped at a fixed position if the target point coordinates are within the stop tolerance range from the time the target point coordinates exceed the short run tolerance coordinates in the direction of travel of the railway vehicle until a set time has elapsed; and a determination result output unit that outputs the determination result of whether or not the railway vehicle has stopped at a fixed position.

[0007] Furthermore, the fixed-position stop detection device according to this disclosure includes: a door image acquisition unit that acquires a door image that includes the passenger boarding / alighting doors on the side of the railway vehicle in the direction of travel; a target point extraction unit that extracts a specific position of the passenger boarding / alighting door from the door image as a target point and determines the coordinates of the target point; a stop tolerance range setting unit that determines the overrun tolerance coordinates from the railway vehicle's overrun tolerance position set in the door image and the short run tolerance coordinates from the short run tolerance position, and sets the range between the overrun tolerance coordinates and the short run tolerance coordinates as the stop tolerance range; a stop position determination unit that determines that the railway vehicle has stopped at a fixed position if the target point coordinates are within the stop tolerance range from the time the target point coordinates exceed the short run tolerance coordinates in the direction of travel of the railway vehicle until a set time has elapsed; and a determination result output unit that outputs the determination result of whether or not the railway vehicle has stopped at a fixed position.

[0008] Furthermore, the fixed-position stop detection device according to this disclosure includes: a platform image acquisition unit that acquires a platform image including a stop reference mark provided on the platform, which is captured from a railway vehicle; a target point extraction unit that extracts the stop reference mark from the platform image as a target point and determines the coordinates of the target point; a stop allowance range setting unit that determines the overrun allowance coordinates from the allowable overrun position of the railway vehicle set on the platform image, and the short run allowance coordinates from the allowable short run position, and sets the area between the overrun allowance coordinates and the short run allowance coordinates as the stop allowance range; a stop position determination unit that determines that the railway vehicle has stopped at a fixed position if the target point coordinates are within the stop allowance range from the time the target point coordinates exceed the short run allowance coordinates in the direction of travel of the railway vehicle until a set time has elapsed; and a determination result output unit that outputs the determination result of whether or not the railway vehicle has stopped at a fixed position.

[0009] Furthermore, the fixed-position stop detection system according to this disclosure comprises a track presence detection sensor that detects when a railway vehicle enters a station, an imaging unit that images the rails on which the railway vehicle is running when the track presence detection sensor detects that a railway vehicle is present at the station, a fixed-position stop detection device according to this disclosure that acquires at least one of a wheel image including the wheels and a door image including the doors from the imaging unit, and an opening / closing control unit that controls the opening and closing of platform doors installed on the station platform based on the determination result of the fixed-position stop detection device.

[0010] Moreover, the fixed-position stop detection method according to the present disclosure includes steps of: acquiring a wheel image including wheels of a railway vehicle; extracting a specific position of a wheel as a target point from the wheel image and obtaining target point coordinates; obtaining overrun allowable coordinates from an allowable position of overrun of the railway vehicle preset on a rail on which the wheels in the wheel image run or on the wheel image, and obtaining short-run allowable coordinates from an allowable position of short-run, and setting a stop allowable range between the overrun allowable coordinates and the short-run allowable coordinates; determining that the railway vehicle has stopped at a fixed position when the target point coordinates are within the stop allowable range from when the target point coordinates exceed the short-run allowable coordinates until a set time elapses in the traveling direction of the railway vehicle; and outputting a determination result as to whether the railway vehicle has stopped at a fixed position.

Advantages of the Invention

[0011] According to the present disclosure, in an image obtained by imaging a railway vehicle, a platform, etc. in the direction of the side surface of the railway vehicle with less disturbance such as shaking, a target point is extracted, and in order to accurately grasp the position of the target point, it is possible to highly accurately determine whether the stop position of the railway vehicle is at a fixed position.

Brief Description of the Drawings

[0012] [Figure 1] It is a block diagram showing an example of the configuration of a fixed-position stop detection system according to Embodiment 1. [Figure 2] It is an explanatory diagram showing an example of the positional relationship between a fixed-position stop detection device, an imaging unit, and a railway vehicle according to Embodiment 1. [Figure 3] It is an explanatory diagram showing an example of the positional relationship between a fixed-position stop detection device, an imaging unit, and a railway vehicle according to Embodiment 1. [Figure 4] It is an explanatory diagram showing an example of a wheel image according to Embodiment 1. [Figure 5] It is an explanatory diagram showing an example of a wheel image according to Embodiment 1. [Figure 6] It is an explanatory diagram showing an example of a wheel image according to Embodiment 1. [Figure 7]This is an explanatory diagram showing an example of a wheel image according to Embodiment 1. [Figure 8] This is an explanatory diagram showing an example of setting the target stopping position of a target point using the fixed-position stopping detection device according to Embodiment 2. [Figure 9] This is a block diagram showing an example of the configuration of the target point extraction unit according to Embodiment 3. [Figure 10] This is an explanatory diagram showing an example of target point learning data according to Embodiment 3. [Figure 11] This is an explanatory diagram showing an example of target point learning data according to Embodiment 3. [Figure 12] This is an explanatory diagram showing an example of target point learning data according to Embodiment 3. [Figure 13] This is a schematic diagram showing a three-layer neural network model according to Embodiment 3. [Figure 14] This flowchart shows the learning flow of the target point extraction unit according to Embodiment 3. [Figure 15] This is a block diagram showing an example of the configuration of the target point extraction unit according to Embodiment 3. [Figure 16] This flowchart shows the estimation flow of the target point extraction unit according to Embodiment 3. [Figure 17] This is a block diagram showing an example of the configuration of the fixed-position stop detection system according to Embodiment 4. [Figure 18] This is a block diagram showing the configuration of the imaging control unit according to Embodiment 4. [Figure 19] This is a block diagram showing the configuration of the imaging control unit according to Embodiment 4. [Figure 20] This is an explanatory diagram showing the position of a railway vehicle when it is determined that it has stopped at a fixed position using the fixed-position stop detection device according to Embodiment 6. [Figure 21] This is an explanatory diagram showing an example of the positional relationship between the fixed-position stop detection device and imaging unit and the stop reference mark according to Embodiment 8. [Figure 22]This flowchart shows the processing routines executed by the fixed-position stop detection device according to Embodiments 1 to 8. [Figure 23] This is a schematic block diagram showing an example of a processing circuit that realizes each function of the fixed-position stop detection device according to Embodiments 1 to 8. [Modes for carrying out the invention]

[0013] The embodiments described herein will be explained below with reference to the drawings. The same or corresponding parts in each drawing are denoted by the same reference numerals. In the description of the embodiments, the descriptions of the same or corresponding parts will be omitted or simplified as appropriate.

[0014] Embodiment 1. The fixed-position stop detection system 1000 according to Embodiment 1 will be described with reference to the figures. Figure 1 is a block diagram showing an example of the configuration of the fixed-position stop detection system 1000. The fixed-position stop detection system 1000 includes an imaging unit 31 that images a railway vehicle, a fixed-position stop detection device 200 that acquires, for example, a wheel image 45 including the wheels 42 from the imaging unit 31 and uses the wheel image 45 to determine whether or not the railway vehicle has stopped at a fixed position, and an opening / closing control unit 32 that controls the opening and closing of platform doors installed on the station platform 44 based on the determination result of the fixed-position stop detection device 200. The fixed-position stop detection device 200 includes a wheel image acquisition unit 11 that acquires a wheel image 45 that includes the wheels 42 of a railway vehicle; a target point extraction unit 12 that extracts a specific position of the wheel 42 from the wheel image 45 as a target point 46 and determines the coordinates of the target point; a stop tolerance range setting unit 13 that determines the overrun tolerance coordinates from the overrun tolerance position of the railway vehicle on the rail 41 on which the wheel 42 in the wheel image 45 runs or on the wheel image 45, and the short run tolerance coordinates from the short run tolerance position, and sets the range between the overrun tolerance coordinates and the short run tolerance coordinates as the stop tolerance range; a stop position determination unit 14 that determines that the railway vehicle has stopped at the fixed position if the target point coordinates are within the stop tolerance range from the time the target point coordinates exceed the short run tolerance coordinates in the direction of travel of the railway vehicle until a set time has elapsed; and a determination result output unit 15 that outputs the determination result of whether or not the railway vehicle has stopped at the fixed position. Here, the fixed position is a predetermined position for stopping, and is the stop tolerance range determined by the overrun tolerance position and the short run tolerance position.

[0015] This fixed-position stop detection device 200 uses a wheel image 45 that shows the rail 41, the railway vehicle body 43, the wheels 42, etc., to distinguish between the target point 46 of the wheel 42 and the permissible position mark 49 attached to the rail 41, and can determine with high accuracy whether or not the railway vehicle has stopped at the fixed position based on the target point 46 and the permissible stopping range.

[0016] Figure 2 is a schematic view of the lead car 50 of a railway vehicle about to stop at a station, seen from above. Figure 3 is a schematic cross-sectional view at the AA section position in Figure 2. In the figure, the X axis represents the direction of travel of the railway vehicle, the Y axis represents the height direction of the railway vehicle, and the Z axis represents the width direction of the railway vehicle. When a railway vehicle enters a station, for example, when a track presence detection sensor installed on the platform 44 detects that the railway vehicle is on a specific track within the station, the imaging unit 31 starts acquiring wheel images 45 of the railway vehicle's wheels 42, and acquires wheel images 45 at specific acquisition intervals. The imaging unit 31 starts imaging at the same timing as the fixed-position stop detection device 200 starts acquiring wheel images 45. As shown in Figure 2, the imaging unit 31 is installed with its imaging field of view fixed in a position such that, for example, the foremost wheel 42 of the lead car 50 of the railway vehicle about to stop at a station is within the field of view of the wheel image 45 when it stops at the fixed position. The imaging unit 31 is installed in the space below the platform 44 to image the wheels 42 of a railway vehicle, for example, as shown in Figure 3.

[0017] Figure 4 shows the time t when a railway vehicle running on rails 41 fixed to sleepers 48 via fixing bolts 47 is about to stop at a station. n This is wheel image 45 at (n is an integer) (the dashed line in the figure represents time t n―1 (This shows the wheel 42 in the image.) Figure 4 shows, for example, the change in position of the first wheel 42 from the front of the leading car 50 over time. The wheel 42 is connected to the railway vehicle body 43, and at least a portion of the contour of the wheel 42 is visible when the wheel 42 is captured in the wheel image 45. The wheel image 45 is acquired so that the wheel 42 is captured above the rail 41, i.e., above the Y-axis. The target point extraction unit 12 detects the contour of at least a portion of the outer diameter of the wheel 42 using a contour detection model, such as the Canny algorithm, from the time the wheel 42 moves to a position where the outer diameter of the wheel 42 can be recognized by the contour detection model. The detected contour is at least one of a continuous line diagram and a collection of discontinuous line diagrams.

[0018] When detecting the outline of the outer diameter of the wheel 42, the target point extraction unit 12 pre-stores the diameter of the outer diameter of the wheel 42, approximates the shape of the outline with a perfect circle, calculates the outer diameter of the outline, and determines that the wheel 42 has been detected when an outline is extracted that matches the calculated outer diameter and the stored outer diameter of the wheel 42. Here, it is determined that the wheel 42 has been detected not only when the calculated outer diameter and the stored outer diameter of the wheel 42 perfectly match, but also when they are within a pre-set error. The target point extraction unit 12 extracts the target point 46 for the first wheel 42 counting from the front of the leading car 50, that is, the first wheel 42 detected after the presence of the train is detected by the presence detection sensor. On the other hand, if the first wheel 42 has not yet been detected in the wheel image 45 after the presence of the train is detected by the presence detection sensor, or if the first wheel 42 has passed through the field of view of the wheel image 45 and is no longer visible, a signal indicating that the first wheel 42 could not be detected is sent to the determination result output unit 15. For the first detected wheel 42 contour, the coordinates of the lowest point in the Y-axis direction and the corresponding X-axis coordinates are determined as the target point coordinates. In Figure 4, time t n―1 The lowest point of the wheel 42's outline is visible in the image, but if the lowest point of the wheel 42 is not visible, the target point 46 is not extracted. The target point coordinates are determined for each acquired wheel image 45.

[0019] Next, we will explain the stop tolerance range set by the stop tolerance range setting unit 13 in order to determine whether or not the target point 46 extracted from the wheel image 45 has stopped at the target stopping position. In the wheel image 45 shown in Figure 4, two rails 41 are shown without overlapping, and two tolerance position marks 491 and 492 are attached to the lower rail 41 (negative direction of the Y axis) in the wheel image 45. Of the two tolerance position marks 491 and 492, the tolerance position mark 491 on the rear side in the direction of travel is set at the allowable short run position for the target point coordinates assuming that the lead car 50 has stopped at the fixed position, and the tolerance position mark 492 on the front side in the direction of travel is set at the allowable overrun position for the target point coordinates assuming that the lead car 50 has stopped at the fixed position. The stop tolerance range setting unit 13 detects the tolerance position marks 491 and 492 using an object detection model such as YOLO (You Only Look Once). Tolerance position marks 491, indicating the tolerance position for short runs, and tolerance position marks 492, indicating the tolerance position for overruns, are attached to the side of the rail 41 for the target point coordinates when the lead car 50 stops at a fixed position. The stop tolerance range setting unit 13 determines the tolerance coordinates for short runs and overruns in the X-axis direction for each of the tolerance position marks 491 and 492, and sets the range between these in the X-axis direction as the stop tolerance range. Alternatively, the stop tolerance range setting unit may also determine the tolerance coordinates for short runs and overruns including the Y-axis coordinates of the uppermost part of the tolerance position marks 491 and 492, and set the range between these as the stop tolerance range.

[0020] time t n―1 Then the lead car 50 will proceed slowly, at time t n This section describes an example of determining whether a railway vehicle has stopped at its designated position when the target point 46 is within the permissible stopping range, regardless of whether the leading vehicle 50 is moving or not. n―1 From time t n During this time, if the target point 46 of the wheel 42 moves in the direction of travel shown in Figure 4 as the railway vehicle moves, the stopping position determination unit 14 determines that the railway vehicle has stopped at the designated position if the target point coordinates are still within the stopping tolerance range after a preset time, for example, 2 seconds, has elapsed from the time when the short run tolerance position was exceeded.

[0021] Here, the track presence detection sensors include, for example, side-range sensors utilizing 2D sensing technology, GPS mounted on the vehicle, and track circuits utilizing short circuits of the wheels 42.

[0022] The imaging unit 31 is, for example, a CCD camera installed in the space below the platform 44, and the imaging interval is, for example, several fps to about 30 fps. The imaging interval is a value stored in the imaging unit 31 or a value input from an external source.

[0023] The wheel image acquisition unit 11 of the fixed-position stop detection device 200 acquires wheel images 45 showing the wheels 42 of the railway vehicle from the imaging unit 31. For example, the wheel images 45 are acquired from the imaging unit 31 via wireless communication. The image acquisition interval is, for example, several fps to about 30 fps.

[0024] If the determination result output unit 15 determines that the railway vehicle has stopped at the designated position, it transmits a signal to the platform door opening / closing control unit 32 indicating that the railway vehicle has stopped at the designated position, and controls the opening and closing of the platform door. Also, as described above, if the determination result output unit 15 receives a signal from the target point extraction unit 12 indicating that it cannot detect the first wheel 42 from the front of the leading car 50, it is certain that the leading car 50 has not stopped within the permitted stopping range, so the determination result output unit 15 transmits a signal to the platform door opening / closing control unit 32, etc., indicating that it has not stopped at the designated position.

[0025] If the opening / closing control unit 32 receives a determination result that the railway vehicle is stopped in the designated position, it sends an open signal to the platform door to open the platform door, and if it receives a determination result that the railway vehicle is not stopped in the designated position, it sends a close signal to the platform door to close the platform door.

[0026] Thus, the fixed-position stop detection device 200 includes a wheel image acquisition unit 11 that acquires a wheel image 45 that includes the wheels 42 of a railway vehicle, a target point extraction unit 12 that extracts a specific position of the wheel 42 from the wheel image 45 as a target point 46 and determines the coordinates of the target point, and a stop tolerance range setting unit that determines the overrun tolerance coordinates from the overrun tolerance position of the railway vehicle pre-set on the rail 41 on which the wheel 42 in the wheel image 45 runs or on the wheel image 45, and the short run tolerance coordinates from the short run tolerance position, and sets the range between the overrun tolerance coordinates and the short run tolerance coordinates as the stop tolerance range. 13, along with a stopping position determination unit 14 that determines that the railway vehicle has stopped at its designated position if the target point coordinates are within the stopping tolerance range from the time the target point coordinates exceed the short-run tolerance coordinates in the direction of travel of the railway vehicle until a set time has elapsed, and a determination result output unit 15 that outputs the determination result of whether or not the railway vehicle has stopped at its designated position, the system extracts the target point 46 from an image of the railway vehicle taken from the side of the railway vehicle where there are few disturbances such as shaking, and accurately grasps the position of the target point 46, thereby enabling high-precision determination of whether or not the stopping position of the railway vehicle is at its designated position.

[0027] As mentioned above, the target point extraction unit 12 may also extract the lowest point of the outer diameter of the wheel 42 as the target point 46. By extracting the lowest point of the wheel 42 as the target point 46, the target point 46 can be extracted without being affected by the shaking of the vehicle body when the railway vehicle is stopped.

[0028] Furthermore, the target point extraction unit 12 may extract the point of contact between the outline of the outer diameter of the wheel 42 and the outline of the rail 41 as the target point 46. If the two rails 41 appear to overlap in the wheel image 45, the imaging unit 31 adjusts the orientation in which it images the rails 41 to detect the rail 41 below the wheel image 45 and the outlines of the two rails 41 using an outline detection model, Hough transform, etc., and extracts the setting of the outline of the rail 41 below the wheel image 45 and the outline of the outer diameter of the wheel 42 as the target point 46. By extracting the point of contact between the outline of the outer diameter of the wheel 42 and the outline of the running surface of the rail 41 as the target point 46, the target point 46 can be easily extracted from the wheel image 45.

[0029] Furthermore, although the example of the target point extraction unit 12 extracting the lowest point of the wheel 42 as the target point 46 has been described, it is also possible to obtain a bounding box including the wheel 42 and extract the center position of the bounding box, etc., as the target point 46. Since no process is performed to extract the lowest point, the target point 46 can be easily extracted from the wheel image 45. Moreover, the target point may be extracted from a location other than the wheel 42 in the railway vehicle. By extracting the target point from a location other than the wheel 42, it is possible to extract the target point even under conditions where the wheel 42 is not visible in the captured wheel image 45. Here, the wheel image 45 includes objects that perform a similar role to the wheel 42.

[0030] Furthermore, while examples have been described in which the acceptable stopping range is set by a predetermined acceptable stopping range in the wheel image 45, or by the acceptable short-run position marks 491 and over-run position marks 492 attached to the side of the rail 41, it is also possible to store the set acceptable short-run coordinates, over-run coordinates, and acceptable stopping range in a fixed imaging field of view, and then set the acceptable stopping range using the acceptable short-run line and over-run line shown in Figure 5, which are provided in the Y-axis direction on the wheel image based on the stored acceptable short-run coordinates and over-run coordinates. For example, Figure 5 shows the time t when a railway vehicle running on a rail 41 fixed to a sleeper 48 attempts to stop at a station.n is the wheel image 45 at (the broken line in the figure indicates the wheel 42 at time t n―1 ). The stop position determination unit 14 determines that the railway vehicle has stopped at a fixed position if the extracted target point 46 moves with the progress of the railway vehicle and is within the stop allowable range even after the elapse of the set time since the target point 46 crossed the short run allowable line between time t n―1 and time t n . Thus, by storing the stop allowable range of the target point 46 of the wheel 42, even when the allowable position marks 49 such as the short run allowable position mark 491 and the overrun allowable position mark 492 on the rail 41 disappear, based on the stop allowable range and the target point 46 of the railway vehicle, it is possible to accurately determine whether or not the railway vehicle has stopped at a fixed position.

[0031] In addition, although the determination result output unit 15 has been shown to output a signal indicating that the railway vehicle has stopped at a fixed position when it is determined that the railway vehicle has stopped at a fixed position, it may output to a notification unit provided in the crew's cabin, the driving command room, etc., the distance between the target point 46 of the stopped railway vehicle and the fixed position, and the direction indicating the overrun side or the short run side. By notifying the distance of the overrun or the short run, the driver can quantitatively grasp the moving amount of the railway vehicle when adjusting the stop position of the railway vehicle, and can assist in stopping the railway vehicle at a fixed position.

[0032] In addition, the target point extraction unit 12 may obtain the number of pixels from the left end of the wheel image 45 to the positions of the target point 46, the short run allowable position, and the overrun allowable position, or the actual lengths in the imaging field of view, as shown in FIG. 6. FIG. 6 is the wheel image 45 at time t when a railway vehicle traveling on the rail 41 fixed to the sleeper 48 via the fixing bolt 47 is about to stop at a station (the broken line in the figure indicates the wheel 42 at time t n ). In FIG. 6, for example, the time change of the position of the first wheel 42 counted from the front of the leading vehicle 50 is shown. n―1 . The target point extraction unit 12 determines the number of images or the length of the real space of the imaging field (e.g., 1500 mm) across the entire width of the acquired wheel image 45, and normalizes the distance from the left edge of the wheel image 45 to the position of the target point 46, the distance from the left edge of the wheel image 45 to the allowable short run position, and the distance from the left edge of the wheel image 45 to the allowable over run position, so that the number of pixels or length of the determined width becomes 100%. The normalized distance from the left edge of the wheel image to the position of the target point 46, the distance from the left edge of the wheel image 45 to the allowable short run position, and the distance from the left edge of the wheel image 45 to the allowable over run position represent the length ratio to the length of the entire width of the wheel image 45. The target stopping position (P2) of the target point 46 after normalization is determined by the target point extraction unit 12. just (Unit: %), the location of target point 46 is L p (Unit: %), the allowable position for short runs is L short (Unit: %), the allowable overrun position is L over (As calculated in %) L short and L over L short ≤L just ≤L over The wheel image 45 is set to satisfy the following condition. just This is a value stored for each station and vehicle type, or a value entered from an external source, L short and L over This is a stored value or a value entered from an external source.

[0033] Here, the target point extraction unit 12 may determine the length ratio from the right edge of the wheel image 45 to the position of the target point 46, the allowable short run position, and the allowable over run position, using the right edge of the wheel image 45 as a reference.

[0034] The stopping position determination unit 14 determines L, which is the length ratio of the extracted target point 46. p The value of increases as the train moves, and at time t n―1 From time t n L between p ≥L short Even after the set time has elapsed since the condition was met, L still remains. short≤L p ≤L over If the following relationship is satisfied, it is determined that the railway vehicle has stopped in its designated position. In this way, by determining the distance from the left or right edge of the wheel image 45 to the position of the target point 46, the allowable short run position, and the allowable overrun position, it becomes unnecessary to select the coordinates of the allowable short run position, the allowable overrun position, etc. from within the image, and the stopping tolerance range can be accurately set by fine-tuning the distance values. Therefore, it is possible to determine with high accuracy whether or not the railway vehicle has stopped in its designated position based on the accurate stopping tolerance range.

[0035] Furthermore, although an example was described in which the fixed-position stop detection device 200 is installed in the space below the station platform 44 together with the imaging unit 31, the fixed-position stop detection device 200 may be installed on the platform 44, in the station office, the train control room, or any other location. Moreover, the fixed-position stop detection device 200 may be installed on the cloud and wheel images 45 may be acquired from the imaging unit 31 via an internet connection. By installing only the imaging unit 31 below the platform 44, even if there are constraints such as a narrow space below the platform 44, the fixed-position stop detection device 200 installed in a remote location can be used to determine with high accuracy whether or not the railway vehicle has stopped at the designated position using the wheel images 45.

[0036] Furthermore, although an example has been described in which only one imaging unit 31 is connected to the fixed-position stop detection device 200, multiple imaging units 31 may be connected to the fixed-position stop detection device 200. By connecting multiple imaging units, it is possible to determine with high accuracy whether or not a railway vehicle entering multiple tracks has stopped at the designated position using a minimum number of fixed-position stop detection devices 200.

[0037] Alternatively, as shown in Figure 7, instead of placing an allowable position mark 49 on the rail 41, it may be possible to determine whether the target point 46 is within the allowable stopping area set in the wheel image 45 using the contact point between the outline of the outer diameter of the wheel 42 and the outline of the rail 41. Figure 7 shows the time change of the position of the first wheel 42 from the front of the leading car 50 of a railway vehicle that is about to stop at a station, and at time tn―1 Wheel 42 is shown as a dotted line, time t n The wheel 42 is shown with a solid line. The stopping tolerance range setting unit 13 divides the wheel image 45 from which the target point 46 has been extracted into eight areas A1 to A8 by setting, for example, seven dividing lines parallel to the Y axis in the direction of the X axis. Of the eight divided areas, for example, the position information of A4 and A5 is stored as the stopping tolerance range. Alternatively, the position information of the area to be set as the stopping tolerance range may be input from an external source. By dividing the wheel image 45 into multiple areas and setting any of the areas as the stopping tolerance range, the work of painting the tolerance position marks 49 on station equipment such as rails 41 after the fixed-position stopping detection device 200 has been installed at the station can be streamlined. Furthermore, the width of the dividing lines in the X-axis direction can be either equal or irregular. By making the spacing of the dividing lines near the acceptable stopping range narrower and the spacing of the dividing lines in the short-run and over-run areas wider, only the minimum amount of dividing line information necessary to set the acceptable stopping range is added to the vehicle image. This reduces the file size of the data files, such as the vehicle image, for which the acceptable stopping range has been set, allowing for smoother stopping position determination processing and enabling high-precision determination of whether or not the railway vehicle has stopped at the correct position.

[0038] Furthermore, although an example was described in which the target point extraction unit 12 extracts the target points 46 and then the stop tolerance range setting unit 13 sets the stop tolerance range, the extraction process of the target points 46 and the setting process of the stop tolerance range may be performed simultaneously, or the target points 46 may be extracted after the stop tolerance range has been set.

[0039] Furthermore, while examples have been described in which the permissible positions for short runs and overruns are determined based on permissible position marks 49 attached to the rail 41 or a range predetermined within the wheel image 45, these may also be determined by stationary sleepers 48, fixing bolts 47 of the rail 41, etc., along with the imaging field shown in Figure 7. The stop permissible range setting unit 13 detects sleepers 48, fixing bolts 47 of the rail 41, etc., and sets the permissible positions for short runs and overruns. Compared to the case where permissible position marks 49 are attached to the rail 41, the possibility of the sleepers 48 and fixing bolts 47 of the rail 41, which form the basis of each permissible position, disappearing is lower, thus reducing the need for periodic repainting of the permissible position marks 49 related to the permissible positions for short runs and overruns.

[0040] Furthermore, although an example has been described in which the target point 46 is set on the first wheel 42 from the front of the leading car 50, the target point 46 may also be set on the second and subsequent wheels 42. In addition, it may be determined whether all of the target points 46 on multiple wheels 42 are within the permissible stopping range.

[0041] Embodiment 2. The fixed-position stop detection device 200 according to Embodiment 2 will be described with reference to the figures. In Embodiment 1, an example was described in which the stop tolerance range was set based on the tolerance position mark 49 attached to the rail 41, but in Embodiment 2, an example will be described in which the stop tolerance range is set based on the type of railway vehicle. The following description will focus on the differences from Embodiment 1, and descriptions of the same or corresponding parts will be omitted as appropriate.

[0042] The stopping tolerance range setting unit 13 sets the target stopping position of the target point 46 based on the vehicle identification information, and determines the overrun tolerance coordinates and short run tolerance coordinates based on the set stopping target position. By setting the stopping tolerance range based on the vehicle identification information, even if the target point 46 of the railway vehicle entering the station differs for each type of vehicle, the stopping tolerance range is flexibly changed according to the type of vehicle, so that it is possible to determine with high accuracy whether or not the railway vehicle has stopped in the correct position.

[0043] Railway vehicles are pre-equipped with identification tags. These identification tags are, for example, RFID (Radio Frequency Identification) tags. An identification antenna installed on the platform 44 detects the identification tags, and an identification reader reads the identification ID and vehicle type information of the leading car 50 from the identification tags via the identification antenna. If the read identification ID matches the identification ID assigned to the fixed-position stop detection device 200 according to this embodiment, the vehicle type information is transmitted to the stop tolerance range setting unit 13. The vehicle type information is, for example, information such as the type of railway vehicle to which the identification tag is attached, such as the 1000 series or 2000 series. The identification antenna, identification reader, and stop tolerance range setting unit 13 are connected by wire or wireless.

[0044] The stopping range setting unit 13 has a vehicle information storage unit inside. The vehicle information storage unit stores vehicle information data that has been input from the outside in advance, associating the vehicle type information of the leading car 50 with the "position of the leading car's wheels". The stopping range setting unit 13 searches the vehicle information storage data stored in the vehicle information storage unit for vehicle information data that matches the vehicle type information of the leading car 50 received from the identification reader, and sets the target stopping position of the target point 46 of the wheel 42 based on the "position of the leading car's wheels" corresponding to the matching vehicle type information of the leading car 50. If the target point extraction unit 12 extracts the lowest point of the outline of the outer diameter of the wheel 42 as the target point 46, the "position of the leading car's wheels" stored in the vehicle information storage data is the lowest point of the outline of the first wheel 42 counting from the front of the leading car 50.

[0045] Figure 8 illustrates an example of setting the target stopping position (P2) at point 46 based on vehicle type information read from an identification tag and vehicle information data stored in the vehicle information storage unit, assuming that the "position of the leading car's wheels" for a railway vehicle stationed at a station is, for example, 1800 mm. Figure 8 is a schematic diagram of the leading car 50 of a railway vehicle stopped at a station, viewed from the side, with a first target stopping position (P1) corresponding to the station predetermined for each station. When a driver stops a railway vehicle at its designated position, the point they compare with P1 at the station is, for example, the center of the crew entrance / exit door 52. When the center of the crew entrance / exit door 52 coincides with P1, the distance from the center of the crew entrance / exit door 52 to the position of the first wheel 42 from the front of the leading car 50 is the "position of the leading car's wheels: 1800 mm" included in the vehicle information storage data. Here, based on the vehicle type information it has read, if the unit 13 recognizes that the "position of the leading vehicle's wheels" is 1800 mm, it sets the target stopping position (P2) of the target point 46 to a position 1800 mm away from P1 in the opposite direction of travel. The stopping tolerance range setting unit 13 then sets the short run tolerance coordinates and the over run tolerance coordinates within the wheel image 45 so as to include P2, and sets the range between the short run tolerance coordinates and the over run tolerance coordinates as the stopping tolerance range.

[0046] Thus, if there are multiple types of railway vehicles entering the station, and the "position of the leading car's wheels" differs for each type, the stopping tolerance range setting unit 13 can efficiently change the stopping tolerance range by correcting the "position of the leading car's wheels" for each type.

[0047] Alternatively, as shown in Figure 8, the vehicle number 53 displayed on the side of the railway vehicle may be read by the identification imaging unit as vehicle identification information, and based on the vehicle information storage data relating the vehicle number 53 to the "position of the leading car's wheels," the "position of the leading car's wheels" of the railway vehicle captured by the identification imaging unit may be recognized from the vehicle number 53 to set the target stopping position of the target point 46. The identification imaging unit is, for example, a CCD camera installed on the platform 44. By setting the target stopping position of the target point 46 based on the vehicle number 53 of the railway vehicle captured by the identification imaging unit, additional modifications to the railway vehicle, such as attaching identification tags to the railway vehicle, are unnecessary, allowing vehicle identification information to be read with an inexpensive configuration and enabling the setting of a stopping tolerance range.

[0048] Furthermore, since the presence of a vehicle at a station can be determined by reading the identification ID from the identification tag using the identification reader, the identification reader may also be used as a vehicle presence detection sensor. By using the identification reader as a vehicle presence detection sensor, it is not necessary to install separate vehicle presence detection sensors when adjusting the permissible stopping range according to the vehicle type, thus minimizing the number of sensors and reducing the installation cost of the fixed-position stopping detection system 1000.

[0049] Furthermore, the vehicle information storage data may include the number of train sets and the number of cars in a railway vehicle. To give a specific example of the number of train sets and the number of cars, for example, a railway vehicle consisting of a 4-car train set and a 6-car train set coupled together would have 2 train sets and 10 cars. By including the number of train sets and the number of cars in the vehicle information storage data, the opening and closing points of the platform doors can be adjusted according to the number of train sets and cars in a railway vehicle entering the station.

[0050] Furthermore, although an example has been described in which the vehicle information storage unit is provided inside the stop allowance range setting unit 13, it may also be provided outside the stop allowance range setting unit 13.

[0051] In the example described, the point that the driver compares to station P1 when stopping a railway vehicle in a fixed position is the center of the crew boarding / alighting door 52, but it may also be the front of the leading car 50. In this case, the distance from the front of the leading car 50 to the position of the first wheel 42 counting from the front of the leading car 50 is stored in the vehicle information storage unit as the position of the leading car 50 included in the vehicle information storage data. Even when applying the fixed-position stopping detection device 200 described in this implementation, in cases where the point that the driver compares to station P1 differs among railway operators when stopping a railway vehicle in a fixed position, it is possible to easily set an appropriate target stopping position for the target point 46 for each fixed-position stopping detection device 200.

[0052] Embodiment 3. The fixed-position stop detection device 200 according to Embodiment 3 will be explained with reference to the figures. Embodiments 1 and 2 described examples where it is easy to extract the target point in the wheel image 45, but Embodiment 3 differs in that it estimates the position of the target point 46 when the wheel image 45 makes it difficult to extract the target point 46. The following explanation will focus on the differences from Embodiments 1 and 2, and explanations of the same or corresponding parts will be omitted as appropriate.

[0053] The target point extraction unit 12 performs supervised learning to learn the target point coordinates using the image patterns of each of the multiple environmental learning images 451, which are captured under at least one of the imaging conditions of sunlight, rainfall, snowfall, and fog, as well as the ground truth coordinate data of the target points 46 associated with the environmental learning images 451. The fixed-position stop detection device 200 then estimates the target point coordinates from the environmental learning images 451 based on the learned target point model. The ground truth coordinate data of the target points 46 may be input to a ground truth coordinate data input unit 33 provided outside the fixed-position stop detection device 200, or it may be stored in the fixed-position stop detection device 200 beforehand.

[0054] The generation of a pre-trained target point model for estimating target point coordinates from wheel images 45 will be described. Figure 9 is a functional block diagram showing the configuration of the target point extraction unit 12. The target point extraction unit 12 includes a target point learning data acquisition unit 121 that acquires image patterns of multiple environmental learning images 451 taken under at least one of the imaging environments of solar radiation, rainfall, snowfall, and fog, as well as ground truth coordinate data of target points 46 associated with the environmental learning images 451, as target point learning data, and a target point learning model generation unit 122 that generates a pre-trained target point model for estimating target point coordinates from wheel images 45 using the target point learning data.

[0055] The target point learning data acquisition unit 121 acquires image patterns of multiple environmental learning images 451 taken at the station under at least one of the imaging conditions of sunlight conditions, rainfall conditions, snowfall conditions, and fog conditions, as well as the ground truth coordinate data of the target points 46 associated with the environmental learning images 451, as target point learning data. The ground truth coordinate data of the target points 46 associated with the environmental learning images 451 is input, for example, by a data input unit provided outside the target point extraction unit 12. This section describes the image patterns of multiple environmental learning images 451 taken under at least one of the following imaging conditions: solar radiation, rainfall, snowfall, and fog. The image patterns may be, for example, multiple environmental learning images 451 with different imaging conditions, such as strong, medium, and weak illuminance for solar radiation conditions, or multiple environmental learning images 451 with different imaging conditions, such as less than 1 mm, 1 mm to less than 5 mm, and 5 mm or more for rainfall. Similarly, for snowfall conditions, multiple environmental learning images 451 with different imaging conditions, such as less than 2 cm of snowfall per hour, 2 cm to less than 8 cm, and 8 cm or more, may be multiple environmental learning images 451 with different imaging conditions, such as less than 50 m of visibility, 50 m to less than 200 m, and 200 m or more for fog. Furthermore, the image patterns may also be environmental learning images 451 with different degrees of insect dispersal.

[0056] When learning about sunlight, an environmental learning image 451 is prepared as learning data, as shown in Figure 10A, where sunlight strongly illuminates the area around the wheel 42, making it difficult to see the wheel 42. Then, as shown in Figure 10B, the target point learning data acquisition unit 121 acquires target point learning data associated with the correct coordinate data of the target point 46. When learning about fog 55, an environmental learning image 451 is prepared as learning data, as shown in Figure 11, 11A, in which fog 55 is covering the area around the wheels 42 and makes it difficult to see the wheels 42. Then, as shown in Figure 11, 11B, the target point learning data acquisition unit 121 acquires target point learning data associated with the correct coordinate data of the target point 46. When learning about snow 56, an environmental learning image 451 is prepared as learning data, as shown in 12A of Figure 12, in which snow 56 is floating around the wheels 42 and the wheels 42 are difficult to see. Then, as shown in 12B of Figure 12, the target point learning data acquisition unit 121 acquires target point learning data associated with the correct coordinate data of the target point 46.

[0057] The target point trained model generation unit 122 trains a target point trained model that estimates the coordinates of a target point using target point training data created based on a combination of an image pattern composed of multiple environment training images 451 with different imaging environments and the ground truth coordinate data of the target point 46. In other words, it generates a trained model that infers the optimal position of the target point 46 from an image pattern composed of multiple environment training images 451 with different imaging environments and the ground truth coordinate data of the target point 46. Here, the target point training data is data that associates an image pattern composed of multiple environment training images 451 with different imaging environments with the ground truth coordinate data of the target point 46.

[0058] The target point trained model generation unit 122 can use any known learning algorithm, such as supervised learning, unsupervised learning, or reinforcement learning. As an example, the case where a neural network is applied will be described.

[0059] The target point trained model generation unit 122 learns the target point coordinates, for example, by supervised learning according to a neural network model. Here, supervised learning is a method in which pairs of input and result (label) data are provided to the target point extraction unit 12, the unit learns features in the training data, and infers the result from the input.

[0060] A neural network consists of an input layer made up of multiple neurons, an intermediate layer (hidden layer) made up of multiple neurons, and an output layer made up of multiple neurons. The intermediate layer can be one or more layers.

[0061] For example, in a three-layer neural network as shown in Figure 13, when multiple inputs are input to the input layer (X1-X3), these values ​​are multiplied by weights W1 (w11-w16) and input to the hidden layer (Y1-Y2). The result is then multiplied again by weights W2 (w21-w26) and output from the output layer (Z1-Z3). This output varies depending on the values ​​of weights W1 and W2.

[0062] The neural network learns the coordinates of target points through so-called supervised learning, according to target point learning data created based on a combination of image patterns consisting of multiple environment learning images 451 with different imaging environments and the ground truth coordinate data of the target points 46.

[0063] In other words, a neural network learns by inputting motion information and movement information into the input layer and adjusting the weights W1 and W2 so that the output from the output layer approaches the correct label.

[0064] The target point trained model generation unit 122 generates and outputs a trained model by performing the training described above.

[0065] The target point trained model storage unit 123 stores the target point trained model output from the target point trained model generation unit 122.

[0066] Next, we will explain the learning process of the target point extraction unit 12 using Figure 14. Figure 14 is a flowchart showing the learning flow of the target point extraction unit 12.

[0067] In step S101, the target point learning data acquisition unit 121 acquires an image pattern composed of multiple environmental learning images 451 with different imaging environments, and the ground truth coordinate data of the target point 46. Although the image pattern composed of multiple environmental learning images 451 with different imaging environments and the ground truth coordinate data of the target point 46 are acquired simultaneously, it is sufficient if the image pattern composed of multiple environmental learning images 451 with different imaging environments and the ground truth coordinate data of the target point 46 can be input in association with each other, and the image pattern composed of multiple environmental learning images 451 with different imaging environments and the ground truth coordinate data of the target point 46 may be acquired at different times.

[0068] In step S102, the target point trained model generation unit 122 learns the target point coordinates and generates a target point trained model by so-called supervised learning, according to training data created based on a combination of an image pattern composed of multiple environmental training images 451 with different imaging environments acquired by the target point training data acquisition unit 121 and the ground truth coordinate data of the target point 46.

[0069] In step S103, the target point trained model storage unit 123 stores the target point trained model for estimating the target point coordinates, which was generated by the target point trained model generation unit 122.

[0070] Next, the use of a pre-trained target point model for estimating the target point coordinates in this embodiment will be described. As shown in Figure 15, the target point extraction unit 12 includes an image data acquisition unit 124 that acquires wheel images 45 from the wheel image acquisition unit 11, and a target point estimation processing unit 125 that uses a pre-trained target point model for estimating the target point coordinates from the wheel images 45 to output the target point coordinates from the wheel images 45 acquired by the image data acquisition unit 124.

[0071] The image data acquisition unit 124 acquires a wheel image 45 from the wheel image acquisition unit 11 in which the wheel 42 is included in the imaging field of view. It is also possible to acquire a wheel image 45 in which the wheel 42 is not clearly visible.

[0072] The target point estimation processing unit 125 estimates the target point coordinates based on the data received from the image data acquisition unit 124 and the target point trained model stored in the target point trained model storage unit 123, and outputs the estimation result to the stop position determination unit 14. By inputting the wheel image 45 acquired by the image data acquisition unit 124 into this target point trained model, the target point coordinates estimated based on the wheel image 45 can be output.

[0073] Next, using Figure 16, we will explain the process for obtaining the target point coordinates using the target point extraction unit 12.

[0074] In step S201, the image data acquisition unit 124 acquires a wheel image 45.

[0075] In step S202, the target point extraction unit 12 retrieves the target point trained model from the target point trained model storage unit 123, inputs the wheel image 45 acquired by the image data acquisition unit 124 into the retrieved target point trained model, and obtains the target point coordinates in the wheel image 45.

[0076] In step S203, the target point extraction unit 12 outputs the target point coordinates obtained by the target point trained model to the stop position determination unit 14. This allows for accurate estimation of the target point coordinates based on the wheel image 45 in which the wheel 42 is not clearly visible.

[0077] The explanation described the case where supervised learning is applied to the learning algorithm used by the target point trained model generation unit 122, but this is not the only case. In addition to supervised learning, reinforcement learning or semi-supervised learning can also be applied to the learning algorithm.

[0078] Thus, in the fixed-position stop detection device 200 according to Embodiment 3, by estimating the target point coordinates using a target point trained model for estimating the target point coordinates, it is possible to accurately estimate the target point coordinates even if the wheel image 45 is unclear, and to determine with high accuracy whether or not the railway vehicle has stopped at a fixed position.

[0079] In this embodiment, the estimated target point coordinates are output using a target point learned model stored in the target point learned model storage unit 123 provided in the fixed-position stop detection device 200. However, a target point learned model may be acquired from an external source, such as another fixed-position stop detection device 200, and the estimated target point coordinates are output based on this learned model. By acquiring a target point learned model from an external source, the fixed-position stop detection system 1000 can be updated to an appropriate learned model even after it has started operating, allowing for more accurate estimation of the target point coordinates and enabling high-precision determination of whether or not the railway vehicle has stopped at the fixed position.

[0080] Embodiment 4. The fixed-position stop detection device 200 according to Embodiment 4 will be explained with reference to the figures. Embodiments 1 to 3 described an example in which the wheel image 45 is acquired under constant imaging conditions for the wheel image 45. Embodiment 4 differs in that it corrects the imaging conditions when capturing the wheel image 45 in an imaging environment where the wheel 42 is expected to be difficult to see in the wheel image 45. The following will mainly explain the differences from Embodiments 1 to 3, and explanations of the same or corresponding parts will be omitted as appropriate.

[0081] As shown in Figure 17, the fixed-position stop detection device 200 includes an imaging control unit 16 that communicates with the imaging unit 31 and controls the imaging conditions for capturing wheel images 45. According to the fixed-position stop detection device 200 of this embodiment, for example, if the sunlight 54 is strong and the target point 46 may become blurred, the exposure time can be shortened or parameters can be set to remove the sunlight 54 as noise. Also, if snow 56 overlaps the target point 46 during snowfall and it may not be possible to image the target point 46, the exposure time can be lengthened or the noise canceling filter strength can be increased. By adjusting the imaging conditions in this way, the imaging conditions can be corrected to be appropriate.

[0082] This section describes the generation of a pre-trained imaging condition model for estimating imaging conditions from climate data. Figure 18 is a functional block diagram showing the configuration of the imaging control unit 16. The imaging control unit 16 includes an imaging condition training data acquisition unit 161 and an imaging condition pre-trained model generation unit 162.

[0083] The imaging condition learning data acquisition unit 161 receives climate data at the time of imaging from the climate data acquisition unit 34, receives stored wheel images associated with the climate data from the wheel image acquisition unit 11, and receives correct imaging condition data, including appropriate imaging conditions for each stored wheel image, from the correct imaging condition data input unit 35. The climate data, stored wheel images, and correct imaging condition data are acquired as data for learning imaging conditions. The correct imaging condition data may be input to the correct imaging condition data input unit 35 located outside the fixed-position stop detection device 200, or it may be stored in advance within the fixed-position stop detection device 200. The imaging condition learning data acquisition unit 161 preferably acquires climate data at the time of imaging from outside the fixed-position stop detection device 200. However, the fixed-position stop detection device 200 may be equipped with a measurement sensor for measuring climate data, and the climate data measured by the measurement sensor may be stored. Climate data may include, for example, the date and time of imaging, solar radiation, rainfall, snowfall, visibility when fog occurs, temperature, etc. Here, the wheel images used for training do not necessarily have to show the wheel 42. The correct imaging condition data associated with the stored wheel images is input, for example, by a data input unit located outside the imaging control unit 16. The correct imaging condition data consists of, for example, the exposure time during imaging, the shutter interval, the noise canceling filter, etc.

[0084] The imaging condition trained model generation unit 162 generates an imaging condition trained model for estimating imaging conditions from wheel images 45 using imaging condition training data.

[0085] The imaging control unit 16 performs supervised learning to learn imaging conditions using climate data acquired when the accumulated wheel images were captured, the accumulated wheel images acquired when the aforementioned climate data was obtained, and appropriate correct imaging condition data for each acquired accumulated wheel image. Then, the imaging control unit 16 estimates the imaging conditions from the climate data based on the learned imaging condition trained model.

[0086] The imaging condition trained model generation unit 162 trains an imaging condition trained model that estimates imaging conditions using imaging condition training data created based on a combination of climate data, accumulated wheel images, and correct imaging condition data. The imaging conditions include exposure time, shutter interval, noise canceling filter, etc., during imaging. In other words, a model trained for imaging conditions is generated that infers the optimal imaging conditions from climate data, wheel images 45 with different climate conditions, and ground truth data for imaging conditions. Here, the data for training the imaging conditions is data in which the climate data, wheel images 45, and ground truth data for imaging conditions are related to each other.

[0087] The learning algorithm used by the imaging condition-learned model generation unit 162 can be any known algorithm, such as supervised learning, unsupervised learning, or reinforcement learning. For example, when a neural network is applied, it is learned in the same way as described in Embodiment 5.

[0088] The imaging condition learned model storage unit 163 stores the imaging condition learned model output from the imaging condition learned model generation unit 162.

[0089] Next, we will explain the process that the imaging control unit 16 learns. Although not shown in the diagram, we will explain by substituting the wording of the steps in Figure 14. In Figure 14, S101 will be read as S301, S102 as 302, S103 as S303, and the phrase "store the target point learned model" in S103 will be read as "store the imaging condition learned model" for the explanation.

[0090] In step S301, the imaging condition learning data acquisition unit 161 acquires climate data, accumulated wheel images, and correct imaging condition data. Although the climate data, accumulated wheel images, and correct imaging condition data are acquired simultaneously, it is sufficient if the climate data, accumulated wheel images, and correct imaging condition data are input in association with each other, and the climate data, accumulated wheel images, and correct imaging condition data may be acquired at different times.

[0091] In step S302, the imaging condition trained model generation unit 162 learns the imaging conditions and generates an imaging condition trained model by so-called supervised learning, according to the training data created based on the combination of climate data acquired by the imaging condition training data acquisition unit 161, accumulated wheel images, and correct imaging condition data.

[0092] In step S303, the imaging condition learned model storage unit 163 stores the imaging condition learned model for estimating imaging conditions, which was generated by the imaging condition learned model generation unit 162.

[0093] Next, the use of a pre-trained target point model for estimating the coordinates of the target point in this embodiment will be described. As shown in Figure 19, the imaging control unit 16 includes a climate data acquisition unit 34 that acquires climate data from outside the fixed-position stop detection device 200, and an imaging condition estimation processing unit 164 that uses a pre-trained imaging condition model for estimating imaging conditions from the climate data to output imaging conditions from the climate data acquired by the climate data acquisition unit 34.

[0094] The climate data acquisition unit 34 acquires solar radiation, rainfall, snowfall, visibility during fog, temperature, etc., from outside the fixed-position stop detection device 200. Alternatively, the fixed-position stop detection device 200 may be equipped with climate data measurement sensors to acquire the data.

[0095] The imaging condition estimation processing unit 164 estimates imaging conditions based on the data received from the climate data acquisition unit 34 and the imaging condition learned model stored in the imaging condition learned model storage unit 163, and outputs the estimation result to the imaging unit 31. By inputting the climate data acquired by the climate data acquisition unit 34 into this imaging condition learned model, it is possible to output imaging conditions estimated based on the climate data.

[0096] Next, we will explain the process for obtaining imaging conditions using the imaging control unit 16. Although not shown in the diagram, we will explain by substituting the wording of the steps in Figure 15. S201 will be read as S401, S202 as 402, S203 as S403, and "Input to the target point learned model" in S202 will be read as "Input to the imaging condition learned model" in the explanation.

[0097] In step S401, the climate data acquisition unit 34 acquires climate data such as solar radiation, rainfall, snowfall, visibility during fog, and temperature.

[0098] In step S402, the imaging control unit 16 obtains an imaging condition learned model from the imaging condition learned model storage unit 163, inputs the climate data acquired by the climate data acquisition unit 34 into the obtained imaging condition learned model, and obtains imaging conditions.

[0099] In step S403, the imaging control unit 16 outputs the imaging conditions obtained from the learned imaging condition model to the imaging unit 31. By adjusting the imaging conditions for the wheel image 45, the target point 46 of the wheel 42 can be clearly imaged.

[0100] The description above describes the case where supervised learning is applied to the learning algorithm used by the imaging condition-learned model generation unit 162, but this is not the only case. In addition to supervised learning, reinforcement learning or semi-supervised learning can also be applied to the learning algorithm.

[0101] Thus, the fixed-position stop detection device 200 according to Embodiment 4 can clearly image the target point 46 of the wheel 42 by optimizing the imaging conditions of the wheel image 45 according to the weather, and accurately determine the coordinates of the target point, thereby enabling high-precision determination of whether or not the railway vehicle has stopped at a fixed position.

[0102] In this embodiment, the estimated target point coordinates are output using a target point learned model stored in the target point learned model storage unit 123 provided in the fixed-position stop detection device 200. However, a target point learned model may be acquired from an external source, such as another fixed-position stop detection device 200, and the estimated target point coordinates are output based on this learned model. By acquiring a target point learned model from an external source, the fixed-position stop detection system 1000 can be updated to an appropriate learned model even after it has started operating, allowing for more accurate estimation of the target point coordinates and enabling high-precision determination of whether or not the railway vehicle has stopped at the fixed position.

[0103] Embodiment 5. The fixed-position stop detection device 200 according to Embodiment 5 will now be described. Embodiments 1 to 4 described examples of determination when the wheel image 45 clearly shows the wheel 42, the railway vehicle body 43, etc., but Embodiment 5 differs in that it detects whether or not there are defects in the acquired wheel image 45 caused by dirt, malfunction of the imaging unit 31, etc. The following will mainly describe the differences from Embodiments 1 to 4, and descriptions of the same or corresponding parts will be omitted as appropriate.

[0104] The fixed-position stop detection device 200 according to this embodiment includes a defect determination unit that determines whether or not a defect has occurred in the wheel image 45. The defect determination unit determines that a defect has occurred in the wheel image 45 if a defect of a size greater than or equal to a defect size threshold has occurred in the target point coordinates, overrun allowable coordinates, short run allowable coordinates, etc., or if the pixel signal is less than or equal to a set pixel signal threshold. The determination result output unit 15 of the fixed-position stop detection device 200 then outputs a defect occurrence signal to a terminal installed in the station office, train control room, etc., if the defect determination unit has determined that a defect has occurred.

[0105] Thus, by including a defect determination unit in the fixed-position stop detection device 200 that determines whether or not a defect has occurred in the wheel image 45, it is possible to remotely monitor whether or not the imaging unit 31 of the fixed-position stop detection system 1000 is operating normally. If the cause of the defect lies in the imaging unit 31, the imaging unit 31 can be subjected to condition-based maintenance.

[0106] Embodiment 6. The fixed-position stop detection device 200 according to Embodiment 6 will be explained with reference to the figures. In Embodiments 1 to 5, even if an overrun or short run occurs, an example was described in which the wheel 42 shown in the wheel image 45 is the first wheel 42 from the front of the leading car 50. However, in Embodiment 6, an example of determination will be described in which, for example, the railway vehicle overruns significantly and the wheel image 45 includes the second and subsequent wheels 42 from the front of the leading car 50. The following explanation will focus on the differences from Embodiments 1 to 5, and explanations of the same or corresponding parts will be omitted as appropriate.

[0107] The fixed-position stop detection device 200 according to this embodiment includes an electromagnetic wave information acquisition unit that acquires electromagnetic wave information emitted from an irradiation source based on the position in which the railway vehicle stops at an overrun-permissible position. The stop position determination unit 14 of the fixed-position stop detection device 200 acquires electromagnetic wave information indicating that the electromagnetic waves are not obstructed by the leading car 50, and determines that the railway vehicle has stopped at the fixed position if the target point 46 is within the permissible stopping range.

[0108] Even if the lead car 50 overruns significantly as shown in Figure 20, and the second and subsequent wheels 42 from the front of the lead car 50 are included in the imaging field of view 57, an electromagnetic wave irradiation unit is provided so that the target point extraction unit 12 does not incorrectly determine that the railway vehicle has stopped at its designated position if the target point coordinates are within the acceptable stopping range. The electromagnetic wave irradiation unit irradiates electromagnetic waves in a direction intersecting the direction of travel of the railway vehicle. Electromagnetic waves include visible light, infrared light, laser light, etc. If the electromagnetic wave irradiation unit has an integrated light-emitting and light-receiving unit, the distance between the railway vehicle body and the electromagnetic wave irradiation unit is detected, and if this distance is detected, it is determined that the railway vehicle has overrun by a greater distance than the distance between the first and second wheels 42 from the front of the lead car 50. If the electromagnetic wave irradiation unit does not have a light receiving unit, an electromagnetic wave light receiving unit is installed on the opposite side of the railway vehicle from the electromagnetic wave irradiation unit. This allows detection that if the railway vehicle significantly overruns the station, the electromagnetic waves reaching the electromagnetic wave receiving unit will be blocked. In this case, as shown in Figure 20, if the shape of the front part of the leading car 50 is streamlined, the electromagnetic wave irradiation unit is installed so as to irradiate electromagnetic waves at the same height as the Y-axis height of the most forward-protruding part.

[0109] In this way, by irradiating electromagnetic waves in a direction intersecting the direction of travel of the railway vehicle, it is possible to correctly determine that the railway vehicle has not stopped in its designated position if it significantly overruns the station and stops there, thereby preventing malfunctions of the platform doors when an overrun occurs.

[0110] Embodiment 7. A fixed-position stop detection device 200 according to Embodiment 7 will now be described. Embodiments 1 to 6 described an example in which a wheel image 45 including the wheels 42 of a railway vehicle is acquired and whether or not the railway vehicle has stopped at a fixed position is determined based on the wheel image 45. Embodiment 7 differs in that a door image including the passenger boarding / alighting door 51 of a railway vehicle is acquired and the target point 46 is extracted and determined within the passenger boarding / alighting door 51 included in the door image. The following will mainly describe the differences from Embodiments 1 to 6, and descriptions of the same or corresponding parts will be omitted as appropriate.

[0111] The fixed-position stop detection device 200 according to this embodiment includes: a door image acquisition unit that acquires a door image that includes the passenger boarding / alighting door 51 on the side with respect to the direction of travel of the railway vehicle; a target point extraction unit 12 that extracts a specific position of the passenger boarding / alighting door 51 from the door image as a target point 46 and determines the coordinates of the target point; a stop tolerance range setting unit 13 that determines the overrun tolerance coordinates from the allowable overrun position of the railway vehicle set in the door image and the short run tolerance coordinates from the allowable short run position, and sets the range between the overrun tolerance coordinates and the short run tolerance coordinates as the stop tolerance range; a stop position determination unit 14 that determines that the railway vehicle has stopped at a fixed position if the target point coordinates are within the stop tolerance range from the time the target point coordinates exceed the short run tolerance coordinates in the direction of travel of the railway vehicle until a set time has elapsed; and a determination result output unit 15 that outputs the determination result of whether or not the railway vehicle has stopped at a fixed position. In other words, if the passenger boarding / alighting door 51 included in the door image consists of two doors, an arbitrary point on the line where the left and right doors are in contact is extracted as the target point 46, and the coordinates of the target point in the direction of travel of the railway vehicle are determined. By determining whether the determined target point coordinates are within the acceptable stopping range, it is possible to determine with high accuracy whether the railway vehicle has stopped in its designated position.

[0112] The door image acquisition unit acquires door images taken from the platform 44 in a direction intersecting the direction of travel of the railway vehicle at specific acquisition intervals and transmits them to the target point extraction unit 12. For example, door images are acquired from the imaging unit 31 via wireless communication.

[0113] The target point extraction unit 12 receives a door image from the door image acquisition unit. If the door image includes a passenger boarding / alighting door 51, the unit also uses an object detection model such as YOLO to detect the passenger boarding / alighting door 51 of the railway vehicle. The unit extracts the target points 46 of the door from the door image and determines the target point coordinates. If the passenger boarding / alighting door 51 consists of two doors, the target point may be any point on the tangent line where the two closed doors meet, or any one of the four corners of the door.

[0114] The stop tolerance range setting unit 13 sets a predetermined area within the captured door image as the stop tolerance range. Alternatively, it may read the allowable overrun position and allowable short run position of the target point 46 each time a determination is made, or it may store the read allowable overrun position and allowable short run position.

[0115] In this way, by acquiring door images and extracting specific positions of the doors as target points 46, it is possible to determine with high accuracy whether or not a railway vehicle has stopped in its designated position, even at stations where there is no space below the platform 44.

[0116] Although the example shown includes obtaining a door image that includes the passenger boarding / alighting door 51, any location other than the passenger boarding / alighting door 51 may be used as the target point 46, as long as it moves along with the train's movement. For example, windows on the side relative to the direction of travel, crew boarding / alighting doors 52, etc., may also be used.

[0117] Embodiment 8. The fixed-position stop detection device 200 according to Embodiment 8 will be explained with reference to the figures. In Embodiments 1 to 7, an example was described in which the target point 46 is located on the railway vehicle side, but in Embodiment 8, the difference is that the stop reference mark 58 provided in the space below the platform 44 is extracted as the target point 46. The following explanation will focus on the differences from Embodiments 1 to 7, and explanations of the same or corresponding parts will be omitted as appropriate.

[0118] The fixed-position stop detection device 200 according to this embodiment includes: a platform 44 image acquisition unit that acquires a platform 44 image including a stop reference mark 58 provided on the platform 44, which is imaged from a railway vehicle; a target point extraction unit 12 that extracts the stop reference mark 58 from the platform 44 image as a target point 46 and determines the target point coordinates; a stop allowable range setting unit 13 that determines the overrun allowable coordinates from the allowable overrun position of the railway vehicle set on the platform 44 image, and the short run allowable coordinates from the allowable short run position, and sets the range between the overrun allowable coordinates and the short run allowable coordinates as the stop allowable range; a stop position determination unit 14 that determines that the railway vehicle has stopped at the fixed position if the target point coordinates are within the stop allowable range from the time the target point coordinates exceed the short run allowable coordinates in the direction of travel of the railway vehicle until a set time has elapsed; and a determination result output unit 15 that outputs the determination result of whether or not the railway vehicle has stopped at the fixed position.

[0119] Figure 21 is a schematic cross-sectional view of a railway vehicle when it has stopped at a designated position, cut in a plane perpendicular to the direction of travel. As shown in Figure 21, the railway vehicle is equipped with an imaging unit 31 and a designated position stop detection device 200. The imaging unit 31 captures an image of the platform 44, including a stop reference mark 58 located in the space below the platform 44. The stop reference mark 58 is positioned so that it appears in the center of the width direction of the platform 44 image when, for example, the railway vehicle has stopped at a designated position at a station.

[0120] The stopping tolerance range setting unit 13 determines the overrun tolerance coordinates and the short run tolerance coordinates from the overrun tolerance position and the short run tolerance position set in the platform 44 image so as to move along with the progress of the railway vehicle, and sets the range between the overrun tolerance coordinates and the short run tolerance coordinates as the stopping tolerance range.

[0121] Thus, since the installation work on the platform 44 side only involves adding a stopping reference mark 58 below platform 44, it is possible to determine with high accuracy whether or not a railway vehicle has stopped in the correct position while keeping modifications to station facilities to a minimum.

[0122] Embodiment 9. The operation of the fixed-position stop detection device 200 described in Embodiments 1 to 8 will now be explained. Figure 22 is a flowchart showing the processing routine executed by the fixed-position stop detection device 200. The fixed-position stop detection device 200 first acquires a wheel image 45 that includes the wheels 42 of the railway vehicle (step S501). Next, it extracts the position of the target point 46 from the wheel image 45 and determines the coordinates of the target point within the wheel image 45 (step S502). Then, it sets the range between the preset overrun allowable position and the short run allowable position as the stop allowable range (step S503). Based on the set stop allowable range and the target point coordinates, it determines whether the target point coordinates are within the stop allowable range (S504). If it is determined in S504 that the target point coordinates are within the stop allowable range (YES in S504), it transmits to the platform door opening / closing control unit 32 that the railway vehicle has stopped at the fixed position (S505). If it is determined in S104 that the target point coordinates are not within the stop allowable range (NO in S504), it notifies the driver's cab that the railway vehicle has not stopped at the fixed position (S506).

[0123] Here, each function of the fixed-position stop detection device 200 is realized by a processing circuit. Figure 23 is a schematic diagram showing an example of a processing circuit that realizes each function of the fixed-position stop detection device 200. The fixed-position stop detection device 200 includes a processor 59, a storage device 60, a communication I / F (interface) 61, etc. For example, a CPU (Central Processing Unit) is used as the processor 59. The storage device 60 transmits and receives data to and from the processor 59 and stores the data. The wheel image 45 captured by the imaging unit 31 is acquired by the wheel image acquisition unit 11 via the communication interface 61. The calculations and determinations of the wheel image acquisition unit 11, target point extraction unit 12, stop tolerance range setting unit 13, and stop position determination unit 14 are performed by the processor 59. The parameters and calculation formulas used for determination are stored in the storage device 60. The determination result is transmitted to the opening / closing control unit 32 via the communication interface 61.

[0124] The processor 59 and memory device 60 may be shared by a single unit, or there may be multiple units. The processor 59 may also be equipped with logic circuits using, for example, an ASIC (Application Specific Integrated Circuit), an IC (Integrated Circuit), a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), and various signal processing circuits. By providing multiple processors 59 of the same or different types, each process may be divided and executed by multiple arithmetic processing units.

[0125] The multiple storage devices 60 include, for example, RAM (Random Access Memory) configured to allow reading and writing of data from the processor 59, ROM (Read Only Memory) configured to allow reading of data from the processor 59, and a hard disk drive (HDD).

[0126] Each function of the fixed-position stop detection device 200 is realized by the processor 59 executing software or program 63 stored in the storage device 60 and cooperating with the hardware. The setting data to be configured for the fixed-position stop detection device 200 may be stored in the storage device 60 as part of the software or program 63, or it may be made available for user input. A non-temporary recording medium 62 on which program 63 for the work monitoring support device is recorded may be distributed and installed in the storage device 60 of the fixed-position stop detection device 200.

[0127] In this way, by obtaining the target point coordinates of the wheel 42 from the acquired wheel image 45 and determining whether or not the target point coordinates are within the acceptable stopping range, a target point 46 can be set and its position accurately determined, thereby enabling high-precision determination of whether or not the stopping position of the railway vehicle is in a fixed position.

[0128] In addition, the steps described include acquiring a wheel image 45 that includes the wheels 42 of a railway vehicle, extracting a specific position of the wheel 42 from the wheel image 45 as a target point, and determining the coordinates of the target point. However, as described in Embodiment 7, the method may also include acquiring a door image that includes the passenger boarding / alighting doors on the side, extracting a target point from the door image, and determining the coordinates of the target point. Furthermore, as described in Embodiment 8, the method may also include acquiring a platform image that includes the stopping reference marks provided on the platform, which are captured from the railway vehicle, extracting a target point from the platform image, and determining the coordinates of the target point.

[0129] While this disclosure describes various exemplary embodiments, the various features, aspects, and functions described in one or more embodiments are not limited to the application of a particular embodiment, but are applicable individually or in various combinations to the embodiments. Therefore, countless variations not illustrated are conceivable within the scope of the art disclosed herein. These include, for example, modifying, adding, or omitting at least one component, or even extracting at least one component and combining it with components from other embodiments.

[0130] The various aspects of this disclosure are summarized below as an appendix. (Note 1) A wheel image acquisition unit that acquires wheel images that include the wheels of a railway vehicle, A target point extraction unit extracts a specific position of the wheel from the wheel image and determines the coordinates of the target point. A stop tolerance range setting unit determines the overrun tolerance coordinates from the overrun tolerance position of the railway vehicle on the rail on which the wheel runs within the wheel image or from the pre-set short-run tolerance position within the wheel image, and sets the range between the overrun tolerance coordinates and the short-run tolerance coordinates as the stop tolerance range. A stop position determination unit determines that the railway vehicle has stopped at a fixed position if, from the time the target point coordinates exceed the short-run allowable coordinates in the direction of travel of the railway vehicle until a set time has elapsed, the target point coordinates are within the stop allowable range, A determination result output unit that outputs a determination result of whether or not the railway vehicle has stopped at the designated position. A fixed-position stop detection device equipped with the following features.

[0131] (Note 2) The target point extraction unit is a fixed-position stop detection device as described in Appendix 1, which extracts the lowest point of the outer diameter portion of the wheel as the target point.

[0132] (Note 3) The target point extraction unit is a fixed-position stop detection device as described in Appendix 1, which extracts the contact point between the outer diameter of the wheel and the rail as the target point.

[0133] (Note 4) It includes a stop-permitted range storage unit that stores the stop-permitted range in the wheel image captured with a fixed imaging field, The stop tolerance range setting unit sets the stop tolerance range received from the stop tolerance range storage unit, as described in any one of the items 1 to 3 of the fixed-position stop detection device.

[0134] (Note 5) The stop tolerance range setting unit sets the target stop position of the target point based on the vehicle identification information of the railway vehicle, and the fixed-position stop detection device according to any one of the appendices 1 to 4 determines the overrun tolerance coordinates and the short run tolerance coordinates based on the target stop position.

[0135] (Note 6) The fixed-position stop detection device according to Appendix 5, wherein the vehicle identification information is at least one of the vehicle number read from the railway vehicle and vehicle type information associated with an identification tag provided on the railway vehicle.

[0136] (Note 7) The aforementioned target point extraction unit is: A target point learning data acquisition unit acquires image patterns of multiple environmental learning images taken under at least one of the imaging conditions of solar radiation, rainfall, snowfall, and fog formation, as well as the ground truth coordinate data of the target point associated with the environmental learning image, as target point learning data. A fixed-position stop detection device according to any one of the appendices 1 to 6, further comprising: a target point trained model generation unit that generates a target point trained model for estimating the target point coordinates from the wheel image using the target point training data.

[0137] (Note 8) The aforementioned target point extraction unit is: An image data acquisition unit that acquires the wheel image from the wheel image acquisition unit, A fixed-position stop detection device according to any one of the appendices 1 to 7, having a target point estimation processing unit that outputs the target point coordinates from the wheel image acquired by the image data acquisition unit, using a target point trained model for estimating the target point coordinates from the wheel image.

[0138] (Note 9) The system includes an imaging control unit that controls the imaging conditions for the wheel image, The imaging control unit, An imaging condition learning data acquisition unit acquires climate data at the time of imaging, accumulated wheel images associated with the climate data, and correct imaging condition data associated with the accumulated wheel images as imaging condition learning data. A fixed-position stop detection device according to any one of the appendices 1 to 8, further comprising: an imaging condition learned model generation unit that generates an imaging condition learned model for estimating the imaging conditions from the wheel image using the imaging condition learning data.

[0139] (Note 10) The system includes an imaging control unit that controls the imaging conditions for the wheel image, The imaging control unit, A climate data acquisition unit that acquires climate data during imaging, An imaging condition estimation processing unit that estimates the imaging conditions from the climate data acquired by the climate data acquisition unit using an imaging condition trained model for estimating the imaging conditions from the climate data, A fixed-position stop detection device according to any one of the appendices 1 to 9, further comprising an imaging condition output unit for controlling the aforementioned imaging conditions.

[0140] (Note 11) The system includes a defect determination unit that determines whether or not there is a defect in the wheel image, The defect detection unit determines that a defect has occurred in the wheel image if, in the wheel image, a defect of a size greater than or equal to the defect size threshold has occurred in at least one of the target point coordinates, the overrun tolerance coordinates, and the short run tolerance coordinates, or if the pixel signal is less than or equal to the set pixel signal threshold. The fixed-position stop detection device according to any one of the appendices 1 to 10, wherein the judgment result output unit outputs a loss occurrence signal when the loss determination unit determines that a loss has occurred.

[0141] (Note 12) The system includes an electromagnetic wave information acquisition unit that acquires electromagnetic wave information emitted from an irradiation source based on the position in which the railway vehicle stops at the overrun allowable position, The fixed-position stop detection device according to any one of the appendices 1 to 11, wherein the stop position determination unit acquires electromagnetic wave information indicating that electromagnetic waves are not obstructed by the leading vehicle, and determines that the railway vehicle has stopped at a fixed position when the target point is within the range of the permitted stopping range.

[0142] (Note 13) The fixed-position stop detection device according to any one of the appendices 1 to 12, wherein the determination result output unit outputs the distance from the stopping position of the target point to the target stopping position, and the direction indicating an overrun or short run, when the stopping position determination unit determines that the railway vehicle is not stopped at the fixed position.

[0143] (Note 14) A door image acquisition unit acquires door images that include the passenger boarding / alighting doors on the side of the railway vehicle in relation to the direction of travel, A target point extraction unit extracts a specific position of the passenger boarding / alighting door from the aforementioned door image and determines the coordinates of the target point, A stop tolerance range setting unit determines the overrun tolerance coordinates from the overrun tolerance position of the railway vehicle set in the door image, and the short run tolerance coordinates from the short run tolerance position, and sets the range between the overrun tolerance coordinates and the short run tolerance coordinates as the stop tolerance range. A stop position determination unit determines that the railway vehicle has stopped at a fixed position if, from the time the target point coordinates exceed the short-run allowable coordinates in the direction of travel of the railway vehicle until a set time has elapsed, the target point coordinates are within the stop allowable range, The aforementioned railway vehicle is a determination result output unit that outputs a determination result of whether or not it has stopped at the designated position. A fixed-position stop detection device equipped with the following features.

[0144] (Note 15) A train occupancy detection sensor that detects when a train enters a station, When the presence detection sensor detects that the railway vehicle is present at the station, the imaging unit captures images of the rails on which the railway vehicle is traveling. A fixed-position stop detection device according to any one of claims 1 to 14, which acquires at least one of a wheel image including the wheel and a door image including the door from the imaging unit, Based on the determination result of the aforementioned fixed-position stop detection device, an opening / closing control unit controls the opening and closing of platform doors installed on the station platform. A fixed-position stop detection system equipped with the following features.

[0145] (Note 16) The steps include obtaining a wheel image that includes the wheels of a railway vehicle, The steps include extracting a specific position of the wheel from the wheel image and determining the coordinates of the target point, The steps include: determining the overrun allowable coordinates from the rail on which the wheel runs within the wheel image or from the allowable overrun position of the railway vehicle predetermined on the wheel image, determining the short run allowable coordinates from the allowable short run position, and setting the range between the overrun allowable coordinates and the short run allowable coordinates as the stop allowable range; The step of determining that the railway vehicle has stopped at a fixed position if, from the time the target point coordinates exceed the short-run allowable coordinates in the direction of travel of the railway vehicle until a set time has elapsed, the target point coordinates are within the stop allowable range, A step of outputting a determination result of whether or not the railway vehicle has stopped at the designated position. A fixed-position stop detection method comprising the following:

[0146] (Note 17) A platform image acquisition unit that acquires platform images including stopping reference marks installed on the platform, which are captured from a railway vehicle, A target point extraction unit extracts the stop reference marks in the platform image as target points and determines the coordinates of the target points, A stop tolerance range setting unit determines the overrun tolerance coordinates from the overrun tolerance position of the railway vehicle set on the platform image, and the short run tolerance coordinates from the short run tolerance position, and sets the range between the overrun tolerance coordinates and the short run tolerance coordinates as the stop tolerance range. A stop position determination unit determines that the railway vehicle has stopped at a fixed position if, from the time the target point coordinates exceed the short-run allowable coordinates in the direction of travel of the railway vehicle until a set time has elapsed, the target point coordinates are within the stop allowable range, The aforementioned railway vehicle is a determination result output unit that outputs a determination result of whether or not it has stopped at the designated position. A fixed-position stop detection device equipped with the following features. [Explanation of symbols]

[0147] 11 Wheel image acquisition unit, 12 Target point extraction unit, 13 Stop tolerance range setting unit, 14 Stop position determination unit, 15 Determination result output unit, 16 Image capture control unit, 31 Image capture unit, 32 Open / close control unit, 33 Correct coordinate data input unit, 34 Climate data acquisition unit, 35 Correct image condition data input unit, 41 Rail, 42 Wheel, 43 Railway vehicle body, 44 Platform, 45 Wheel image, 46 Target point, 49 Tolerable position mark, 50 Leading car, 51 Passenger boarding / alighting door, 52 Crew boarding / alighting door, 53 Vehicle number, 54 Sunlight, 55 Fog, 56 Snow, 57 Image capture field of view, 58 Stop reference mark, 200 Fixed position stop detection device, 1000 Fixed position stop detection system

Claims

1. A wheel image acquisition unit that acquires wheel images that include the wheels of a railway vehicle, A target point extraction unit extracts a specific position of the wheel from the wheel image and determines the coordinates of the target point. A stop tolerance range setting unit determines the overrun tolerance coordinates from the overrun tolerance position of the railway vehicle on the rail on which the wheel runs within the wheel image or from the pre-set short-run tolerance position within the wheel image, and sets the range between the overrun tolerance coordinates and the short-run tolerance coordinates as the stop tolerance range. A stop position determination unit determines that the railway vehicle has stopped at a fixed position if, from the time the target point coordinates exceed the short-run allowable coordinates in the direction of travel of the railway vehicle until a set time has elapsed, the target point coordinates are within the stop allowable range, A determination result output unit that outputs a determination result of whether or not the railway vehicle has stopped at the designated position. A fixed-position stop detection device equipped with the following features.

2. The fixed-position stop detection device according to claim 1, wherein the target point extraction unit extracts the lowest point of the outer diameter portion of the wheel as the target point.

3. The fixed-position stop detection device according to claim 1, wherein the target point extraction unit extracts the contact point between the outer diameter portion of the wheel and the rail as the target point.

4. It includes a stop-permitted range storage unit that stores the stop-permitted range in the wheel image captured with a fixed imaging field, The fixed-position stop detection device according to claim 1, wherein the stop allowable range setting unit sets the stop allowable range received from the stop allowable range storage unit.

5. The fixed-position stop detection device according to claim 1, wherein the stop tolerance range setting unit sets the target stop position of the target point based on the vehicle identification information of the railway vehicle, and determines the overrun tolerance coordinates and the short run tolerance coordinates based on the stop tolerance position.

6. The fixed-position stop detection device according to claim 5, wherein the vehicle identification information is at least one of the vehicle number read from the railway vehicle and vehicle type information associated with an identification tag provided on the railway vehicle.

7. The aforementioned target point extraction unit is, A target point learning data acquisition unit acquires image patterns of multiple environmental learning images taken under at least one of the imaging conditions of solar radiation, rainfall, snowfall, and fog formation, as well as the ground truth coordinate data of the target point associated with the environmental learning image, as target point learning data. The fixed-position stop detection device according to claim 1, further comprising: a target point trained model generation unit that generates a target point trained model for estimating the target point coordinates from the wheel image using the target point training data.

8. The aforementioned target point extraction unit is, An image data acquisition unit that acquires the wheel image from the wheel image acquisition unit, The fixed-position stop detection device according to claim 1, further comprising a target point estimation processing unit that outputs the target point coordinates from the wheel image acquired by the image data acquisition unit, using a target point trained model for estimating the target point coordinates from the wheel image.

9. The system includes an imaging control unit that controls the imaging conditions for the wheel image, The imaging control unit, An imaging condition learning data acquisition unit acquires climate data at the time of imaging, accumulated wheel images associated with the climate data, and correct imaging condition data associated with the accumulated wheel images as imaging condition learning data. The fixed-position stop detection device according to claim 1, further comprising: an imaging condition learned model generation unit that generates an imaging condition learned model for estimating the imaging conditions from the wheel image using the imaging condition learning data.

10. The system includes an imaging control unit that controls the imaging conditions for the wheel image, The imaging control unit, A climate data acquisition unit that acquires climate data during imaging, An imaging condition estimation processing unit that estimates the imaging conditions from the climate data acquired by the climate data acquisition unit using an imaging condition trained model for estimating the imaging conditions from the climate data, The fixed-position stop detection device according to claim 1, further comprising an imaging condition output unit for controlling the imaging conditions.

11. The system includes a defect determination unit that determines whether or not there is a defect in the wheel image, The defect detection unit determines that a defect has occurred in the wheel image if, in the wheel image, a defect of a size greater than or equal to the defect size threshold has occurred in at least one of the target point coordinates, the overrun tolerance coordinates, and the short run tolerance coordinates, or if the pixel signal is less than or equal to the set pixel signal threshold. The fixed-position stop detection device according to claim 1, wherein the judgment result output unit outputs a defect occurrence signal when the defect determination unit determines that a defect has occurred.

12. The system includes an electromagnetic wave information acquisition unit that acquires electromagnetic wave information emitted from an irradiation source based on the position in which the railway vehicle stops at the overrun allowable position, The fixed-position stop detection device according to claim 1, wherein the stop position determination unit acquires electromagnetic wave information indicating that electromagnetic waves are not obstructed by the lead vehicle, and determines that the railway vehicle has stopped at a fixed position when the target point is within the range of the permissible stopping range.

13. The fixed-position stop detection device according to claim 1, wherein the determination result output unit outputs the distance from the stopping position of the target point to the target stopping position, and the direction indicating an overrun or short run, when the stopping position determination unit determines that the railway vehicle is not stopped at the fixed position.

14. A door image acquisition unit acquires door images that include the passenger boarding / alighting doors on the side of the railway vehicle in relation to the direction of travel, A target point extraction unit extracts a specific position of the passenger boarding / alighting door from the aforementioned door image and determines the coordinates of the target point, A stop tolerance range setting unit determines the overrun tolerance coordinates from the overrun tolerance position of the railway vehicle set in the door image, and the short run tolerance coordinates from the short run tolerance position, and sets the range between the overrun tolerance coordinates and the short run tolerance coordinates as the stop tolerance range. A stop position determination unit determines that the railway vehicle has stopped at a fixed position if, from the time the target point coordinates exceed the short-run allowable coordinates in the direction of travel of the railway vehicle until a set time has elapsed, the target point coordinates are within the stop allowable range, The aforementioned railway vehicle is a determination result output unit that outputs a determination result of whether or not it has stopped at the designated position. A fixed-position stop detection device equipped with the following features.

15. A train occupancy detection sensor that detects when a train enters a station, When the presence detection sensor detects that the railway vehicle is present at the station, the imaging unit captures images of the rails on which the railway vehicle is traveling. A fixed-position stop detection device according to any one of claims 1 to 14, which acquires at least one of a wheel image including the wheel and a door image including the door from the imaging unit, Based on the determination result of the aforementioned fixed-position stop detection device, an opening / closing control unit controls the opening and closing of platform doors installed on the station platform. A fixed-position stop detection system equipped with the following features.

16. The steps include obtaining a wheel image that includes the wheels of a railway vehicle, The steps include extracting a specific position of the wheel from the wheel image and determining the coordinates of the target point, The steps include determining the overrun allowable coordinates from the overrun allowable position of the railway vehicle on the rail on which the wheel runs within the wheel image or from the pre-set overrun allowable position on the wheel image, and determining the short run allowable coordinates from the short run allowable position, and setting the range between the overrun allowable coordinates and the short run allowable coordinates as the stopping allowable range, The step of determining that the railway vehicle has stopped at a fixed position if, from the time the target point coordinates exceed the short-run allowable coordinates in the direction of travel of the railway vehicle until a set time has elapsed, the target point coordinates are within the stop allowable range, A step of outputting a determination result of whether or not the railway vehicle has stopped at the designated position. A fixed-position stop detection method comprising the following:

17. A platform image acquisition unit that acquires platform images including stopping reference marks installed on the platform, which are captured from a railway vehicle, A target point extraction unit extracts the stop reference marks in the platform image as target points and determines the coordinates of the target points, A stop tolerance range setting unit determines the overrun tolerance coordinates from the overrun tolerance position of the railway vehicle set on the platform image, and the short run tolerance coordinates from the short run tolerance position, and sets the range between the overrun tolerance coordinates and the short run tolerance coordinates as the stop tolerance range. A stop position determination unit determines that the railway vehicle has stopped at a fixed position if, from the time the target point coordinates exceed the short-run allowable coordinates in the direction of travel of the railway vehicle until a set time has elapsed, the target point coordinates are within the stop allowable range, The aforementioned railway vehicle is a determination result output unit that outputs a determination result of whether or not it has stopped at the designated position. A fixed-position stop detection device equipped with the following features.

Images