JP6717666B2 - Inspection image generator - Google Patents

Description

The present invention relates to an inspection image generation device that generates a captured image for a visual inspection of a railway vehicle.

Traditionally, railway organizations regularly inspect the appearance of railway cars for abnormalities. In the visual inspection, the inspector dives under the vehicle or climbs on the roof, and it is checked mainly by visual inspection whether there is any abnormality. If an abnormality is found, the relevant departments will be contacted immediately and repairs will be arranged. Such inspection work requires a lot of trouble and is dangerous for the inspector.

Therefore, conventionally, it has been proposed to photograph a railway vehicle using a photographing device and inspect the appearance of the railway vehicle from the photographed image (see, for example, Patent Documents 1 and 2). The technique of Patent Document 1 is for photographing a specific portion of a railway vehicle from a predetermined angle with a dedicated camera, and comparing the photographed image with a reference image to determine whether there is an abnormality. The technology of Patent Document 2 determines the presence or absence of an abnormality by photographing a train running on a track, synthesizing an image of the entire train, setting a predetermined inspection area in the image, and comparing the image with a normal train image. To do. The image of the entire train is created by cutting and pasting a plurality of images obtained by high-speed continuous shooting from the side of the train.

On the other hand, an inspection system using a line scan camera is generally known in the field of appearance inspection of industrial products and the like. The line scan camera captures an image of a moving subject together with one line of an optical image and continuously captures the captured one line of captured images to obtain a two-dimensional image. According to the line scan camera, it is possible to continuously shoot a subject long in one direction. In addition, since the imaging process of converting an optical image into an electric signal is performed for each one-dimensional optical image, it is possible to obtain a very high division ability as compared with an area camera that performs two-dimensional imaging.

JP-A-5-143714 JP, 2013-53875, A

The above-mentioned conventional proposals for inspecting the appearance of a railway vehicle using an image capturing device have some problems. For example, in the inspection device of Patent Document 1, since an image of a railway vehicle is taken from a predetermined angle with a dedicated camera for each specific part, there is a problem that a large amount of imaging equipment is required and the cost increases. In addition, there is a problem that the imaging process becomes very complicated because the inspection region is individually imaged. As the number of inspection points on railway vehicles increases, the cost and complexity increase accordingly.

Further, with the inspection device of Patent Document 2, although the imaging process is relatively easy, it is difficult to obtain an image of a train with high dividing ability and little distortion, and the inspection accuracy is lower than that of visual inspection. There is a problem called. As shown in Patent Document 2, when high-speed continuous shooting is performed while a train is running, a portion that appears in the center of one captured image and a lateral position that is off the center due to variations in the moving speed of the train and the shooting timing. Each time you take a picture, it will be different from the part shown in. In a general captured image, the central part of the image displays the shape of the part viewed straight, but the lateral part displays the shape of the part viewed diagonally from left to right, resulting in a difference in shape. For this reason, in a trained image obtained by synthesizing such images, shape distortion partially appears, and the location where the shape distortion appears is different for each shooting and is not constant. The distortion of such a shape can be reduced by photographing the train from a distance, but in that case, the dividing ability of the photographed image becomes extremely low.

On the other hand, by using the line scan camera, it is possible to avoid that the shape viewed from the left and right sides is included in the image. However, a line scan camera generally targets a subject in a shooting environment, such as an industrial product that is uniformly illuminated and conveyed at an accurate speed in a line production process. It was Then, a photographed image with high resolution without distortion has been obtained by photographing in such an environment.

When an appearance inspection is performed by using a captured image, it is convenient to take an image when a railway vehicle enters a vehicle depot, for example, because the time and effort of moving the vehicle only for the inspection can be saved. However, under such conditions, it is difficult to move the railway vehicle accurately at a constant speed. Further, the ground on which the rail is laid has a strong portion and a relatively soft portion. Therefore, when the railway vehicle travels on the rail, the railway vehicle is slightly displaced vertically due to its own weight. Furthermore, since a suspension such as an air spring is provided between the bogie and the vehicle body, the vehicle body also moves up and down due to the action of the suspension. Therefore, if no measures are taken, it is difficult to obtain a captured image with high resolution and little distortion even if the railway vehicle running in this way is captured by a line scan camera.

SUMMARY OF THE INVENTION It is an object of the present invention to provide an inspection image generating apparatus capable of obtaining a captured image of a railroad vehicle traveling on a rail with no irregular distortion and a high dividing ability for visual inspection.

The inspection image generating apparatus of the present invention, in order to achieve the above object,
A line scan type photographing means for obtaining a photographed image by vertically scanning a railroad vehicle moving on a track,
A reference device having a display surface showing a reference position in the up-down direction and fixed to the railroad track at a position overlapping with a shooting location of the shooting means;
Distortion correcting means for correcting the distortion of the shape of the photographed image caused by the vertical movement of the railway vehicle,
The distortion correction means,
Rail distortion correction means for shifting the photographed image vertically for each part so that a line indicating the reference position continuously reflected in the photographed image becomes linear.
Of the photographed image, except for the image of the region where the bogie of the railway vehicle and the rail are imaged, the image of the region where the body of the railway vehicle and the accessories fixedly connected to the vehicle body are imaged Vehicle body distortion correction means for shifting in the vertical direction,
It is characterized by having.

According to such a configuration, the line-scan type photographing means can obtain a photographed image of the railway vehicle having a high resolution and showing the shape when viewed from a fixed direction at any location. Further, the distortion correction means corrects the distortion of the shape of the captured image due to the sinking of the rail and the distortion of the shape of the captured image due to the vertical vibration of the vehicle body due to the action of the suspension from the captured image. Therefore, it is possible to provide a photographed image of a railway vehicle that is suitable for visual inspection without irregular distortion.

Here, the vehicle body distortion correction means,
Of the captured image, a mask means for hiding a region in which a bogie and a rail of a railway vehicle are imaged, and extracting an image of a region in which a vehicle body of a railway vehicle and an accessory fixedly connected to the vehicle body are imaged,
A distorted image division unit that divides the image extracted by the mask unit into rectangular image elements that are long in the vertical direction;
Image correction means for correcting the vertical shift by comparing the rectangular image element with the reference image of the railroad vehicle,
A mask releasing means for restoring the image of the area hidden by the masking means to the image corrected by the image correcting means,
Should be included.
With such a configuration, it is possible to cleanly correct the distortion of the captured image caused by the vertical vibration of the vehicle body due to the action of the suspension such as the air spring.

Further, the inspection image generating apparatus according to the present invention is
Further comprising speed expansion/contraction correction means for correcting the variation in the aspect ratio of each part of the captured image due to the moving speed of the railway vehicle,
The speed expansion/contraction correction means,
Characteristic portion extracting means for extracting the characteristic portion of the railway vehicle by performing edge extraction of the photographed image,
A stretchable image dividing unit that divides the captured image into rectangular image elements that are long in the vertical direction,
Aspect ratio calculation means for comparing the reference profile showing the reference shape of the characteristic portion of the railway vehicle with the characteristic portion extracted by the characteristic portion extraction means to calculate the deviation of the aspect ratio of each part of the captured image. When,
An image expansion/contraction means for expanding/contracting the image elements divided by the expansion/contraction image division means in the horizontal direction so that the deviation of the aspect ratio calculated by the aspect ratio calculation means is eliminated.
Should be included.
With such a configuration, it is possible to cleanly correct the distortion of the image due to the moving speed of the railway vehicle.

Further, the inspection image generating apparatus according to the present invention may be configured to perform the correction processing of the captured image in the order of the rail distortion correction means, the speed expansion/contraction correction means, and the vehicle body distortion correction means.
According to this configuration, first, the rail distortion correction unit corrects the entire area of the image in the vertical direction. Therefore, when the correction processing of the speed expansion/contraction correction unit is performed, the aspect ratio of each part in the captured image can be easily calculated. And it can be performed with high accuracy. Therefore, the distortion of the captured image caused by the moving speed can be corrected neatly. Further, the rail distortion correction means and the speed expansion/contraction correction means perform the vertical correction and the horizontal correction for the entire area of the image, and then the correction processing of the vehicle body distortion correction means is performed. It is possible to easily and accurately separate the car from the carriage. As a result, the distortion of the image caused by the vertical vibration of the vehicle body due to the suspension such as the air spring can be corrected more beautifully.

Furthermore, the inspection image generating apparatus according to the present invention is
The display surface of the reference unit further includes a lightness display section showing lightness of a plurality of gradations,
It is preferable to further include a brightness correction unit that corrects the brightness of the captured image based on the brightness of a plurality of gradations of the brightness display unit included in the captured image.
According to this configuration, it is possible to suppress the variation in the brightness for each captured image by the brightness correction using the brightness display unit of the reference device. Further, one reference device can provide both an index for rail distortion correction and an index for brightness correction.

According to the present invention, it is possible to provide an inspection image generation apparatus capable of obtaining a captured image of a railroad vehicle traveling on a rail with a high dividing ability without irregular distortion for visual inspection.

1 is a block diagram showing an overall configuration of a railway vehicle appearance inspection device according to an embodiment of the present invention. It is a figure which shows the example of arrangement|positioning of some components of the visual inspection apparatus of embodiment. It is a front view which shows a reference|standard. It is a flow chart which shows the procedure of the inspection processing performed by the inspection processing device. 4A and 4B are diagrams illustrating an example of rail distortion correction processing executed in step S2 of FIG. 4, where FIG. 6A is a photographed image diagram before correction and FIG. 4B is a photographed image diagram after correction. It is explanatory drawing which shows an example of the brightness correction process performed by step S3 of FIG. 5 is a flowchart showing a detailed procedure of a speed expansion/contraction correction process executed in step S4 of FIG. 4. 6A and 6B are diagrams illustrating an example of speed expansion/contraction correction processing, FIG. 7A is an image view of a reference profile showing a reference shape of a characteristic portion of a railway vehicle, FIG. It is a picked-up image figure after correction. 6 is a flowchart showing a detailed procedure of a vehicle body distortion correction process executed in step S5 of FIG. 4. 6A and 6B are diagrams illustrating an example of a vehicle body distortion correction process, in which FIG. 7A is a photographed image diagram before correction, FIG. 9B is a mask image diagram, and FIG. 9C is a process of detecting a deviation amount of image elements divided into rectangular shapes. Explanatory drawing, (d) is explanatory drawing of the process which shifts an image element.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
FIG. 1 is a block diagram showing an overall configuration of a railway vehicle appearance inspection device according to an embodiment of the present invention. FIG. 2 is a diagram showing an arrangement example of some components of the appearance inspection apparatus of the embodiment. FIG. 3 is a front view showing the reference device.
The visual inspection apparatus 1 according to the embodiment of the present invention is an apparatus that generates a captured image of a railway vehicle suitable for visual inspection and inspects the external appearance of the railway vehicle based on the captured image. The captured image generating means included in the visual inspection apparatus 1 corresponds to an example of the inspection image generating apparatus according to the present invention.

As shown in FIGS. 1 and 2, the visual inspection device 1 includes a vehicle proximity detection device 11, a photographing device 12, a background plate 14, an image processing device 15, a reference device 17, and an inspection processing device 20. Among these, the vehicle proximity detection device 11, the image capturing device 12, the background plate 14, the image processing device 15, and the reference device 17 are, for example, as shown in FIG. It is placed by the side. The inspection processing device 20 is arranged indoors and is connected to the vehicle proximity detection device 11 and the image processing device 15 by wire or wirelessly. The visual inspection apparatus 1 includes two sets of imaging devices 12 arranged on one side and the other side of the railroad track so that the left side and the right side of the railway vehicle can be imaged.

The vehicle proximity detection device 11 detects, for example, an optical sensor or an electric sensor using a track circuit that the railway vehicle enters the photographing section, and outputs a detection signal to the photographing device 12. Further, the vehicle proximity detection device 11 has a receiver of an RFID (radio frequency identifier), communicates with the RFID mounted on the railway vehicle to read the vehicle number information of the railway vehicle, and uses this information as the inspection processing device 20. Send to. The inspection processing device 20 can specify the vehicle type from the vehicle number information.

The image capturing device 12 has a line scan camera 13 as an image capturing unit, and captures an image of a railway vehicle traveling on a track from the side by a line scan method. The line scan camera 13 is a line scan type imaging device that repeats unidirectional scanning, that is, line-shaped imaging at a constant frequency. The line scan camera 13 repeats line-shaped image pickup in a direction intersecting the moving direction with respect to an image pickup target moving in a predetermined direction, and sequentially outputs line-shaped image pickup data. By arranging a plurality of line-shaped image pickup data that are sequentially output in a two-dimensional direction, a plane-shaped spread image can be obtained. The line-shaped image pickup means an image pickup having a large number of pixels in the vertical direction and a single pixel in the horizontal direction, for example, image pickup of 2000 pixels in the vertical direction×1 pixel in the horizontal direction. Hereinafter, line-shaped imaging is referred to as “line scan”, and the direction of the pixel row imaged by line scanning is referred to as “line scan direction”.

The line scan camera 13 is set such that the line scan direction is a direction that intersects with the line, such as the vertical direction. The photographing device 12 automatically starts photographing based on the proximity signal from the vehicle proximity detection device 11. The imaging device 12 continuously captures the line-shaped imaging data and transmits the imaging data to the image processing device 15. In line-scan photography, when the object to be photographed moves at a constant speed in a direction orthogonal to the line-scan direction, a photographed image with high resolution without distortion can be obtained. On the other hand, if the object to be photographed has variations in speed or displacement in the line scan direction, these appear as distortions in the shape in the photographed image.

The background plate 14 is a structure having a background display surface of a single color and a single brightness. The background plate 14 is arranged at a position facing the image capturing device 12 with the line in between so that the background display surface overlaps the position where the line scan camera 13 scans the line. The background plate 14 is an area in which the other side of the vehicle, such as the lower part of the vehicle body or the periphery of the bogie, is reflected when the railway vehicle enters the railroad track, and is reflected in the image captured by the line scan camera 13 as the background.

The image processing device 15 expands the line-shaped image data continuously captured by the line scan camera 13 into the frame memory 16 to generate the image data of the entire railway vehicle which is long in the track direction. Send to 20. The captured image of the railroad vehicle includes brightness variations due to ambient brightness. In addition, in this captured image, the shapes of multiple systems are caused by variations in the moving speed of the railway vehicle, vertical vibration of the railway vehicle due to sinking of the rail, and vertical vibration of the vehicle body due to the action of the air spring (suspension) of the railway vehicle. Includes distortion.

As shown in FIG. 3, the reference unit 17 has a display surface 17a in which the brightness of a plurality of gradations is shown stepwise, and is fixedly connected to the rail 90 so as not to interfere with the railway vehicle. The reference unit 17 is arranged so that the display surface 17a overlaps with the shooting location of the line scan camera 13. The display surface 17a can display the reference position in the vertical direction by the boundary line of the brightness of a plurality of gradations as well as the display of the brightness.

The inspection processing device 20 includes a CPU (central processing unit) 21 that executes a control program, a RAM (random access memory) 22 that provides a memory space for work to the CPU 21, and an interface 23 that exchanges signals with external devices. And a storage device 24 that stores a control program and control data. The storage device 24 stores a visual inspection program 24A as a control program for performing a visual inspection of the railway vehicle based on the captured image. In addition, the storage device 24 stores a normal image database (normal image storage means) 24B in which appearance images of normal railway vehicles that serve as a comparison reference at the time of appearance inspection are stored, and a captured image showing abnormal appearance of the railway vehicles. An abnormal image database (abnormal image storage means) 24C to be stored is provided.

Next, the inspection processing executed by the inspection processing device 20 will be described.
FIG. 4 is a flowchart showing the procedure of the inspection process executed by the inspection processing apparatus.
The inspection process is started when the image processing device 15 receives the captured image of the railway vehicle from the inspection processing device 20. When the inspection process is started, the inspection processing device 20 first inputs the captured image of the railway vehicle from the image processing device 15 (step S1).

When the captured image is input, the inspection processing device 20 performs a plurality of correction processes (steps S2 to S5) for correcting the distortion of the captured image. Details of these correction processes will be described later. By each of these correction processes, the captured image of the railway vehicle in which the variations in the captured image due to the variable elements of the railway vehicle and the capturing environment are corrected and the distortion is corrected is obtained.
These correction processes are realized by the cooperation of the hardware such as the CPU 21, the RAM 22, and the storage device 24, and the software included in the appearance inspection program 24A. The configuration that realizes the correction processing in steps S2, S3, S4, and S5 corresponds to an example of the rail distortion correction unit, the brightness correction unit, the speed expansion/contraction correction unit, and the vehicle body distortion correction unit according to the present invention, respectively. The configuration that realizes the correction processing in steps S2 and S5 corresponds to an example of the distortion correction means according to the present invention.

When the correction process is completed, the inspection processing device 20 executes a loop process (steps S6 to S11 in FIG. 3) for inspecting the presence or absence of abnormality based on the captured image. When shifting to the loop processing, the inspection processing device 20 first sequentially selects each inspection target area from among a plurality of preset inspection target areas (step S6).
The area to be inspected is set in advance so as to surround a portion to be inspected for visual inspection, such as a bogie of a railway vehicle, each part in the bogie, a cock of an air passage provided at a lower portion of a vehicle body, a brake device, a lever of a brake device, and the like. ing. The coordinate information indicating the inspection target area is stored in, for example, the coordinate information database 24D provided in the storage device 24, and the inspection processing device 20 reads the information from this and selects the inspection target areas one by one.

When one inspection target area is selected, the inspection processing device 20 extracts an image of this inspection target area from the captured image (step S7), and an image of a normal railway vehicle stored in the normal image database 24B (hereinafter, (Referred to as “normal image”) (step S8). The normal image database 24B stores one or a plurality of normal railway vehicle images for each vehicle type, and the inspection processing device 20 determines, in step S8, one normal image of the same vehicle type as the inspection target railway vehicle. Is read from the normal image database 24B and compared. In the comparison process, for example, the difference in hue and brightness of each pixel between the captured image and the normal image is digitized, and the sum of the differences in all pixels in the comparison target area is calculated as the distance value. It should be noted that when the difference between each pixel is quantified, the difference between lightness or hue may be squared to obtain a difference value, or the difference value may be calculated by weighting according to the lightness or hue. The method can be changed appropriately. In addition, before the comparison, the process of eliminating the small positional deviation between the two may be performed before performing the comparison process, or the positions to be compared may be shifted little by little and a plurality of comparisons may be performed to compare the minimum distance value. It may be adopted as a result.

Then, the inspection processing device 20 determines whether or not the comparison with all the normal images of the same vehicle type accumulated in the normal image database 24B is completed (step S9), and if not yet, returns to step S8, and next. The normal loop image is read and the comparison process is repeated to continue the loop process. On the other hand, if completed, the inspection processing apparatus 20 exits the loop processing and shifts the processing to the next step. By the loop processing of steps S8 and S9, the captured image of one inspection target area is compared with all the normal images of the same vehicle type stored in the normal image database 24B.

When the comparison with all the normal images is completed, the inspection processing apparatus 20 compares the distance value obtained by a plurality of image comparisons with the threshold value to determine whether there is an abnormality in the inspection target area (step S10). ). The threshold value for determining the presence or absence of abnormality is set to be larger than the average distance value calculated by comparison between normal images, and normality is a distance value that rarely occurs in comparison between images. It Such a threshold value is appropriately set for each inspection target area, and is stored in the determination threshold value database 24E provided in the storage device 24.

In the determination process of step S10, the smallest distance value among the plurality of distance values obtained by the plurality of image comparisons may be used for comparison with the threshold value. By such a determination process, among the plurality of normal images accumulated in the normal image database, the normal image having the closest image in the inspection area and the captured image are compared, and based on this comparison, the abnormal image is detected. The presence or absence will be determined.

Even if the railroad vehicles are of the same type, they may not be completely the same because, for example, the arrangement of parts such as cables is slightly different and normal fading occurs over time. Therefore, even when the inspection target areas are compared, a small difference in a normal range is excluded by performing comparison using the normal images of a plurality of the same vehicle types as described above and determining whether there is an abnormality. Therefore, it is possible to highly accurately determine whether or not there is an abnormality in the inspection target. If the portion that determines the presence or absence of an abnormality is minute and the difference between the images is relatively small, by setting a comparatively small area for comparison, even if there is a difference in the normal range. It is possible to judge with high accuracy without missing the presence or absence of abnormality.

When the distance value is smaller than the threshold value and is determined to be normal as a result of the determination process in step S10, the inspection processing apparatus 20 advances the process to step S12. On the other hand, when the distance value is equal to or greater than the threshold value and is determined to be abnormal, the inspection processing device 20 registers the captured image of the railroad vehicle determined to be abnormal in the abnormal image database 24C (step S11). When registering, data indicating the corresponding inspection target area may be added so that the inspection target area determined to be abnormal can be known. In addition, when it is determined that an abnormality has occurred in a plurality of inspection target areas of the same captured image, one captured image and a plurality of inspection target areas determined to be abnormal are registered so that the same captured image is not registered twice. The information to be shown and the information to be shown may be accumulated.

By accumulating captured images that have been determined to be abnormal, after that, by analyzing these captured images, the deterioration over time is analyzed, the tendency of abnormalities for each vehicle type is analyzed, and various abnormalities are analyzed. be able to. Then, an efficient maintenance system can be realized according to these analysis results.
In the abnormality registration process of step S11, the inspection processing device 20 may perform a notification process of notifying an employee of the abnormality. In the notification process, the inspection target area determined to be abnormal may be displayed on the captured image of the railway vehicle in a distinguishable manner to notify the inspection target area.

When it is determined that the inspection process is normal or abnormal, the inspection processing apparatus 20 determines whether or not the processes for all the inspection target areas are completed (step S12). If not yet, the process returns to step S6 to continue the loop process. On the other hand, if the processing is completed, the loop processing is exited and the processing moves to the next step. Through the loop processing of steps S6 to S12 described above, the presence/absence of abnormality of the railway vehicle is determined for a plurality of inspection target areas in the captured image.

After exiting the loop processing, the inspection processing apparatus 20 determines whether the results of image comparison of all the inspection target areas are normal (step S13), and if not, terminates this inspection processing. On the other hand, if all are normal, the inspection processing device 20 determines whether the captured image satisfies the registration condition in the normal image database 24B (step S14), and if satisfied, the captured image together with the vehicle type information in the normal image database. It is registered in 24B (step S15). The captured image to be registered may be only the image after the correction process or both images before and after the correction process. By accumulating the picked-up images determined to be normal, the number of normal images in the normal image database 24B is increased by stacking the inspection times. As a result, various methods for determining the presence/absence of abnormality can be incorporated, and inspection with higher accuracy can be performed.

As the registration condition of the captured image, for example, the condition that the maximum value or the average value of the distance values calculated by the image comparison in step S8 is within a normal range, or the captured image of the same vehicle model stored in the normal image database 24B Various conditions may be applied, such as the condition that the number has not reached the upper limit.
When the normal image is registered, the inspection processing device 20 ends the inspection process of FIG. If it is determined in step S14 that the registration condition is not satisfied, the inspection process of FIG.

Subsequently, a plurality of correction processes executed in steps S2 to S5 of FIG. 4 will be described in detail.
<Rail distortion correction processing>
FIG. 5 is a diagram illustrating an example of rail distortion correction processing executed in step S2 of FIG. FIG. 5A shows a photographed image before correction, and FIG. 5B shows a photographed image after correction. Note that, in FIGS. 5A and 5B, the distortion of the aspect ratio due to the speed variation of the railway vehicle is omitted.

The rail distortion correction processing is processing for correcting the shape distortion of the captured image due to the sinking of the rail 90 when the rail vehicle is passing. The ground on which the rail 90 is laid has a strong portion and a relatively weak portion. Therefore, when the railcar travels on the rail 90, the rail 90 sinks due to the weight of the railcar, and the railcar is slightly displaced vertically. When the railway vehicle moves up and down during shooting, the captured image of the railway vehicle is distorted. The shape distortion due to the sinking of the rail 90 mainly occurs in the portion of the carriage 81 directly supported by the rail 90.

The line scan camera 13 performs a line scan so as to include the display surface 17a of the reference device 17. Therefore, as shown in FIG. 5A, the captured image of the railway vehicle includes an image of the display surface 17a that is reflected as if it was stretched from the start end to the end. Further, since the reference device 17 is fixed to the rail 90, when the rail 90 is depressed, the reference device 17 is similarly depressed. Therefore, the shape distortion due to the sinking of the rail 90 appears as the distortion of the line of the reference unit 17. Therefore, the correction of this shape distortion is performed by shifting the pixel row of 1 pixel in the vertical direction in the line scan direction so that the line indicating the reference point of the display surface 17a becomes a straight line from the start end to the end of the captured image. To be achieved. Subsequently, a specific processing procedure will be described.

When shifting to the rail distortion correction process, the inspection processing device 20 first performs an edge extraction process for extracting an edge of the display surface 17a of the reference device 17 from the captured image. Since the display surface 17a has a portion where the brightness changes stepwise, the edge of this portion can be extracted with high sensitivity and high accuracy.
Next, the inspection processing device 20 sets a reference straight line L0 (FIG. 5A) having a constant height in the captured image, and a pixel line having a width of 1 pixel, which is long in the vertical direction, and the edge portion of the display surface 17a is the reference straight line. Slide it so that it overlaps L0. The inspection processing apparatus 20 shifts such pixel rows from the start end to the end of the captured image. This completes the rail distortion correction process.

As shown in FIG. 5B, in the captured image after the correction, the lines of the rail and the display surface 17a of the reference unit 17 are corrected to be linear, and along with this, the shape of the portion of the bogie 81 of the railway vehicle is mainly formed. The distortion is corrected. In FIG. 5B, the shape distortion remains in the vehicle body 80. This is because a suspension such as an air spring is interposed between the carriage 81 and the vehicle body 80, and It is caused by relative vertical vibration. This shape distortion is corrected by the correction processing described later.

<Brightness correction>
FIG. 6 is an explanatory diagram showing an example of the brightness correction processing executed in step S3 of FIG.
The brightness correction process is a process of correcting variations in the brightness of a captured image due to the brightness of the environment at the time of shooting. During the visual inspection of the railway vehicle, the inspection processing device 20 compares the photographed image and the normal image for each pixel and calculates the correlation value to determine whether there is an abnormality. Therefore, it is preferable that the brightness variation of the captured image is small. Even if the appearance inspection is performed by comparing the captured image with the normal image by edge extraction and edge comparison, the brightness correction of the captured image is effective in order to suppress variations in sensitivity and accuracy of edge extraction. ..

The inspection processing device 20 performs brightness correction using the image of the display surface 17a of the reference device 17 reflected in the captured image. As shown in FIG. 6A, the inspection processing device 20 holds, as control data, the brightness value of the display surface 17a obtained when the shooting environment has standard brightness. Since the display surface 17a displays the luminosity of a plurality of gradations, the inspection processing apparatus 20 holds the luminance values of gradations 1 to 5 respectively corresponding to these.

When shifting to the brightness correction processing, as shown in FIG. 6B, the inspection processing device 20 first obtains the brightness values of gradation 1 to gradation 5 of the display surface 17a in the captured image. The photographed image is represented in gray scale, and white from maximum brightness to black at minimum brightness is divided into 256 levels. The maximum brightness value is "255" and the minimum brightness value is "0". Next, as shown in FIG. 6C, luminance values (solid line graph line) of gradation 1 to gradation 5 in the captured image and gradation 1 to gradation obtained with standard brightness. 5 is compared with the luminance value of 5 (broken line in the graph) to obtain a correction gamma curve as shown in FIG. Next, the inspection processing device 20 develops the gamma curve into a look-up table and corrects the brightness of all pixels of the captured image according to the look-up table. This completes the brightness correction process.

<Speed expansion/contraction correction processing>
FIG. 7 is a flowchart showing a detailed procedure of the speed expansion/contraction correction processing executed in step S4 of FIG. 8A and 8B are views for explaining an example of the speed expansion/contraction correction processing. FIG. 8A is an image view of a reference profile showing a reference shape of a characteristic portion of a railway vehicle, and FIG. 8B is a captured image view before correction. (C) is a photographed image diagram after correction.
The speed expansion/contraction correction processing is processing for correcting the shape distortion of the railway vehicle due to the variation in the moving speed of the railway vehicle. If the moving speed of the railway vehicle is high, the aspect ratio of the railway vehicle in the captured image becomes relatively small in the horizontal direction, and if the moving speed of the railway vehicle is slow, the horizontal ratio becomes relatively large. If the moving speed of the railcar changes during shooting, the aspect ratio of the railcar changes within one captured image.

As shown in the flowchart of FIG. 7, when the process proceeds to the speed expansion/contraction correction process, first, the inspection processing device 20 extracts an edge of the captured image to extract a geometrical characteristic portion of the railway vehicle (step S21: characteristic portion). (Corresponding to the extracting means), for example, the profile line P1 of the railway vehicle is created from the extracted characteristic portion (step S22). As shown in FIG. 8(b), the profile line P1 is a continuous representation of the characteristics of the shape of the railcar from the beginning to the end, and is not particularly limited, but as an example, a predetermined edge extracted feature portion is specified. It is possible to apply a line that connects the points of.
Subsequently, the inspection processing device 20 divides the captured image into a plurality of image elements for each rectangular region f that is long in the line scan direction from the start end to the end (step S23: corresponds to a stretchable image dividing unit). Hereinafter, this image element will be referred to as a “strip image element”. The rectangular area f may be, for example, an area of 5 to 10 horizontal pixels×all vertical pixels.

After the division, the inspection processing device 20 compares the profile line P1 with the reference profile line RP for each strip image element, and calculates the deviation of the aspect ratio of the strip image element from the difference in the inclination degree of the profile line P1 (step S24: Corresponds to aspect ratio calculation means). The reference profile line RP is a profile line that is obtained from image data of a railway vehicle without distortion that serves as a comparison reference. The inspection processing device 20 previously uses, as control data, image data of a railroad vehicle without distortion, which serves as a comparison reference, and data of a reference profile line RP representing characteristics of the shape of the railroad vehicle, as shown in FIG. 8A. Holding The image data of the comparison reference and the reference profile line RP are held for each vehicle model of the railway vehicle.

After calculating the deviation of the aspect ratio, the inspection processing apparatus 20 expands/contracts the strip image element in the horizontal direction so as to eliminate the deviation (step S25: corresponds to image expanding/contracting means). The expansion/contraction processing is performed by thinning out a pixel row having a width of one pixel in the line scan direction or adding a pixel row. At this time, the image is made continuous between a pair of adjacent pixel rows. Interpolation processing may be added.
After performing the expansion/contraction process of the strip image elements of the entire area f, the inspection processing apparatus 20 performs a combining process of sequentially attaching the strip image elements after the expansion/contraction process (step S26). This completes the speed expansion/contraction correction processing. As a result, as shown in FIG. 8C, a captured image of the railway vehicle having a uniform aspect ratio can be obtained.

<Vehicle distortion correction processing>
FIG. 9 is a flowchart showing the detailed procedure of the vehicle body distortion correction process executed in step S5 of FIG. 10A and 10B are diagrams illustrating an example of the vehicle body distortion correction process. FIG. 10A is a photographed image diagram before correction, FIG. 10B is a mask image diagram, and FIG. 10C is a shift amount of image elements divided into rectangular shapes. 2D is an explanatory diagram of the detection processing of FIG.

The vehicle body distortion correction processing is processing for correcting the shape distortion of the captured image due to the vibration generated in the vehicle body 80 (see FIG. 10A) of the railway vehicle by the action of the suspension such as the air spring. Since the suspension is provided between the bogie 81 of the railway vehicle and the vehicle body 80, the shape distortion of the photographed image due to the action of the suspension does not appear on the bogie 81 or the rail portion, and the vehicle body 80 and the lower portion of the vehicle body 80 are not affected. It appears on the part of the attachment 82 that is fixedly connected.
When the process shifts to the vehicle body distortion correction process, the inspection processing device 20 first performs a mask process on the captured image using the mask image m1 in FIG. 10B (step S31: corresponding to a mask unit). The mask image m1 is provided in advance so as to hide the carriage 81 and the portion below the rail 90, and is held in the inspection processing device 20 as control data.

Then, the inspection processing apparatus 20 divides the masked photographed image into rectangular regions that are long in the line scan direction to generate a plurality of image elements (step S32: corresponding to a distorted image dividing unit). The area to be divided is similar to the area f shown in FIG. 8B, and hereinafter, the image element generated by the division is referred to as a “strip image element”.
After dividing the captured image into strip image elements, the inspection processing apparatus 20 executes a loop process (steps S33 to S37) in which the process is repeated one by one for each of the plurality of divided strip image elements. When shifting to the loop processing, the inspection processing device 20 selects one strip image element (step S33), and cuts out a rectangular image g2 of vertical h pixels×horizontal w pixels as shown in FIG. 10C (step S33). S34), the cut-out rectangular image g2 is compared with the rectangular image Rg of the same position of the comparison reference railcar image (step S35). The rectangular area g2 may be selected so as to include the boundary line L2 between the masked portion and the non-masked portion of the mask image m1. Since this rectangular image g2 includes, for example, an edge indicating the lower edge of the vehicle body 80, by comparing this edge with the rectangular images g2 and Rg, the deviation amount of the rectangular image g2 from the comparison reference is calculated. To be done.

Next, as shown in FIG. 10D, the inspection processing apparatus 20 fixes the mask portion of one strip image element based on the boundary line L2 of the mask portion of the mask image m1 and only the non-mask portion g1. The shift amount is moved (step S36: equivalent to image correction means). At the time of this movement, there is a shortage of pixels at the upper or lower end of the moving area, but the inspection processing device 20 compensates for the shortage of pixels by performing interpolation processing such as filling in the pixels of the same brightness as the shadow portion.
After performing the process of correcting the deviation amount for one strip image element, the inspection processing apparatus 20 determines whether the process has been completed for all strip image elements (step S37). If not, the process proceeds to step S33. Return and continue loop processing. By repeating the loop processing, the correction processing of all the strip image elements in the captured image is achieved. When the processing is completed, the loop processing is exited and the processing proceeds to the next step.

When moving to the next step, the inspection processing apparatus 20 performs a combining process so that all the processed strip image elements are attached in order (step S38), and further releases the mask by the mask image m1 (step S39: Equivalent to the mask release means). Then, the vehicle body distortion correction process ends. Through such processing, a photographed image of the railway vehicle in which the shape distortion of the vehicle body 80 and the accessory 82 below the vehicle body is corrected is obtained.

As described above, according to the generation function of the captured image of the railway vehicle of the inspection processing apparatus 20 of the present embodiment, first, the line scan camera 13 has high resolution, and when the line scanning camera 13 is seen from a certain direction at any position. A captured image of the railroad vehicle whose shape is shown can be obtained. Furthermore, by the correction function of the captured image (steps S2 to S5 in FIG. 4), the captured image is distorted from the captured image due to the sinking of the rails, and the vertical vibration of the vehicle body due to the action of the suspension captures the image. The distortion of the shape of the image and the distortion of the shape of the captured image due to the moving speed of the railway vehicle are corrected, and further, the variation in the brightness of the captured image due to the brightness of the capturing environment can be suppressed. Therefore, it is possible to provide a photographed image of a railway vehicle that is suitable for visual inspection without irregular distortion.

Further, according to the function of generating the captured image of the railway vehicle of the inspection processing apparatus 20 of the present embodiment, first, the rail distortion correction processing (step S2 in FIG. 4) is performed to correct the entire area of the image in the vertical direction. Therefore, in the velocity expansion/contraction correction process (step S4 in FIG. 4), the aspect ratio of each part in the captured image can be easily and accurately calculated. Therefore, the distortion of the shape of the captured image caused by the moving speed of the railway vehicle can be corrected neatly. Further, the vehicle body distortion correction processing (step S5 in FIG. 4) is performed after the vertical distortion correction and the horizontal distortion correction are performed on the entire area of the captured image by the rail distortion correction processing and the speed expansion/contraction correction processing. It is possible to easily and accurately separate the vehicle body and the carriage from the captured image. As a result, the distortion of the image caused by the vertical vibration of the vehicle body due to the suspension such as the air spring can be corrected more beautifully.

Although the embodiments of the present invention have been described above, the present invention is not limited to the above embodiments. For example, in the above-described embodiment, the method of capturing the left and right side surfaces of the railway vehicle to generate the captured image for the visual inspection has been described. However, for example, the upper portion of the railway vehicle is captured obliquely downward, or the lower portion of the railway vehicle is captured. You may generate|occur|produce image|photographed diagonally upward and perform the visual inspection of a vehicle upper part or a vehicle lower part, and produce|generate the imaged image. Further, in the above embodiment, an example of a specific procedure of each correction process has been described, but the details of the correction process can be appropriately changed without departing from the spirit of the invention.

Further, in the above embodiment, an example in which the normal image database 24B and the abnormal image database 24C are provided in the storage device 24 in the inspection processing device 20 has been shown. However, the normal image database 24B and the abnormal image database 24C may be provided in different storage devices and connected to the inspection processing apparatus 20 (computer) by, for example, a LAN.
Further, in the above-described embodiment, the example in which the inspection processing device that performs the image correction process and the abnormality detection is arranged apart from the imaging unit has been described. However, the inspection processing device may be arranged beside the line similarly to the photographing means. In that case, the function of the image processing device (15) is incorporated in the inspection processing device (20) to integrally configure both. You may.

1 Appearance inspection device 11 Vehicle proximity detection device 13 Line scan camera (imaging means)
14 background plate 15 image processing device 17 reference device 17a display surface 20 inspection processing device (distortion correction means, rail distortion correction means, vehicle body distortion correction means, speed expansion/contraction correction means, brightness correction means)
21 CPU
22 RAM
23 interface 24 storage device 24A appearance inspection program f rectangular area RP reference profile line P1 profile line m1 mask image

Claims (5)

A line scan type photographing means for obtaining a photographed image by vertically scanning a railroad vehicle moving on a track,
A reference device having a display surface showing a reference position in the up-down direction and fixed to the railroad track at a position overlapping with a shooting location of the shooting means;
Distortion correcting means for correcting the distortion of the shape of the photographed image caused by the vertical movement of the railway vehicle,
The distortion correction means,
Rail distortion correction means for shifting the photographed image vertically for each part so that a line indicating the reference position continuously reflected in the photographed image becomes linear.
Of the photographed image, except for the image of the region where the bogie of the railway vehicle and the rail are imaged, the image of the region where the body of the railway vehicle and the accessories fixedly connected to the vehicle body are imaged Vehicle body distortion correction means for shifting in the vertical direction,
An image generation apparatus for inspection, comprising: The vehicle body distortion correction means,
Of the captured image, a mask means for hiding a region in which a bogie and a rail of a railway vehicle are imaged, and extracting an image of a region in which a vehicle body of a railway vehicle and an accessory fixedly connected to the vehicle body are imaged,
A distorted image division unit that divides the image extracted by the mask unit into rectangular image elements that are long in the vertical direction;
Image correction means for correcting the vertical shift by comparing the rectangular image element with the reference image of the railroad vehicle,
A mask releasing means for restoring the image of the area hidden by the masking means to the image corrected by the image correcting means,
The inspection image generating apparatus according to claim 1, further comprising: Further comprising speed expansion/contraction correction means for correcting the variation in the aspect ratio of each part of the captured image due to the moving speed of the railway vehicle,
The speed expansion/contraction correction means,
Characteristic portion extracting means for extracting the characteristic portion of the railway vehicle by performing edge extraction of the photographed image,
A stretchable image dividing unit that divides the captured image into rectangular image elements that are long in the vertical direction,
Aspect ratio calculation means for comparing the reference profile showing the reference shape of the characteristic portion of the railway vehicle with the characteristic portion extracted by the characteristic portion extraction means to calculate the deviation of the aspect ratio of each part of the captured image. When,
An image expansion/contraction means for expanding/contracting the image elements divided by the expansion/contraction image division means in the horizontal direction so that the deviation of the aspect ratio calculated by the aspect ratio calculation means is eliminated.
The inspection image generating apparatus according to claim 1 or 2, further comprising: The inspection image generation apparatus according to claim 3, wherein the rail image distortion correction unit, the speed expansion/contraction correction unit, and the vehicle body distortion correction unit perform the correction process of the captured image in this order. The display surface of the reference unit further includes a lightness display unit showing lightness of a plurality of gradations,
The brightness correction unit that corrects the brightness of the captured image based on the brightness of a plurality of gradations of the brightness display unit included in the captured image, further comprising: a brightness correction unit. The image generation device for inspection according to [4].

Images