JP2012015697A - Pantagraph detection device and method thereof - Google Patents

Abstract

PROBLEM TO BE SOLVED: To perform calculation to detect a pantagraph on the roof of a passing train at a high speed while keeping accuracy.SOLUTION: In a pantagraph detection method for arranging a pantagraph detection sensor 40 and a monitoring camera 30 in the upper part of a track 2 where a train 1 passes, photographing the pantagraph 1a by the monitoring camera 30 when the pantagraph detection sensor 40 detects the pantagraph 1a, performing image processing for input images which are photographed by the monitoring camera 30, and then, keeping only images showing the pantagraph 1a, an image processor 50 performs orientation code matching with respect to the input images, so as to obtain robustness in brightness change due to sunshine change, etc., and also to stably determine presence determination of the pantagraph in the images.

Description

  The present invention relates to a pantograph detection apparatus and method. More specifically, an apparatus and method for detecting a pantograph of a passing train from an image taken by a monitoring camera installed above a track enables detection of even a dark image with high accuracy.

As a device for photographing the state of a passing train on the roof, a device is generally used in which a camera and illumination are installed above the track, a recording device is installed indoors, and recording is performed in conjunction with a vehicle approach alarm device. (Non-Patent Document 1).
Based on the train approach signal generated by the vehicle approach warning device when the train approaches, this device turns on the lighting and starts recording, and records an image during the period when the train approach signal is generated. The photograph is taken on the roof of the train passing by.

Several proposals have been made for a vehicle approach warning device that generates a train approach signal that serves as a reference for the start and end of imaging.
In Patent Document 1, a vibration sensor is directly attached to a rail using a magnet, and a train approach signal is generated by detecting vibration when a railway vehicle approaches.

In Patent Document 2, a train current detection sensor is attached to a rail via a magnet, the output of the sensor is connected to an amplifier built in the arithmetic processing unit, and an alarm is sounded based on the result processed by the arithmetic processing unit.
Patent Document 3 includes a transmission unit and a reception unit, transmits an exploration signal to the rail by the transmission unit, and receives a reflected signal reflected by the approaching track vehicle by the reception unit, thereby detecting the approach of the train. .

On the other hand, the following proposal was performed as a device which photographs the state of the passing train on the roof (Patent Document 4).
In other words, by detecting the train that entered the image of the surveillance camera by image processing and recording the image during the period when the train exists in the image, the image on the roof of the passing train is efficiently captured and stored. did.

In addition, by detecting the train that entered the image of the surveillance camera by object shape inspection applying image processing, and recording the image of the period when the train exists in the image, the image on the roof of the passing train Were taken and stored efficiently.
Furthermore, by detecting the train by the approach detection process using the technology of the intruder detection device, and recording the image of the period when the train is present in the image, it is possible to efficiently capture the image on the roof of the passing train And stored.

Japanese Patent Laid-Open No. 2001-39304 “Vehicle Approach Alarm Device and Vehicle Failure Diagnosis Device Utilizing Data Output from Vibration Data Collection Device Generated when Railway Car Runs” JP 2002-337687 “Vehicle approach warning device” Japanese Patent Laid-Open No. 9-132141 "Railway Vehicle Approach Warning Device" Japanese Patent Application Laid-Open No. 2010-12084 “Passing train photographing device by image processing”

F.ULLAH, S.KANEKO and S. IGARASHI, "Orientation Code Matching for Robust Object Search", IEICE Trans.lnf. & Syst., Vol.E84-D, No. & pp.999-1006, 2001

The vehicle approach warning devices disclosed in Patent Documents 1, 2, and 3 detect an approaching train at an early stage. For this reason, in the passing train imaging device that starts and ends the recording based on the train approach signal generated by the vehicle approach warning device, there are many imaging times other than the time during which the train actually passes through the field of view of the camera. Exists.
In this case, a lot of image storage area of the recording device is used for unnecessary image data that has not been photographed for the train, and an image showing a pantograph that is an important monitoring target as a rooftop device is searched. It takes time and effort to do so.

Patent Document 4 is an apparatus that detects an approach of a train by image processing and captures an image on a passing train roof. Since these devices record images during the period when the train is present in the images, many of the recording devices are compared to the passing train imaging device based on the vehicle approach warning device of Patent Documents 1, 2, and 3. The image storage area can be greatly saved.
However, in order to search for an image in which a pantograph that is an important monitoring target as a rooftop apparatus is shown, time and labor are required as in Patent Documents 1, 2, and 3.

The pantograph detection device according to claim 1 of the present invention that solves the above-described problem is that a surveillance camera is installed above a track through which a train passes, and the lower part is photographed by the surveillance camera, and the photographed input image is processed. In the pantograph detection device that stores only an image in which a pantograph appears in the input image, direction code matching is performed on the input image, and when the degree of matching is higher than a preset direction code search threshold, And an image processing device for determining that a pantograph is shown.
The pantograph detection device according to claim 2 of the present invention that solves the above problem is the pantograph detection device according to claim 1, wherein the image processing device varies the direction code search threshold according to the brightness of the input image. It is characterized by.

The pantograph detection device according to claim 3 of the present invention that solves the above-described problems is the pantograph detection device according to claim 1 or 2, wherein the image processing device performs pantograph detection by the direction code verification in the input image. As a search range, an edge histogram is created using the shape feature of the pantograph, and the edge histogram is set to a range equal to or greater than a preset threshold value.
The pantograph detection device according to a fourth aspect of the present invention for solving the above-mentioned problems is the pantograph detection device according to the third aspect, wherein the image processing device performs preprocessing of the edge histogram on the input image. Image quality correction is performed.

The pantograph detection device according to claim 5 of the present invention that solves the above problem is the pantograph detection device according to claim 4, wherein the image processing device performs contrast enhancement by linear density conversion of a density distribution as the image quality correction. It is characterized by that.
The pantograph detection device according to claim 6 of the present invention for solving the above problem is the pantograph detection device according to claim 5, wherein the image processing device includes a constant region at the center of the image including the pantograph when performing the contrast enhancement. The density value histogram is used.

The pantograph detection device according to claim 7 of the present invention for solving the above-mentioned problems is the pantograph detection device according to claim 1, wherein a pantograph detection sensor is also installed above the track, and the pantograph detection sensor is a pantograph on the train. The pantograph is photographed by the monitoring camera when the camera is detected.
The pantograph detection method according to claim 8 of the present invention that solves the above-described problem is that a surveillance camera is installed above a track through which a train passes, the lower part is photographed by the surveillance camera, and the photographed input image is image-processed. In the pantograph detection method for storing only an image in which a pantograph is shown in the input image, the image processing performs direction code matching on the input image, and has a higher fitness than a preset direction code search threshold. Sometimes it is determined that a pantograph is shown in the input image.

The pantograph detection method according to claim 9 of the present invention that solves the above-described problem is the pantograph detection method according to claim 8, wherein the direction code collation changes the direction code search threshold according to the brightness of the input image. It is characterized by.
The pantograph detection method according to claim 10 of the present invention that solves the above problem is the pantograph detection method according to claim 8 or 9, wherein the search range for performing pantograph detection by the direction code verification in the input image is the pantograph. An edge histogram is created using the shape feature, and the edge histogram is set in a range equal to or larger than a preset threshold value.

A pantograph detection method according to an eleventh aspect of the present invention for solving the above-described problem is the pantograph detection method according to the tenth aspect, wherein image quality correction is performed on the input image as a pre-process for creating the edge histogram. Features.
A pantograph detection method according to a twelfth aspect of the present invention that solves the above problem is characterized in that, in the pantograph detection method according to the eleventh aspect, contrast enhancement is performed by linear density conversion of a density distribution as the image quality correction.

A pantograph detection method according to a thirteenth aspect of the present invention for solving the above-described problem is the pantograph detection method according to the twelfth aspect, wherein when performing the contrast enhancement, a density value histogram in a fixed region in the center of the image including the pantograph is used. It is characterized by using.
The pantograph detection method according to claim 14 of the present invention that solves the above problem is the pantograph detection method according to claim 8, wherein a pantograph detection sensor is also installed above the track, and the pantograph detection sensor is a pantograph on the train. The pantograph is photographed by the monitoring camera when the camera is detected.

  According to the inventions according to claims 1 and 8, since the direction code collation is performed in order to store only the image in which the pantograph which is an important monitoring target as the rooftop device is reflected, the brightness fluctuation due to the change in sunlight or the like. Therefore, the presence / absence of the pantograph in the image can be determined more stably.

  According to the second and ninth aspects of the invention, since the direction code search threshold is varied depending on the brightness of the input image in the direction code matching, the pantograph can be detected with high accuracy even in a dark image.

  According to the third and tenth aspects of the invention, since the edge histogram is created using the shape characteristics of the pantograph, the search range for detecting the pantograph can be limited to a range where the pantograph may exist. There are few false detections and high speed pantograph detection can be performed.

  According to the inventions according to claims 4 and 11, image quality correction (noise removal by median filter, contrast enhancement by linear density conversion of density distribution, etc.) is performed on the input image as preprocessing for creating the edge histogram. In addition, pantograph detection with few false detections can be performed even in a dark image.

  According to the fifth and twelfth aspects of the present invention, the contrast enhancement is performed by linear density conversion of the density distribution as the image quality correction, so that a clear edge can be extracted even for a dark image with a small density difference.

  According to the inventions according to claims 6 and 13, when performing contrast enhancement by linear density conversion of density distribution, the density value histogram in a certain region in the center of the image including the pantograph is used, thereby affecting the background portion of the image edge. In contrast, contrast enhancement can be performed.

  According to the inventions according to claims 7 and 14, since the pantograph is photographed by the surveillance camera when the pantograph detection sensor detects the pantograph, most of the images in which the pantograph is shown are processed until all the images are processed. The time is early. Therefore, the image storage area of the storage device can be used efficiently.

It is the schematic which shows the apparatus structure which concerns on Example 1 of this invention. It is a graph which shows transition of a bright input image and pantograph detection collation in the image. It is a graph which shows transition of a pantograph detection collation in a dark input image and the same image. It is a graph which shows linear density | concentration conversion. It is a graph which shows determination of the density range before conversion in linear density conversion. It is explanatory drawing of a median filter. It is explanatory drawing of an input image and a horizontal edge image. It is explanatory drawing of a horizontal edge image and an edge histogram. It is explanatory drawing of a bright image, a horizontal edge image, and an edge histogram. It is explanatory drawing of the horizontal edge image and edge histogram of a dark image. It is a flowchart which shows the flow of a pantograph process detection. It is the schematic which shows the apparatus structure which concerns on Example 2 of this invention.

  The basic concept which is the premise of the present invention is as follows.

First, when the pantograph detection sensor detects the pantograph, the pantograph is photographed by the surveillance camera and the image is recorded.
However, the pantograph detection sensor may react to a vehicle roof structure other than the pantograph or some flying object, and erroneously capture an image without the pantograph.

Therefore, whether or not a pantograph exists for the recorded image is determined by image processing, and only the image showing the pantograph is stored.
Whether or not a pantograph exists is determined by detecting a pantograph from an image by orientation code matching (OCM: Orientation Code Matching) of Non-Patent Document 1 to perform robust matching against brightness change and hiding. it can.

Direction code collation is a method of inspecting the degree of matching between image data by coding the strength direction of each image portion and collating the codes of the reference image and the input image.
Therefore, since only the image in which the pantograph is shown can be stored, the image storage area of the storage device can be used more efficiently than in Patent Document 4, and the time and labor for searching for the image in which the pantograph is shown can be saved. It can be omitted.

However, if the image to be shot is dark overall, such as when shooting at night, the difference in shading is small and the noise increases, so an accurate direction code can be obtained when coding the intensity direction of the shading of the image. No more pixels.
Therefore, there is a problem that even if the direction code collation robust to the brightness fluctuation is performed, the degree of matching with the pantograph reference image is reduced, and the pantograph cannot be detected.

In addition, when a vehicle roof structure that is similar to the pantograph in the intensity direction of light and shade is reflected, there is a problem that many cases are erroneously detected that the pantograph is reflected by direction code verification.
Hereinafter, an embodiment for solving such a problem will be described with reference to the drawings.

(Method using pantograph detection sensor)

  In this embodiment, as shown in FIG. 1, a pantograph detection sensor 40 composed of a distance sensor or the like, a monitoring camera 30, and a lighting device 20 are installed above a track 2 through which a train 1 passes. When the pantograph detection sensor 40 detects the pantograph 1a on the train 1, the lower part is photographed by the monitoring camera 30 and the image of the pantograph 1a is recorded. The input image is stored in an image storage area of a storage device (not shown).

However, there are cases where the pantograph detection sensor 40 reacts to a vehicle roof structure other than the pantograph 1a or some flying object and takes an image without the pantograph 1a.
Therefore, the image processing device 50 determines whether or not the pantograph 1a exists for the recorded image, and stores only the image showing the pantograph 1a.
The image processing device 50 detects the pantograph 1a from the input image by direction code verification.

Direction code verification is a method of inspecting the degree of matching between image data by coding the intensity direction of each part of the image and comparing the codes of the reference image and the input image. This is a robust verification method.
That is, a directional code image (referred to as a “reference code image”) obtained by encoding the pantograph reference image with the intensity of the image portion is created, and further, the pantograph verification including the position of the pantograph part and the range of the pantograph part in the pantograph reference image Create data.

  Next, a direction code image (referred to as “input code image”) of the input image is created, and in the preset search range (referred to as “pantograph search range”) A, the direction code comparison is performed in the pantograph reference image. The image data of the pantograph portion are compared with each other, and a pantograph detection position and a direction code collation value (referred to as “pantograph detection collation value”) at the pantograph detection position are obtained. That is, the pantograph search range A is a range in which pantograph detection is performed by direction code matching in the input image.

The larger the pantograph detection collation value, the higher the degree of matching with the pantograph reference image.
Finally, the pantograph detection collation value is compared with a preset threshold value Th _cm (referred to as “direction code search threshold value”), and if the degree of matching is high, it is determined that the pantograph 1a is present in the inspected image.

In this embodiment, in the case of a dark image, the direction code search threshold value Th _ocm used for the determination of the fitness is set to the problem that the fitness with the pantograph reference image decreases and the pantograph 1a cannot be detected according to the basic concept. The problem is solved by varying the brightness of the input image.
This is because the suitability of dark images generally decreases compared to the transition of the suitability of bright images.

2A and 3A show a bright input image and a dark input image taken by the surveillance camera 30, respectively. FIGS. 2B and 3B show a bright input image, respectively. The graph which projected the pantograph detection collation value to the Y coordinate when searching in the pantograph search range A in a dark input image is shown.
As shown in FIGS. 2A, 2B and 3B, 3B, the Y coordinate where the pantograph 1a is present has a large pantograph detection collation value, which is a sharp mountain, and is suitable for dark images. The degree decreases and the mountain becomes smaller overall. The Y coordinate is a coordinate along the traveling direction of the train 1, that is, a ordinate.

Equation (1) shows an equation for _{obtaining the} direction code search threshold Th _ocm .
Here, mode _ocm is the mode value (= value with the highest appearance frequency) of the fitness in the pantograph search range.
In the portion where the pantograph 1a does not exist, since the pantograph detection collation value is small and the appearance frequency is high within the pantograph search range A, it can be considered that the mode value mode _ocm represents the fitness of the normal portion where the pantograph 1a does not exist.

W _ocm is a weighting constant that determines the adjustment amount of the direction code search threshold, and ave is the average luminance value of the input image.
That is, as shown in FIG. 2 (a) (b), as the image is bright (= higher ave is large) direction code search threshold Th _ocm is set largely apart value from the normal value, on the contrary, FIG. 3 ( as shown in a) (b), as the image is dark (= higher ave is small) direction code search threshold Th _ocm is set to a value close to the normal value.

In this way, by changing the direction code search threshold Th _{ocm according} to the brightness of the image, the pantograph detection collation value of the pantograph portion becomes larger than the direction code search threshold Th _ocm regardless of the brightness of the image, and even a dark image has a pantograph 1a can be detected.
Th _ocm = mode _ocm + avexW _ocm (1)

In the present embodiment, the pantograph search range A at the time of direction code matching is automatically limited to the area where the pantograph 1a is supposed to exist as follows. The problem that a case where an object other than the pantograph such as the object 1b is erroneously detected is seen is prevented.
That is, the pantograph portion has a shape characteristic that horizontally long objects are concentrated (= there are many horizontal lines) and the boundary is clear (= the edge strength is large).

It can be considered that the pantograph search range A can be automatically set by using such a shape feature of the pantograph 1a. However, in a dark image, (1) there is little contrast and (2) there is a lot of noise, so there is a clear edge. Cannot be extracted, and the shape feature of the pantograph 1a cannot be captured.
Therefore, a clear edge can be extracted by performing the following preprocessing (image quality correction) on the input image.

First, (1) is solved by performing contrast enhancement by linear density conversion of the density distribution.
The linear density conversion is an operation of converting density values so that the relationship between the density values before conversion and the density values after conversion is linear. As shown in FIG. What has been distributed in the density range is expanded to a density range of 0 to 255 after conversion.

For this reason, when linear density conversion is performed on a dark image, the density difference between adjacent pixels becomes larger and the edge strength increases.
Note that the density ranges a to b before conversion are determined by obtaining the minimum density value and the maximum density value for the input image.

At this time, in order to prevent the density ranges a to b from being set wider than the appearance due to the influence of a very small number of density values such as the lighting device 20 in a dark image, the degree of improvement in contrast is weakened. As shown in FIG. 5, the density value in which the number of pixels in the density value histogram does not satisfy a certain ratio of the total number of pixels is ignored.
Further, in order to prevent the density value of the background portion at the edge of the image from affecting the density value histogram, the density value histogram is created by limiting the range to a certain area in the center of the image where the pantograph 1a is displayed.

Next, (2) is solved by removing noise by a median filter.
The median filter is a process of replacing the median value with the target pixel when the density values in the local region of nxn around the target pixel are sequentially sorted as shown in FIG.
As a result, noise can be removed without blurring the edges.

For the automatic setting of the pantograph search range A, an image obtained by extracting only the horizontal line pixels having the gradient direction of 90 ° or 270 ° as shown in FIG. Called edge image).
First, as shown in FIG. 8A, a region B (referred to as a “horizontal edge count region”) B whose horizontal width = image width and vertical width = pantograph 1a is scanned in the vertical direction with respect to the horizontal edge image. I will do it.

In the pantograph portion, a large number of horizontal edge pixels having high edge strength exist in the horizontal edge count region B.
On the other hand, the structure on the vehicle roof is often not one of the horizontally long, the horizontally long object is not concentrated in one place, or has many similar colors. Does not exist so much or edge strength is small.

Therefore, for the evaluation, an edge histogram is used in which the average value of the horizontal edge intensity in the area is projected for each Y coordinate at the center of the horizontal edge count area B as shown in FIG.
In the region where the pantograph 1a is present, the number of horizontal edge pixels is large and the horizontal edge strength is also high, so the edge strength average value is large.

Therefore, when the edge intensity average value of the edge histogram is a Y coordinate equal to or _larger than a preset threshold Th _{hist, the} pantograph 1a is considered to exist and is set to the pantograph search range A.
At this time, the edge intensity average value of the dark image shown in FIG. 10 is generally smaller than that of the bright image shown in FIG.

Therefore, as in the case of the fitness, the threshold value Th _hist is obtained by changing the threshold Th _{hist according to} the brightness of the input image according to the equation (2).
Here, mode _hist is the mode value of the edge intensity average value in the edge histogram, W _hist is a weight constant that determines the adjustment amount of the threshold Th _hist , and ave is the average luminance value of the input image.
As in the case of Expression (1), the threshold Th _hist is set closer to the normal value as the image becomes darker.

As described above, by automatically setting the pantograph search range A at the time of direction code matching only to the area where the pantograph 1a is supposed to exist, most areas of the vehicle roof structure that may be erroneously detected. Deviates from the pantograph search range A in which direction code matching is performed.
Accordingly, false detection is reduced, accuracy is improved, and at the same time, the number of collations during direction code collation can be reduced, so that collation processing can be speeded up.
Th _hist = mode _hist + ave × W _hist (2)

About the above image processing, the image processing apparatus 50 performs as follows according to the flowchart shown in FIG.
(1) Pantograph search range A is obtained by the following processes (a) to (force). (A) Linear density conversion is performed on the input image, contrast enhancement is performed (step S1),
(A) The image created in (A) is subjected to noise removal by applying a median filter (step S2),
(C) Edge detection is performed on the image created in (a) (step S3).

(D) A horizontal edge image is created by leaving only the horizontal line pixels whose gradient direction is 90 degrees or around 270 degrees with respect to the image created in (c) (step S4).
(E) The horizontal edge count area B is scanned in the vertical direction with respect to the image created in (d), and the average value of the horizontal edge intensity in the area is obtained. The obtained average value is projected for each Y coordinate at the center of the horizontal edge count region B, and an edge histogram is created (step S5).
(F) The pantograph search range A is obtained using the edge histogram created in (e) and the threshold value calculated from the brightness of the input image (step S6).

(2) A direction code image of the input image is obtained, and an input code image is created (step S7).
(3) Direction code verification is performed using the pantograph search range A obtained in (1) and (2), the input code image, and the prepared reference code image (step S8).
(4) The pantograph detection collation value obtained in (3) is compared with the threshold value calculated based on the input image brightness to determine the fitness, that is, the presence or absence of the pantograph 1a (step S9).

(Method without using pantograph detection sensor)
The configuration of this embodiment is shown in FIG.
In this embodiment, the input image acquisition method is different from that of the first embodiment.

In the first embodiment, the pantograph detection sensor 40 is installed above the track 2 through which the train 1 passes, and the input image is acquired when the sensor 40 reacts. However, in this embodiment, the pantograph detection sensor 40 is not used. Similarly to Patent Document 4, an image taken by the monitoring camera 30 is taken as an input image.
That is, in this embodiment, as shown in FIG. 12, the monitoring camera 30 and the lighting device 20 are installed above the track 2 through which the train 1 passes, the lower part is photographed by the monitoring camera 30, and the image is recorded. .

For the input image photographed by the monitoring camera 30, the pantograph 1a is detected from the image by direction code matching as in the first embodiment. The specific contents are as shown in FIG.
In the present embodiment, since the pantograph detection sensor 40 can be omitted, a simple apparatus configuration can be achieved at a lower cost than in the first embodiment.

In the present embodiment, since the surveillance camera 30 always captures the lower part, many images of the train 1 on which the pantograph 1a is not reflected are also captured.
However, since the pantograph detection can be performed on all the captured images by using the pantograph detection method described in the first embodiment, only the image showing the pantograph 1a can be finally stored.

  The timing for detecting the pantograph 1a may be the following two patterns. In either case, only an image showing the pantograph 1a can be finally stored, so an image showing the pantograph can be stored. Manual search can be saved.

・ Pantograph detection is performed on images immediately after shooting. In the case of an image showing the pantograph 1a, it is stored in the image storage area of the recording device.
・ Pantograph detection is performed after all the captured images are stored in the image storage area of the recording device. If the image does not show the pantograph 1a, the image is deleted from the image storage area.

  However, there are more images to be detected in the present embodiment than in the first embodiment (in the first embodiment, most images of the train 1 in which the pantograph 1a is shown, but in this embodiment, the train 1 that passes through Since all the images are processed), the time until all the images are processed is longer than that in the first embodiment.

  INDUSTRIAL APPLICABILITY The present invention can be widely used industrially as a technique for detecting a pantograph of a passing train from an image taken by a surveillance camera installed above the track.

DESCRIPTION OF SYMBOLS 1 Train 1a Pantograph 2 Track 20 Illuminating device 30 Surveillance camera 40 Pantograph detection sensor 50 Image processing apparatus A Pantograph search range B Horizontal edge count area

Claims (14)

  A pantograph detection device that stores only an image in which a pantograph is shown in the input image by installing a monitoring camera above the track through which the train passes, photographing the lower side with the monitoring camera, and processing the captured input image The image processing apparatus includes: an image processing device that performs direction code matching on the input image and determines that a pantograph appears in the input image when the fitness is higher than a preset direction code search threshold. A pantograph detection device.   The pantograph detection device according to claim 1, wherein the image processing device varies the direction code search threshold according to the brightness of the input image.   The pantograph detection device according to claim 1 or 2, wherein the image processing device creates an edge histogram using a shape characteristic of the pantograph as a search range for performing pantograph detection by the direction code matching in the input image, The pantograph detection device, wherein the edge histogram is set in a range equal to or greater than a preset threshold value.   4. The pantograph detection apparatus according to claim 3, wherein the image processing apparatus performs image quality correction on the input image as a pre-process for creating the edge histogram.   5. The pantograph detection device according to claim 4, wherein the image processing device performs contrast enhancement by linear density conversion of a density distribution as the image quality correction.   6. The pantograph detection device according to claim 5, wherein when performing the contrast enhancement, the image processing device uses a density value histogram in a fixed region in the center of the image including the pantograph.   The pantograph detection device according to claim 1, wherein a pantograph detection sensor is also installed above the track, and the pantograph is photographed by the monitoring camera when the pantograph detection sensor detects a pantograph on the train. Pantograph detection device.   A pantograph detection method in which a surveillance camera is installed above a track through which a train passes, the lower part is photographed by the surveillance camera, and only the image in which the pantograph is shown in the input image is stored by performing image processing on the photographed input image The image processing is characterized in that the input image is subjected to direction code collation, and it is determined that a pantograph is shown in the input image when the fitness is higher than a preset direction code search threshold. Pantograph detection method to do.   9. The pantograph detection method according to claim 8, wherein the direction code collation changes the direction code search threshold according to the brightness of the input image.   10. The pantograph detection method according to claim 8 or 9, wherein an edge histogram is created using a shape characteristic of the pantograph for a search range in which pantograph detection is performed by directional code matching in the input image, and the edge histogram is set in advance. A pantograph detection method, characterized by being set in a range equal to or greater than the threshold value.   11. The pantograph detection method according to claim 10, wherein image quality correction is performed on the input image as a pre-process for creating the edge histogram.   12. The pantograph detection method according to claim 11, wherein contrast enhancement is performed by linear density conversion of density distribution as the image quality correction.   13. The pantograph detection method according to claim 12, wherein when performing the contrast enhancement, a density value histogram in a fixed region in the center of the image including the pantograph is used.   9. The pantograph detection method according to claim 8, wherein a pantograph detection sensor is also installed above the track, and the pantograph is photographed by the monitoring camera when the pantograph detection sensor detects a pantograph on the train. Pantograph detection method to do.

Images