JP2006250573A - Rail joint plate fastening bolt fallout detection device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a rail joint plate fastening bolt fallout detection device for detecting fallout of a bolt for a rail joint plate in the traveling state of an inspection car. <P>SOLUTION: In this device, the shutter of a camera is released according to a detection signal of a joint detector just when the camera comes to a rail joint and a static image from the camera is collected, and therefore an image of the joint plate can be surely collected as a static image even if a visual field is limited. Moreover, since an illumination optical system illuminates the rail joint plate at an angle at which a shadow of a flange of the rail joint plate falls on the joint plate, an image of the shadow of the flange existing at the position of the joint plate at the time of a bolt fallout and an insertion hole for the fastening bolt are shown in the collected static image whose visual field is small. The bolt fallout can be detected from the static image at a high speed by determining whether or not there is an image corresponding to the insertion hole for the bolt on the basis of the image of the shadow of the flange. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a rail joint plate fastening bolt drop detection device, and more particularly, to a rail joint plate fastening bolt that can detect a rail joint plate fastening bolt drop in a running state of a track inspection vehicle. The present invention relates to a dropout detection device.

Railway rails expand and contract according to the ambient temperature. For this reason, a gap (joint) is provided to connect the rails to each other. The connection is made by fixing the joint plate with bolts by sandwiching the end of the rail from the front and back sides with the joint plate.
In order to prevent trains from getting on the rails, damage to the rails, etc., whether or not the joints are at appropriate intervals is regularly inspected by a track inspection vehicle (hereinafter referred to as inspection vehicle) or inspection vehicle. When it is not within the range of the predetermined reference width, the joint interval is adjusted.
A rail joint plate detection device that is mounted on a test vehicle for such inspection and that detects the joint plate itself by irradiating the position of the joint plate with laser light is known (Patent Document 1).
Further, as a technique for detecting a bolt drop of a bolt that fixes a rail to a sleeper, a rail fastener drop-off detection device using a laser displacement meter is known (Patent Document 2).
JP 2001-280918 A JP-A-11-172606

In the joint portion where there is a gap, the wheel of the vehicle falls, so that vibration easily occurs, and the fastening bolt of the joint plate is seen to fall off. Normally, for example, the rails are joined with a structure in which the joint plate is fastened with a plurality of bolts, for example, four or six, and the joint plate sandwiches the ends of the rails from both the front and back sides, so one bolt is missing. Even if it occurs, it does not lead to a major problem, but it is certain that the connection state will loosen, and in the event of an accident and the point of riding, it detects bolt disengagement in the running state of the inspection vehicle There is a request for equipment.
However, at present, there is no device for detecting the drop of the fastening bolt of such a rail joint plate.
FIG. 9 is a cross-sectional explanatory view of the rail joint plate in a state where the rails are fastened. As shown in this figure, fastening bolts 2 (hereinafter referred to as bolts 2) for fastening the joint plate 1 stop the joint plate 1 on the side surfaces of the connecting rails 3 and 4. The joint plate 1 has a flange 3a protruding in the horizontal direction at the top, which is in contact with the back surface of the head 3b of the rails 3 and 4 (representing the head on the rail 3 side). Therefore, when viewed from above, a part of the head 2a of the bolt 2 is hidden, or only a part of the head 2a is visible in the nut 2b. Accordingly, when the presence / absence of the bolt 2 is optically detected from above, it is difficult to discriminate, and it is difficult to detect the bolt dropout in the running state of the inspection vehicle. For this reason, techniques such as Patent Document 1 and Patent Document 2 cannot be directly applied to detection of bolt dropout of a seam plate. At present, the actual situation is that the detection of bolt dropout of the joint plate relies on human hands. In the figure, 2c is a washer and 3c is a bolt insertion hole.

The object of the present invention is to solve the above-mentioned problems of the prior art and meet the above-mentioned demand, and can detect the drop of the rail joint plate bolt in the running state of the inspection vehicle. An object of the present invention is to provide a falling detection device for a fastening bolt of a plate.

In order to achieve such an object, the feature of the rail joint plate fastening bolt drop detection device of the present invention is that the rail joint plate of the inspection vehicle that detects the fastening bolt missing of the rail joint plate connecting the rails to each other is provided. In the fastening bolt dropout detection device,
A non-contact type joint detector that is provided on the rail of the inspection vehicle and detects the rail joint, and a rail joint plate that is provided rearward with respect to the traveling direction of the inspection vehicle with respect to the joint detector in the inspection vehicle. A camera for capturing a still image, an irradiation optical system that illuminates the rail joint plate at an angle at which the shadow of the rail joint plate flange falls on the rail joint plate, and the camera in accordance with the detection signal of the joint detector. A control device for taking a still image from the camera by turning off the shutter of the camera at the timing of the corresponding position;
This is to detect whether an image corresponding to the insertion hole of the fastening bolt is present in the still image within a predetermined range of the image of the rail joint plate determined on the basis of the image of the shadow of the flange in the still image.

By the way, as described above, since a rail joint plate detection device that detects the joint plate itself is known as a conventional technique, simply detecting the joint plate and capturing the image with a camera to detect a bolt drop. It is possible. However, there is a problem in that an image cannot be obtained unless the field of view is widened even if the seam plate is simply detected and an image is captured by a running inspection vehicle. This is because the inspection vehicle swings up, down, left and right, and the vehicle shifts outward on the curve. If an image with a wide field of view is obtained as an image portion, and the image of the seam plate is searched for from among them and the bolt dropout is detected, the processing load becomes very large as image processing and the reliability of detection is also low. Become.
Therefore, the present invention has conceived of obtaining an image mainly composed of a seam plate having a field of view as narrow as possible by detecting the position of the seam portion itself of the seam plate as described above.
In the present invention, as described above, since the camera releases the shutter at the timing when the camera arrives at the rail joint according to the detection signal of the joint detector, the still image from the camera is collected. Even if it is restricted, the image of the seam plate can be reliably collected as a still image. Moreover, since the irradiation optical system illuminates the rail joint plate at an angle where the shadow of the rail joint plate flange falls on the rail joint plate, the collected still image with a narrow field of view becomes an image mainly composed of the joint plate, When there is dropout, an image portion of the shadow of the flange at the position of the rail joint plate and an image portion of the insertion hole of the fastening bolt are displayed. Therefore, by determining whether there is an image corresponding to the insertion hole of the bolt with reference to the image of the shadow of the flange, it is possible to detect the bolt-out state from the still image at high speed.
As a result, it is possible to easily detect the drop of the bolt of the rail joint plate in the traveling state of the inspection vehicle.

FIG. 1 is an explanatory diagram of a drop detection device for a fastening bolt of a rail joint plate of an inspection vehicle and a joint plate image pickup device thereof according to an embodiment to which the rail joint plate fastening bolt drop detection device of the present invention is applied. 2 is an explanatory diagram of a rail joint detector, FIG. 3 is an explanatory diagram of a detection circuit and a data processing device of the rail joint detector, FIG. 4 is an explanatory diagram of a detection waveform of the rail joint detector, and FIG. FIG. 6 is an explanatory diagram schematically illustrating an example of an image obtained by binarizing a captured still image, and FIG. 7 is a diagram illustrating brightness with respect to the Y coordinate of the binarized image illustrated in FIG. 6. FIG. 8 is a flowchart of image processing for bolt hole detection. The same components as those in FIG. 9 are denoted by the same reference numerals in the respective drawings.
In FIG. 1 (a), 10 is a test vehicle equipped with a rail joint plate fastening bolt drop detection device 5 (hereinafter referred to as a fastening bolt drop detection device 5), 10a is its carriage, and 10b is a wheel. It is. The fastening bolt dropout detection device 5 includes a rail joint detector 6, a joint plate image pickup device 7, and a data processing device 8.
The rail joint detector 6 includes an electromagnetic sensor 11, a case 12, and a detection circuit 20 as shown in FIGS. The electromagnetic sensor 11 built in the case 12 is composed of two air-core coils 11a and 11b which are wound in opposite directions to each other and have a triangular winding shape with one side being adjacent. The case 12 in which this is housed is made of a non-magnetic material such as synthetic resin, and is fixed in a completely sealed state filled with resin by shifting the centers Oa and Ob of the air-core coils 11a and 11b. As shown in (a), a predetermined distance from the surface of the head of the rail 3 or the rail 4 from the carriage 10a of the inspection vehicle 10 via the bracket 13, for example, about 3 cm, corresponds to one of the two rails. Installed. And the joint 1a which exists between the rail 3 and the rail 4 of Fig.1 (a) is detected.

As shown in FIG. 2, since the air core coils 11a and 11b are wound in opposite directions, the detection signals are also in opposite phases. Therefore, even if noise is added to the signal, it is canceled out. Furthermore, although the width of the head 3b of the rails 3 and 4 is usually about 65 mm, as shown in FIG. 1, the center Oa of the air-core coil 11a of the electromagnetic sensor 11 fixed by the case 12 is It is arranged corresponding to the center line O of the head 3b, and the center Ob of the air-core coil 11b is arranged so as to correspond to the rails 3 and 4 with an offset of about 10 mm from the center line O of the head 3b of the rail. ing.
The reason why the center Ob of the air-core coil 11b is shifted outward from the center line O is because the rail has a curve, and when the inspection vehicle 10 is shifted outward at this time, the center Ob is more than necessary. This prevents the air-core coils 11a and 11b from being detached and the detection signal of the electromagnetic sensor 11 from being obtained. In addition, the shift | offset | difference to the outer side of the inspection vehicle 10 is blocked | prevented by the flange of the wheel 10b, and does not shift | deviate large.

As shown in FIG. 1 (b), the joint board image capturing device 7 includes a joint board imaging camera 71 and an illumination optical system 72, and the joint board imaging camera 71 is installed on the lower side of the carriage 10a. The side surface of the joint plate 1 is imaged from the inside of the rails 3 and 4 obliquely from above at an angle θ as an elevation angle from the track surface of θ = 20 ° to 40 ° with respect to 1. The illumination optical system 72 illuminates the side surface of the joint plate 1 from the inside of the rails 3 and 4 obliquely upward at an angle φ as an elevation angle from a track surface having a larger φ = 50 ° to 70 °. The joint plate imaging camera 71 and the illumination optical system 72 are installed inside a housing 73 fixed to the carriage 10 a of the inspection vehicle 10.
Since two rails 3 and 4 are laid in a predetermined track width, as shown in FIG. 1 (a) and FIG. 1 (b), the joint plate imaging camera 71 is connected to each rail. A total of four illumination optical systems 72 are provided for each rail, and three illumination optical systems 72 are provided for each rail, for a total of six.
Each joint board imaging camera 71 is, for example, a CCD camera that obtains a still image of one screen (one frame) 640 × 480 pixels (VGA) as a captured image. Generated for each of the rails.
Here, what is important is the irradiation angle φ of the illumination optical system 72. The illumination angle φ is an angle at which the shadow of the flange 3a of the joint plate 1 overlaps the bolt insertion hole 3c (see FIG. 9) of the bolt 2 or falls between them. Thereby, for example, if the shadow of the flange 3a is applied to the bolt portion of the joint plate 1, and if the bolt is pulled out to form a hole, the image portion where the bolt insertion hole 3c is black as it is (see reference numeral 16 in FIG. 6). ) Can be obtained. Therefore, it is possible to detect the bolt dropout from the captured image based on the shadow image of the flange 3a. That is, when there is a black pixel block (see reference numeral 19 in FIG. 7) within a certain range from the shadow image portion of the flange 3a of the joint board 1 (see reference numeral 15 in FIG. 6) as a reference, the joint board It is possible to detect as one bolt hole. Thereby, it can be determined at high speed whether there is a bolt omission in the joint plate 1 in a state where the inspection vehicle 10 travels.

FIG. 6 is an explanatory diagram schematically showing an example of an image obtained by binarizing a captured still image. Reference numeral 14 denotes one screen of a binarized image captured by the joint plate imaging camera 71. In this image 14, the shadow image portion of the flange 3 a of the joint plate 1 is 15, and the image portion of the bolt insertion hole of the joint plate 1 is 16. The heads 3b of the rails 3 and 4 and the side surfaces of the joint plate 1 have a strong reflected light, so that they are at the white level (data “1”), and the others are at the black level (data “0”). It becomes. 17a is a bright upper background and 17b is a dark lower background.
Reference numeral 18 denotes a noise image portion of reflected light as noise appearing in the shadow image portion 15 of the flange 3 a of the joint plate 1.
In the binarized image 14, the image portion 16 of the bolt insertion hole appears as a set of black level pixels of a predetermined number of pixels. Detection is based on the image portion 15.
By the way, if detecting the bolt omission from the entire screen of the collected image, the load of image processing increases, and various noises are added to the image, so let's detect the image portion 16 of the bolt insertion hole with high accuracy. If so, the noise removal processing takes time. In addition, it takes time to determine the image portion 16 of the bolt insertion hole. Therefore, here, the position of the shadow image portion 15 of the flange 3a is detected, and it is determined whether or not there is an image portion 16 of the bolt insertion hole in a range corresponding to the bolt hole from now on. To detect bolt holes at high speed. This process will be described later.

FIG. 3 is an explanatory diagram of a detection circuit and a data processing device of the rail joint detector, in which the detection circuit 20 receives differential amplifiers 21a and 11b that receive signals from air core coils 11a and 11b wound in opposite directions, respectively. 21b, a bias circuit 22 for supplying a bias current to the air-core coils 11a and 11b, and signals from the differential amplifiers 21a and 21b from A / D conversion circuits (A / D) 23a and 23b according to a predetermined distance pulse PL Thus, the detection signal is converted into a digital value via the A / D 23a, 23b and sent to the data processing device 8.
The bias circuit 22 includes resistors R1 to R4 and a constant voltage power supply circuit 22a, and is a circuit for supplying a bias current to the air core coils 11a and 11b.
In the rail joint 1a, the inductance of the air-core coils 11a and 11b changes according to the joint plate 1 and the gap between the rail and the rail, and a voltage corresponding to the inductance is generated at the terminals of the air-core coils 11a and 11b. Detection signals corresponding to the differential amplifiers 21a and 21b can be obtained.
The distance pulse PL is a pulse signal generated according to the rotation of the wheel 10b, and is a pulse signal having a wavelength (period) of 45 mm / P generated by the distance pulse generation circuit 24 and generated once for 45 mm travel.
The data processing device 8 includes an MPU 81, a memory 82, an interface 83, an image memory 84, a bus 85 for connecting them, a distance pulse PL, and a differential amplifier that is A / D converted in accordance with the distance pulse PL. The signals 21a and 21b are received as measurement data via the interface 83, and the difference between them is calculated.
The detection circuit 20, the distance pulse generation circuit 24, and the data processing device 8 are mounted on the inspection vehicle 10.
In addition, two joint plate imaging cameras 71 provided on each rail are connected to the data processing device 8 via A / Ds 25a and 25b, respectively. Further, the shutters of the four seam board imaging cameras 71 are operated from the data processing device 8 via the interface 83, and the still images of each seam board imaging camera 71 are A / D converted as 8-bit 256 gradation still images. Each of the four images is transferred to a predetermined area of the image memory 84 and stored therein.
Each of the two images corresponding to each rail is picked up with a part overlapping more than the bolt insertion hole.

FIG. 4A shows measurement data in the data processing device 8 in which the digital values of the differential amplifiers 21a and 21b are graphed. BLa is measurement data of the differential amplifier 21a and BLb is measurement data of the differential amplifier 21b. It is. The portion indicated by the long circle is a detection signal of the rail joint position (position of the joint 1a), and it is due to the joint plate 1 that there are irregular waves in the waveform before and after that. The horizontal axis represents the number of samples by the distance pulse PL, and the vertical axis represents the voltage [mV]. The reason why the waveform of BLb is small is that the air-core coil 11b is displaced from the rail center O by 10 mm. The number of samples by the distance pulse PL on the horizontal axis corresponds to the travel distance of the inspection vehicle 10.
And in the data processor 8, the result of calculating the difference BLa-BLb of these measurement data is the graph BLa-BLb of FIG. 4B. Thereby, the detection signal DL of the rail joint position can be obtained as a detection signal having a large waveform. However, since the inspection vehicle 10 swings up, down, left and right, the detection waveform fluctuates greatly.
Therefore, if a moving average obtained by setting a predetermined sample interval and adding a predetermined number of samples for each sample data and erasing the first one sample data is obtained, the characteristic as shown in FIG. A graph (BLa-BLb) m can be obtained.
DL is a rail joint position detection signal. As shown in FIG. 5 (b), the rail joint position signal DP shown in FIG. 5 (c) can be obtained by setting the slice level TH (threshold value) to this and obtaining the detection signal DL.
If the gate including the rail joint position (position of the joint 1a) is set in a predetermined range from the distance pulse PL and the rail joint position signal DP is extracted, various kinds of signals generated during the traveling of the inspection vehicle are obtained. Noise is removed, and the rail joint position (position of the joint 1a) can be detected with high accuracy.
In order to perform the above processing, as shown in FIG. 3, the memory 82 of the data processing device 8 stores a difference calculation program for measurement data BLa and BLb, a moving average value calculation program, a threshold setting / rail joint position detection program, and the like. Are stored, and the MPU 81 sequentially executes these programs.

The data processing device 8 counts the distance pulse PL received from the distance pulse generation circuit 24 in the soft counter area set in the memory 82 when the rail joint position signal DP is detected from the measurement data of the rail joint detector 6. When the count value corresponding to the predetermined distance D (see FIG. 1) is reached, that is, after the joint 1a (see FIG. 1 (a)) is detected, the inspection vehicle 10 is set to the predetermined distance D. When the vehicle travels, a control signal for releasing the shutter of each joint board imaging camera 71 is generated via the interface 83, and one frame image obtained by each joint board imaging camera 71 is collected and stored in each image memory 84. Corresponding to the plate imaging camera 71, it is stored as four still images. Two still images of one frame corresponding to each rail, a total of four images, are binarized with a predetermined threshold value, are taken into the work area 82e of the memory 82, and correspond to each rail of one frame stored in the work area 82e. An image portion 15 in FIG. 6 called a shadow of the flange 3a is obtained from the captured image (original still image) of the joint plate 1, a reference line is determined therefrom, and the position of the reference line is determined from the binarized image as shown in FIG. It is determined whether or not the image portion 16 of the bolt insertion hole is within the range that should exist.
In the memory 82, in order to perform the above processing separately from the above-described difference calculation program for measurement data BLa, BLb, moving average value calculation program, threshold setting / rail joint position detection program, etc. A plate image collection program 82a, a binarization processing program 82b, a reference line detection program 82c on a shadow image, a joint plate bolt hole detection program 82d, and the like are stored, and a work area 82e is provided.
When the MPU 81 detects the rail joint position signal DP from the measurement data of the rail joint detector 6, the MPU 81 calls and executes the joint board image collection program 82a. When this is executed by the MPU 81, the program 81a starts counting the distance pulse PL, and the count value corresponds to the distance D between the rail joint detector 6 and the joint plate image pickup device 7. When the shutters of the four seam board imaging cameras 71 are turned off, four still images of one frame of 8-bit 256 gradations are provided via the A / D 25a, 25b, two corresponding to each rail. Collected as a digital value in the image memory 84. Then, the binarization processing program 82b is called.

When this is executed by the MPU 81, the binarization processing program 82b binarizes each image for one screen stored in the image memory 84 with a predetermined threshold value, and displays the upper background and lower background shown in FIG. Are stored in the work area 82 e of the memory 82.
The threshold value of the binarization processing is 640 pixels for one horizontal line in each collected still image with the vertical direction of the rails 3 and 4 as the Y direction and the X direction parallel to the rails 3 and 4 as the horizontal line. The brightness is integrated, and the integrated brightness is calculated corresponding to each pixel (Y coordinate value) in the Y direction corresponding to the horizontal line, and stored in the work area 82e as data. Then, the minimum value of the minimum brightness level that appears first among the minimum values of the integrated values in the Y direction is obtained and divided by 640, and the minimum value LP is obtained as the average minimum minimum value for one pixel of one horizontal line. calculate. Then, a light level minimum minimum value LP + α slightly higher than this is used as a threshold value and binarized on the basis thereof to generate binarized image data.
Thereby, the binarized image 14 as shown in FIG. 6 is obtained.
That is, when the brightness of 640 pixels for one line in the horizontal direction is integrated, the first minimum minimum value LP is the region with the highest black level of one horizontal line in the shadow image portion 15 of the flange 3a in FIG. One horizontal line, which corresponds to one horizontal line at a position indicated by 15a in FIG. Accordingly, if the minimum minimum value LP + α is set as a threshold value, the shadow image area 15 of the flange 3a in FIG. 6 can be dropped to the black level, and the other values in FIG. The converted image 14 can be obtained. However, at this time, the lower background 17b is also at the black level.
This is performed for each of the four still images in the image memory 84, and four binarized images 14 are stored in predetermined areas of the work area 82e.
Note that + α is selected as a value for dropping the shadow image portion 15 (shadow image of the flange 3a) to the black level.

When this is executed by the MPU 81, the reference line detection program 82c on the shadow image reads the binarized image from the work area 82e and weights the white level with “0” for black and “255” for white. Weighting processing is performed to calculate the brightness of 640 pixels (one horizontal line) integrated for 640 pixels (one horizontal line) for each pixel in the Y direction (each Y coordinate), Furthermore, by dividing by 640, data of the average value of brightness per pixel of each horizontal line is calculated, and this is weighted data (reference line detection weighted with the white level corresponding to each pixel in the Y direction). Reference line detection data) is stored in the work area 82e. FIG. 7 shows the reference line detection data as a graph. 7 is an explanatory diagram of the brightness characteristic with respect to the Y coordinate of the binarized image of FIG. 6. In FIG. 7, the position of the brightness A that is the first minimum minimum value is the minimum value Lmin. Corresponds to the pixel position.
As shown in FIG. 6, the shadow image portion (reference line band) 15 of the flange 3a is a band mainly composed of a black level between a white level band and a white level band when viewed as a large frame. Is appearing. Therefore, a graph of brightness characteristics as shown in FIG. 7 is obtained. The reason why the pixel position of the minimum value Lmin does not become the black level “0” is because the reflection region 18 as noise exists.
In this reference line detection data, the level of 1/3 of the maximum brightness 255 and the minimum brightness minimum value A is calculated as the slice level (threshold) THa of the brightness of the binarized image 14 by the following formula. To do.
THa = (255−A) / 3 + A
This is the pixel position corresponding to the shadow image portion (reference line band) 15 of the flange 3a when the noise image portion 18 in FIG. 6 is deleted. Noise is removed by this process, and the pixel position S of the Y coordinate of the reference line is determined using the lower line of the image portion (reference line band) 15 as the reference line. Here, the joint plate bolt hole detection program 82d is called.

When this is executed by the MPU 81, the seam plate bolt hole detection program 82d defines the area corresponding to the brightness portion of the slice level THa as the reference line band, that is, the shadow image portion 15 of the flange 3a shown in FIG. 7 is a black level block region 19 (corresponding to the image portion 16) from the position of the Y coordinate S to the reference line S on the side (the pixel position S in the Y direction) as a reference or continuously from this position. It is sequentially searched from the pixel position S in the Y direction behind the slice level THa in FIG.
That is, if the number of pixels having the same level of brightness less than 255 and equal to or higher than the brightness of the slice level THa is continuously set to a certain number, for example, 1 mm per pixel, 6 pixels to 8 pixels It is determined whether or not there is an image portion 16 of the bolt insertion hole by determining whether or not there is a region 19. When there is a certain range in the Y direction from the pixel position S of the Y coordinate corresponding to the reference line and the range of the number of Q pixels, it is determined that it is a bolt insertion hole.
Note that the same level of brightness is obtained by setting the brightness of ± β for the brightness in the range equal to or higher than the brightness of the slice level THa. However, the upper limit is the brightness 255, and the lower limit is the brightness of the slice level THa.

FIG. 8 is an overall flowchart of image processing for bolt hole detection performed by the MPU 81 executing the above processing programs.
Here, the rail joint detection process and the bolt hole detection process are executed in parallel as task processes. Further, in the bolt hole detection processing, two binarized still images are processed in parallel for each rail by task processing at the same time, and detection processing is performed for each.
First, the rail joint detection process on the left side of the drawing will be described.
In the rail joint detection process, the detection signal from the rail joint detector 6 is read (step 101), the difference calculation process between the measurement data BLa and BLb is performed (step 102), and the moving average value (BLa−BLb) m is calculated. A moving average calculation process is performed (step 103), a threshold value setting / rail joint position call is made, a threshold value TH is set, and whether or not there is a rail joint is determined by (BLa−BLb) m> TH ( Step 104).
When the determination result of step 104 is YES, the detection position of the detection signal DL is externally measured as seam position measurement data corresponding to the traveling distance of the test vehicle 10 calculated according to the number of samples by the distance pulse PL. It is stored in a storage device (not shown) or the like (step 105). Then, the bolt hole detection processing program is called to start the bolt hole detection processing (step 106). Next, it is determined whether or not the process is finished (step 107). If NO, the process returns to step 101, and the process from step 101 to step 107 is circulated and executed unless the process is finished.
Of the above, the process of step 105 and the process of step 106 are reversed, and step 106 may be in front of step 105.

When the rail joint 1a is detected in step 104, the bolt through hole detection processing program is called in step 106 and the bolt through hole detection processing on the right side of the drawing is entered.
In the bolt hole detection processing, a counter is set and counting of the distance pulse PL is started (step 201), and it is determined whether or not the count value of the distance pulse PL has reached a predetermined value K (step 202). Returning to step 201, the distance pulse PL is counted after waiting for the next distance pulse PL. In this loop, it is determined from the predetermined value K whether the test vehicle 10 has traveled the distance D between the rail joint detector 6 and the joint plate image pickup device 7 shown in FIG.
When the number of the distance pulses PL reaches the predetermined value K, it becomes YES in the determination step 202, it is detected that the inspection vehicle 10 has traveled the distance D, and the MPU 81 calls and executes the joint plate image collection program 82a. To do. Therefore, the shutters of the four joint plate imaging cameras 71 are turned off, the four images of the joints are stored in the image memory 84, and four four-screen images are collected (step 203). The binarization processing program 82b is called and executed, binarization processing is performed on the four still images with the minimum minimum value LP + α as a threshold value, and four binarized images 14 shown in FIG. Each area 82e is stored (step 204). From step 204, parallel detection processing is performed on two still images corresponding to each rail as described above. Hereinafter, processing for one still image will be described. Similar processing is performed for other devices at the same time.

Next, the MPU 81 calls and executes the reference line detection program 82c on the shadow image to generate reference line detection data corresponding to each pixel in the Y direction weighted to the brightness (step 205), and the maximum brightness. The brightness level 255 and the level of 1/3 of the minimum brightness minimum value A are set as the slice level (threshold value) THa of the brightness of the binarized image 14 and below the shadow image portion (reference line band) 15 of the flange 3a. The pixel position of the reference line on the side is detected as the pixel position S (step 206). Then, pixels having a brightness of less than 255 and having the same degree of brightness above the slice level THa are continuously, for example, there are pixels continuing from the pixel position S in the range of 6 pixels to 8 pixels. Whether or not (step 207). This determines whether it is a bolt insertion hole. If there are two or more bolt insertion holes, the range of 6 pixels to 8 pixels increases according to the number of holes. In the case of two, the range is 8 pixels or more and 10 pixels, and in the case of 3 or more, the range is 12 pixels or more and 15 pixels. When the joint plate is 6-point bolted, there is no further missing in one image, and usually there is only one tightening bolt missing, so the above determination is made. Can be appropriately selected.
Here, when NO, the pixel position S is updated to S = S + 1 (step 208), it is determined whether S> = Q, and it is determined whether or not the maximum pixel position Q of the Y coordinate of the joint plate 1 has been exceeded. (Step 209), if NO, return to Step 207.
When the determination in step 207 is YES, it is determined that a bolt insertion hole has been detected, and is added to the measurement data of the detection position of the detection signal DL stored in step 105 as a bolt insertion hole detection flag. The position is recorded in an external storage device or the like (step 210), and the bolt hole detection process is terminated.
At this time, the bolt hole detection process ends. However, unless the rail joint detection process ends, the bolt hole detection process starts again when the rail joint is detected in step 104.
In this way, the bolt hole detection of the joint plate 1 is performed, and the data is stored in the external storage device.
Note that the seam plate 1 may be numbered and managed from the inspection start point.
Further, according to the range determination in the above-described step 207, even if there are two or more bolt insertion holes in the joint plate 1, the brightness level falls within the range below the brightness 255 and above the brightness of the slice level THa. In this case, detection is possible.

  By the way, in the embodiment, it is determined whether or not there is a black level block region 19 (corresponding to the image portion 16) in the reference line detection data according to the Y coordinate shown in FIG. However, for the black level block area 19, the number of pixels in this area obtained from the XY coordinates corresponding to the block area 19 from the original still image in the image memory 84 or the binarized still image stored in the work area 82e ( It is also possible to determine whether or not there is a bolt hole depending on whether the number of pixels (area) corresponds to the bolt hole or not.

As described above, in the embodiment, two cameras are used for one rail. However, this is an example in which the length of the joint plate 1 is long. May be. If it is longer, three units may be provided. In either case, it is not necessary to increase the field of view of one camera. In addition, images collected by a plurality of cameras may be combined into a single still image corresponding to the rail, and a bolt hole detection process may be performed on this.
In the embodiment, an electromagnetic sensor using the first and second air-core coils arranged adjacent to each other and wound in opposite directions as the rail joint detector is used. Of course, if it is a non-contact type, a laser type displacement sensor and other optical sensors can be used. In addition, the shape of the air core coil of the electromagnetic sensor in the embodiment is an example, and it is needless to say that it may be a circular air core coil. Further, it is not always necessary to detect the rail joint (joint 1a) by program processing.
Furthermore, in the embodiment, the case 12 is a resin-filled sealed type, but this may be a partially opened case.
Furthermore, in the embodiment, an example is shown in which the shadow of the flange of the joint plate overlaps with the bolt hole, but these do not necessarily have to overlap. What is necessary is just to have a shadow of the flange of the seam plate as a retrieval reference line for the image portion of the bolt hole.

FIG. 1 (a) is an explanatory view of a detection device for detecting a fastening bolt of a rail joint plate of a test vehicle according to one embodiment to which the rail joint plate fastening bolt detection device of the present invention is applied, FIG. 1 (b). These are explanatory drawings of a joint board image pick-up device. FIG. 2 is an explanatory diagram of a rail joint detector. FIG. 3 is an explanatory diagram of a detection circuit and a data processing device of the rail joint detector. FIG. 4 is an explanatory diagram of a detection waveform of the rail joint detector. FIG. 5 is an explanatory diagram of the joint detection process. FIG. 6 is an explanatory diagram schematically illustrating an example of an image obtained by binarizing a captured still image. FIG. 7 is an explanatory diagram of the brightness characteristic with respect to the Y coordinate of the binarized image of FIG. FIG. 8 is a flowchart of image processing for bolt hole detection. FIG. 9 is a cross-sectional explanatory view of the rail joint plate in a state where the rails are fastened.

Explanation of symbols

1 ... Seam plate, 1a ... Seam, 2 ... Bolt,
2a ... bolt head, 2b ... nut,
3, 4 ... rail, 3a ... flange, 3b ... rail head,
3c ... Bolt insertion hole,
5 ... Drop detection device for fastening bolt of rail joint plate,
6 ... Rail joint detector, 7 ... Joint plate image pickup device,
8 ... Data processing device, 10 ... Inspection vehicle,
11 ... Electromagnetic sensor,
11a, 11b ... air-core coils,
12 ... Case, 13 ... Bracket,
20 ... detection circuit, 21a, 21b ... differential amplifier,
22a ... constant voltage power supply circuit,
23a, 23b, 25a, 25b ... A / D conversion circuit (A / D),
24 ... Distance pulse generation circuit,
71 ... Seam plate imaging camera, 72 ... Illumination optical system,
81 ... MPU, 82 ... memory,
82a ... Seam board image collection program,
82b ... Background removal / binarization processing program,
83c ... Reference line detection program on shadow image,
82d ... Seam plate bolt hole detection program,
82e ... work area, 83 ... interface,
84: Image memory, 85: Bus.

Claims (5)

In the drop detection device for the fastening bolt of the rail joint plate of the track inspection vehicle for detecting the fastening bolt missing of the rail joint plate connecting the rail and the rail,
A non-contact type seam detector provided on the rail of the track inspection vehicle for detecting the rail seam;
A camera for taking a still image of a joint plate of the rail provided rearward with respect to the traveling direction of the trajectory inspection vehicle in the trajectory inspection vehicle;
An irradiation optical system for irradiating the rail joint plate at an angle at which the shadow of the flange of the rail joint plate falls on the rail joint plate;
A controller for taking the still image from the camera by opening the shutter of the camera at a timing when the camera has reached a position corresponding to the rail joint according to a detection signal of the joint detector;
A rail for detecting whether or not an image corresponding to the insertion hole of the fastening bolt is present in the still image within a predetermined range of the image of the rail joint plate determined on the basis of the shadow image of the flange in the still image. A device for detecting a drop in the fastening bolt of the joint plate.   The illumination optical system is mounted on the track inspection vehicle, and the joint plate is configured to emit light that shadows the flange from the inside of the rail at a selected angle of 50 ° to 70 ° with respect to the rail. 2. The rail joint plate fastening bolt drop detection device according to claim 1, wherein the predetermined range is a range based on an image line below a shadow image of the flange.   In the still image, the vertical direction perpendicular to the rail is defined as the Y direction, the horizontal direction parallel to the rail is defined as the X direction, and the brightness obtained by integrating the brightness of pixels for one line in the X direction is defined as the Y direction. A value calculated by corresponding to each one pixel, obtaining a minimum minimum value that appears first among the minimum values of the integrated values in the Y direction, and dividing the minimum minimum value by the number of pixels for one line in the horizontal direction. 3. The rail joint according to claim 2, wherein the still image is binarized with a predetermined threshold calculated on the basis of the rail, and an image corresponding to the insertion hole of the fastening bolt is detected in the binarized image. A device for detecting the falling bolts of plate fastening bolts. The seam detector is provided in the track inspection car corresponding to the upper part of the head of the rail, the first and second air-core coils arranged adjacent to each other and wound in opposite directions,
A case containing these first and second air-core coils,
A detection circuit that obtains first and second detection signals corresponding to changes in inductance of the air-core coils by causing a bias current to flow through the first and second air-core coils,
4. The rail joint plate fastening bolt drop detection device according to claim 3, wherein the detection signal of the joint detector is obtained based on a difference in level between the first detection signal and the second detection signal.   The case is of a sealed type, and the first and second air core coils are air core coils having a triangular winding shape, each of which is disposed adjacent to each other, and the first, The center of one of the second air-core coils is shifted outward from the center of the head of the rail, and the detection signal of the joint detector is the first detection signal and the second detection signal. The apparatus for detecting a drop of a fastening bolt of a rail joint plate according to claim 4, which is determined based on a moving average value of a difference in level.

Images