Train monitoring system and train monitoring method

Abstract

To provide a train monitoring system capable of suppressing the influence of deterioration of visibility of camera images on monitoring work.SOLUTION: A train monitoring system according to this example includes cameras 11 (A) and 11 (B) attached to the side of the train body and photographing the same monitoring area from different directions, an image analysis device 17 that executes object detection processing for detecting an object existing in the monitoring area for each of a plurality of camera images from these cameras, and a monitor 13 that displays any one of the plurality of camera images, and on the camera image displayed on the monitor 13, a marker 22 indicating the position of the object detected by the object detection processing for each of the plurality of camera images is superimposed.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a train monitoring system that monitors trains using camera images.

Background Art

[0002] In the operation of railway vehicles, on lines with few passengers and short train formations in operation, single-person operation by one driver is adopted. Amid the trend of a decreasing railway workforce due to population dynamics associated with a declining working population, each railway operator is adopting a policy of increasing the number of lines and the number of cars in the formation to which single-person operation is applied. However, increasing the number of cars in the formation increases the burden of safety confirmation work for a single driver, raising concerns that accidents are likely to occur due to oversight.

[0003]

[0004] Conventionally, to support single-person operation, a system for checking the boarding and alighting status of passengers using cameras and monitors or mirrors installed on the platforms of each station has been used. There is also a vehicle-side camera system in which cameras and monitors are installed on the train side, which has been spreading in recent years because it does not require maintenance of infrastructure facilities.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0006] ​In recent years, further development has been underway regarding a train monitoring system that captures the surroundings of a train (mainly near the doors) using cameras attached to each vehicle of the train and displays the footage on a monitor in the train's driver's cab. For example, a system has been developed that analyzes camera images to detect objects (including people) and superimposes the detection results on the camera image on the monitor.

[0007] However, when applying object detection by analyzing camera images to a train monitoring system, there are problems as described in the following (1) and (2). (1) In image analysis for detecting specific objects from images, including deep learning, it is impossible to achieve a 100% detection rate. Also, since the specification conditions of the train monitoring system are outdoors, it is difficult to obtain camera images suitable for image analysis for reasons such as (a) the lighting conditions for the subject are sparse, (b) the target subjects are diverse, (c) the background varies depending on the stop station, and (d) blurring occurs in the image due to a slow shutter speed when the vehicle is moving.

[0008] (2) In the on-vehicle camera, one of the severe conditions is the irradiation of sunlight such as the setting sun. When strong light enters from a specific angle, it may not be possible to generate a clear image at the camera imaging unit. Also, when sunlight enters with rain, snow, dirt, etc. adhering to the front part of the camera, it further causes a decrease in visibility. This not only leads to a decrease in the detection rate by image analysis but also makes it difficult for crew members and drivers to make visual judgments.

[0009] The present invention has been made in view of the above-described conventional circumstances, and an object thereof is to provide a train monitoring system capable of suppressing the influence on the monitoring operation due to a decrease in the visibility of camera images.

Means for Solving the Problems

[0010] To achieve the above object, a train monitoring system according to an aspect of the present invention is configured as follows. That is, the train monitoring system according to the present invention includes a plurality of cameras attached to the side surface of the train body, each of which captures the same monitoring area from different directions, an analysis device that executes an object detection process for each of the plurality of camera images obtained by the plurality of cameras to detect an object existing in the monitoring area, and a monitor that displays any one of the plurality of camera images. A marking body indicating the position of the object detected by the object detection process for each of the plurality of camera images is superimposed on the camera image displayed on the monitor, which is a feature of the present invention.

[0011] Here, in the train monitoring system according to the present invention, the position where the marking body is superimposed on the camera image displayed on the monitor can be determined based on the result of converting the position of the object in the coordinate system of the camera image in which the object is detected by the object detection process into the coordinate system of the camera image displayed on the monitor.

[0012] Further, in the train monitoring system according to the present invention, coordinate system conversion can be performed using the height of the two cameras that captured the camera image in which the object is detected by the object detection process and the camera image displayed on the monitor with respect to the station platform, the distance between the two cameras, and the vertical and horizontal field angles of the two cameras.

[0013] Further, in the train monitoring system according to the present invention, when there is a camera that is backlit with respect to sunlight among the plurality of cameras, the camera image obtained by a camera other than the backlit camera can be displayed on the monitor.

[0014] Further, in the train monitoring system according to the present invention, the backlit camera can be identified using the horizontal shooting range of each camera, the angle in the track direction of the train, and the incident angle of sunlight.

[0015] In addition, in the train monitoring system according to the present invention, the angle of the traveling direction of the train can be calculated based on the position information of the train at the timing when the train enters the station platform and the position information of the train at the timing when the train stops at the station platform.

[0016] Moreover, a train monitoring method according to another aspect of the present invention is configured as follows. That is, a plurality of cameras attached to the side surface of the train body photograph the same monitoring area from different directions, and an analysis device executes an object detection process for detecting an object existing in the monitoring area for each of a plurality of camera images obtained by the plurality of cameras. A monitor displays any one of the plurality of camera images, and a marking body indicating the position of the object detected by the object detection process for each of the plurality of camera images is superimposed on the camera image displayed on the monitor.

Effect of the Invention

[0017] According to the present invention, it is possible to provide a train monitoring system capable of suppressing the influence on the monitoring operation due to the deterioration of the visibility of the camera image.

Brief Description of the Drawings

[0018]

Figure 1

Figure 2A

Figure 2B

Figure 3A

Figure 3B

Figure 4

Figure 5

Figure 6A

Figure 6B

Figure 7

Figure 8

Figure 9

Embodiments for Carrying Out the Invention

[0019] An embodiment of the present invention will be described with reference to the drawings. FIG. 1 shows a configuration example of a train monitoring system according to an embodiment of the present invention. Here, a three-car train will be described as an example, but the number of vehicles forming the train is arbitrary.

[0020] In FIG. 1, each vehicle 10-1 to 10-3 of the train has, on both sides of the vehicle body side surface, a camera 11(A) facing the train traveling direction and a camera 11(B) facing the opposite direction, and is provided with a network switch 12 which is a relay device for constructing an IP network in the train. The cameras 11(A) and 11(B) installed on the same side surface of the same vehicle are arranged to capture the same monitoring area from different directions. In this example, the camera 11(A) installed on the rear side of the vehicle captures the monitoring area (mainly near the door) from the viewpoint of the train traveling direction, and the camera 11(B) installed on the front side of the vehicle captures the same monitoring area from the viewpoint opposite to the train traveling direction.

[0021] In the driver's cabs located in the vehicles 10-1 and 10-3 at both ends of the train, two monitors 13(A) and 13(B) are installed. The monitor 13 displays the video of one of the two cameras 11(A) and 11(B) on the door opening side of each vehicle during specific events such as door opening. Specifically, the display area of the monitor 13 is divided into a plurality of areas, and split display is performed by assigning the plurality of camera videos (one camera video for each vehicle) taken by the cameras of each vehicle to each area for display.

[0022] In this way, by splitting and displaying one camera video for each vehicle, the platform area corresponding to the length of the train formation can be entirely displayed on the monitor 13, and a comprehensive safety check of the platform on the door opening side can be carried out before door closing operation or vehicle startup. Note that if two or more camera videos of each vehicle are split and displayed, the resolution of the camera videos will decrease, making it difficult for the driver to visually recognize objects. Also, since there is a limit to the display size of the monitor 13, one camera video for each vehicle is being displayed. In this example, it is assumed that only the monitor 13 of the leading vehicle (for example, vehicle 10-1) in the train traveling direction is operated, but the monitor 13 of the last vehicle 10-3 can also be additionally operated.

[0023] Also, in the driver's cabs located in the vehicles 10-1 and 10-3 at both ends of the train, a control device 14 is also installed to control the display ON / OFF etc. of the monitor 13 based on the vehicle information 15 obtained from the vehicle side or the upper-level system. Examples of the vehicle information 15 include a door opening signal and a speed detection signal. The door opening signal is a signal indicating that the door has been opened, and also includes information on the opened side (in this specification, the left side with respect to the train traveling direction is defined as the "mountain side" and the right side as the "sea side"). The speed detection signal is a signal indicating that the speed of the train is equal to or higher than a predetermined value (for example, 5 km / h). Also, the control device 14 also has an interface for acquiring position information (specifically, latitude and longitude information) and time information from a GPS (Global Positioning System) antenna 16 installed on the train.

[0024] While the train is stopped at the station, the display of the monitor 13 is turned on, and the control device 14 controls the operation of the monitor 13 so that the display of the monitor 13 is turned off as the train starts to move. As a result, since the camera image is not displayed on the monitor 13 during the running of the train, the driver can concentrate on driving the train. In this example, the display of the monitor 13 continues not only when the train is completely stopped but also when the speed is such that it is not recognized as running. Then, when the speed of the train reaches a predetermined value (for example, 5 km / h), the display of the monitor 13 is stopped.

[0025] Furthermore, in the monitoring system of this example, an image analysis device 17 for analyzing the camera image of each of the vehicles 10-1 to 10-3 is installed in each vehicle. During the display period of the camera image by the monitor 13, the image analysis device 17 acquires the camera image from the two cameras 11(A) and 11(B) on the door opening side and executes an object detection process for detecting an object (including a person) to be detected existing in the monitoring area (mainly near the door).

[0026] The object detection process is realized by using, for example, deep learning technology. As an example, the object detection process can perform processing on a per-image basis by utilizing a generally known detection model such as Yolo, and can identify the position of the object in the image and the type of the object. When an object is detected by the object detection process, the image analysis device 17 transmits an analysis result notification including the coordinates and type information of the object to the monitor 13.

[0027] When the monitor 13 receives the analysis result notification, it highlights the camera image in which the object was detected. The highlighting is realized by surrounding the corresponding camera image with a frame 21 of a predetermined color (for example, red) as shown in FIG. 2A, or by superimposing (overlaying) a frame 22 surrounding the detected object on the camera image as shown in FIG. 2B. Also, other forms of marking bodies such as markers or arrows indicating the detected object may be superimposed on the camera image. As a result, it becomes easier for the driver to recognize dangerous events on the station platform.

[0028] Here, the detection results of the image analysis device 17 include "correct detection" when an object is correctly detected, "undetected" when the object could not be detected despite its presence, and "false detection" when it is determined that the object exists despite its absence. In an operation where a person such as a driver makes the final judgment on the safety of the platform, reducing undetected cases is an important issue. As one method of reducing such undetected cases, there is a method of notifying the detection results using a plurality of cameras that photograph the same monitoring area.

[0029] In this example, two cameras 11(A) and 11(B) facing each other on the same vehicle photograph the same monitoring area from different directions. FIG. 3A is an example of a camera image of the monitoring area photographed by camera 11(A), and FIG. 3B is an example of a camera image of the same monitoring area photographed by camera 11(B). Object detection processing is executed for both camera images photographed by these two cameras, and the results are alarm-notified using OR logic. Thereby, even when undetected occurs in one camera image, if an object is detected in the other camera image, it is possible to reduce undetected cases for the entire system.

[0030] Also, since cameras 11(A) and 11(B) are installed facing each other, when sunlight is incident on one camera, that is, in the case of backlight illumination conditions, the other camera has frontlight illumination conditions. Since the visibility of the camera image is high under frontlight illumination conditions, dynamically changing the camera for the object detection target and the display target according to the direction of sunlight serves as a countermeasure against backlight from sunlight. The train monitoring system in this example is equipped with a GPS antenna 16 and can acquire time information and position information, so these information will be used to control the switching of the target camera.

[0031] Hereinafter, (1) a method of notifying the detection results using two cameras and (2) a method of discriminating between frontlight and backlight conditions will be described.

[0032] (1) Regarding the method of notifying the detection results using two cameras In object recognition using a detection model such as Yolo that uses deep learning, the coordinate information of an object on an image can be obtained. By using this coordinate information, as shown in FIG. 2B, a marking body (frame 22) indicating the detected object can be superimposed and displayed on the camera image in which the object is detected.

[0033] Here, for example, during the display of the camera image of camera 11(B), it is assumed that an object was not detected in the object detection process related to camera 11(B), but was detected in the object detection process related to camera 11(A). In this case, since the coordinate information of the object cannot be obtained in the object detection process related to camera 11(B), as it is, the marking body cannot be superimposed and displayed on the camera image of camera 11(B). Therefore, when an object is detected only in the object detection process related to camera 11(A), it is necessary to devise a way to superimpose and display the marking body indicating the object on the camera image of camera 11(B). Therefore, the coordinate information of the object obtained in the object detection process related to camera 11(A), that is, the position information of the object in the coordinate system of the camera image of camera 11(A), is converted into the coordinate system of the camera image of camera 11(B).

[0034] Here, an approximate coordinate conversion method will be described under the conditions shown in FIGS. 4 and 5. It is assumed that the mounting positions of cameras 11(A) and 11(B) are at a height H from the platform surface. Also, it is assumed that the detected object P (subject) exists on the platform surface at a position d in the sleeper direction and l in the rail direction from camera 11(A). A Here, the height direction is defined as the Y direction, the "rail direction" (that is, the track direction of the train) is defined as the Z direction, and the "sleeper direction" (that is, the horizontal direction orthogonal to the train traveling direction) is defined as the X direction.

[0035] In this case, assuming that the distance between cameras 11(A) and 11(B) is a known value L, the distance l in the rail direction from camera 11(A) to object P B can be expressed by the following (Equation 1). l B = L - l A …(Equation 1)

[0036] First, referring to FIG. 4, the coordinate transformation in the vertical direction (Y direction) in the camera image will be described. Here, the angle e of the vertical component in the straight line connecting the camera 11(A) and the object P A can be approximately expressed by the following (Equation 2). e A = tan -1 (l A / H) …(Equation 2)

[0037] Also, when the vertical field angle of the camera 11(A) is φ, assuming that the upper boundary line of the field angle is horizontal, the angle φ between the lower boundary line of the field angle and the straight line connecting the camera 11(A) and the object P A can be approximately expressed by the following (Equation 3). φ A = e A -(π / 2 - φ) …(Equation 3)

[0038] Here, the resolution of the camera image is set to VGA (640×480). Also, as shown in FIG. 6A, the coordinates of the object P in the camera image of the camera 11(A) are (x A , y A ), and as shown in FIG. 6B, the coordinates of the object P in the camera image of the camera 11(B) are (x B , y B ). These coordinate values are the positions in pixel units from the upper left of the image.

[0039] In this case, the vertical position y A of the object P in the camera image of the camera 11(A) can be expressed by the following (Equation 4). y A = 480×(φ - φ A ) / φ …(Equation 4) Also, by using (Equation 2) and (Equation 3), the position y A can be expressed by the following (Equation 5). y A = 480×{φ - (tan -1 (l A / H) - (π / 2 - φ))} / φ …(Equation 5)

[0040] Similarly, the vertical position y of the object P in the camera image of the camera 11(B) B can be expressed by the following (Equation 6) and (Equation 7). Here, f b is the angle between the boundary line of the lower surface of the viewing angle of the camera 11(B) and the straight line connecting the camera 11(B) and the object P. y B = 480×(φ - φ B ) / V …(Equation 6) y B = 480×{φ - (tan -1 (l B / H) - (π / 2 - φ))} / φ …(Equation 7) Therefore, from (Equation 1), (Equation 5), and (Equation 7), y B can be expressed using y A .

[0041] Next, with reference to FIG. 5, the coordinate transformation in the horizontal direction (X direction) in the camera image will be described. Here, let the horizontal viewing angle of the camera 11(A) be θ, and the viewing angle range into which the vehicle enters be θ t , and the angle between the side surface of the vehicle and the object P be θ A . In this case, θ A can be expressed by the following (Equation 8) using the X-direction distance d from the side surface of the vehicle to the object P. Note that θ t depends on the method of mounting the camera on the vehicle but is assumed to be known. θ A = tan -1 (d / l A ) …(Equation 8)

[0042] Also, the horizontal position x A of the object P in the camera image can be expressed by the following (Equation 9). x A = 640×(θ A + θ t ) / θ …(Equation 9) By substituting (Equation 8) into this, the position x A can be expressed by the following (Equation 10). x A = 640×{tan -1 (d / l A) + θ t} / θ … (Equation 10)

[0043] Similarly, the horizontal position x of the object P in the camera image of camera 11(B) B can be expressed by the following (Equation 11) and (Equation 12). x B = 640×(θ B + θ t ) / θ … (Equation 11) x B = 640×{tan -1 (d / l B ) + θ t} / θ … (Equation 12) Therefore, from (Equation 1), (Equation 10), and (Equation 12), x B can be expressed using x A .

[0044] As described above, the position information (x A , y A ) of the object P in the coordinate system of the camera image of camera 11(A) can be converted into the position information (x B , y B ) of the coordinate system of the camera image of camera 11(B). Utilizing this, it becomes possible to superimpose the marking body 22 B indicating the object P detected from the camera image of camera 11(A) onto the camera image of camera 11(B).

[0045] Here, it is assumed that the lower part of the object P is in contact with the platform surface (that is, the lower side of the marking body indicating the object P exists on the platform surface). In this case, the position of the lower side of the marking body 22 B in the camera image of camera 11(B) can be calculated by performing coordinate transformation on the position of the lower side of the marking body 22 A in the camera image of camera 11(A). Also, the width w B of the marking body 22 B (the length in the X direction) in the camera image of camera 11(B) is the marking body 22 BBy specifying the positions of both ends of the lower side through coordinate transformation, it can be calculated. Also, the marking body 22 in the camera image of camera 11(B) B The height h B (the length in the Y direction) of the marking body 22 in the camera image of camera 11(B) B The width w B and the marking body 22 in the camera image of camera 11(A) A The height h A and width w A Based on the ratio of, it can be calculated (that is, h B =w B ×h A / w A ). Thus, the marking body 22 indicating the object P detected from the camera image of camera 11(A) can be superimposed at an appropriate position in the camera image of camera 11(B) B .

[0046] (2) Method for discriminating frontlight and backlight conditions The frontlight and backlight conditions are determined by factors such as weather, time, and the track direction of the train on the platform. It is assumed that the weather information uses the weather forecast information transmitted from the ground facilities. Also, the incident angle of sunlight in each time zone is assumed to be stored in the database within the system. Regarding the track direction of the train at each station, although it can be solved by creating a database for each station, here, a general track direction identification method using GPS is used.

[0047] Referring to Fig. 7, the method for calculating the track direction of the train will be described. First, the position information T 1 (x 1 ,y 1 ) of the train at the timing when the train enters the platform is measured and recorded. The timing when the train enters the platform can be obtained from the platform detection system using an ultrasonic sensor or the like. Furthermore, the position information T 2 (x 2 ,y 2 ) of the train at the timing when the train stops at the platform is measured and recorded. The position information T 1, T 2 For example, the latitude and longitude information obtainable through the GPS antenna 16 can be used.

[0048] These position information Ts 1 , T 2 can be used to identify the track direction of the train in terms of azimuth. That is, the angle θ1 of the track direction of the train (the inclination in terms of azimuth when based on the east direction) can be expressed by the following (Equation 13). θ1 = tan -1 {(y 1 - y 2 ) / (x 1 - x 2 )} …(Equation 13)

[0049] There are also stations where the platform has a curved shape. At such stations, if the position information is measured at the timing when the train enters the platform and the timing when it stops, and the track direction of the train is calculated, the result may be different from the actual angle of the track direction. For such stations, it is possible to cope by delaying the timing of measuring the position information T. That is, at the timing when both the condition that the train has entered the platform and the condition that the speed of the train is 3 km / h or less (AND condition) are satisfied, by measuring the position information T of the train 1 and reducing the measurement interval of the position information, it becomes possible to calculate an angle close to the actual angle of the track direction. 1

[0050] Next, in order to analyze the incident situation of sunlight on the camera, it is necessary to identify the incident angle of sunlight at each time. Here, a method of simply calculating the incident angle of sunlight only by two-dimensional calculation will be described with reference to FIG. 8. As conditions, the time from sunrise to sunset is defined as Ts, and the time elapsed from sunrise at the current time is defined as Tn. When the sunrise timing is 0° (degrees) and the sunset timing is 180°, the incident angle θ2 of sunlight at the current time can be expressed by the following (Equation 14). θ2 = (Tn / Ts) × 180° …(Equation 14) ​

[0051] Here, as shown in FIG. 9, let the angle between the boundary line (the side far from the train) of the horizontal shooting range on the platform side of the camera 11(B) and the train traveling direction be θh. In this case, the condition for sunlight to enter the camera 11(B) is as follows in Equation (15). θ1 + θ2 < θh …(Equation 15)

[0052] According to this condition, using the train's position information and the date and time information as variables, it is possible to determine the direct incidence situation of sunlight on the camera, so the presence of a camera with backlight can be detected. Therefore, for example, when the camera 11(B) is detected as a camera with backlight, by selecting the camera image of the other camera, that is, the camera 11(A), as the display target on the monitor 13, it is possible to display a highly visible camera image on the monitor 13.

[0053] As described above, the train monitoring system in this example includes cameras 11(A) and 11(B) attached to the side surface of the train body and photographing the same monitoring area from different directions, an image analysis device 17 that executes object detection processing for detecting an object existing in the monitoring area for each of a plurality of camera images obtained by these cameras, and a monitor 13 that displays any one of the plurality of camera images. The camera image displayed on the monitor 13 is configured such that a marking body 22 indicating the position of the object detected by the object detection processing for each of the plurality of camera images is superimposed.

[0054] Therefore, even when the visibility of one of the camera images of the camera 11(A) or the camera 11(B) decreases, it is possible to detect an object existing in the monitoring area by the object detection processing for the other camera image and superimpose the marking body 22 on the camera image of the monitor 13. Thereby, it is possible to suppress the influence on the monitoring operation due to the decrease in the visibility of the camera image.

[0055] In addition, when there is a camera (for example, camera 11(B)) that is backlit with respect to sunlight among cameras 11(A) and 11(B) in the train monitoring system of this example, the camera images from cameras other than the backlit camera (for example, camera 11(A)) are configured to be displayed on monitor 13.

[0056] Therefore, when the visibility of the camera image of one of cameras 11(A) or 11(B) is reduced due to backlighting by sunlight, the camera image of the other camera with good visibility can be displayed on monitor 13. For this reason, even in a situation where backlighting such as the setting sun is a concern, it is possible to continue the monitoring operation using a camera image with good visibility.

[0057] Here, in the above description, the same monitoring area is photographed from different directions by two cameras 11(A) and 11(B), but it may be photographed by three or more cameras. This makes it possible to reduce detection misses of objects more. Also, a camera image with better visibility can be displayed on the monitor.

[0058] In addition, in the above description, the calculation of coordinate conversion is performed as needed when detecting an object, but a coordinate conversion table may be prepared in advance by calculating and setting the correspondence relationship of each coordinate on the platform surface between camera images, and the coordinate conversion may be performed using this coordinate conversion table.

[0059] Note that the processing such as coordinate conversion and determination of frontlight / backlight conditions may be executed by image analysis device 17, may be executed by monitor 13, or may be executed by other devices. Also, the device that executes these processes is realized, for example, when a computer equipped with hardware resources such as a processor and a memory executes a program for realizing each function according to the present invention by the processor.

[0060] The embodiments of the present invention have been described above. However, these embodiments are merely examples and do not limit the technical scope of the present invention. The present invention can take various other embodiments and can undergo various modifications such as omission and substitution without departing from the gist of the present invention. These embodiments and their modifications are included in the scope and gist of the invention described in this specification and the like, and are also included in the invention described in the claims and its equivalent scope.

[0061] Furthermore, the present invention can be provided not only as the devices as described in the above description and the systems constituted by these devices, but also as the methods executed by these devices, programs for realizing the functions of these devices by a processor, storage media for storing such programs in a computer-readable manner, and the like.

Industrial Applicability

[0062] The present invention can be used in a train monitoring system for monitoring trains by camera images.

Explanation of Signs

[0063] 10-1 to 10-3: Vehicles, 11(A), 11(B): Cameras, 12: Network switch, 13(A), 13(B): Monitors, 14: Control device, 15: Vehicle information, 16: GPS antenna, 17: Image analysis device

Claims

1. A plurality of cameras attached to the side surface of a train body, each of which captures the same monitoring area from different directions, an analysis device that executes object detection processing for detecting an object existing in the monitoring area for each of a plurality of camera images obtained by the plurality of cameras, a monitor that displays any one of the plurality of camera images, a marking body indicating the position of an object detected by the object detection processing for each of the plurality of camera images is superimposed on the camera image displayed on the monitor, A train monitoring system, wherein a position at which the marking body is superimposed on the camera image displayed on the monitor is determined based on a result of converting the position of the object in the coordinate system of the camera image in which the object is detected by the object detection processing into the coordinate system of the camera image displayed on the monitor.

2. In the train monitoring system according to Claim 1, the coordinate system conversion is performed using the heights of two cameras that capture the camera image in which the object is detected by the object detection processing and the camera image displayed on the monitor with respect to the station platform, the distance between the two cameras, and the vertical and horizontal viewing angles of the two cameras. A train monitoring system characterized by that.

3. In the train monitoring system according to Claim 1 or Claim 2, When there is a camera that is backlit with respect to sunlight among the plurality of cameras, a camera image obtained by a camera other than the backlit camera is displayed on the monitor. A train monitoring system characterized by that.

4. In the train monitoring system according to Claim 3, A train monitoring system characterized by identifying the backlit camera using the horizontal shooting range of each camera, the angle in the track direction of the train, and the incident angle of sunlight.

5. In the train monitoring system according to Claim 4, An angle in the track direction of the train is calculated based on the position information of the train at the timing when the train enters the station platform and the position information of the train at the timing when the train stops at the station platform. A train monitoring system characterized by that.

6. A plurality of cameras attached to the side surface of a train body capture the same monitoring area from different directions, an analysis device executes object detection processing for detecting an object existing in the monitoring area for each of a plurality of camera images obtained by the plurality of cameras, The monitor displays any one of the plurality of camera images, on the camera image displayed on the monitor, an indication body indicating the position of an object detected by the object detection process for each of the plurality of camera images is superimposed, a position where the indication body is superimposed on the camera image displayed on the monitor is determined based on a result of converting the position of the object in the coordinate system of the camera image in which the object is detected by the object detection process into the coordinate system of the camera image displayed on the monitor, a train monitoring method characterized by that.

Images