JP7170414B2 - Track identification device - Google Patents

Description

Embodiments of the present invention relate to track identification devices.

2. Description of the Related Art Conventionally, there is known a technique of detecting a track on which a railway vehicle travels from a captured image obtained by a camera by subjecting the captured image to image processing using a predetermined algorithm. As such a conventional algorithm, for example, among a plurality of preset parabolic railroad candidates, a candidate having the largest feature value set as a measure representing railroad-likeness is sequentially searched for each predetermined area. is known to detect tracks that form a single path.

"Rail Extraction for Driver Support in Railways", Conference: Intelligent Vehicles Symposium (IV), 2011 IEEE

The conventional algorithm as described above calculates and compares all the feature values of a plurality of parabolic candidates, so the amount of calculation tends to increase and the processing tends to take time. In addition, in the upper (deeper) region of the captured image, the width of the railroad track is smaller, and the area to be searched for railroad candidates is also smaller. .

A track identification device according to an embodiment includes an image acquisition unit, a feature amount extraction unit, and a track detection unit. The image acquisition unit acquires a captured image obtained by capturing an area in the traveling direction of the railway vehicle. The feature quantity extracting unit extracts a feature quantity representing the likeness of a railroad from the captured image. The track detection unit detects a track from the captured image based on the feature amount. The track detection section includes an initial detection section and a secondary detection section. The initial detection unit identifies, based on the feature amount, a line segment that is most likely to be a railway line extending between the lower end and the upper end of an initial area preset on the lower end side of the captured image, and identifies the identified line segment as one of the railway lines. detected as the initial part of the part. The secondary detection unit has the upper end of the initial portion as a starting point, and a plurality of line segments connecting a plurality of candidate points selected from within a search area set above the starting point and the upper end of the detected railroad track. Among the candidates, one line segment candidate with the highest railroad-likeness is determined based on the feature amount, the determined one line segment candidate is detected as a midway portion of a portion of the railroad, and the upper end of the midway portion is detected as follows. By repeatedly executing the process of setting the starting point, the midway portion is repeatedly detected until the upper end of the already detected track reaches a predetermined position.

FIG. 1 is an exemplary and schematic block diagram showing a schematic configuration of a vehicle system including a track identification device according to an embodiment. FIG. 2 is an exemplary and schematic block diagram showing the functions of the vehicle system including the track identification device according to the embodiment. FIG. 3 is an exemplary schematic diagram showing an example of a captured image according to the embodiment; FIG. 4 is an exemplary schematic diagram for explaining an initial detection unit according to the embodiment; FIG. 5 is an exemplary and schematic diagram for explaining a secondary detection unit according to the embodiment; FIG. 6 is an exemplary schematic diagram for explaining an interpolation detection unit according to the embodiment; FIG. 7 is an exemplary schematic diagram for explaining an example of interpolation curve approximation that can be performed by the interpolation detection unit according to the embodiment; FIG. 8 is an exemplary schematic diagram showing an example of approximated interpolation curves according to the embodiment. FIG. 9 is an exemplary and schematic flow chart showing a series of processes executed by the vehicle system including the track identification device according to the embodiment. FIG. 10 is an exemplary and schematic flow chart showing details of a track detection process according to the embodiment.

Hereinafter, embodiments will be described based on the drawings. The configurations of the embodiments described below and the actions and results (effects) brought about by the configurations are merely examples, and are not limited to the following descriptions.

<Embodiment>
First, the configuration of the embodiment will be described.

FIG. 1 is an exemplary and schematic block diagram showing a schematic configuration of a vehicle system 100 including a track identification device 120 according to an embodiment. The vehicle system 100 is a system installed in a railroad vehicle RV to realize safe driving of the railroad vehicle RV running along the railroad track RL. In addition, in FIG. 1, the traveling direction of the railway vehicle RV is illustrated as D. As shown in FIG.

As shown in FIG. 1 , the vehicle system 100 has an imaging device 110 , a track identification device 120 , an obstacle detection device 130 and a notification device 140 . The imaging device 110 is connected to the track identification device 120 and the obstacle detection device 130 . The obstacle detection device 130 is also connected to the track identification device 120 and the notification device 140 .

The imaging device 110 is a camera that captures an area (a range indicated by R0) in the traveling direction D of the railroad vehicle RV. In the embodiment, the installation position and installation angle of the imaging device 110 are set so that the railroad RL is reflected in the captured image obtained by the imaging device 110 .

The track identification device 120 detects the track RL from the captured image obtained by the imaging device 110 (details will be described later).

The obstacle detection device 130 obtains information about obstacles (observation objects that require attention) existing around the railroad RL based on the captured image obtained by the imaging device 110 and the detection result of the track identification device 120. (For example, the presence or absence of an obstacle, the distance to the obstacle, etc.) is detected.

The notification device 140 notifies the passenger (driver, etc.) of the railway vehicle RV of the detection result of the obstacle detection device 130 . The notification method may be a visual method using an image, a auditory method using voice, or a tactile method using vibration or the like. Alternatively, a combination of these methods may be used.

Here, in the embodiment, both the trajectory identification device 120 and the obstacle detection device 130 are configured as computers having hardware such as processors and memories. Therefore, the trajectory identification device 120 and the obstacle detection device 130 can realize functions as described below by executing a predetermined computer program stored in a memory or the like by their respective processors.

FIG. 2 is an exemplary and schematic block diagram showing functions of the vehicle system 100 including the track identification device 120 according to the embodiment. In addition, in the embodiment, each function shown in FIG. 2 may be realized only by dedicated hardware (circuit).

As shown in FIG. 2 , the track identification device 120 has an image acquisition section 121 , a feature amount extraction section 122 , a track detection section 123 and a storage section 124 . Further, the obstacle detection device 130 has a detection area acquisition section 131 and an obstacle detection section 132 as functional configurations.

The image acquisition unit 121 acquires the captured image obtained by the imaging device 110 . For example, the image acquisition unit 121 acquires an image 300 as shown in FIG. 3 below as a captured image.

FIG. 3 is an exemplary schematic diagram showing an example of a captured image according to the embodiment; An image 300 shown in FIG. 3 includes, as an example, a track 301 configured by a pair of rails 301a and 301b. As shown in FIG. 3, on the image 300, the width of the railroad track 301 (the distance between the pair of rails 301a and 301b) increases from the lower side (the near side) to the upper side (the far side) due to perspective. gradually becomes smaller.

Returning to FIG. 2 , the feature quantity extraction unit 122 extracts a feature quantity set as a scale representing the railroad RL-likeness from the captured image acquired by the image acquisition unit 121 . As the feature amount, for example, a gradient of luminance in a predetermined direction based on luminance may be used. The predetermined direction is, for example, a direction that intersects the direction in which the line RL extends in the captured image (for example, the horizontal direction in the example shown in FIG. 3).

The track detection unit 123 detects the track RL from the captured image acquired by the image acquisition unit 121 based on the feature quantity extracted by the feature quantity extraction unit 122 .

By the way, conventionally, as an algorithm for detecting a railroad track RL from a captured image, for example, among a plurality of preset parabolic railroad track candidates, a candidate having the largest feature value set as a measure representing the likelihood of a track RL is selected. is sequentially searched for each predetermined area to detect the line RL that constitutes a single route.

Such a conventional algorithm calculates and compares all the feature values of a plurality of parabolic candidates, so the amount of calculation tends to increase and the processing tends to take a long time. In addition, in the upper (deeper) area of the captured image, the width of the line RL becomes smaller, and the area to be searched for the candidate for the line RL becomes smaller, so the accuracy decreases due to the influence of local noise and the like. It's easy to do.

Therefore, the track detection unit 123 according to the embodiment detects the track RL from the captured image at a higher speed and with a higher accuracy by using a configuration described below.

That is, the track detection unit 123 according to the embodiment is configured to detect the track RL from the captured image at a higher speed and with higher accuracy. an initial detection unit 123a and a secondary detection unit 123b that detect the portion up to the line segment as a set of line segments, and an interpolation detection unit 123c that detects the remaining portion of the line RL based on an interpolation curve described later. .

FIG. 4 is an exemplary and schematic diagram for explaining the initial detection unit 123a according to the embodiment. As shown in FIG. 4, the initial detection unit 123a according to the embodiment includes a line extending between the lower end and the upper end of regions 401a and 401b as initial areas set in advance on the lower end side of an image 300 as a captured image. A line segment with the highest likelihood is specified based on the feature quantity, and the specified line segment (see the thick line) is detected as an initial part of the railroad track 301 . Information about the initial area and information about the detected track RL are stored in the storage unit 124 (see FIG. 2) of the track identification device 120 . Then, in the embodiment, starting from the upper ends P1a and P1b of the line segment specified as the initial portion, the secondary detection unit 123b performs the processing described below.

FIG. 5 is an exemplary and schematic diagram for explaining the secondary detection unit 123b according to the embodiment. As shown in FIG. 5, the secondary detection unit 123b according to the embodiment first uses the upper ends P1a and P1b of the initial portion as starting points, and a search area 501a set above the upper end P1a as the starting point. A plurality of candidate points P2a, P2a′, and P2a″ as candidates for the end point with respect to the upper end P1a as the starting point are selected from within the search area 501b set above the upper end P1b as the starting point, A plurality of candidate points P2b, P2b', and P2b'' are selected as end point candidates for the top edge P1b as the starting point.

Then, the secondary detection unit 123b selects one line segment having the highest likelihood of the line RL among the plurality of line segment candidates connecting the plurality of candidate points selected as described above and the detected upper end of the line RL. Candidates are determined based on the feature amount, and one determined line segment candidate is detected as a part of the middle portion of the railroad RL. For example, in the example shown in FIG. 5, the thick line extending between the upper end P1a of the initial portion and the candidate point P2a at the upper end of the search area 501a and the upper end P1b of the initial portion and the candidate point P2b at the upper end of the search area 501b A thick line extending in between is determined as a line segment candidate having the highest likelihood of being on the railroad RL, and is detected as an intermediate portion.

Then, the secondary detection unit 123b repeats the same process with the upper end of the midway portion detected above as the next starting point. By repeatedly executing the detection of a part (next middle part) of the railroad RL based on a plurality of candidate points selected from the search area, until the upper end of the detected railroad RL reaches a predetermined position, the middle part is repeatedly detected. In FIG. 5, illustration of the search areas after the search areas 501a and 501b is omitted for the sake of simplification.

Here, in the embodiment, the point with the highest likelihood of the railroad RL is, for example, the feature amount of a part of the railroad RL that has already been detected (initial part in the example shown in FIG. 5) and the search area (shown in FIG. In the example, it is identified based on at least one of (such as the average value of) the feature amount of 501a and 501b). According to such a configuration, considering at least one of the degree of similarity with the detected line RL and the line RL-likeness in the search area, it is possible to appropriately determine the point with the highest line-likeness. can be done.

In addition, in the embodiment, the search areas (501a and 501b in the example shown in FIG. 5) are captured images (in the example shown in FIG. In the example shown, it is determined based on position and/or angle on 300). According to such a configuration, considering the continuity with the most recently detected railroad track RL, the area that is highly likely to contain the candidate point of the next railroad track RL is accurately determined as the next search area. can do.

In addition, in the embodiment, the search areas (501a and 501b in the example shown in FIG. 5) have the left-right direction of the captured image (300 in the example shown in FIG. 5) as the longitudinal direction. In the example, it is set in a rectangular shape with the vertical direction of 300) as the lateral direction. According to such a configuration, even when the track RL is curved, it is possible to set a search area having a shape that easily includes the entire track RL and that is easy to handle in calculation.

Note that the detection by the secondary detection unit 123b is performed until the detected upper end of the railroad RL reaches the height of the vanishing point of the railroad RL in the captured image (300 in the example shown in FIG. 5), that is, the entire railroad RL. may be repeated until the detection of is completed, but in the embodiment, as an example, at the timing when the detected upper end of the line RL reaches a predetermined position, the detection by the secondary detection unit 123b is started. It is assumed that the processing is switched to detection by the interpolation detection unit 123c as described above.

FIG. 6 is an exemplary and schematic diagram for explaining the interpolation detection unit 123c according to the embodiment. In the example shown in FIG. 6, the line RL up to the height of the upper ends PNa and PNb has already been detected by the secondary detection unit 124b, and the line RL after the height of the upper ends PNa and PNb is detected by the interpolation detection unit 123c. It represents the situation of being detected.

As shown in FIG. 6, the interpolation detection unit 123c first extracts points (vanishing point candidates) as vanishing point candidates of the railroad RL from a vanishing point area 600 set on the upper end side of an image 300 as a captured image. P600 to P602 are identified based on the feature amount of the image 300 (of the vanishing point area 600).

Then, the interpolation detection unit 123c acquires a plurality of interpolation curves that interpolate a region between the plurality of points P600 to P602 specified as described above and the detected upper ends PNa and PNb of the railroad RL. In the example shown in FIG. 6, curves La1 and La2 as interpolation curves based on point P601 as a vanishing point candidate, and curves La2 and Lb2 as interpolation curves based on point P602 as a vanishing point candidate. exemplified. In the example shown in FIG. 6, it goes without saying that there can be vanishing point candidates other than the points P601 and P602, and interpolation curves other than the curves La1, Lb1, La2, and Lb2.

Then, the interpolation detection unit 123c selects one interpolation curve (an interpolation curve based on the point P600, not shown in FIG. 6) having the highest likelihood of being on the railway line among the plurality of interpolation curves acquired as described above. to detect the undetected portion of the line RL. According to such a configuration, a portion of the line RL that has not been detected by the initial detection section 123a and the secondary detection section 123b can be detected at higher speed and with higher accuracy based on the interpolation curve.

Here, in the embodiment, in order to further speed up the calculation, the interpolation detection unit 123c does not detect the interpolation curve as it is as (a part of) the line RL, but the interpolation curve is easy to handle in calculation. A plurality of line segments can be approximated and the plurality of line segments can be detected as (part of) the line RL.

FIG. 7 is an exemplary and schematic diagram for explaining an example of interpolation curve approximation that can be performed by the interpolation detection unit 123c according to the embodiment, and FIG. 8 is an approximated interpolation curve according to the embodiment. 1 is an exemplary and schematic diagram showing an example of . In the examples shown in FIGS. 7 and 8, La and Lb are bisected at points Pxa and Pxb of a given height (reference Lx). In the examples shown in FIGS. 7 and 8, the undetected line RL located above the upper ends PNa and PNb of the detected line RL is the line segment connecting the points Pxa and Pxb and the upper ends PNa and PNb. , and line segments connecting the points Pxa and Pxb and the ends of the rails 301a and 301b forming the railroad track RL.

7 and 8 show an example in which a polygonal line approximated by halving the interpolation curve is detected as the undetected line RL. A polygonal line that divides and approximates may be detected as the undetected line RL. 7 and 8 show an example in which a polygonal line is formed by a plurality of line segments approximating an interpolation curve. You may have to line up.

7 and 8, the point P701 corresponding to the vanishing point of the line RL and the upper ends PNa and PNb of the middle portion of the line RL detected by the secondary detection unit 123b are interpolated by interpolation curves. Examples are given. However, the embodiment does not implement the detection of the middle portion by the secondary detection unit 123b, and the point P701, the upper ends P1a and P1b of the initial portion detected by the initial detection unit 123a (see FIGS. 5 and 6), may be interpolated using an interpolation curve.

Thus, the track detection unit 123 (the initial detection unit 123a, the secondary detection unit 123b, and the interpolation detection unit 123c) according to the embodiment detects the track RL from the captured image by predetermined image processing.

Returning to FIG. 2, the detection area acquisition unit 131 determines a detection area to detect (monitor) an obstacle based on the captured image obtained by the imaging device 110 and the detection result of the track identification device 120. get. The detection area is, for example, a preset area around the railroad track RL.

Then, the obstacle detection unit 132 monitors the detection area acquired by the detection area acquisition unit 131, and obtains information about the obstacle (for example, the presence or absence of the obstacle, the distance to the obstacle, etc.) from the detection area. detect.

Next, the control operation of the first embodiment will be explained.

FIG. 9 is an exemplary and schematic flow chart showing a series of processes executed by the vehicle system 100 including the track identification device 120 according to the embodiment. The processing flow shown in FIG. 9 is repeatedly executed at a predetermined control cycle while the railway vehicle RV is running, for example.

In the processing flow shown in FIG. 9 , first, in S901, the image acquisition unit 121 of the track identification device 120 acquires the imaged image obtained by the imaging device 110 . As described above, the captured image obtained here includes the railroad track RL.

Then, in S902, the feature quantity extraction unit 122 and the track detection unit 123 of the track identification device 120 execute the processing shown in FIG. To detect.

FIG. 10 is an exemplary and schematic flow chart showing details of the detection process of the railroad track RL according to the embodiment.

In the processing flow shown in FIG. 10 , first, in S1001, the feature amount extraction unit 122 extracts a feature amount as a scale representing the railroad RL-likeness from the captured image acquired in S901.

Then, in S1002, the initial detection unit 123a of the railroad detection unit 123 detects line segments in the initial area of the captured image acquired in S901 using the algorithm described above based on the feature amount extracted in S1001. It is detected as part of the RL (initial part) (see FIG. 4).

Then, in S1003, the secondary detection unit 123b of the line detection unit 123 uses the upper end of the detected line RL as a starting point, and selects the next search area set above the starting point as a candidate for the end point with respect to the starting point. A plurality of candidate points are selected based on the feature amount extracted in S1001.

Then, in S1004, the secondary detection unit 123b acquires a plurality of line segment candidates connecting the detected upper end of the track RL and the plurality of candidate points selected in S1003.

Then, in S1005, the secondary detection unit 123b determines one line segment candidate that most resembles the railroad RL from the plurality of line segment candidates identified in S1004 based on the feature amount extracted in S1001. One line segment candidate is detected as a part (middle part) of the line RL (see FIG. 5).

In S1006, the secondary detection unit 123b determines whether or not the detection of the railroad RL has been completed up to a predetermined position by the processing in S1005. ) is reached.

If it is determined in S1006 that the detection of the railroad RL has not been completed up to the predetermined position, the process returns to S1003. On the other hand, if it is determined in S1006 that the detection of the railroad RL has been completed up to the predetermined position, the process proceeds to S1007.

In S1007, the interpolation detection unit 123c of the railroad detection unit 123 selects a plurality of vanishing point candidates as vanishing point candidates of the railroad RL on the captured image from the vanishing point area set on the upper end side of the captured image. It specifies based on the extracted feature amount (see FIG. 6).

Then, in S1008, the interpolation detection unit 123c acquires a plurality of interpolation curves that interpolate the area between each of the plurality of candidate points identified in S1007 and the upper end of the detected railroad track (middle portion) ( See Figure 6).

Then, in S1009, the interpolation detection unit 123c determines one interpolation curve that most resembles the railroad RL among the plurality of interpolation curves acquired in S1008 based on the feature amount extracted in S1001, and determines the determined one interpolation curve. , the undetected line RL is detected. At this time, the interpolation detecting unit 123c approximates the interpolated curve having the highest likelihood of the line RL with a plurality of line segments, and detects a straight line or polygonal line formed by the plurality of line segments as an undetected line RL. Good (see Figures 7 and 8).

When the process of S1009 ends, the process returns to the process flow shown in FIG. That is, when the process of S1009 ends, the process of S902 shown in FIG. 9 ends, and the process proceeds to S903.

In S903, the detection area acquisition unit 131 of the obstacle detection device 130 acquires a detection area to detect an obstacle based on the detection result of S902. As described above, the detection area acquired in this S903 is, for example, a preset range area around all the railroad tracks RL on which the railway vehicle RV may run.

Then, in S904, the obstacle detection unit 132 of the obstacle detection device 130 monitors the detection area acquired in S903, and obtains information about the obstacle (for example, whether or not there is an obstacle and whether or not the obstacle exists) from the detection area. distance, etc.).

Then, in S905, the notification device 140 notifies the passenger (driver, etc.) of the railway vehicle RV of the detection result in S904. As described above, the notification method may be a method that appeals to vision using an image, a method that appeals to hearing using sound, or a method that appeals to touch using vibration or the like. It may be a method, or a method in which these are used in combination.

As described above, the track identification device 120 according to the embodiment includes the image acquisition unit 121 that acquires a captured image obtained by capturing an image of the area in the traveling direction of the railway vehicle RV, and the track RL-likeness from the captured image. and a line detection unit 123 for detecting the line RL from the captured image based on the feature amount. The line detection unit 123 includes an initial detection unit 123a, and a secondary detection unit 123b. The initial detection unit 123a identifies, based on the feature amount, a line segment that extends between the lower end and the upper end of an initial area set in advance on the lower end side of the captured image and has the highest likelihood of being a line RL. Detect as an initial part of a part of RL. The secondary detection unit 123b uses the upper end of the initial portion as a starting point, and connects a plurality of candidate points selected from within a search area set above the starting point and the upper end of the detected track RL. Among the line segment candidates, one line segment candidate that is most likely to be on the railway RL is determined based on the feature amount, the determined one line segment candidate is detected as a middle portion of a part of the railway RL, and the middle portion of the middle portion is detected. By repeatedly executing the process of setting the upper end to the next starting point, the midway portion is repeatedly detected until the upper end of the detected line RL reaches a predetermined position. According to such a configuration, the line RL is detected as a set of line segments that are easy to handle in calculation, compared to the conventional algorithm that detects the line RL as a set of parabolas, for example. High-speed and more accurate detection is possible.

In addition to the above-described initial detection unit 123a and secondary detection unit 123b, the track detection unit 123 according to the embodiment detects the disappearance of the track RL on the captured image from the vanishing point area set on the upper end side of the captured image. Identify multiple vanishing point candidates as point candidates based on the feature amount, and obtain multiple interpolation curves that interpolate the area between the identified multiple vanishing point candidates and the upper end of the detected middle part. and an interpolation detection unit 123c for detecting an undetected portion of the line RL based on the interpolation curve having the highest line RL-likeness among the acquired plurality of interpolation curves. According to such a configuration, by calculating a predetermined interpolation curve, it is possible to detect the railroad track RL from the captured image at a higher speed and with a higher degree of accuracy.

Note that the line detection unit 123 according to the embodiment may have a configuration including only the initial detection unit 123a and the interpolation detection unit 123c, without including the secondary detection unit 123b. In this case, the interpolation detection unit 123c identifies one point corresponding to the vanishing point of the railroad RL located on the upper end side of the captured image based on the feature amount, and detects a plurality of points having the same height as the one identified point. Obtaining a plurality of interpolation curves that interpolate at least part of an area between the candidate point and the upper end of the initial portion, and based on the interpolation curve having the highest railway RL-likeness among the obtained plurality of interpolation curves, At least part of the rest of the line RL other than the initial part is detected. With such a configuration as well, it is possible to detect the line RL from the captured image at a higher speed and with a higher degree of accuracy.

Although the embodiment of the present invention has been described above, the above embodiment is merely an example and is not intended to limit the scope of the invention. The above embodiments can be implemented in various forms, and various omissions, replacements, and modifications can be made without departing from the scope of the invention. The above embodiments and their modifications are included in the scope and gist of the invention, and are included in the scope of the invention described in the claims and equivalents thereof.

120 track identification device 121 image acquisition unit 122 feature amount extraction unit 123 track detection unit 123a initial detection unit 123b secondary detection unit 123c interpolation detection unit

Claims (6)

an image acquisition unit that acquires a captured image obtained by capturing an area in the traveling direction of the railway vehicle;
A feature quantity extracting unit that extracts a feature quantity representing railroad-likeness from the captured image;
a track detection unit that detects the track from the captured image based on the feature amount;
with
The line detection unit is
A line segment that extends between the lower end and the upper end of an initial area preset on the lower end side of the captured image and that has the highest likelihood of being a line is specified based on the feature amount, and the specified line segment is one of the lines. an initial detection unit that detects as an initial portion of the unit;
Using the upper end of the initial portion as a starting point, among a plurality of line segment candidates connecting a plurality of candidate points selected from within a search area set above the starting point and the detected upper end of the track, One line segment candidate having the highest rail-likeness is determined based on the feature amount, the determined one line segment candidate is detected as a midway portion of a portion of the railroad, and the upper end of the midway portion is detected as follows. a secondary detection unit that repeatedly detects the midway portion until the detected upper end of the track reaches a predetermined position by repeatedly executing the process of setting the line as a starting point;
a trajectory identifier. The secondary detection unit sets the next search area based on at least one of the most recently detected position and angle of the part of the railroad track on the captured image.
A track identification device according to claim 1. The secondary detection unit determines a point with the highest rail-likeness within the search area based on an average value of the feature values on the plurality of line segment candidates.
The track identification device according to claim 1 or 2. The search area is set in a rectangular shape with the horizontal direction of the captured image as the longitudinal direction and the vertical direction of the captured image as the lateral direction,
A track identification device according to any one of claims 1 to 3. The line detection unit is
A plurality of vanishing point candidates as vanishing point candidates of the railroad track on the captured image are specified based on the feature quantity from the vanishing point area set on the upper end side of the captured image, and the specified vanishing points are specified. Obtaining a plurality of interpolation curves that interpolate the area between the point candidate and the upper end of the detected railroad track, and based on the interpolation curve having the highest rail-likeness among the plurality of obtained interpolation curves, The track identification device according to any one of claims 1 to 4, further comprising an interpolation detection unit that detects an undetected portion of the track. The interpolation detection unit detects, as the undetected portions, a plurality of line segments that approximate the interpolation curve with the highest line-likeness.
A track identification device according to claim 5.

Images