Forward-looking system and forward-looking method

Abstract

This invention provides a forward-facing monitoring system that, based on images captured by a surveillance camera monitoring the area in front of the train, can determine the distance to the track ahead of the moving train within a small margin of error, regardless of the type of track. [Solution] The system includes a calculation module 204 that calculates distance based on the track width that the track occupies in the captured image. If the track in front of the train is a curved track, the calculation module calculates the distance to a predetermined area based on the attributes of the curved track (radius of curvature, cant).

Description

Technical Field

[0001] The present invention relates to a forward monitoring system and a method thereof, and more particularly, to a forward monitoring system and a method thereof for estimating the distance to a predetermined area of a track in front of a train based on a captured image of a monitoring camera that captures the front of the train.

Background Art

[0002] In recent years, driverless operation of railways has attracted attention. Although driverless operation has already been realized in subways and new transportation systems running on elevated structures, there are no practical examples on lines where people may enter, such as lines with level crossings or lines within vehicle depots where workers are present (general lines).

[0003] In order to expand driverless operation to general lines, a system for monitoring the front of a train in motion is important. Each of Patent Documents 1 to 3 discloses a monitoring camera that captures the front of a train.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0005] To detect obstacles ahead of a train traveling on a track and determine their location, the distance from the surveillance camera to the track ahead of the train is estimated based on images captured by the surveillance camera. However, the inventors have found that even with a monocular camera, it is difficult to estimate the actual distance to any point on the track displayed in the captured image. Therefore, the present invention aims to provide a forward monitoring system and method that can determine the distance to any area on the track, regardless of the type of track, based on an image of the area ahead of a moving train captured by a surveillance camera, with a small error margin. [Means for solving the problem]

[0006] To achieve the above objective, the present invention provides a forward-looking system comprising a surveillance camera for photographing the front of a train and an image processing device, wherein the image processing device estimates the distance to a predetermined area on the track in front of the train based on an image captured by the surveillance camera, and further comprises a feature extraction module for extracting the track from the captured image and a calculation module for calculating the distance based on the track width occupied by the track in the captured image, wherein the calculation module determines the distance to the predetermined area based on the attributes of the curved track when the track in front of the train is a curved track.

[0007] Furthermore, the present invention provides a forward monitoring method for monitoring the area in front of a train by estimating the distance to a predetermined area of ​​the track in front of the train based on images captured by a surveillance camera that photographs the area in front of the train, characterized in that when the track is extracted from the captured images and the distance is calculated based on the track width occupied by the track in the captured images, if the track in front of the train is a curved track, the distance to the predetermined area is determined based on the attributes of the curved track. [Effects of the Invention]

[0008] As described above, according to the present invention, even if the track in front of the train is a curved track, the distance to a predetermined area of ​​the track in front of the train can be accurately determined based on the image captured by the surveillance camera monitoring the area in front of the train, just as in the case of a straight track. [Brief explanation of the drawing]

[0009] [Figure 1] This is a side view of the lead car of a train equipped with a forward-facing monitoring system. [Figure 2A] This figure shows an example of an image captured by a surveillance camera along a straight line. [Figure 2B] This figure shows an example of an image captured by a surveillance camera of a trajectory consisting of a circular arc and a transition curve. [Figure 3] Figure 3 shows the orbital width along the u-axis direction of the screen coordinate system for orbits within the monitoring area. [Figure 4] Figure 4 illustrates the transformation from the screen coordinate system to the 3D world coordinate system. [Figure 5] Figure 5 shows the principle of calculating the target distance. [Figure 6] This is a hardware block diagram of an image processing device. [Figure 7] This is a functional block diagram of the image processing device 30. [Figure 8] This is a functional block diagram illustrating the process of correcting the target distance. [Figure 9] Figure 9 illustrates the correction of the estimated distance based on the radius of curvature in a circular trajectory. [Figure 10] Figure 10 illustrates the correction of the estimated distance based on the amount of displacement in a circular trajectory. [Figure 11] Figure 11 shows a clothoid curve and its parameters. [Figure 12] Figure 12 illustrates the correction of the estimated distance based on the cant quantity in a circular trajectory. [Figure 13] Figure 13 illustrates the correction of the estimated distance based on the cant quantity in a circular trajectory.

Embodiment for Carrying out the Invention

[0010] Next, an embodiment of the front monitoring system according to the present invention will be described. FIG. 1 is a side view of the leading car 1 of a train equipped with the front monitoring system 10. The leading car 1 travels on the track 50 along the direction of reference numeral 16. The front monitoring system 10 includes a front monitoring camera 20, an image processing device 30, and a data communication path 21 connecting these. Reference numeral 22 of the front monitoring camera 20 indicates the position of the camera viewpoint, and reference numeral 24 indicates the line-of-sight direction of the front monitoring camera.

[0011] The front monitoring camera 20 is mounted inside the front of the leading car 1 such that its viewpoint 22 becomes the driver's viewpoint. The line-of-sight direction 24 of the front monitoring camera is set to overlook the track ahead from the front of the train, similar to the driver's line of sight. Reference numeral 26 indicates the monitoring area (or monitoring point) of the track 50 by the front monitoring camera. This monitoring area is provided in a predetermined area of the track in front of the train. The front monitoring camera 20 continuously outputs an image including the monitoring area 26 of the track 50.

[0012] As the front monitoring camera 20, for distance measurement to an object in the image, a high-resolution and low-latency monocular camera is suitable. As the front monitoring camera 20, for example, a plurality of types for short-distance monitoring, medium-distance monitoring, and long-distance monitoring may be mounted on the train 1. These plurality of types of monitoring cameras have different internal parameters such as focal length and angle of view.

[0013] The image processing device 30 calculates the distance to the monitoring area 26 of the track existing within the angle of view of the front monitoring camera 20 based on the image of the track 50 included in the captured image. The monitoring area 26 may be appropriately set according to the purpose of monitoring as the distance from the front of the train 1. For example, if the detection of an intruder within the building limit area is the purpose of monitoring, the monitoring area 26 can be set as an area including from 200 meters to 600 meters, which is the braking distance of a railway train. The image processing device 30, or the train operation support system device that has received the calculated distance, monitors for intruders into the building limit area by continuously monitoring this area.

[0014] The track 50 is composed of a straight track and a curved track. The curved track is composed of an arc track and a transition curve track between the straight track and the arc track. The transition curve track exists to prevent the deterioration of the riding comfort when the train moves from the straight track to the arc track or when the train moves from the arc track to the straight track. FIG. 2A is an example of an image captured by the front monitoring camera 20 of the straight track 50M. FIG. 2B is an example of an image captured by the front monitoring camera 20 of the track 50 composed of the arc track 50B and the transition curve track 50A.

[0015] As shown in FIG. 3, the image processing device 30 calculates the distance from the camera viewpoint 22 to the monitoring area 26 based on the track width (rail spacing) W along the u-axis direction of the screen coordinate system between the tracks 50, 50 in the monitoring area 26. Therefore, the image processing device 30 converts the screen coordinate system into a three-dimensional world coordinate system where the actual track 50 exists (FIG. 4). Any point (u, v) on the projection plane (screen coordinate system 20B) of the captured image is projected onto a straight line 25 passing through the camera origin (camera viewpoint 22) in the world coordinate system 20C (x, y, z). In the camera coordinate system 20A, the Z-axis corresponds to the depth direction (line-of-sight direction). The captured image may be defined in a normalized screen coordinate system 20D with the center O of the image coordinates as the origin instead of the screen coordinate system 20B (u, v) with the upper left of the image as the origin.

[0016] The image processing device 30 cannot determine the distance from the camera viewpoint 22 to the monitoring area 26 (hereinafter referred to as the target distance) using only screen coordinates (u, v). However, since the actual track gauge size is a predetermined eigenvalue, the target distance can be calculated and estimated by counting the number of pixels that make up the track gauge W (Figure 3) in the image.

[0017] Figure 5 shows the principle of calculating the target distance. C is the projection center of the forward-facing camera 20 (camera viewpoint 22). Reference numeral 22A denotes the screen projection plane in the u direction of the screen coordinate system 20B. A and B are the track gauges in the x direction of the world coordinate system. Y is the distance (target distance) from the camera viewpoint 22 to the monitoring area 26 along the line of sight, i.e., in the y direction of the world coordinate system. a and b indicate the track gauges of the monitoring area 26 in the image. h is counted by the number of pixels in the track gauge. The y coordinate of the screen projection plane 22A is equal to the focal length (f) of the camera.

[0018] Since multiple parameters have the relationship H / h = Y / f, and the focal length is known as an intrinsic parameter of the camera, the image processing device 30 can calculate the target distance (Y) based on the number of pixels (=h) and the length of the track (eigenvalue: H). Note that the method in Figure 5 assumes that the track is straight.

[0019] Next, the details of the image processing device 30 will be described. Figure 6 is a hardware block diagram of the image processing device 30. The image processing device 30 includes a CPU and other processors 104, a main memory 106, an auxiliary memory 108, an input interface 112, and an output interface 114, all of which are interconnected via an internal communication line 102 such as a bus.

[0020] The processor 104 controls the operation of the entire image processing unit 30. The main memory 106 is composed of, for example, volatile semiconductor memory and is used as the work memory of the processor 104. The auxiliary storage device 108 is composed of a large-capacity non-volatile storage device such as a hard disk drive, SSD, or flash memory and is used to retain various programs and data for a long period of time. The image processing program 1003a, which processes images captured by the forward-facing camera 20 and is stored in the auxiliary storage device 108, is loaded into the main memory 106 when the image processing unit 30 is started or when necessary, and is executed by the processor 104.

[0021] The image processing program 1003a may be recorded on a non-temporary recording medium, read from the non-temporary recording medium by a media reader, and loaded into the main memory 106. Alternatively, the image processing program 1003a may be acquired from an external computer via a network and loaded into the main memory 106. The input interface 112 is used to capture images taken by the forward-facing camera 20. The output interface 114 is used to output the distance from the forward-facing camera 20 to the track, calculated by the processor 104, to a driver assistance system or the like.

[0022] The database (DB) 1003b of the auxiliary storage device 108 stores various information necessary for train operation. The stored information includes, for example, train operation curves and track information linked to train position information. Train operation curves are data that railway operators use to plan efficient train operation by continuously graphing information such as elapsed time, speed, curves, and gradients according to changes in the running position. Track information includes, for example, the distinction between straight tracks, circular tracks, and transition curve tracks, track length, operating speed, speed limit, average speed, and in the case of circular tracks, the radius of curvature and cant amount as attribute information, and in the case of transition curve tracks, the displacement amount as attribute information. The cant amount may be calculated by the calculation module 204 described later based on the attribute information instead of being stored in the database 1003b. Attribute information refers to the characteristics of the track.

[0023] Figure 7 shows an example of a functional block diagram of the image processing device 30. The processor 104 realizes this functional block by executing the image processing program 1003a. The feature extraction module 200 extracts the trajectory position from the image captured by the monocular camera, which is the forward-looking camera 20, using, for example, an object detection model consisting of a semantic segmentation algorithm.

[0024] The coordinate transformation module 202 converts the extracted track positions from screen coordinates to world coordinates, as explained in Figure 5. The calculation module 204 calculates the size of the track in the image based on the number of pixels that make up the track in the image. The calculation module 204 then calculates the target distance based on this value.

[0025] The train position detection module 208 calculates the train's position by integrating the values ​​from the speedometer mounted on the train. Alternatively, the train's position may be calculated by obtaining the train's current latitude, longitude, and altitude in real time using a GNSS receiver mounted on the train. Alternatively, the train's position may be estimated by continuously monitoring the shape of the track, captured by a forward-facing camera, from the time of launch. The train position detection module 208 outputs the calculated train position to the calculation module 204. The calculation module 204 refers to the database 1003b based on the train position and determines the type of track where the monitoring area exists.

[0026] If the trajectory is a straight line, the calculation module 204 determines the target distance (e), calculated according to the model in Figure 5, as the estimated distance. If the trajectory is a curve other than a straight line, the model described in Figure 5 does not hold, and therefore the number of pixels in the track gauge does not correspond to the target distance. This will be explained later. Therefore, the calculation module 204 first calculates the target distance (e) according to the model in Figure 5 based on this number of pixels, and then corrects it. Note that a module is a function realized by the processor executing a program, and may be replaced with other terms such as means, unit, element, etc. Modules may be realized by dedicated hardware such as integrated circuits.

[0027] Figure 8 is a functional block diagram illustrating the process of correcting the target distance (e). The calculation module 204 reads the attribute information of the curved track by referring to the database 1003b based on the detected position of the train (S1). Furthermore, the calculation module 204 calculates or obtains the cant amount from the database based on the attribute information of the curved track (S2), and recalculates the target distance (e) taking the cant amount into consideration (300).

[0028] The calculation module 204 corrects the target distance (e) (302) by referring to a data table pre-stored in the database (S3) based on the calculated target distance (e), one or more attribute pieces of information, and the cant, and determines the corrected target distance (d) as the estimated distance. If the trajectory is curved, the number of pixels in the track does not accurately correspond to the target distance, so it is necessary to correct the target distance calculated according to the model in Figure 5. If the trajectory is straight, the calculation module 204 outputs the calculated target distance (304:e) as is.

[0029] Next, we will explain the operation of correcting the target distance. As shown in Figure 5, if the track 50 is a straight line, the direction of the track gauge of the track 50 is parallel to the u-axis of the camera coordinate system, so the target distance can be determined without error based on the number of pixels of the track gauge in the track image. On the other hand, if the monitoring area 26 is on a curve relative to the forward monitoring camera 20, that is, if the forward monitoring camera 20 is on a straight track and the track in which the monitoring area exists is curved in a circular arc, then as shown in Figure 9, the direction of the track gauge rotates (θ) in the global coordinate system, so the premise for when the track is a straight line no longer holds, and estimation errors occur in the calculation result of the target distance.

[0030] In Figure 9, although the track gauge size is actually a line segment DB, the track gauge size based on the monitoring image is counted as AB (>DB). Therefore, the calculation module 204 corrects the target distance as follows.

[0031] The error rate is 1 - (e / d). Here, "e" is the target distance calculated based on a model where the trajectory is a straight line, as explained in Figure 5, and d is the true distance corrected to eliminate the error. This error rate is This can be expressed as 1-(DB / AB)=1-cosθ.

[0032] In Figure 9, ω is the angle from the line of sight 24 of the forward-facing camera to B on the track 50, r is the radius of curvature, and w is the track gauge (straight section) on the screen. In Figure 9, the radius of curvature r is shown as the radius of curvature of the inner rail of the curve, but it could also be the radius of curvature of the outer rail, or the radius of curvature of the centerlines of the left and right rails. These can be converted to each other by addition or subtraction if the size of the track is known. cosθ = RE / RB (θ < (π / 2)) RE = r + (w / 2) - d * tan(ω) cosθ=(r+(w / 2)) / (rd / r*tan(ω)) ≒1-(d / r)*tan(ω)∵r≫w / 2 The error rate is (d / r)*tan(ω), so 1-(e / d)=(d / r)*tan(ω), and multiplying both sides by d and rearranging, ([(tan(ω)] / r)*d 2 ) - d + e = 0.

[0033] ω is uniquely determined by the screen coordinate u, and if r is known, then tan(ω) / r is independent of d and e. Therefore, solving it as a quadratic equation in d,

number

[0034] This correction suppresses fluctuations in the target distance even when the trajectory of the forward-facing surveillance camera 20 changes from a straight line to a circular arc (curve). As a result, the surveillance system can continue monitoring for obstacle intrusion into the building clearance area.

[0035] In most cases, a transition curve is provided between the straight track 50A and the arc section 50B. As shown in Figure 10, when moving from the straight track 50A through the transition curve to the arc section 50B, there is a displacement Δr between the line of sight direction and the starting end of the arc section. Therefore, when determining the distance from the viewpoint C to the monitoring area 26 of the arc section 50B, it is necessary to pay attention to this displacement (transition amount: Δr). Figure 10 shows the tracks of the arc section 50B and the straight section 50A, and the illustration of the transition curve is omitted. The transition curve is composed of a clothoid curve, and the transition amount Δr can be calculated from the clothoid coefficient.

[0036] Figure 11 shows the clothoid curve and its parameters. 50C is the clothoid curve, and the transition curve section consisting of the clothoid curve 50C leads to the circular arc section 50B. The parameters of the clothoid curve shown in the figure have the following relationships.

number

[0037] In Figure 10, the error rate considering the displacement Δr is cosθ = d / r*tan(ω) - (Δr / r), and the quadratic equation in terms of d is:

number

[0038] Therefore, d is

number

[0039] In addition to the radius of curvature and the amount of displacement mentioned above, the superelevation (cant) is another factor that should be used to correct the target distance (e). Superelevation refers to the slope applied so that the outer track of a curve is higher than the inner track, and the superelevation is the difference in height between the two tracks.

[0040] The calculation module 204 uses the train's position information as a key to refer to database 1003b, read the radius of curvature (r[m]), track width (W[m]), and average speed (V[km / h]), and estimates the cant amount (C[m]).

number

[0041] Furthermore, if the cant values ​​for each curve are included in database 1003b, then estimation of the cant values ​​by calculation is unnecessary, and the cant values ​​themselves can be referenced from database 1003b.

[0042] The calculation module 204 recalculates the target distance using the acquired cant quantity in the following procedure.

[0043] First, the image is rotated to compensate for the camera's roll rotation caused by cant. Figure 12 illustrates the method of image rotation. 401 is the image before rotation, and 402 is the image after rotation. The rotation angle is calculated using the estimated cant amount C and the trajectory width G. It can be calculated using tan(C / G). Alternatively, if the image contains objects that are known to be standing perpendicular to the ground, such as utility poles or support posts, the rotation angle can be determined so that these objects are perpendicular.

[0044] There are vehicles with a tilting body, which tilt the body to improve ride comfort when passing through curves. If the vehicle has a tilt, in addition to the camera roll rotation due to the cant, image rotation may be performed to compensate for the rotation due to the tilt. In this embodiment, the vehicle is not a vehicle with a tilting body.

[0045] 403 is the center of rotation. Since we need to simulate roll rotation due to cant, the center of rotation needs to be shifted downwards from the center of the image by the camera's height relative to the track. The amount of shift required can be determined using the camera's focal length f, vertical field of view V, vertical pixel count, and camera installation height al. In that case, The center of rotation of the image is (al) / (2f(tanV / 2)) pixels, which is shifted downwards from the center of the image. In the case of a vehicle with a tilting body, the position of the camera relative to the axis of rotation of the tilting body must be considered, but the center of rotation of the image can be found using a similar approach.

[0046] 404 is the line representing the center of the u-coordinate. 404 is a straight line perpendicular to the center of 401. In the rotated image 402, when determining the angle ω on the track, assume that 404 is the line of sight 24 of the forward-facing surveillance camera.

[0047] Next, the target distance (e) is recalculated considering the cant amount using the rotated image 402. Figure 13 illustrates the method for recalculating the target distance (e) in the rotated image 402.

[0048] First, we search for pairs of points A and B on the left and right rails that have the same target distance Y. In Figure 13, let 501 be point B and 502 be point A. Draw a vertical line down from 503 and let point 503 be A', which has the same v coordinate as B. The length of line segment AA' is the cant, and the length of AB is the track gauge. Also, since points A, A', and B have the same target distance Y, the ratio of the lengths of the line segments on the screen can be considered the same as the ratio of the actual lengths. Therefore, if the cant is C meters, the track gauge is H meters, the number of pixels of AA' on the screen is c, and the number of pixels of AB is h, then C / H = c / h holds. Since C and G are known, we just need to find point A such that h / c = H / C for point B.

[0049] We can start point A at a different rail than point B, but with the same v-coordinate as point B. Then, we can calculate the h / c value while moving point A upwards along the rail one pixel at a time. When point A is at its initial position, c / h=0, and as point A moves upwards in the image, c / h increases, so at a certain point h / c≧H / C. This point is the desired position of point A.

[0050] Using the above method, a point A that satisfies the conditions can be found for any point B. Considering the line segment AB as the track gauge on the screen, the target distance (e) can be calculated using the relationship H / h=Y / f shown in Figure 5, with the number of pixels h in AB. The correction of the target distance considering curvature is given by the aforementioned formula.

number

[0051] The calculation module 204 can also perform the calculation to find d from e based on a table that stores the results of pre-calculated values ​​for the curve attributes and ω. This table uses the uncorrected target distance (e), radius of curvature, clothoid coefficient corresponding to the displacement amount, and cant amount as explanatory variables, and the corrected target distance (d) as the dependent variable. Train operators create this table in advance and store it in DB1003b. By using this table, the image calculation module 204 can quickly obtain the distance to a predetermined area of ​​the track in front of the moving train.

[0052] According to the forward monitoring system 10 and its method, even if the track in front of the train is a curved track, the distance to a predetermined area of ​​the track in front of the train can be determined based on images taken by the forward monitoring camera, with minimal errors due to curves and cant, or in other words, with high accuracy, just as in the case of a straight track.

[0053] To summarize the embodiments described above, the following invention is presented. The first forward monitoring system comprises a surveillance camera 20 that photographs the area in front of the train 1 and an image processing device 30, wherein the image processing device 30 is a forward monitoring system 10 that estimates the distance to a predetermined monitoring area 26 of the track 50 in front of the train based on the image captured by the surveillance camera, wherein the image processing device 30 comprises a feature extraction module 200 that extracts the track from the captured image and a calculation module 204 that calculates the distance based on the track width occupied by the track in the captured image, wherein the calculation module 204 is configured to calculate the distance to the predetermined area based on the attributes of the curved track in front of the train.

[0054] The second forward monitoring system is characterized in that, in the first forward monitoring system, the calculation module 204 calculates the distance to a predetermined area based on the track width occupied by the curved track in the captured image, the target distance to a predetermined area calculated based on an eigenvalue predetermined as the track width of the track, the radius of curvature and the cant amount as attribute information of the circular arc track as a curved track, and the displacement amount as attribute information of the transition curved track between the circular arc track and the straight track.

[0055] The third forward monitoring system is characterized in that, in any one of the forward monitoring systems described above, the calculation module 204 corrects the target distance based on the radius of curvature and the amount of displacement.

[0056] The fourth forward-looking system is characterized in that, in any one of the forward-looking systems described above, the calculation module 204 recalculates the target distance based on the cant quantity.

[0057] The fifth forward monitoring system is characterized in that, in any one of the forward monitoring systems described above, the calculation module 204 corrects the distance to a predetermined region calculated based on the track width occupied by the curved track in the captured image and an eigenvalue predetermined as the track width of the track, based on the attributes of the curved track.

[0058] The sixth forward monitoring system is characterized in that, in any one of the forward monitoring systems described above, the calculation module 204 calculates the distance to a predetermined area based on at least one of a plurality of parameters as attributes.

[0059] The seventh forward monitoring system is characterized in that, in any one of the forward monitoring systems described above, the calculation module 204 calculates the distance to a predetermined region based on the respective attributes of the circular arc trajectory as a curved trajectory and / or the transition curve trajectory between the circular arc trajectory and the straight trajectory.

[0060] The eighth forward monitoring system is characterized in that, in any one of the forward monitoring systems described above, the calculation module 204 calculates the distance to a predetermined region based on the radius of curvature and / or the cant amount as attribute information of the circular trajectory.

[0061] The ninth forward monitoring system is characterized in that, in any one of the forward monitoring systems described above, the calculation module 204 calculates the distance to a predetermined region based on the displacement amount as attribute information of the transition curve trajectory.

[0062] The tenth forward-looking system is characterized in that, in any one of the forward-looking systems described above, the calculation module 204 and the surveillance camera are monocular cameras.

[0063] Furthermore, the forward monitoring method is characterized in that, in monitoring the area in front of the train by estimating the distance to a predetermined area of ​​the track in front of the train based on images captured by a surveillance camera that photographs the area in front of the train, the track is extracted from the captured images, and when calculating the distance based on the track width that the track occupies in the captured images, the distance to the predetermined area is calculated based on the attributes of the curved track in front of the train.

[0064] Although the present invention has been described above using embodiments, it goes without saying that the technical scope of the present invention is not limited to the scope described in the above embodiments. It will be obvious to those skilled in the art that various modifications or improvements can be made to the above embodiments. Furthermore, it is clear from the claims that such modified or improved forms may also be included in the technical scope of the present invention. [Explanation of Symbols]

[0065] 10: Forward-looking system, 20: Surveillance camera, 26: Surveillance area, 30: Image processing device, 50: Orbit, 200: Feature extraction module, 204: Computation module

Claims

1. A forward-facing monitoring system comprising a surveillance camera for photographing the front of a train and an image processing device, wherein the image processing device estimates the distance to a predetermined area of ​​the track in front of the train based on the image captured by the surveillance camera, The aforementioned image processing device is A feature extraction module for extracting the trajectory from the aforementioned captured image, and The system includes a calculation module that calculates the distance based on the trajectory width that the trajectory occupies in the captured image, The calculation module is characterized in that it calculates the distance to the predetermined area based on the attributes of the curved track in front of the train. Forward monitoring system.

2. The aforementioned calculation module is The target distance to the predetermined region is calculated based on the track width of the curved track in the captured image and the eigenvalues ​​predetermined as the track width of the track, and Based on the radius of curvature and cant amount as attributes of the circular arc trajectory as the curved trajectory, and at least one of the transition amount as an attribute of the transition curve trajectory between the circular arc trajectory and the straight trajectory, The method is characterized by calculating the distance to the predetermined region. The forward monitoring system according to claim 1.

3. The aforementioned calculation module is The objective distance is corrected based on the radius of curvature and the amount of displacement, characterized in that The forward monitoring system according to claim 2.

4. The calculation module is characterized by recalculating the target distance based on the cant quantity. The forward monitoring system according to claim 2.

5. The aforementioned calculation module is The method is characterized by correcting the target distance to the predetermined region, which is calculated based on the track width of the curved track in the captured image and an eigenvalue predetermined as the track width of the track, based on the attributes of the curved track. The forward monitoring system according to claim 1.

6. The aforementioned calculation module is The distance to the predetermined region is calculated based on at least one of the multiple parameters that are considered attributes. The forward monitoring system according to claim 1.

7. The aforementioned calculation module is The distance to the predetermined region is calculated based on the respective attributes of the circular arc trajectory as the curved trajectory, and / or the transition curve trajectory between the circular arc trajectory and the straight trajectory. The forward monitoring system according to claim 5.

8. The calculation module is characterized by calculating the distance to the predetermined region based on the radius of curvature and / or the cant amount as attributes of the circular trajectory. The forward monitoring system according to claim 7.

9. The calculation module is characterized by calculating the distance to the predetermined region based on the displacement amount as an attribute of the transition curve trajectory. The forward monitoring system according to claim 7 or 8.

10. The aforementioned surveillance camera is characterized by being a monocular camera. The forward monitoring system according to claim 1.

11. In a forward monitoring method for monitoring the area in front of a train by estimating the distance to a predetermined area of ​​the track in front of the train based on images captured by a surveillance camera that photographs the area in front of the train, A forward monitoring method characterized in that, when extracting the track from the captured image and calculating the distance based on the track width occupied by the track in the captured image, the distance to the predetermined area is calculated based on the attributes of the curved track in front of the train.

Images