Track detection system

Abstract

To provide a track detection system capable of estimating the shape of a track in an area in which the image quality of a video camera deteriorates or an image cannot be acquired.SOLUTION: In a track detection system including imaging means provided in a vehicle traveling on a railroad track to acquire a video in front of the vehicle in a traveling direction, vehicle position detection means for detecting a position of the traveling vehicle, a database storing information on various railroad lineside facilities provided along the railroad track, and an arithmetic unit for performing image processing of the video acquired by the imaging means to enable detection of at least a portion of the track, in the database, the information on the lineside facilities are associated with position information to be stored, and the arithmetic unit acquires the information on the lineside facilities within a prescribed range in front of the vehicle from the database by using the position information detected by the vehicle position detection means, and estimates a track shape in front of the vehicle in a vehicle traveling direction on the basis of the acquired information on the lineside facilities and detected information on the track.SELECTED DRAWING: Figure 4

Description

[Technical field]

[0001] The present invention relates to a track detection system that detects the tracks of a railway track, and in particular to a track detection system that estimates the shape of the tracks ahead in the direction of travel based on images taken of the area ahead of the vehicle by an imaging device (video camera) mounted on the railway vehicle. [Background technology]

[0002] For safe and smooth operation of railway vehicles, it is necessary to quickly detect obstacles ahead of the vehicle and slow down or stop the vehicle. In order to quickly detect obstacles ahead of the vehicle, it is also important to understand the shape of the track ahead of the vehicle. 2. Description of the Related Art Conventionally, there has been an invention relating to a vehicle front monitoring device that monitors the front of a vehicle using an image captured by a video camera and detects and displays an obstacle, as described, for example, in Patent Document 1.

[0003] In addition, an invention has been proposed that detects the shape of tracks and track branches from images taken of the area ahead of the railway vehicle's direction of travel, and identifies the track on which the vehicle will travel among multiple tracks that the vehicle may potentially travel beyond the branch, by utilizing already detected track information (Patent Document 2). Furthermore, an invention has been proposed that uses two types of cameras, a wide-angle camera and a telephoto camera, as a forward monitoring device to narrow down the track pattern of the rail vanishing point and eliminate erroneous detection of points that should be imaged (Patent Document 3). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2019-221115 A [Patent Document 2] JP 2019-218022 A [Patent Document 3] JP 2015-49740 A Summary of the Invention [Problem to be solved by the invention]

[0005] When monitoring the area in front of a train using a video camera, image quality can deteriorate depending on the shooting environment, such as at night, in a tunnel, or in bad weather, and accurate track detection can be difficult due to special track configurations such as railroad crossings and forks. In addition, if the track ahead curves sharply and is hidden behind a building or has a downward slope, it is impossible to capture an image of that area, making track detection impossible.

[0006] However, neither of the inventions in Patent Document 1 nor Patent Document 2 discloses a technique for estimating the shape of tracks in areas where image quality has deteriorated or where an image cannot be acquired. In addition, while the invention in Patent Document 3 can narrow down the track pattern of the rail vanishing point, it has the problem that it uses two types of cameras, a wide-angle camera and a telephoto camera, which places a heavy burden on the image processing device and takes a long time to produce processing results.

[0007] The present invention has been made against the background described above, and aims to provide a railway track detection system that can estimate the shape of railway tracks in areas where the image quality of the video camera is degraded or where images cannot be acquired. Another object of the present invention is to provide a track detection system that can estimate the shape of tracks in areas where the image quality of the video camera is degraded or images cannot be acquired, and can quickly recognize obstacles in front of the vehicle. [Means for solving the problem]

[0008] In order to solve the above problems, the present invention relates to A track detection system including an imaging means provided on a vehicle traveling on a railroad track for capturing an image of a road ahead in the direction of travel, a vehicle position detection means for detecting the position of the traveling vehicle, a database storing information on various wayside facilities provided along the railroad track, and a computing device capable of detecting at least a portion of the track by image processing the image captured by the imaging means, The database stores information about the lineside facilities linked to location information, The calculation device is configured to obtain information on the wayside facilities within a specified range ahead of the vehicle from the database using the position information detected by the vehicle position detection means, and to estimate the track shape ahead in the direction of vehicle travel based on the obtained wayside facility information and track detection information obtained by image processing.

[0009] According to the track detection system configured as described above, even in an area where the image quality of the imaging means (video camera) is degraded or an image cannot be acquired, the shape of the track can be estimated by using information from wayside facilities. In addition, since the shape of the track in an area where the image quality of the imaging means is degraded or an image cannot be acquired can be estimated, the range to be focused on in the image acquired from the imaging means can be narrowed down, and obstacles in front of the vehicle can be quickly recognized.

[0010] Here, it is preferable that the information on wayside facilities includes at least the curvature and / or gradient of the track ahead in the vehicle's travel direction, position information on railroad crossings, and position information on junctions. The image quality of the image pickup means (video camera) mounted on the vehicle for capturing images of the area ahead in the traveling direction is degraded or images cannot be captured in the areas where the track ahead is curved or there is a crossing or a branch point ahead. Therefore, with the above-mentioned configuration, it is possible to estimate the shape of the track when the track ahead is curved or there is a crossing or a branch point ahead.

[0011] Furthermore, preferably, the calculation device is configured to sequentially calculate the cross-correlation between one result and the other result detected by image processing of the image acquired by the imaging means while shifting one result in parallel in the vertical and horizontal directions based on the result of the estimation and the result of the detection, using a predetermined formula, to determine the amount of movement when the correlation is highest, and to output the result with the smallest amount of movement as the track detection result. With this configuration, the track geometry to be output can be determined based on the track geometry estimated using information from lineside facilities and the track geometry detected by image processing, thereby improving the accuracy of the track geometry.

[0012] Also, preferably, the calculation device is configured to calculate the time required to obtain information about the wayside facilities ahead of the vehicle from the database, and to add the distance traveled during the calculated time to the position detected by the vehicle position detection means, thereby correcting the vehicle's own position. With this configuration, the vehicle's own position is corrected taking into account the time required to obtain information about the wayside facilities from the database, so that the image area of ​​the wayside facilities reflected in the image obtained by the imaging means can be accurately grasped, thereby improving the detection accuracy of the track shape. Effect of the Invention

[0013] The track detection system according to the present invention can estimate the shape of the track in an area where the image quality is degraded or where it is not possible to acquire an image. Also, by estimating the shape of the track in an area where the image quality is degraded or where it is not possible to acquire an image and recognizing the travel route ahead, it is possible to quickly recognize obstacles ahead of the vehicle. [Brief description of the drawings]

[0014] [Figure 1] 1 is a system configuration diagram showing an embodiment of a railroad detection system according to the present invention; [Diagram 2] 13(A) and 13(B) are images showing an example in which it is difficult to recognize the shape of the track in the image captured by a video camera mounted on a vehicle. [Diagram 3] 13A and 13B are diagrams showing example images in which a track shape is estimated by the track detection system of an embodiment in an image in which the track shape is difficult to recognize. [Figure 4] 10 is a flowchart illustrating an example of a procedure of a track detection process in the track detection system of the embodiment. [Diagram 5] 1A is an explanatory diagram showing an example of a curved track from above, and FIG. 1B is an explanatory diagram showing an enlarged view of a point PM and a next point PM+1 in a curvature portion of the curved track. [Figure 6] FIG. 1 is an explanatory diagram showing an example of a track with an elevation difference as viewed from the side. [Figure 7] (A) is an explanatory diagram showing an example of a mask area and interpolation processing ahead of the tracks, (B) is an explanatory diagram showing an example of a mask area when there is a branch point ahead of the tracks, and (C) is an explanatory diagram showing an example of interpolation processing in the mask area of ​​(B). [Figure 8] The diagram shows the procedures for two methods of estimating track shape using information about lineside facilities and displaying it with high accuracy on camera images. (A) is an explanatory diagram of the procedure when database reference results are used as preprocessing for track detection from images, and (B) is an explanatory diagram of the procedure when the track detection results from images are compared with the database reference results to make a comprehensive judgment. [Figure 9] 5 is a flowchart showing an example of a procedure for a process of calculating a position of a vehicle in the railroad detection system according to the embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] An embodiment of a track detection system according to the present invention will now be described with reference to the drawings, in which: Fig. 1 is a block diagram showing an example of the configuration of a track detection system mounted on a railway vehicle (train). As shown in Fig. 1, a track detection system 10 according to this embodiment includes a video camera 11 that is installed at the front end of the leading car with its lens facing forward and captures the image ahead in the running direction, a calculation device (image processing device) 12 that processes the image signal from the video camera 11, a speed generator 13 that detects the number of rotations of the wheels to calculate the running speed and the vehicle's own position, an on-board coil 14 that receives position correction information from a transponder ground coil, a communication device 15 that receives information transmitted from the traffic control system and the safety system via digital radio and track circuits, a GPS device 16 that receives GPS signals from GPS satellites, and a database (storage device) 17 that stores information used by the calculation device 12 when processing the image signal. In addition, a monitor 18 installed in the driver's seat for displaying the image captured by the video camera 11 may be included.

[0016] Database 17 stores information about railway trackside facilities (maintenance facility information) such as railroad crossings, junctions, tunnels, stations, track curvature, and gradient of the line on which the vehicle is traveling, linked to location information (distance in kilometers) and stored in a format such as a CSV file. The database for storing the wayside facility information may be provided in a server or the like on the ground side, and the vehicle may transmit position information (distance in kilometers) as search information to the wayside side to sequentially obtain the necessary wayside facility information. The database 17 may also store the ID (identification number) or train number of the vehicle.

[0017] The arithmetic unit 12 is composed of a microprocessor (MPU), a non-volatile storage device such as a ROM that stores programs executed by the MPU, and a readable / writable storage device such as a RAM, and has a function of calculating its own position (kilometers) based on a signal from the tachograph generator 13, and a function of converting latitude and longitude information obtained from a GPS signal into kilometer distance information. In addition, since the kilometer distance calculated based on the signal from the tachograph generator 13 and the latitude and longitude information obtained from the GPS signal contains an error, the arithmetic unit 12 has a function of correcting its own position using position correction information received from the transponder ground coil by the on-board coil 14.

[0018] Along railway lines, safety systems are installed that have the function of detecting the position of a running train based on signals from track circuits, etc., and transmitting signals from ATC ground equipment to trains and controlling signals and points to prevent trains from getting too close to preceding trains. In addition, existing railway systems also include traffic control systems that have the function of detecting the position and route of trains, monitoring whether they are running according to the timetable, and providing route information to the safety system. Therefore, the track detection system 10 of this embodiment can grasp the route on which the vehicle is traveling and the direction of travel by acquiring information on the vehicle's position from the safety system or traffic control system using the calculated or corrected information on the vehicle's position or the vehicle ID or train number via the communication device 15, and can read in advance information on wayside facilities ahead in the direction of travel from the database 17. Also, information on the vehicle's route and direction of travel may be stored in advance in the vehicle database 17 without going through the communication device 15.

[0019] Next, functions of the railroad track detection system of this embodiment will be described. When monitoring the front of a vehicle using a video camera, if there is a tunnel ahead as shown in Figure 2 (A), the shape of the nearby tracks can be detected, but it is difficult to detect the shape of tracks further away. The same is true when driving at night with the headlights on or in the rain. Also, as shown in Figure 2 (B), if the front of the track is greatly curved and hidden by a building or other object, it is not possible to obtain an image of that area, and track detection may not be possible. The same is true when there is a downward slope ahead.

[0020] Furthermore, in situations where there are junctions or railroad crossings ahead of the tracks, the camera image will capture track materials and equipment that are easily mistaken for rails in addition to the rails to be detected, such as guard rails, tongue rails, and lead rails at junctions, and steel paving plates and guard rails at railroad crossings, making it impossible to perform correct track detection. In the above-mentioned cases, the railway detection system of this embodiment estimates the railway shape by eliminating the influence of track materials and facilities based on wayside facility information acquired from a database. Specifically, for example, when the image quality ahead of the tracks is poor or there are areas without images, as in Figures 2(A) and (B), the shape of the tracks is estimated based on lineside facility information obtained from a database, and the area to be focused on is narrowed down using the forward image and the estimated shape or the tracks detected in the image, as shown by the dashed lines in Figures 3(A) and (B), thereby correctly recognizing obstacles.

[0021] A specific method for estimating the track geometry ahead in the vehicle travel direction based on wayside facility information will be described below. Fig. 4 shows a flowchart of the overall procedure of track detection processing including track geometry estimation processing. The track detection process in Fig. 4 is started when the track detection system 10 is powered on. When the system is powered on, power is supplied to the video camera 11, which starts capturing images. The video camera 11 may start capturing images in conjunction with the movement of a train.

[0022] 4 is started, the calculation device (image processing device) 12 executes a process of calculating the position of the vehicle (step S1). The position of the vehicle can be calculated from the position of the departure station (e.g., kilometers) and the traveled distance calculated based on a signal from the speed generator 13.

[0023] Next, using the position information calculated in step S1, equipment information along the track within a predetermined range (for example, up to 1 km ahead) ahead of the vehicle in the direction of travel is obtained from database 17 (step S2). Next, video data is read from video camera 11 (step S3), and a process is executed to detect the shape of the track in the video (step S4). At this time, based on the equipment information obtained in step S2, for example, the image area of ​​the equipment in the image is masked to perform the process of detecting the shape of the track. This makes it possible to suppress erroneous detection of the track.

[0024] Thereafter, it is determined whether or not the shape of the track has been detected throughout the entire range of the image, that is, over the long distance (step S5), and if it is determined that the shape of the track has been detected over the long distance (Yes), the process proceeds to step S7. On the other hand, if it is determined in step S5 that the shape of the track could not be detected (No), the process proceeds to step S6, where a process of estimating the track shape ahead is executed. The details of the process of estimating the track shape ahead will be described later.

[0025] Next, a process is executed to recognize obstacles on the track using the forward image and the track shape estimated in step S6 or the track detected in the image in step S5 (step S7). After that, the process returns to step S1 and the above process is executed repeatedly. Here, the obstacle recognition process can be realized by utilizing the known image processing technology described in the above-mentioned Prior Art Document 1 and the like, and therefore a detailed explanation will be omitted.

[0026] Next, a specific method of estimating the shape of the track ahead by the track detection system of this embodiment will be described with reference to Figs. 5 to 7. Of these, Fig. 5(A) is a schematic diagram equivalent to a plan view of a curved track seen from above, which can be represented using track curvature information in a database. Also, Fig. 6 is a schematic diagram equivalent to a side view of a track with an elevation difference seen from the side, which can be represented using track gradient information in a database. In Fig. 5(A), the dashed line is the center line of the track. In the track geometry estimation process, first, as shown in Fig. 5(A), the camera position is set as the origin O, and points P0, P1, P2, P3...PN are taken from there in the direction of travel at intervals of a predetermined width Δd, and the curvatures k0 to kN and gradients sθ to sN at points P0 to PN are found by referring to the trackside facility information in the database.

[0027] In the case of Figure 5(A), the curve shape of the track is determined by the curvatures k0 to kN at each point P0 to PN ahead of the origin O in the direction of travel. Figure 5(B) shows an enlarged view of a point PM and the next point PM+1 on the curved portion of the track. If the radius of curvature at point PM is RM, then RM = 1 / kM. Here, the distance between PM and PM+1 (step width Δd) is, strictly speaking, the length along the arc, but if Δd is small, the length of the arc between PM and PM+1 can be considered to be equal to the length of the straight line connecting both points. If the interior angle between points PM and PM+1 at the center of curvature is θ as in Figure 5(B), then the horizontal movement amount Δt from point PM to point PM+1 is given by the following equation Δt=Δd·sin(θ / 2)≒Δd· (θ / 2) It can be calculated as follows.

[0028] Furthermore, when it is determined from the lineside facility information that the point PM-PM+1 is a branch point, the horizontal movement amount Δt from the point PM to the point PM+1 is calculated by the following equation using the angle θB determined from the number of the turnout. Δt=Δd· (θ / 2)+Δd· (θB / 2) It can be calculated as follows. By repeating the above calculation from the origin O to point PN, the curve of the center line of the track on the horizontal plane can be determined. Then, by performing coordinate transformation on this, the estimated track shape can be drawn on the camera image as shown in Figure 3(B).

[0029] Next, a method for estimating the track geometry taking into account the track gradient will be described with reference to FIG. When the track ahead has a gradient, it is necessary to estimate the track shape based on the gradient at the camera position, i.e., the origin O. Specifically, when the gradient s0 at the origin O is negative, i.e., a downhill slope, and the gradient sM at the forward point PM is 0, the track ahead appears to be an uphill slope when viewed from the camera position. In the above case, as shown in Figure 6, if the distance from the optical axis of the camera at point PM to the orbital plane is hM, the distance from the optical axis to the orbital plane at point PM+1, hM+1, is calculated by the following equation: hM+1=hM+(sM-s0) Δd It can be calculated as follows.

[0030] By repeating the above calculation from the origin O to point PN, the shape of the track surface on a vertical plane based on the optical axis of the camera can be determined. Then, by performing coordinate transformation on this, the position of the track on the camera image can be shown. The shape of the track surface can be determined in a similar manner when the gradient at the camera position, i.e., the origin O, is 0 and the track ahead is inclined. In addition, by combining the track shape on the horizontal plane explained using FIG. 5 and the track shape on the vertical plane explained using FIG. 6, it is possible to estimate a track shape in which the gradient changes as it curves. Furthermore, even if there is a tunnel ahead, such as that shown in Figure 2(A), the track shape can be estimated by referring to lineside facility information, and the estimated track shape can be drawn on top of the camera image, as shown in Figure 3(A).

[0031] Next, a method for estimating the track shape when there is a railroad crossing or a branch point ahead of the vehicle will be described with reference to FIG. When the system references the wayside facility information in the database and determines that there is a railroad crossing or junction ahead of the vehicle and its location, it masks the image area of ​​the railroad crossing or junction on the camera image. In the case of a railroad crossing, it then performs an interpolation process to draw lines connecting the edges of the rails detected in the areas before and after the masked area MA, as shown in Figure 7(A). It then performs a process to calculate a line connecting the centers of the left and right edges for each rail and determine the track shapes P1 and P2.

[0032] In addition, if there is a junction ahead of the vehicle, information on the type of switch and its shape known from the lineside facility information is used to perform an interpolation process to draw lines connecting the rail edges determined by image processing in the areas before and after the masked area MA, as shown in Figure 7(B), and in the same manner as above, a process is performed to determine the track shapes P1, P2; P3, P4 as shown in Figure 7(C). If there is a station ahead of the vehicle, the track shape can be determined by masking the image area of ​​the platform and other facilities near the tracks and then performing image processing.

[0033] Next, a method will be described in which the track geometry is estimated with high accuracy using the trackside facility information as described above, and the track detection results obtained by processing the images from the camera are matched with the actual track geometry. Possible methods for estimating track geometry using wayside facility information and matching it with camera images with high accuracy include a method (first method) in which the estimation process of track geometry using wayside facility information is performed as pre-processing of track detection by image processing of camera images, as shown in Fig. 8(A), and a method (second method) in which the track position is corrected by comparing the track detection results by image processing of camera images with the track geometry estimation results using wayside facility information, as shown in Fig. 8(B). The two methods will be explained in order below.

[0034] (1st method) When detecting tracks from camera images using image processing, the track position is identified by searching within the image to find the edges of the rails. In such cases, for example, under conditions where the light from the headlights only illuminates the area in front and the edges of the rails become unclear as the distance increases, it is difficult to identify the track position in the distance where the light from the headlights does not reach, and non-track facilities, buildings, and noise components may be mistakenly detected as tracks.

[0035] Therefore, as described above, it is possible to obtain wayside facility information from a database based on vehicle position information, and use information on the track curvature and gradient contained in the wayside facility information to estimate at which vertical and horizontal coordinate position in the image the track ahead is located, so that by narrowing down the estimated range and searching for the rail edges, it is possible to improve detection accuracy by determining the track shape by compensating for the blurred edges even if the image quality of the imaging means is poor. Also, because it is possible to estimate the shape of the track, it is possible to narrow down the range to be focused on in the image acquired from the imaging means, so that obstacles ahead of the vehicle can be quickly recognized.

[0036] When setting the search range, it is advisable to set appropriate margins in both the vertical and horizontal directions, taking into consideration camera installation errors and image blurring caused by vehicle swaying while traveling.Then, the expected track inclination is set with respect to the vertical coordinate axis (Y-axis) of the set search range, and the rail edges are searched for in the image using the estimated position and inclination as constraints to determine the track shape.This makes it possible to detect the track position in the image with high accuracy.

[0037] (Second method) The first method described above is effective in reducing false positives and detection failures caused by unclear rail edges. However, when the track ahead is hidden by trackside buildings on a sharp curve or there is a downward slope ahead, the distant track is not reflected in the image in the first place, making it impossible to detect the track from the camera image alone. Therefore, the second method is to identify the overall shape of the track ahead by simultaneously performing a process of detecting tracks through image processing of camera images and a process of estimating the track shape using information on lineside facilities in a database.

[0038] In addition, after carrying out the two processes in parallel, it is necessary to align the results of track detection using images (A) and the results of track shape estimation using a database (B). For alignment, a rough alignment (first stage) is first performed using the image of the area near the vehicle in the camera image (the lower part of the image) as the reference, and then a more precise alignment (second stage) is performed using the results of the entire area detected by image processing. In the first stage of alignment, result B is translated vertically and horizontally so that the center and width of the tracks in the lower part of the image (the foreground area) are consistent between result A and result B.

[0039] Next, in the second stage of alignment, alignment is performed using cross-correlation between the entire detection result of result A and the corresponding part of result B. Specifically, using the horizontal and vertical positions of the first stage alignment results as a reference, result B is moved up, down, left, and right, and the cross-correlation between result A and result B' after the movement is calculated. First, the tracks are detected from the captured camera image using image processing techniques such as edge detection. The track detection results are expressed as binary data (whether there is a track or not). On the other hand, a geographic information engine is defined as an engine that, when given the current location information, outputs information on the curvature and gradient of the track ahead in the direction of travel. The obtained curvature and gradient information is used to reconstruct the 3D shape of the left and right rails of the track, and this is then converted by viewpoint transformation into the track shape as seen from the driver's cab, resulting in an estimated track shape. The estimated track shape results are also expressed, for example, as binary data.

[0040] In order to measure the similarity between the two types of binary data obtained as described above, it is common to evaluate them using a correlation coefficient. Methods for calculating correlation include SSD (sum of squared differences), SAD (sum of absolute differences), and NCC (normalized cross correlation). If the result of detection by image processing is A and the result of estimating the track shape is B, and the coordinates from the origin of the image are expressed as (x, y), and the pixel value at position (x, y) in A is expressed as A(x, y) (similar for B), then for SSD, the following formula (Equation 1) is used:

number

number

[0041] Here, since vibrations generally occur in moving vehicles, the range of image A captured by a camera attached to the vehicle fluctuates vertically and horizontally over time. On the other hand, the track shape estimation result B does not take into account the vibrations of the vehicle, and is an image taken from the viewpoint of looking in the direction of travel from the center of the track. Therefore, consider image B' obtained by translating B by Δx in the horizontal direction and Δy in the vertical direction, and the correlation coefficient between A and B' is given by the following equation (Equation 3):

number

[0042] Next, a specific example of the self-location recognition process in step S2 of the flowchart in Fig. 4 will be described. Fig. 9 shows a flowchart of the procedure of the self-location recognition process. When this process starts, first, the elapsed time is calculated from the corrected time of the previous position (in kilometers) (step S21). Then, it is determined whether the elapsed time is equal to or longer than a predetermined time (e.g., 60 seconds) (step S22). If it is determined that the elapsed time is equal to or longer than the predetermined time (Yes), the process proceeds to the next step to determine whether the GPS accuracy is within a predetermined range (e.g., within 10 m) (step S23). If it is determined that the GPS accuracy is within the predetermined range (Yes), the latitude and longitude information is acquired from the GPS device (step S24).

[0043] Next, the current position (kilometres) is calculated based on the signal from the speed generator 13 (step S25), and the calculated position information is used to refer to the database to obtain geographic information (step S26). Then, the kilometres contained in the geographic information are obtained (step S27), and the time required to refer to the geographic information database is calculated (step S28). Next, it is determined whether or not the kilometres have been successfully obtained from the database (step S29). If it is determined that the acquisition was successful (Yes), the process proceeds to the next step, where the distance traveled during the time required to refer to the database is added to the kilometres calculated in step S25 to correct the vehicle's own position (kilometres) (step S30). After that, the correction time of the vehicle's own position (kilometres) is updated, and the process ends (step S31).

[0044] According to the above process, the vehicle's own position (km) is corrected by taking into account the distance traveled during the time required to refer to the database, so that highly accurate self-position information can be maintained. In addition, although not shown in Fig. 7, when the on-board unit 14 acquires position information (km) from a transponder ground unit, the vehicle's own position (km) may be corrected based on the acquired position information.

[0045] Although the embodiment of the present invention has been described above, the present invention is not limited to the above embodiment, and various modifications and changes are possible. For example, in the above embodiment, the arithmetic device 12 constituting the on-board system grasps the position information of the vehicle itself, acquires wayside facility information based on the position information, and detects the railroad track to be the route, but since the position of each vehicle can also be grasped by a safety system on the ground side, the system may be configured so that the arithmetic device provided on the ground side acquires wayside facility information based on the position information of each vehicle, detects the railroad track to be the route, and transmits the railroad track detection result to the corresponding vehicle.

[0046] In addition, in the above embodiment, an example was described in which a video camera was used as an imaging means for acquiring images of the area ahead in the direction of travel of the vehicle, but the imaging means is not limited to a video camera and may be a still camera capable of continuously taking still images, or the like. Furthermore, in the above embodiment, information on track facilities is obtained from a database and used to detect tracks in an image, but it is also possible to obtain train operation information or position information of other trains from the database, recognize images of oncoming trains reflected in the image, and mask the area of ​​the image to perform track detection processing. In addition, in the above embodiment, the shape of the track in areas where the camera image quality is degraded or images cannot be acquired may be estimated, and the shape of the track may be displayed on the monitor 18 in the driver's seat of the train to alert the driver and provide driving assistance. [Explanation of symbols]

[0047] 10. Railroad Track Detection System 11 Video Camera 12 Calculation device (image processing device) 13 Speed ​​generator 14 Car 15. Communications Equipment 16 GPS device 17 Database 18 Monitors

Claims

1. A track detection system comprising: an imaging means provided on a vehicle traveling on a railroad track for capturing an image of a road ahead in the direction of travel; a vehicle position detection means for detecting the position of the traveling vehicle; a database storing information on various wayside facilities provided along the railroad track; and a computing device capable of processing the image captured by the imaging means to detect at least a portion of the track, The database stores information about the lineside facilities linked to location information, The computing device uses position information detected by the vehicle position detection means to obtain information on the wayside facilities within a predetermined range ahead of the vehicle from the database, and estimates the track shape ahead in the direction of vehicle travel based on the obtained wayside facility information and track detection information obtained by image processing.

2. The railway detection system according to claim 1, characterized in that the information on the wayside facilities includes at least the curvature and / or gradient of the tracks ahead in the direction of vehicle travel, location information of railroad crossings, and location information of branching points.

3. The arithmetic unit detects tracks from the image acquired by the imaging means, The track detection system according to claim 2 , wherein the track detection result is combined with an estimated track shape ahead in the vehicle travel direction, so that the result can be used for recognition processing of an obstacle ahead of the vehicle.

4. The railroad detection system according to claim 3, characterized in that the calculation device is configured to sequentially calculate the cross-correlation between one result of the estimation and the other result of the detection performed by image processing of the image acquired by the imaging means while shifting the other result in parallel in the vertical and horizontal directions based on the result of the estimation and the result of the detection performed by the image processing of the image acquired by the imaging means, to obtain the amount of movement when the correlation is highest, and to output the result with the smallest amount of movement as the railroad detection result.

5. The track detection system according to any one of claims 1 to 4, characterized in that the calculation device is configured to calculate the time required to obtain information about the wayside facilities ahead of the vehicle from the database, and to add the distance traveled during the calculated time to the position detected by the vehicle position detection means to correct the vehicle's own position.

Images