JP2023023692A - Obstacle detection device - Google Patents

Abstract

To provide an obstacle detection device that stably can detect an obstacle to a railway vehicle which is travelling on a track, with high accuracy.SOLUTION: An obstacle detection device (1), provided in a railway vehicle (2) that runs on a track in which overhead wires which are electric wires are arranged at an upper side thereof, comprises: a plurality of cameras (3); a data processing part (11) that generates camera image data from videos taken with the plurality of cameras; a distance calculating part (12) that calculates a distance from the camera image data, and generates and outputs distance data; an electric wire recognizing part (13) that recognizes electric wires from the camera image data, and generates and outputs electric wire position data; and a separation measuring part (14) that measures a separation value between the railway vehicle and the electric wire, using the electric wire position data generated by the electric wire recognizing part and the distance data generated by the distance calculating part, and generates and outputs a separation value from the electric wire, where the data processing part (11) detects an obstacle on the basis of the separation value from the electric wire.SELECTED DRAWING: Figure 2

Description

TECHNICAL FIELD The present invention relates to an obstacle detection device provided in a railway vehicle.

2. Description of the Related Art For stable operation of railroad vehicles, development of devices for detecting obstacles in front of or behind railroad vehicles running on tracks (rails) is underway.
In particular, Patent Literature 1, which is an example of conventional technology, discloses a technique of measuring the distance to an obstacle using a predetermined construction gauge area.

In addition, in Non-Patent Document 1, which is an example of the conventional technology, a technique called graph cut, which uses a penalty term for optimization so that the estimated distance of a pixel of interest does not differ greatly from the estimated distances of the surrounding pixels, is described. is disclosed.
In addition, Patent Document 2, which is an example of the prior art, discloses an image processing technology that makes it possible to separate trolley wires from objects other than trolley wires in a line sensor image and reduce erroneous detection of trolley wire worn surfaces. disclosed.

V. Kolmogorov and R. Zabih, "Multi-camera scene reconstruction via graph cuts", ECCV (European Conference on Computer Vision), Vol3, 2002, pp. 82-96.

Japanese Patent No. 6524529 JP 2013-15336 A

However, in Patent Document 1, which is the conventional technology, a three-dimensional position can be calculated by calculating parallax between stereo cameras by block matching processing.
However, there is a problem that the overhead wire in the image in front of the running railway vehicle is thin, and it is sometimes difficult to perform accurate parallax calculation by block matching.

SUMMARY OF THE INVENTION The present invention has been made in view of the above, and an object of the present invention is to provide a technology capable of stably detecting obstacles on railroad vehicles running on tracks with high accuracy.

One aspect of the present invention that solves the above-described problems and achieves the object is an obstacle detection device provided in a railroad vehicle that runs on a track on which overhead wires that are electric wires are arranged, comprising: a plurality of cameras; A data processing unit that generates camera image data from images acquired by a plurality of cameras, and a distance that calculates a distance from the camera image data by local pixel matching and graph cutting, and generates and outputs distance data. a calculation unit, an electric wire recognition unit that recognizes electric wires from the camera image data, generates and outputs electric wire position data, electric wire position data generated by the electric wire recognition unit, and distance data generated by the distance calculation unit; a distance measuring unit that measures a distance value between the railway vehicle and the electric wire using , generates and outputs a distance value from the electric wire, and It is an obstacle detection device that detects an obstacle based on a value.

Alternatively, one aspect of the present invention that solves the above-described problems and achieves the object is an obstacle detection device provided in a railroad vehicle that runs on a track on which overhead wires that are electric wires are arranged, comprising: a plurality of cameras; a data processing unit for generating camera image data from images acquired by the plurality of cameras; a distance calculation unit for calculating a distance from the camera image data, generating and outputting distance data; and grayscale top hat processing. and a wire recognition unit that recognizes wires from the camera image data, generates and outputs wire position data, and the wire position data generated by the wire recognition unit and the distance data generated by the distance calculation unit, by combining processing. and a distance measuring unit that measures a distance value between the railway vehicle and the electric wire using and generates and outputs a distance value from the electric wire, and An obstacle detection device that detects an obstacle based on a distance value.

Alternatively, one aspect of the present invention that solves the above-described problems and achieves the object is an obstacle detection device provided in a railroad vehicle that runs on a track on which overhead wires that are electric wires are arranged, comprising: a plurality of cameras; a data processing unit for generating camera image data from images acquired by the plurality of cameras; a distance calculation unit for calculating a distance from the camera image data, generating and outputting distance data; Using a wire recognition unit that recognizes a wire and generates and outputs wire position data, the wire position data generated by the wire recognition unit, and the distance data generated by the distance calculation unit, the railway vehicle and a distance measuring unit that measures a distance value from the electric wire, generates and outputs a distance value from the electric wire, and the data processing unit detects an obstacle based on the distance value from the electric wire. It is an obstacle detection device.

ADVANTAGE OF THE INVENTION According to this invention, the obstruction of the rail vehicle which runs a track|orbit can be detected stably with high precision.

FIG. 1 is a top view showing a railway vehicle equipped with an obstacle detection device according to an embodiment. FIG. 2 is a functional block diagram showing the configuration of the obstacle detection device according to the embodiment. FIG. 3 is a flowchart showing an obstacle detection method performed by the obstacle detection device according to the embodiment. FIG. 4 is a diagram showing camera images from videos acquired by a plurality of cameras 3 in the embodiment. FIG. 5 is a diagram in which electric wire detection areas are set for the camera image shown in FIG. FIG. 6 is a diagram showing the result of negative/positive inversion processing performed on the electric wire detection area shown in FIG. FIG. 7 is a diagram showing the result of performing labeling processing on FIG. FIG. 8 is a diagram showing the result of performing the merging process on FIG.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments for carrying out the present invention will be described with reference to the accompanying drawings.
However, the present invention is not to be construed as limited by the description of the following embodiments.

<Embodiment 1>
FIG. 1 is a top view showing a railway vehicle 2 equipped with an obstacle detection device 1 according to this embodiment.
A railway vehicle 2 shown in FIG. 1 runs on a track 4 in the direction of travel.
A railway vehicle 2 shown in FIG. 1 includes a plurality of cameras 3 fixed to the head of the vehicle, and camera image data acquired by the plurality of cameras 3 is sent to the obstacle detection device 1 .
The obstacle detection device 1 monitors the front of the railroad vehicle 2 using the camera image data, and detects obstacles in front of the railroad vehicle 2 .
Here, trees are assumed as obstacles.

Note that the plurality of cameras 3 may be fixedly provided at the rear of the vehicle instead of at the front of the vehicle.
When a plurality of cameras 3 are provided at the rear of the vehicle, the obstacle detection device 1 monitors the rear of the railroad vehicle 2 using the camera image data acquired by the plurality of cameras 3, and monitors the rear of the railroad vehicle 2. Detect obstacles.

FIG. 2 is a functional block diagram showing the configuration of the obstacle detection device 1 according to this embodiment.
The obstacle detection device 1 shown in FIG. 2 includes a data processing section 11, a distance calculation section 12, an electric wire recognition section 13, and a distance measurement section .
The data processing unit 11 generates camera image data from images acquired by the plurality of cameras 3 .
The distance calculator 12 calculates a distance from the input camera image data, generates distance data, and outputs the generated distance data.
The electric wire recognition unit 13 recognizes electric wires from the input camera image data, generates and outputs electric wire position data.
Using the wire position data generated by the wire recognition unit 13 and the distance data generated by the distance calculation unit 12, the distance measurement unit 14 measures the distance value between the railway vehicle 2 and the wire, and calculates the distance from the wire. Generate and print a value.
The distance value from the electric wire is sent to the data processing unit 11, and the data processing unit 11 detects obstacles based on the distance value from the electric wire. Notify the person, etc.

FIG. 3 is a flowchart showing an obstacle detection method performed by the obstacle detection device 1 according to this embodiment.
As shown in FIG. 3, when the obstacle detection device 1 starts processing, the distance calculation unit 12 calculates the distance from the camera image data to generate distance data (distance calculation processing S101), and the electric wire recognition unit 13 detects the camera image data. A wire is recognized from the image data to generate wire position data (wire recognition processing S102). is generated (distance measurement processing S103), and the processing ends.

FIG. 4 is a diagram showing camera images from videos acquired by a plurality of cameras 3 in this embodiment.
When the obstacle detection apparatus 1 according to the present embodiment detects an obstacle from the camera images shown in FIG.

First, the distance calculator 12 calculates a distance as described below, generates distance data, and outputs the data.
If the relative positions and orientations of the multiple cameras 3 are known in advance, the distance from each of the multiple cameras 3 to the obstacle can be calculated by the principle of triangulation.
Note that when calculating the distance, it is necessary to specify which pixel in the image captured by the other camera corresponds to the pixel of interest in the image captured by one camera.
As such a method, a technique called block matching is widely used, in which an area similar to pixels surrounding a specific pixel in an image is searched from images of other cameras.
Block matching is effective when there is a characteristic texture around the pixel of interest in the image. It can be difficult.
In order to solve such a problem, there is a technique called graph cut, which uses a penalty term for optimization so that the estimated distance of a pixel of interest does not differ greatly from the estimated distances of surrounding pixels.
Such a technique is disclosed in Non-Patent Document 1, for example.
The distance calculation unit 12 can perform distance calculation with high accuracy by using the above-described block matching and graph cutting.

In this way, the distance calculation unit 12 performs block matching, which is local pixel matching, and graph cutting, so that distance calculation can be performed with high accuracy without performing edge detection, binarization processing, and cluster detection. can be done.

The electric wire recognition unit 13 recognizes electric wires as described below, recognizes electric wires in each of the plurality of cameras 3, and generates and outputs electric wire position data.

First, the wire recognition unit 13 sets the wire detection area for the camera image.
FIG. 5 is a diagram in which electric wire detection areas are set for the camera image shown in FIG.
In FIG. 5, the railcar 2 is running on the right railroad track in the camera image. ing.

Next, the electric wire recognition unit 13 performs negative-positive inversion processing on the luminance value of the electric wire detection area of the set camera image, and performs grayscale top-hat processing and binarization processing to make the area darker than the surrounding area (positive image). , and extract thin objects.
Such a process is disclosed, for example, in US Pat.
FIG. 6 is a diagram showing the result of negative/positive inversion processing performed on the electric wire detection area shown in FIG.
In FIG. 6, portions corresponding to the electric wires shown in FIG. 5 are indicated by white lines.

Next, the electric wire recognition unit 13 performs labeling processing.
FIG. 7 is a diagram showing the result of performing labeling processing on FIG.
In FIG. 7, six regions are each labeled.
The numbers in the circled numbers shown in FIG. 7 indicate the area numbers.
According to the labeling process, the extracted regions can be handled as a cluster by clustering the regions where the four neighboring regions of the attention region are connected with the same label.
Although the electric wire in the camera image is cut off due to noise or the like or structures such as curvilinear fittings or utility poles, the labeling process enables the process of combining a plurality of areas.

Next, the electric wire recognition unit 13 performs connection processing on the data that has been subjected to the labeling processing.
FIG. 8 is a diagram showing the result of performing the merging process on FIG.
8, the first and second regions shown in FIG. 7 are combined, the third and fifth regions shown in FIG. 7 are combined, and the fourth region shown in FIG. and the sixth region.
The process of combining multiple regions is performed twice using the angular difference and distance after linear approximation between labeling regions.
In the first combining process, the threshold values of the angular difference and the distance are set small in order to combine regions that have been cut off due to noise or the like.
In the second merging process, the angular difference and distance thresholds are set large in order to merge regions that have been discontinued due to structures such as curvilinear pulls or utility poles.
In the merging process, if there are multiple candidates whose angular difference and distance are within the thresholds, the candidate with the shortest distance between labeling regions is merged.
Data is interpolated by linear interpolation between a plurality of areas that are combined in the combining process.

Next, the electric wire recognition unit 13 associates electric wires detected between the camera images acquired by each of the plurality of cameras 3 .
The distance from the camera 3 is calculated according to the principle of triangulation for all combinations between a plurality of electric wires detected from each image.
Here, excluding combinations in which the average X coordinate and Y coordinate do not fall within the range of the electric wire existing region set in advance, and when there are a plurality of combinations for one electric wire, the u direction shown in FIG. A combination in which the order of is the same is adopted.
In this manner, the electric wire recognition unit 13 recognizes electric wires and generates electric wire position data.

In this way, the wire recognizing unit 13 can acquire the position of the wire through the grayscale top hat processing and the combining processing.

The distance measurement unit 14 uses the distance data (x, y, z) generated by the distance calculation unit 12 and the wire position data (x _trolley , y _trolley , z _trolley ) generated by the wire recognition unit 13 to calculate the following Calculate the distance value from the wire position by the formula (1).
Although the distance value dist in the xy plane is calculated in the following formula (1), the distance value dist is similarly calculated in the yz plane and the zx plane.
Also, when a plurality of electric wires are detected, the distance value from the outermost electric wire is calculated.

Then, the data processing unit 11 sets a preset management value as a threshold, and detects the existence of an obstacle when the distance value calculated by the above formula (1) is equal to or less than the threshold.

In this way, the distance measuring unit 14 can calculate the distance value from the electric wire, and can detect obstacles.

In addition, the present invention is not limited to the above-described embodiments, and includes various modifications in which components are added, deleted, or converted to the above-described configuration.

1 obstacle detection device 11 data processing unit 12 distance calculation unit 13 electric wire recognition unit 14 distance measurement unit 2 railway vehicle 3 camera 4 track

Claims (3)

An obstacle detection device provided in a railroad vehicle running on a track on which an overhead wire, which is an electric wire, is arranged above,
multiple cameras and
a data processing unit that generates camera image data from images acquired by the plurality of cameras;
a distance calculation unit that calculates a distance from the camera image data by local pixel matching and graph cutting, and generates and outputs distance data;
an electric wire recognition unit that recognizes an electric wire from the camera image data and generates and outputs electric wire position data;
Using the wire position data generated by the wire recognition unit and the distance data generated by the distance calculation unit, the distance value between the railway vehicle and the wire is measured, and the distance value from the wire is generated. and a distance measuring unit that outputs the
The obstacle detection device, wherein the data processing unit detects an obstacle based on a distance value from the electric wire. An obstacle detection device provided in a railroad vehicle running on a track on which an overhead wire, which is an electric wire, is arranged above,
multiple cameras and
a data processing unit that generates camera image data from images acquired by the plurality of cameras;
a distance calculation unit that calculates a distance from the camera image data and generates and outputs distance data;
a wire recognition unit that recognizes wires from the camera image data by grayscale top hat processing and combination processing, and generates and outputs wire position data;
Using the wire position data generated by the wire recognition unit and the distance data generated by the distance calculation unit, the distance value between the railway vehicle and the wire is measured, and the distance value from the wire is generated. and a distance measuring unit that outputs the
The obstacle detection device, wherein the data processing unit detects an obstacle based on a distance value from the electric wire. An obstacle detection device provided in a railroad vehicle running on a track on which an overhead wire, which is an electric wire, is arranged above,
multiple cameras and
a data processing unit that generates camera image data from images acquired by the plurality of cameras;
a distance calculation unit that calculates a distance from the camera image data and generates and outputs distance data;
an electric wire recognition unit that recognizes an electric wire from the camera image data and generates and outputs electric wire position data;
Using the wire position data generated by the wire recognition unit and the distance data generated by the distance calculation unit, the distance value between the railway vehicle and the wire is measured, and the distance value from the wire is generated. and a distance measuring unit that outputs the
The obstacle detection device, wherein the data processing unit detects an obstacle based on a distance value from the electric wire.

Images