JP2024049413A - Line abnormality detection system and line abnormality detection method - Google Patents

Abstract

[Problem] To provide a wire abnormality detection system that detects wires with higher accuracy and in a shorter time. [Solution] A wire abnormality detection system 10 is composed of line sensor cameras 11a, 11b, a memory unit, an electric train line fitting detection unit, a wire detection unit, and a wire abnormality detection unit. The line sensor cameras 11a, 11b photograph wires (suspension wires 4, trolley wires 5, hangers 6) from above the vehicle roof 2 of a moving vehicle 1. The lane fitting detection unit detects fittings (hangers 6) attached to the wires from the image data. The wire detection unit calculates a likelihood indicating the degree to which fittings can be connected in a straight line, and detects the wires (suspension wires 4, trolley wires 5) from the image data based on the likelihood. The wire abnormality detection unit detects abnormalities in the wires (suspension wires 4, trolley wires 5) from the wire detection information and the wire image. [Selected Figure] Figure 1

Description

The present invention relates to a line abnormality detection system and a line abnormality detection method.

Conventionally, maintenance and inspection work has been carried out on overhead electric railway equipment (trolley wires, suspension wires, and auxiliary suspension wires; hereafter referred to as "lines"), with inspectors visually checking for fraying or breaks in the twisted wires that make up the lines. However, there are limitations to visual inspection, and there is a risk of abnormalities being overlooked.

Patent Document 1 discloses a technology for detecting contact wires from line sensor camera images. Patent Document 2 discloses a technology for combining line sensor camera images and range lasers to extract overhead and suspended wires and perform three-dimensional measurements.

JP 2009-040215 A JP 2015-017882 A

However, in Patent Document 1, the continuity of the overhead lines is evaluated by edge detection processing of images captured by a line sensor camera, but there is a problem in that image processing cannot reliably detect the wires.

In addition, in Patent Document 2, multiple image processing steps must be performed on the image from the line sensor, which results in heavy processing due to the need for three-dimensional measurement, and the processing time is therefore long.

The present invention aims to provide a line abnormality detection system and a line abnormality detection method that can detect lines with higher accuracy and in a shorter time.

In order to solve the above problems, the present invention employs the following means.
In other words, the wire abnormality detection system of the present invention is a wire abnormality detection system that detects abnormalities in wires including at least suspension wires and trolley wires used for operating railway vehicles, and is characterized by comprising an imaging unit that photographs the wires and metal fittings attached to the wires, a trolley wire fitting detection unit that detects the metal fittings from image data photographed by the imaging unit, a wire detection unit that detects an image area including the wires from the image data based on the positions of the detected metal fittings and on a likelihood indicating the degree to which the metal fittings can be connected in a straight line, and a wire abnormality detection unit that detects abnormalities in the wires from the image area including the wires.

The wire abnormality detection method of the present invention is a wire abnormality detection method for detecting an abnormality in a wire including at least a suspension wire and a contact wire for the operation of a railway vehicle, and is characterized by including the steps of photographing the wire and metal fittings attached to the wire with an imaging unit, detecting the metal fittings from image data photographed by the imaging unit, detecting an image area including the wire from the image data based on a likelihood indicating the degree to which the metal fittings can be connected in a straight line based on the position of the detected metal fittings, and detecting an abnormality in the wire from the image area including the wire.

This invention makes it possible to detect lines more reliably and in a shorter time.

1 is a perspective view showing an electric railway facility in which a wire abnormality detection system 10 according to a first embodiment of the present invention is used. FIG. 2 is a block diagram showing a functional configuration of the line abnormality detection system 10 according to the first embodiment. 4 is a flowchart for explaining the operation of the line abnormality detection system 10 according to the first embodiment. 2 is a conceptual diagram for explaining a line detection process by the line abnormality detection system 10 of the first embodiment. FIG. FIG. 2 is a conceptual diagram for explaining a method for calculating a joint likelihood of main parts by the line abnormality detection system 10 according to the first embodiment. 10 is a conceptual diagram for explaining a method for calculating a joint likelihood of a remainder rectangle by the line abnormality detection system 10 according to the first embodiment. FIG. 1 is a conceptual diagram for explaining rectangular metal fitting connection information obtained by the line abnormality detection system 10 according to the first embodiment. FIG. 1 is a conceptual diagram for explaining a method for allocating line rectangles by the line abnormality detection system 10 according to the first embodiment. FIG. FIG. 11 is a perspective view showing an electric railway facility in which a wire abnormality detection system 10 according to a second embodiment of the present invention is used. FIG. 11 is a block diagram showing a functional configuration of a line abnormality detection system 10 according to a second embodiment. 10 is a flowchart for explaining the operation of the line abnormality detection system 10 according to the second embodiment. FIG. 11 is a schematic diagram showing a metal fitting rectangle of a compound section in which the line abnormality detection system 10 according to the second embodiment is used. FIG. 11 is a perspective view showing an electric railway facility in which a wire abnormality detection system 10 according to a third embodiment of the present invention is used. FIG. 11 is a block diagram showing a functional configuration of a line abnormality detection system 10 according to a third embodiment. 10 is a flowchart for explaining the operation of the line abnormality detection system 10 according to the third embodiment. FIG. 13 is a conceptual diagram for explaining a joint likelihood calculation method using laser data by the line anomaly detection system 10 according to the third embodiment.

Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.
The object of the present invention is to enable automatic line detection in images acquired from a camera mounted on the roof of a railway vehicle, and to reduce the labor required for maintenance and inspection work on line by performing abnormality detection on the line-detected images in subsequent processing.

A. First embodiment Fig. 1 is a perspective view showing electric railway equipment in which a wire abnormality detection system 10 according to a first embodiment of the present invention is used. In Fig. 1, a vehicle 1 is traveling in the direction of travel indicated by the arrow. In the electric railway equipment, overhead poles 3 are arranged at predetermined intervals along the track on which the vehicle 1 travels, and a suspension wire 4 is suspended from the overhead poles 3. The suspension wire 4 supports a contact wire 5 by a hanger 6 (metal fitting).

In this first embodiment, the wire abnormality detection system 10 is composed of two line sensor cameras (imaging units) 11a, 11b installed on the vehicle roof 2 and an analysis device 12. The line sensor cameras 11a, 11b are installed so that their imaging surfaces are in the direction of the sleepers. In other words, the line sensor cameras 11a, 11b photograph the wires (suspension wire 4, trolley wire 5) and hangers 6 (metal fittings) from diagonally downward directions on the left and right sides of the vehicle. The reason two line sensor cameras 11a, 11b are used is to take images from the left and right directions in case there is a problem such as not being able to take images for some reason, but the wire abnormality detection system 10 can function with just one camera.

Note that Figure 1 only shows one wire (main line; suspension wire 4, contact wire 5), but in reality, there may also be wires such as crossover wires that cross the main line diagonally. In the following processing, we assume that there are wires such as the crossover wires included in addition to the main line wires, and explain this as a "remainder rectangle."

Figure 2 is a block diagram showing the functional configuration of the wire abnormality detection system 10 according to the first embodiment. As described above, the wire abnormality detection system 10 is composed of line sensor cameras 11a, 11b and an analysis device 12. The analysis device 12 is composed of a memory unit 13, an electric train line metal fitting detection unit 14, a wire detection unit 15, and a wire abnormality detection unit 16. Image data captured by the line sensor cameras 11a, 11b is stored in the memory unit 13. The line sensor cameras 11a, 11b continuously capture images, and from then on, analysis processing is performed with each space between overhead poles as one unit.

The overhead contact wire fittings detection unit 14 detects fittings (hangers 6) from the image data stored in the memory unit 13, and stores the detected fittings in the memory unit 13 as fittings detection information. The wire detection unit 15 detects wires (suspension wires 4, trolley wires 5) from the fittings detection information stored in the memory unit 13, and stores the detected wires in the memory unit 13 as wire detection information. The wire abnormality detection unit 16 detects abnormalities in the wires (suspension wires 4, trolley wires 5) from the wire detection image assigned based on the wire detection information stored in the memory unit 13, and stores the abnormality determination (result) in the memory unit 13.

Figure 3 is a flowchart for explaining the operation of the wire abnormality detection system 10 according to the first embodiment. First, the electric train line fitting detection unit 14 detects images of fittings (hangers 6) by performing electric train line fitting detection processing on the image data 20 between the overhead poles stored in the memory unit 13, and obtains fitting detection information 21 that links the fitting detection position (xmin, ymin, xmax, ymax) with a label name (step S10). Note that the method of electric train line fitting detection processing is not particularly limited. For example, it may be an object detection model using deep learning such as SSD, YOLO, Faster R-CNN, M2Det, etc.

Here, FIG. 4 is a conceptual diagram for explaining the wire detection process by the wire abnormality detection system 10 of the first embodiment. As shown in the figure, the image data is one unit per pole. In the case of the image data shown in FIG. 4, the case where two wires (crossing wires) are photographed is shown. The thick line in the figure is the hanger 6 (metal fitting) that suspends the trolley wire 5. In the wire detection process, all hangers 6 (metal fittings) in the image are detected, a metal fitting rectangle 6a that surrounds each hanger 6 (metal fitting) is set, and the coordinates of the metal fitting rectangle 6a are obtained as the metal fitting detection position (xmin, ymin, xmax, ymax). In addition, in the wire detection process, a label name is set to indicate which metal fitting each hanger 6 (metal fitting) corresponds to and to distinguish which metal fitting in the image it is. In this case, the label name is "hanger i" (i = 0, 1, ..., n).

Next, the wire detection unit 15 receives metal fitting detection information 21 that links the metal fitting detection positions (xmin, ymin, xmax, ymax) and label names of all hangers 6 (metal fittings) included between the overhead poles, connects metal fitting rectangles 6a that it has determined to be on the same overhead line (metal fitting rectangle 6a joining process: steps S14, S16, S18 below), and finally assigns a wire rectangle (wire rectangle assignment process: step S20 below), thereby detecting wires and obtaining wire detection information 22 (step S12).

To explain in more detail, the line detection unit 15 first registers the "hanger" label rectangle for the metal fitting rectangle 6a from the metal fitting detection information 21 as a metal fitting rectangle to be used for line detection (step S14). Next, the line detection unit 15 combines the metal fitting rectangles of the main parts (straight parts) (step S16). Here, the combination of the metal fitting rectangles of the main parts (straight parts) will be explained.

FIG. 5 is a conceptual diagram for explaining a method of calculating the likelihood of joining main parts by the line abnormality detection system 10 according to the first embodiment. In the joining of main parts (linear parts), the likelihood of joining is calculated from the positional relationship between each of the joining candidates. As shown in FIG. 5, when the metal fitting rectangle "0" and the metal fitting rectangle "1" are considered as joining candidates, the likelihood of joining is calculated using three parameters, namely, the deviation distance D _x indicating the distance of the deviation of the center of gravity of the metal fitting rectangle, the distance in the line direction D _y indicating the distance between the metal fitting rectangles, and the angle θ indicating the angle of the deviation of the center of gravity, and the deviation distance likelihood L _Dx , the distance likelihood in the line direction L _Dy , and the angle likelihood L _θ are calculated by weighting and adding up each likelihood from the following formulas (1) to (4). Since the main line (traveling line) is depicted linearly in the vertical direction of the image, the elements of the deviation distance D _x and the distance D _y in the line direction are emphasized. Then, the rectangles with the maximum total likelihood L are joined and registered as the joining source (parent rectangle) and the joining destination (child rectangle). Therefore, the smaller the deviation distance _Dx , the closer the distance _Dy in the line direction is to the specified installation interval, and the closer the angle θ is to 0, the larger the total likelihood becomes.

Next, the line detection unit 15 combines the excess rectangles other than the main part (metal fitting rectangles other than the main part) (step S18). Here, we will explain how to combine the excess rectangles (diagonal parts) with metal fitting rectangles.

Fig. 6 is a conceptual diagram for explaining a method for calculating the likelihood of joining of the remainder rectangle by the line anomaly detection system 10 according to the first embodiment. In joining of the remainder rectangle, the joining likelihood is calculated from the positional relationship of the target fitting rectangle (the fitting rectangle "1" in Fig. 6, the fitting rectangle "0" above it, and the fitting rectangle "2" below it) among the fitting rectangles that were not joined as the main part. As shown in Fig. 6, the joining likelihood is calculated using three parameters: the line direction distance D _y1 indicating the distance between the fitting rectangle "0" and the fitting rectangle "1", the line direction distance D _y2 indicating the distance between the fitting rectangle "1" and the fitting rectangle "2", and the differential angle θ d indicating the angle formed by the line connecting the centers of gravity of the fitting rectangle "0" and the fitting rectangle "1" and the line connecting the centers of gravity of the fitting rectangle "1 _" and the fitting rectangle "2". The differential angle likelihood L _θd , the line direction distance likelihood L _Dy2 , and the line direction distance likelihood L _Dy1 are weighted and summed according to the following formulas (5) to (8). The remaining rectangles are not linearly distributed in the vertical direction of the image, and are considered to be lines that are the above-mentioned crossovers or the like photographed at an angle, so the element of the difference angle _θd is emphasized.Then, the metal fitting rectangle with the largest total likelihood L is registered as the connection source (parent rectangle) and connection destination (child rectangle).Therefore, the smaller the difference angle _θd is, and the closer the line direction distance _Dy1 and the line direction distance _Dy2 are to the specified installation interval, the greater the total likelihood will be.

Figure 7 is a conceptual diagram for explaining the metal fitting rectangle connection information by the wire anomaly detection system 10 according to the first embodiment. By combining the metal fitting rectangles of the main parts (straight parts) and the remainder rectangles (metal fitting rectangles) described above, as shown in Figure 7, metal fitting rectangles "0", "2", "4", "6", and "8" are combined as the main parts, and metal fitting rectangles "1", "3", "5", "7", and "9" are combined as remainder rectangles. The combination of the main parts is referred to as combined group G1, and the combination of the remainder rectangles is referred to as combined group G2.

Next, the line detection unit 15 performs an allocation process to allocate line rectangles containing lines from the combined metal fitting rectangles (step S20). More specifically, the allocation process of line rectangles is performed for each combined group based on the combination information of the metal fitting rectangles. The allocation of line rectangles to combined group G1 shown in FIG. 7 is described below.

Figures 8(a) and (b) are conceptual diagrams for explaining a method for allocating stripe rectangles by the stripe anomaly detection system 10 according to the first embodiment. As shown in Figure 8(a), areas are divided using the rectangle height position as a boundary, and an approximation formula passing through the coordinates of the center of gravity of the rectangle is found for each area. For the topmost and bottommost areas, an approximation formula for a linear function is found from the coordinates of the nearest pair of rectangles. For the other areas, an approximation formula for a quadratic function is found from the coordinates of the three previous and following rectangles.

Next, as shown in Figure 8(b), the line rectangle size is determined using the larger of the two metal fitting rectangles closest to the top of the image coordinate system, and the center coordinates of the line rectangle are set to the approximate formula found in Figure 8(a), and line rectangles (square dashed rectangles) are assigned starting from the top. The final line rectangle is adjusted by adding an amount that makes it difficult to cut. The position of the next rectangle assignment line can also be adjusted by [rectangle width x shift rate (0 to 1)].

Finally, the line abnormality detection unit 16 performs a line abnormality detection process by inputting the line detection image assigned based on the line detection information 22, performing anomaly detection to detect abnormalities in the lines, and finally outputting an anomaly determination result (step S22). There are no particular limitations on the method of anomaly detection here, but it is possible to use, for example, image recognition using a deep learning object recognition model such as VAE, GAN, ResNET, DeepSAD, or CS-FLOW.

According to the first embodiment described above, the electric train wire fitting detection unit 14 detects fittings attached to the wires from images acquired by the line sensor cameras 11a and 11b mounted on the vehicle roof 2 of the vehicle 1, the wire detection unit 15 detects the wires from the fitting detection information, and the wire abnormality detection unit 16 detects wire abnormalities from the wire detection information and the wire image. This makes it possible to detect wires with higher accuracy and in a shorter time, thereby realizing labor savings in maintenance and inspection work for wires.

In addition, according to the first embodiment described above, the line detection unit 15 calculates the likelihood that metal fittings can be connected in a straight line, and detects lines from image data based on the likelihood. This makes it possible to detect not only the main line, but also lines that intersect with the main line with greater accuracy and in a shorter time, thereby realizing labor savings in maintenance and inspection work for lines.

B. Second Embodiment Next, a second embodiment of the present invention will be described.
9 is a perspective view showing an electric railway facility in which a wire anomaly detection system 10 according to a second embodiment of the present invention is used. The second embodiment is a process for a compound catenary type section. As shown in FIG. 9, the compound catenary type is a system in which an auxiliary catenary 7 is added between a suspension wire 4 and a contact wire 5, the suspension wire 4 supports the auxiliary catenary 7 with a dropper 8, and the auxiliary catenary 7 supports the contact wire 5 with a hanger 6.

Figure 10 is a block diagram showing the functional configuration of the line abnormality detection system 10 according to the second embodiment. Note that parts corresponding to the functional configuration of the line abnormality detection system 10 according to the first embodiment shown in Figure 1 are given the same reference numerals and will not be described. In the second embodiment, a metal fitting identification unit 17 is newly provided. The metal fitting identification unit 17 acquires metal fitting detection information that identifies the above-mentioned hanger 6 and dropper 8 as different metal fittings. Therefore, the line detection unit 15 detects lines for each metal fitting rectangle based on the metal fitting detection information identified for each metal fitting.

Figure 11 is a flowchart for explaining the operation of the line abnormality detection system 10 according to the second embodiment. The operation of the line abnormality detection system 10 according to the second embodiment is basically the same as that of the first embodiment, but the metal fitting identification unit 17 adds an identification process based on the type of metal fitting, and the line detection unit 15 performs a line detection process according to the type of metal fitting.

First, the electric train line fitting detection unit 14 detects images of the fittings (hanger 6, dropper 8) by performing electric train line fitting detection processing on the image data 20 between the overhead poles stored in the memory unit 13, and obtains fitting detection information 21 that links the fitting detection position (xmin, ymin, xmax, ymax) with the label name (step S30). Next, the fitting identification unit 17 obtains fitting detection information 24 that identifies the hanger 6 and dropper 8 as different fittings from the fitting detection information 21 (step S32). Since the hanger 6 and dropper 8 are different sizes, the width of the fitting rectangle is used as information to identify the hanger 6 and dropper 8.

Figure 12 is a schematic diagram showing metal fitting rectangles in a compound section in which the wire anomaly detection system 10 according to the second embodiment is used. As shown in Figure 12, the rectangle width of the hanger 6 rectangle and the metal fitting rectangle of the dropper 8 are significantly different. There are no particular limitations on the classification method, but for example, two-class classification based on the metal fitting rectangle width size using k-means may be performed.

Next, the wire detection unit 15 receives the metal fitting detection information 24, which is linked to the metal fitting detection positions and label names of all metal fittings (hangers 6 and droppers 8) included between the overhead poles, connects the metal fitting rectangles that are determined to be on the same overhead line, and finally performs wire detection by assigning a wire rectangle, and obtains the wire detection information 25 (step S34). Steps S36 to S42 shown in the figure are similar to steps S14 to S20 in the first embodiment, except that there are different metal fittings (hangers 6 and droppers 8) that have been identified, so their explanation is omitted. In the metal fitting rectangle combining process, only rectangles with the same label (same metal fittings) are considered as candidates for combining, which makes it possible to improve the accuracy of wire detection in the compound catenary section. Therefore, in this second embodiment, the wire detection information 25 is obtained as the wire rectangle obtained by combining hangers 6 and the wire rectangle obtained by combining droppers 8.

Finally, the line abnormality detection unit 16 performs a line abnormality detection process by inputting a line detection image assigned based on line detection information 25 including a line rectangle obtained by connecting hangers 6 together and a line rectangle obtained by connecting droppers 8 together, performing anomaly detection to detect line abnormalities, and finally outputting an anomaly determination result 26 (step S44). As with the first embodiment, the method of anomaly detection here is not particularly limited, but for example, image recognition using an object recognition model of deep learning may be used.

In the second embodiment described above, the metal fittings identification unit 17 is provided, but this is not limiting. The electric train line metal fittings detection unit 14 detects and labels the hangers 6 and droppers 8 as metal fittings ("hanger" and "dropper"), and the wire detection unit 15 registers the hangers 6 and droppers 8 according to the labels. In this way, the electric train line metal fittings detection unit 14 and the wire detection unit 15 may share the functions of the metal fittings identification unit 17.

According to the second embodiment described above, in a compound catenary section, metal fittings attached to the wires are detected from images acquired by the line sensor cameras 11a and 11b mounted on the vehicle roof 2 of the vehicle 1, different metal fittings are classified, the wires are detected from the metal fitting detection information for each classified metal fitting, and abnormalities in the wires are detected from the wire detection information and the wire images. This makes it possible to detect the suspension wires, auxiliary suspension wires, and contact wires used in the compound catenary system with greater accuracy and in a shorter time, thereby realizing labor-saving maintenance and inspection work for the wires.

In addition, according to the first embodiment described above, the likelihood indicating the degree to which metal fittings can be connected in a straight line is calculated, and the wires are detected from the image data based on the likelihood. This makes it possible to detect the suspension wires, auxiliary suspension wires, and contact wires used in the compound catenary system with greater accuracy and in a shorter time, not only for the main line but also for wires that cross the main line, thereby realizing labor-saving maintenance and inspection work for the wires.

C. Third Embodiment Next, a third embodiment of the present invention will be described.
Fig. 13 is a perspective view showing an electric railway facility in which a wire anomaly detection system 10 according to a third embodiment of the present invention is used. Note that parts corresponding to those in Fig. 1 are given the same reference numerals and their explanations are omitted. In the second embodiment, in addition to the line sensor cameras 11a and 11b, a range laser 18 is additionally disposed on the vehicle roof 2. The range laser 18 acquires the distance to the wires (suspension wires 4 and trolley wires 5) and metal fittings (hangers 6) and the elevation angle relative to the wires (suspension wires 4 and trolley wires 5) and metal fittings (hangers 6) as laser data.

Figure 14 is a block diagram showing the functional configuration of the wire abnormality detection system 10 according to the third embodiment. Note that parts corresponding to those in Figure 2 are given the same reference numerals and their explanations are omitted. The range laser 18 acquires the distance to the wires (suspension wire 4 and trolley wire 5) and the elevation angle relative to the wires (suspension wire 4 and trolley wire 5) as laser data and stores them in the memory unit 13. Therefore, the wire detection unit 15 detects the wires for each metal fitting rectangle based on the metal fitting detection information classified for each metal fitting and the laser data.

Figure 15 is a flowchart for explaining the operation of the line abnormality detection system 10 according to the third embodiment. The operation of the line abnormality detection system 10 according to the third embodiment is basically the same as that of the first embodiment, but the processing in the line detection unit 15 is line detection processing using metal fitting detection information and laser data.

First, the electric train line fitting detection unit 14 detects images of fittings (hangers 6) by performing an electric train line fitting detection process on the image data 20 between the overhead poles stored in the memory unit 13, and obtains fitting detection information 21 that links the fitting detection position with the label name (step S50).

Next, the wire detection unit 15 receives the metal fitting detection information 21, which links the metal fitting detection positions and label names of all metal fittings (hangers 6) included between the overhead poles, and the laser data 30 acquired by the range laser 18, and connects the metal fitting rectangles that it has determined to be on the same overhead line, and finally performs wire detection by assigning wire rectangles, and acquires wire detection information 25 (step S52).

In the third embodiment, steps S54 to S60 shown in the figure are similar to steps S14 to S20 in the first embodiment, but we will explain a method of calculating the likelihood using laser data 30 that is different from that in the first embodiment.

16(a) to (d) are conceptual diagrams for explaining a method of calculating a likelihood of combination using the laser data 30 by the line anomaly detection system 10 according to the third embodiment. In the main part of the combination in step 56, a likelihood term using the laser data 30 is added. As shown in FIG. 16(a) to (d), the likelihood using the laser data 30 is calculated in the following manner. First, as shown in FIG. 16(a), it is confirmed that a laser point exists in the metal fitting rectangle "0" that is the combination source (if it does not exist, the likelihood is 0), and as shown in FIG. 16(b), a linear approximation equation is calculated between the metal fitting rectangle "1" that is the combination candidate. Next, as shown in FIG. 16(c), a laser angle θ _l based on the linear approximation equation and a rectangle angle θ based on a straight line connecting the center of gravity points of the metal fitting rectangles "0" and "1" are calculated, and as shown in FIG. 16(d), a difference angle θ _sl between the laser angle θ _l and the rectangle angle θ is calculated. Then, a laser likelihood _L1 indicating the degree of deviation between a straight line connecting the centers of gravity of the metal fitting rectangle and a straight line expressed by a linear approximation equation by the laser is calculated from the following equation (9).

The laser likelihood _L1 is calculated for all channels of the input laser data 30, and the highest total likelihood L is used (lasers are managed by channel by tracking the same overhead line). The total value obtained by weighting and adding up the laser likelihood _L1 , the deviation distance likelihood _Ldx of the first embodiment, the distance likelihood _Ldy in the line direction, and the angle likelihood _Lθ is the total likelihood L with the candidate rectangle to be combined (Equation (10)). Since more input information is input than in the first embodiment, more accurate line detection is possible. Similarly, line detection is performed using the laser data 30 for combining the remaining rectangles in step S58. In addition, the same process as in the first and second embodiments is performed for the allocation of line rectangles in step S60.

Finally, the line abnormality detection unit 16 performs a line abnormality detection process by inputting the line detection image assigned based on the line detection information 31, performing anomaly detection to detect abnormalities in the lines, and finally outputting an anomaly determination result 32 (step S64). As with the first and second embodiments described above, the method of anomaly detection here is not particularly limited, but for example, image recognition using an object recognition model of deep learning may be used.

According to the third embodiment described above, metal fittings attached to the lines are detected from images acquired by the line sensor cameras 11a and 11b mounted on the vehicle roof 2 of the vehicle 1, and a likelihood indicating the degree to which the metal fittings can be connected in a straight line is calculated from the metal fitting detection information and the laser data 30 acquired by the range laser 18. The lines are detected from the image data based on the likelihood, and abnormalities in the lines are detected from the line detection information and the line image. This makes it possible to detect the lines with greater accuracy and in a shorter time, thereby realizing labor savings in maintenance and inspection work for the lines.

REFERENCE SIGNS LIST 1 Vehicle 2 Vehicle roof 3 Catenary pole 4 Suspension wire 5 Trolley wire 6 Hanger 7 Auxiliary suspension wire 8 Dropper 10 Wire abnormality detection system 11a, 11b Line sensor camera 12 Analysis device 13 Memory unit 14 Catenary wire fitting detection unit 15 Wire detection unit 16 Wire abnormality detection unit 17 Fitting identification unit 18 Range laser 20 Image data 21 Fitting detection information 22, 25, 31 Wire detection information 23, 26, 32 Abnormality determination result 24 Fitting detection information (identified)
30 Laser Data

Claims (7)

A wire abnormality detection system that detects an abnormality in a wire including at least a suspension wire and a contact wire for operating a railway vehicle,
An imaging unit that captures an image of the wire and a metal fitting attached to the wire;
a train line fitting detection unit that detects the fittings from image data captured by the imaging unit;
a line detection unit that detects an image area including the line from the image data based on a likelihood indicating a degree to which the metal fittings are likely to be connected to each other in a straight line based on the detected positions of the metal fittings;
a line abnormality detection unit that detects the line abnormality from an image region including the line;
A line abnormality detection system comprising: The wire includes an auxiliary suspension wire in addition to the suspension wire and the contact wire, and the metal fitting includes at least a hanger that supports the contact wire and a dropper that supports the auxiliary suspension wire,
A metal fitting identification unit is further provided for identifying the hanger and the dropper from the image data.
the line detection unit detects an image area including the line from the image data by distinguishing between the hanger and the dropper;
2. The line abnormality detection system according to claim 1 . the line detection unit detects an image area including the line from the image data based on a total likelihood of a deviation distance likelihood indicating a deviation of the centers of gravity of at least two metal fittings included in the image data, a distance likelihood in a line direction corresponding to the distance between the two metal fittings, and an angle likelihood corresponding to an angle of the deviation of the centers of gravity of the two metal fittings.
3. The line abnormality detection system according to claim 1 or 2. the line detection unit detects an image area including the line from the image data based on a total likelihood obtained by summing a distance likelihood in a line direction corresponding to a distance between a first fitting and a second fitting, a distance likelihood in a line direction corresponding to a distance between the second fitting and a third fitting, and a differential angle likelihood corresponding to an angle formed by a line connecting the centers of gravity of the first fitting and the second fitting and a line connecting the centers of gravity of the second fitting and the third fitting, among at least three fittings included in the image data;
2. The line abnormality detection system according to claim 1 . A range finder unit that obtains distances and elevation angles to the wire and the metal fitting,
the line detection unit detects an image area including the line from the image data based on a likelihood indicating a degree to which the metal fittings are likely to be joined in a straight line, based on the position of the detected metal fitting and the distance and elevation angle acquired by the range measurement laser unit;
2. The line abnormality detection system according to claim 1 . The line detection unit includes:
an image area including the line is detected from the image data based on a total likelihood obtained by adding up a laser likelihood corresponding to the degree of deviation between an angle formed by a straight line connecting the centers of gravity of at least two metal fittings included in the image data and an angle formed by a laser light from the range measurement laser unit, a deviation distance likelihood indicating the deviation of the centers of gravity of the at least two metal fittings, a distance likelihood in a line direction corresponding to the distance between the at least two metal fittings, and an angle likelihood corresponding to the angle of the deviation of the centers of gravity of the at least two metal fittings;
6. The line abnormality detection system according to claim 5. A wire abnormality detection method for detecting an abnormality in a wire including at least a suspension wire and a contact wire for operation of a railway vehicle, comprising:
Photographing the wire and the metal fittings attached to the wire with an imaging unit;
Detecting the metal fitting from image data captured by the imaging unit;
detecting an image area including the line from the image data based on a likelihood indicating a degree to which the metal fittings are likely to be connected to each other in a straight line based on the detected positions of the metal fittings;
detecting an abnormality of the streak from an image region including the streak;
A line abnormality detection method comprising:

Images