Track inspection device equipped with a relaxation determination system

Abstract

An object to be inspected having a plurality of fasteners is imaged as video data, the fasteners are detected, and a fastener loosening determination system is provided that can more quickly determine the loosening of the fasteners. 【Solution means】A system M for determining the loosening of fasteners at the fastening part of a structure, comprising: an imaging means 1 for photographing the structure to be inspected; a fastener detection means 21 for detecting fasteners from the video data photographed by the imaging means 1; and a loosening determination means 22 for determining the loosening state of the fasteners detected by the fastener detection means 21. The fastener detection means 21 uses an AI (Artificial Intelligence) that has been trained to detect fasteners using reference video data photographed in the same field of view as the video data in advance, detects fasteners from the video data, and generates still image data recording the fasteners. The loosening determination means 22 determines the loosening state of the fasteners by comparing with reference still image data obtained by imaging in advance a state where the fastening by the fasteners is in good condition.

Description

Technical Field

[0001] The present invention relates to a relaxation determination system for a fastener that fastens at least two or more members to each other, a track inspection device equipped with the relaxation determination system, and a track inspection method.

Background Art

[0002] Conventionally, in order to determine the fastening state (whether there is looseness) of fasteners such as bolts and nuts used for fixing at least two or more members to each other, it is generally known to mark a mark called a mating mark on the fastener in advance.

[0003] Specifically, a straight line is marked with a marker or the like so as to straddle the fastener and the structure fastened by the fastener. When the fastener loosens, a deviation occurs between the straight line marked on the fastener and the straight line marked on the structure. By visually checking the deviation of this mating mark, it is determined that the fastener is loose.

[0004] The relaxation determination method using the mating mark has the advantage that an operator can visually determine the relaxation state of the fastener and can easily perform the inspection. However, when inspecting a structure with a large number of fastening parts, it is necessary to determine the relaxation state of the fasteners by a large number of personnel, which is a drawback in that it requires a great deal of time and labor.

[0005] Therefore, Patent Document 1 and Patent Document 2 disclose a system for determining the mating mark marked on a fastener by image processing. Both Patent Document 1 and Patent Document 2 are provided with imaging means, and compare the image of the fastener taken by this imaging means with the fastener in a state where the fastening state is good recorded in advance, and determine the relaxation state from the change in the angle of the mating mark.

Prior Art Documents

Patent Documents

[0006]

Patent Document 1

[0007] The relaxation determination systems described in Patent Document 1 and Patent Document 2 can save the labor of visually checking each fastener one by one by determining the relaxation state of the fastener through image processing, and can obtain a uniform determination result without being affected by the judgment of the operator.

[0008] However, since the relaxation determination systems described in Patent Document 1 and Patent Document 2 perform a comparison with a pre-stored image of a good fastening state, it is necessary to photograph each fastener to be detected by an imaging means one by one. Therefore, when the number of fasteners to be detected is large, it is too laborious to perform the photographing with a single imaging means, or the high cost due to providing a plurality of imaging means corresponding to each fastener is cited as a problem.

[0009] In addition, since the relaxation determination systems described in Patent Document 1 and Patent Document 2 have the fastener as the detection target, only the fastener is photographed by the imaging means and the relaxation determination is performed by image processing, so it is impossible to detect other defects of the structure provided with the fastener.

[0010] The present invention has been made in view of the above problems, and photographs an inspection target having a plurality of fasteners as moving image data, and detects the fasteners from the moving image data to more quickly determine the relaxation of the fasteners, and when the inspection target has a defect other than the relaxation of the fasteners, it is possible to detect the defect based on the previously captured moving image data. A relaxation determination system for fasteners mu equipped with a track inspection device place is provided for the purpose. [Means for Solving the Problems]

[0011] In order to solve the above-mentioned conventional problems, An inspection device (V) for a track, which is a vehicle that travels automatically on the track and is equipped with a loosening determination system for a fastener that fixes the rails constituting the track of the train. The inspection device for the track includes a base frame (31) made of a metal angle material or a pipe material, a base plate (32) laid on the base frame 31, axles (33) provided at the front and rear of the base frame respectively, wheels (34) connected to both ends of the two axles, and support columns (361, which extend upward and rearward obliquely from the vicinity of the front left and right wheels.361), a mounting part (362) installed on the upper part of the support part, a handle part (37) for performing operation provided on the mounting part, on the base plate of the carriage part (3) composed of these, an electric motor (4) for driving the wheels, a control part (5) for controlling the electric motor, an information processing terminal (2) functioning as the fastening tool detection means and the relaxation determination means described later, and a seat part (8) for an operator to board are provided. Rail imaging means (11) fixed in front of and above the left and right two wheels constituting the front wheels of the carriage part and capable of photographing the fastening tools to be inspected and the periphery of the track as moving image data as the vehicle moves forward from above the rail, and a rail illumination (61) arranged above the rail imaging means and illuminating downward, forward imaging means (12) fixed to the mounting part with the lens facing the traveling direction and capable of photographing the front of the traveling as moving image data, and a front illumination (62) arranged beside the forward imaging means and illuminating the front, and a terminal device held by the operator. The relaxation determination system is configured by these, and the relaxation determination system uses an AI (Artificial Intelligence) that has learned to detect the fastening tools from reference moving image data photographed in the same viewing angle as the moving image data in advance by the rail imaging means and the forward imaging means, and is capable of executing a fastening tool extraction process for detecting the fastening tools from the moving image data photographed by the imaging means and generating still image data recording the fastening tools. The relaxation state of the fastening tools is determined by comparing the still image data with reference still image data photographed in a state where the fastening by the fastening tools is in good condition. When relaxation is recognized, relaxation information (A) is generated and transmitted to the terminal device to notify the operator. The fastening tool detection means also uses an AI that has learned to detect different points by comparing the moving image data with the reference moving image data, and detects defects generated in the periphery of the track and the front part of the traveling determined by comparing the fastening tools and the reference moving image data from the moving image data obtained by the rail imaging means and the forward imaging means, and is further capable of executing a defect detection process for transmitting the defect information (B) to the terminal device to notify the operator. It is characterized by the following.

Advantages of the Invention

[0014] According to the present invention, a plurality of fasteners provided in a structure to be inspected are photographed as moving image data by an imaging means, and still image data in which the fasteners are recorded is generated and detected using AI from the acquired moving image data, and the loosening determination of the detected fasteners is performed by image processing, so that the loosening determination of the fasteners can be performed quickly.

[0015] Further, when detecting a fastener from moving image data and extracting it as still image data, AI compares it with reference moving image data in which the detection of the fastener has been learned in advance. Therefore, the different parts of the two moving image data determined by comparing the moving image data at the time of inspection with the reference moving image data, that is, the differences can be recognized and detected by AI as defects.

Brief Description of the Drawings

[0016]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Embodiment for Carrying Out the Invention

[0017] The gist of this invention is a system for determining the loosening of fasteners at the fastening parts of a structure, which comprises imaging means for photographing the structure to be inspected, fastener detection means for detecting fasteners from the moving image data photographed by the imaging means, and loosening determination means for determining the loosening state of the fasteners detected by the fastener detection means. The fastener detection means uses an AI (Artificial Intelligence) that has been trained to detect the fasteners from the moving image data using reference moving image data photographed at the same angle of view as the moving image data in advance, generates still image data recording the fasteners, and the loosening determination means determines the loosening state of the fasteners by comparing the still image data with reference still image data that has previously imaged a state where the fastening by the fasteners is in good condition.

[0018] Hereinafter, with reference to the accompanying drawings, an embodiment of a fastener loosening determination system M, an orbit inspection device V provided with the fastener loosening determination system M, and an orbit inspection method P according to the present invention will be described. The following embodiments are an example of embodying the present invention and do not limit the technical scope of the present invention.

[0019] [1. About the fastener loosening determination system] First, the fastener loosening determination system M according to the present invention will be described with reference to the drawings.

[0020] As shown in FIG. 1, the fastener loosening determination system M according to this embodiment mainly includes imaging means 1 for photographing the structure to be inspected as moving image data D1, fastener detection means 21 for detecting fasteners VN from the moving image data D1 photographed by the imaging means 1 and generating still image data G1, and loosening determination means 22 for determining the loosening state of the fasteners VN from the still image data G1 generated by the fastener detection means 21.

[0021] The structure to be inspected includes a plurality of fasteners VN. For example, it includes buildings equipped with a plurality of high-strength bolts, or train tracks that have bolts and nuts for connecting rails. In this embodiment, the fastener VN is exemplified by being composed of bolts and nuts. However, the fastener relaxation determination system M according to the present invention is not limited to bolts and nuts as the objects of relaxation determination, and any object that can be visually confirmed (judgment is possible by image processing) and whose abnormal conditions such as relaxation state and detachment can be judged can be the object of determination. Further, the fastener VN and the structure fastened by the fastener VN are marked with a straight alignment mark PM straddling the fastener VN and the rail TR as shown in FIGS. 3 and 4 in advance, taking the track TO as an example.

[0022] The imaging means 1 is composed of an imaging device such as a video camera equipped with an imaging element such as a CCD (Charge Coupled Device). The imaging means 1 photographs a plurality of fasteners VN provided in the structure in a determined order. The photographed moving image data D1 is transmitted to the fastener detection means 21.

[0023] The fastener detection means 21 performs a process of extracting and generating each fastener VN recorded in the moving image data D1 photographed by the imaging means 1 as still image data G1. This fastener detection means 21 uses AI (Artificial Intelligence) that has learned fastener detection. Specifically, the shape characteristics of the fastener VN are learned in advance by reference moving image data D2 obtained by photographing the fasteners VN provided in the structure. Then, a process is performed to extract and write out, as still image data G1, those that match the shape characteristics of the fastener VN learned from the moving image data D1 photographed by the imaging means 1 during inspection.

[0024] More specifically, as shown in FIG. 2, the fastener detection means 21 includes a video receiving unit 211 that receives the video data D1 transmitted from the imaging means 1, a reference video data storage unit 212 that stores the reference video data D2 previously captured by the imaging means 1, a fastener extraction unit 213 that detects the fastener VN from the video data D1 and extracts and generates it as still image data G1, and a still image transmission unit 214 that transmits the generated still image data G1 to the relaxation determination means 22.

[0025] The video receiving unit 211 receives the video data transmitted from the imaging means 1. The video data to be received includes the reference video data D2 captured by the imaging means 1 in advance at the same viewing angle as during the inspection for the fastener VN of the structure, and the video data D1 captured by the imaging means 1 during the inspection.

[0026] The reference video data storage unit 212 stores the reference video data D2 among the video data received by the video receiving unit 211. Also, although it will be described in detail later, since the video data D1 captured during the inspection is also used for AI learning, the video data D1 captured during the inspection is also stored in this reference video data storage unit 212 after the inspection.

[0027] The fastener extraction unit 213 can mainly perform four processes: a reference video data learning process 213a, a fastener extraction process 213b, a defect detection process 213c, and a learning data storage process 213d by the learning AI.

[0028] The reference video data learning process 213a is a process of learning the shape features of the fastener VN provided in the structure captured by the imaging means 1 from the reference video data D2 to correctly recognize the fastener VN, and recognizing the parts other than the fastener VN (for example, the design surface of the structure, etc.) recorded in the reference video data D2 as normal states.

[0029] The AI that constitutes the fastener detection means 21 can extract and generate the fastener VN as still image data G1 from the video data D1 captured during the inspection and transmitted in real time by the fastener extraction process 213b, which will be described next, according to the learning content of this reference video data learning process 213a. Also, by learning the content recorded in the reference video data D2, it becomes possible to recognize and process the differences as defects in the defect detection process 231c by comparing with the video data D1 captured during the inspection and transmitted in real time.

[0030] The fastener extraction process 213b is a process of extracting and generating only the fastener VN as still image data G1 from the video data D1 received by the video receiving unit 211 when actually performing the inspection. This extracted and generated still image data G1 is used to determine whether the fastener VN reflected in the still image data G1 is loose by comparing it with the reference still image data G2 through image processing, which will be described in detail later.

[0031] The defect detection process 213c is a process of detecting defects by comparing the video data D1 received by the video receiving unit 211 when actually performing the inspection with the normal state of the structure learned in advance from the reference video data D2. Specifically, if the normal state of the design surface of the structure is being learned, damage such as scratches and dents on the design surface can be recognized as a difference from the reference video data D2, and the difference can be detected as a defect.

[0032] The defects detected by the defect detection process 213c are transmitted as defect information B to the terminals held by the workers performing the inspection or the terminals held by the workers observing the inspection work remotely.

[0033] The learning data storage process 213d is a process of storing the video data D1 for which an inspection has been performed and its results together with the reference video data D2 for use in AI learning. In AI learning, the higher the amount of data for learning, the higher the accuracy of the results. Therefore, by storing the video data D1 and defect information B used in each inspection and using the stored information as the learning target by the reference video data learning process 213a, the accuracy of the fastener extraction process 213b and the defect detection process 213c can be improved. This video data D1 is stored by the reference video data storage unit 212.

[0034] The still image transmission unit 214 is for transmitting the still image data G1 generated from the video data D1 by the fastener extraction process 213b of the fastener extraction unit 213 to the relaxation determination means 22.

[0035] As shown in FIGS. 3 and 4, the relaxation determination means 22 compares the still image data G1 received from the fastener detection means 21 with the reference still image data G2 of the fastener VN in a state where the fastening by the fastener VN stored in the relaxation determination means 22 in advance is good, that is, a state where there is no deviation in the mating mark PM, and determines the relaxation state of the fastener VN provided in the inspection target.

[0036] As shown in FIG. 1, the relaxation determination means 22 is composed of a still image reception unit 221, a reference still image storage unit 222, an image comparison unit 223, and a relaxation notification unit 224.

[0037] The still image data G1 transmitted by the still image transmission unit 214 of the fastener detection means 21 is received by the still image reception unit 221 and transmitted to the image comparison unit 223 of the relaxation determination means 22.

[0038] The reference still image memory unit 222 stores reference still image data G2 in a state where the fastening by the fastener VN is in good condition. The state where the fastening is in good condition means that, as shown in FIGS. 3 and 4, the alignment marks PM previously marked on the structure so as to straddle the fastener VN appear in a straight line without shifting. This reference still image data G2 is a previously captured image of the fastener VN. For example, the still image data G1 of the fastener VN generated by the fastener detection means 21 or an image captured by another imaging means not included in the configuration of the fastener relaxation determination system M is used.

[0039] The image comparison unit 223 compares the still image data G1 of the fastener VN generated by the fastener detection means 21 with the reference still image data G2 when performing an inspection to determine the relaxation state of the fastener VN. The fastener VN for which relaxation is recognized as a result of the comparison transmits the information as relaxation information A to the relaxation notification unit 224. The relaxation information A transmitted to the relaxation notification unit 224 is information such as the still image data G1 of the fastener VN for which relaxation is recognized and which number of the still image data G1 the still image data G1 was determined by the image comparison unit 223.

[0040] The relaxation notification unit 224 notifies the operator who is performing the relaxation inspection of the fastener VN of the relaxation information A received from the image comparison unit 223. The notification of the relaxation information is transmitted to the terminal device held by the operator performing the inspection or the terminal device held by the operator observing the inspection work from a distance.

[0041] The fastener detection means 21 and the relaxation determination means 22 are provided, for example, in an information processing terminal 2 such as a PC having a configuration in which a CPU (Central Processing Unit) as an arithmetic processing device that executes various arithmetic processes and controls, a storage device such as a RAM (Random Access Memory) and a ROM (Read Only Memory), an input / output device such as an input / output interface for data input / output, and peripheral circuits such as a clock circuit are connected by a bus or the like.

[0042] [Regarding the track inspection device equipped with the fastener relaxation determination system] Next, a track inspection device V equipped with a fastening member relaxation determination system M used when the structure to be inspected is a track TO for train running will be described with reference to the drawings. In the present embodiment, the track inspection device V described below is self-propelled by including an electric motor 4, but it is not necessarily required to be self-propelled. For example, it may be configured to travel on the track TO by being towed by a vehicle capable of running and perform inspections.

[0043] As shown in FIG. 6, the track inspection device V mainly includes a carriage unit 3 having wheels 34 connected to an axle 33, an electric motor 4 that outputs a rotational driving force to the axle 33 through a drive transmission means, a control unit 5 that electrically controls the electric motor 4, a plurality of imaging means 1 provided on the carriage unit 3, an illumination unit 6 disposed near each imaging means 1, and a terminal protection unit 7 that houses an information processing terminal 2 inside that receives the moving image data D1 captured by the imaging means 1 and performs a relaxation determination of the fastening member VN, and a seat unit 8 for an operator to board.

[0044] The carriage unit 3 is composed of a base frame 31 made of metal angle members or pipe materials, a base plate 32 laid on the base frame 31, axles 33 provided at the front and rear of the base frame 31, and wheels 34 connected to both ends of each of the two axles 33. That is, the carriage unit 3 is a four-wheel vehicle having two structures of axles 33 and wheels 34 connected to both ends of the axles 33 in the front and rear. Hereinafter, the axle 33 and the wheel 34 located in front of the carriage 4 are referred to as the front wheel FW, and the axle 33 and the wheel 34 located in the rear are referred to as the rear wheel RW.

[0045] Each of the four wheels 34 is provided with a wheel detachment prevention portion 45 having a diameter larger than that of the wheel 34 body on the carriage 4 side. The wheel detachment prevention portion 45 can prevent the detachment of the wheel 34 body rotating on the rail TR by being located inside the two rails TR constituting the track TO when the carriage 4 travels on the track TO.

[0046] A front frame 36 extending upward is provided on the front side of the base frame 31. Specifically, it consists of a support column portion 361 extending upward and rearward from the vicinity of the left and right wheels 34 constituting the front wheels FW, a ladder-shaped installation portion 362 installed on the upper part of the support column portion 361, and an auxiliary column portion 363 extending downward from the installation portion 362 and fixed to the base plate. Further, a handle portion 37 for performing operations such as speed adjustment and stop of the trolley 4 is provided on the installation portion 362.

[0047] The electric motor 4 is installed on the base plate 32 of the trolley 4 and transmits the output rotational driving force to at least one of the front wheels FW and the rear wheels RW through driving transmission means (not shown). Due to this rotational driving force, the wheels 34 rotate and the trolley 4 can run by itself. The electric motor 4 is supplied with power by a power supply unit (not shown). This power supply unit may use, for example, a storage battery.

[0048] The control unit 5 is installed on the base plate and controls the start and stop of the electric motor 4, the speed adjustment of the trolley, etc. Also, it is connected through the handle portion 37 and electrical wiring (not shown), and the operator can perform operations such as speed adjustment and braking at hand.

[0049] The imaging means 1 consists of a rail imaging means 11 for imaging the fasteners VN connecting the rails TR and a front imaging means 12 for imaging the traveling direction of the trolley 4.

[0050] As shown in FIG. 6, the rail imaging means 11 is fixed in front of and above the left and right two wheels 34 constituting the front wheels FW by a support arm 38 connected to the base frame 31. That is, the rail imaging means 11 is arranged at a position where it can image the rail TR on which the wheel 34 body travels from above.

[0051] The front imaging means 12 is provided on the installation portion 362 of the front frame 36 and is fixed so that the imaging lens faces the traveling direction of the trolley 4.

[0052] The video data D1 captured by the track imaging means 11 and the front imaging means 12 is transmitted to the information processing terminal 2. When transmitting this video, the information processing terminal 2 and each imaging means 1 may be connected by electrical wiring, or may be connected by wireless communication.

[0053] The lighting unit 6 is disposed in the vicinity of each imaging means 1. Specifically, it consists of a track lighting 61 disposed above the track imaging means 11 and illuminating downward, and a front lighting 62 disposed beside the front imaging means 12 and illuminating forward.

[0054] Incidentally, the inspection of the train track TO is carried out at night after the train operation has ended. Therefore, it is necessary to ensure illumination during night work. Thus, the track TR photographed by the track lighting 61 is illuminated to surely perform the photography by the track imaging means 11, and the front of the carriage 4 is illuminated by the front lighting 62 so that the work can be carried out safely.

[0055] The terminal protection part 7 is in the shape of a rectangular box and is installed on the base plate 32. It houses the information processing terminal 2 equipped with the fastener detection means 21 and the relaxation determination means 22 inside, and can protect the information processing terminal 2 from impacts, rain, etc. Also, the information processing terminal 2 housed inside can transmit the information from the relaxation determination means 22 to an external terminal (a terminal held by an operator or a terminal operated by a supervisor monitoring the work) by wireless communication.

[0056] Furthermore, the information processing terminal 2 is equipped with a GPS (Global Positioning System) function, so that in the inspection of the long track TO to be inspected, the current position can always be notified to the external terminal. This position information can also be collated with the video data D1 obtained from the front imaging means 12.

[0057] [3. About the method of inspecting the track] Next, a method P for inspecting a track for train running using the track inspection device V will be described with reference to the drawings. As shown in FIGS. 3 to 5, the track TO to be inspected is previously marked with a mating mark PM for determining the relaxation of the fastener VN at the fastening part.

[0058] The track inspection method P is roughly divided into two steps: an AI learning step S1 that takes pictures of the track TO to be inspected in advance and learns the detection of fasteners VN and the state without abnormalities of the track TO, and a track inspection step S2 that actually uses the track inspection device V to travel on the track TO and inspects whether there is any loosening of the fasteners VN of the track TO or any abnormalities around the track TO.

[0059] As shown in FIG. 7, the AI learning step S1 is a preparatory step to be carried out before inspection, and consists of a reference video data shooting step S1-1, a reference video data storage step S1-2, and a fastener detection learning step S1-3.

[0060] The reference video data shooting step S1-1 is a step of shooting the track TO to be inspected in advance. Specifically, the track inspection device V is installed on the track TO to be inspected, and an operator rides on it and travels the same route as the actual inspection. At this time, the two rail imaging means 11 mounted on the track inspection device V take pictures from above the two rails TR. At this time, the rail imaging means 11 is arranged to take pictures of the rail TR from above by the track inspection device V, and when traveling the same route, the shooting angle can inevitably be the same each time. The shooting data taken by the rail imaging means 11 is transmitted as reference video data D2 to the fastener detection means 21 of the information processing terminal 2 and stored in the reference video data storage unit 212.

[0061] In addition, the front imaging means 12 may perform imaging simultaneously for the purpose of verification work with the position information obtained from the GPS function provided in the information processing terminal 2. In this case, the imaging data captured by the front imaging means 12 is transmitted to the functional part having the GPS function of the information processing terminal 2, and the position information obtained from the GPS is associated with the imaging data captured by the front imaging means 12, which can be used to identify the location where the fastener VN is loose or the location where a defect is found during inspection. Also, the front imaging means 12 can obtain normal state imaging data by transmitting it to the fastener detection means 21 in the same manner as the imaging data captured by the track imaging means 11, and can be used for detecting defects in the subsequent defect detection process 213c. At this time, the imaging data obtained from the front imaging means 12 is transmitted to the fastener detection means 21 as the reference video data D2 together with the imaging data of the track imaging means 11 and stored in the reference video data storage unit 212.

[0062] The reference video data storage step S1-2 is a step of storing the reference video data D2 transmitted from the imaging means 1 by the fastener detection means 21 of the information processing terminal 2. Specifically, the video receiving unit 211 of the fastener detection means 21 receives the reference video data D2 transmitted from the imaging means 1. The received reference video data D2 is stored by the reference video data storage unit 212.

[0063] The fastener detection learning step S1-3 is a step of performing learning for detecting the fastener VN from the reference video data D2 stored in the reference video data storage unit 212. Specifically, it is performed by causing the learning AI constituting the fastener detection means 21 to recognize the shape of the fastener VN.

[0064] The fastener detection learning process S1-3 is performed by the fastener extraction unit 213 of the fastener detection means 21, and this process becomes the reference video data learning process 213a described above. That is, it is a process of teaching the learning AI to correctly detect the fastener VN photographed in the reference video data D2 by recognizing the shape of the fastener VN and extract and generate it as the still image data G1. At this time, the generated still image data G1 may be transmitted to the relaxation determination means 22 as the reference still image data G2 used when performing relaxation determination in a later process. At this time, the reference still image data G2 transmitted to the relaxation determination means 22 is image data in a state where the mating mark PM previously marked on the fastening part appears in a straight line without shifting, as shown in FIGS. 3 and 4.

[0065] Also, at the same time as detecting the fastener VN, the state around the rail TR, which is a part other than the fastener VN recorded in the reference video data D2, is also taught to the AI as a normal state. Also, when the imaging data captured by the front imaging means 12 is included in the reference video data D2, the AI is taught from this imaging data as a normal state in front of the track inspection device V. By this learning, the AI can recognize the normal state that serves as the reference for defect detection in the defect detection process 213c.

[0066] By finishing the AI learning process S1 described above, the track TO to be inspected in advance is photographed, and the process of preparing for inspection to learn the detection of the fastener VN and the non-abnormal state of the track TO is completed. Next, the track inspection process S2 will be described, in which the track inspection device V is actually used to travel on the track TO and inspect whether there is any relaxation of the fastener VN on the track TO or any abnormality around the track TO.

[0067] As shown in FIG. 8, the track inspection process S2 consists of five processes: a video data acquisition process S2-1, a still image data generation process S2-2, a fastener relaxation determination process S2-3, a defect detection process S2-4, and a relaxation and defect notification process S2-5.

[0068] The video data acquisition step S2-1 is a step of acquiring video data D1 obtained by photographing the rail TR and the front of the travel while the track inspection device V travels on the track TO to be inspected by the imaging means 1. Specifically, the track inspection device V is installed on the track TO to be inspected, and an operator rides on it and travels on the track TO. While traveling, a video obtained by photographing the rail TR by the rail imaging means 11 and a video obtained by photographing the front of the travel and the inspection state by the front imaging means 12 are acquired as the video data D1.

[0069] At this time, the route for inspection is determined in advance, and it is performed so as to coincide with the route in the reference data photographing step S1-1. By doing so, it becomes possible to compare with the normal state learned by the AI from the photographed data taken in the reference video data photographing step S1-1, and in the inspection of the track TO, not only the loosening determination of the fastener VN but also the defects near the rail TR and the travel route can be detected.

[0070] Also, in the video data acquisition step S2-1, the front imaging means 12 also simultaneously photographs the front during inspection. The photographed data taken by the front imaging means 12 is stored as a recorded video during inspection or used for defect detection in the defect detection step S2-4 described later in the same manner as the video data D1.

[0071] The video data D1 photographed by the imaging means 1 is transmitted in real time to the information processing terminal 2 mounted on the track inspection device V. The information processing terminal 2 sequentially processes the received video data D1 in the next step.

[0072] The still image data generation step S2-2 is a step of detecting only the fastener VN from the video data D1 acquired in the video data acquisition step S2-1 and generating still image data G1. Specifically, the video data D1 transmitted from the rail imaging means 11 is received by the video receiving unit 211 of the fastener detection means 21 and sent to the fastener extraction unit 213. The fastener extraction unit 213 detects the fastener VN from the intermittently transmitted video data D1 and generates still image data G1 so as to cut out the detected fastener VN.

[0073] The process of generating this still image data G1 is performed as a fastener extraction process 213b by the fastener extraction unit 213 of the fastener detection means 21. That is, the fastener detection means 21 has learned in advance by the reference video data learning process 213a to determine whether or not it is a fastener VN from the shape, so that the fastener VN can be detected in real time from the video data D1 intermittently received from the rail imaging means 11, and the still image data G1 can be extracted and generated. The generated still image data G1 is transmitted to the relaxation determination means 22.

[0074] The fastener relaxation determination step S2-3 is a step of comparing the reference still image data G2 stored in advance in the reference still image storage unit 222 of the relaxation determination means 22 with the still image data G1 generated in the still image data generation step S2-2 to determine whether or not relaxation is recognized in the fastener VN. This process is performed by the image comparison unit 223 of the relaxation determination means 22.

[0075] As shown in FIGS. 3 and 4, the reference still image data G2 is image data in which the mating mark PM marked on the structure so as to straddle the fastener VN appears as a straight line. The state in which relaxation of the fastener VN is recognized compared with this reference still image data G2 is a state in which a deviation has occurred in the mating mark PM due to the rotation of the fastener VN, as shown in FIG. 5.

[0076] As shown in FIG. 8, the defect detection step S2-4 is a step of detecting defects occurring in the track TO recorded in the video data D1 and its surroundings and the front part of the travel by comparing the pre-stored reference video data D2 with the video data D1 acquired in the video data acquisition step S2-1. Specifically, the AI compares the reference video data D2 captured by the rail imaging means 11 and the front imaging means 12 in the reference video data capturing step S1-1 with the video data D1 captured by the rail imaging means 11 and the front imaging means 12 in the video data acquisition step S2-1. When a difference is recognized in the comparison, the location where the difference occurs is detected as a defect. The defect detection step S2-4 is executed as a defect detection process 213c by the fastener extraction unit 213 of the fastener detection means 21.

[0077] In the defect detection step S2-4, any object can be compared with the reference video data D2 as long as the comparison is made between the video captured at the same viewing angle as the video learned by the pre-trained AI as a reference. Defects can be detected by the comparison.

[0078] In the relaxation and defect notification step S2-5, when the relaxation of the fastener VN is recognized in the fastener relaxation determination step S2-3, the relaxation information A generated and the defect information B obtained in the defect detection step S2-4 are transmitted to the terminal possessed by the worker for notification. The relaxation information A is transmitted to the terminal possessed by the worker by the relaxation notification unit 224 of the relaxation determination means 22. The defect information B is transmitted to the terminal possessed by the worker by the fastener detection means 21.

[0079] The track inspection method P uses the track inspection device V equipped with the fastener relaxation determination system M, so that the inspection work of the track TO, which was conventionally visually inspected by workers for the deviation of the alignment mark PM in a human wave tactic, can be quickly performed while riding and driving on a vehicle capable of traveling on the track TO. That is, it is possible to shorten the time related to the inspection and eliminate human errors.

[0080] In addition, by providing the fastener detection means 21 with a learning AI, the fastener VN can be detected in real time from the video data D1 captured during the inspection and generated as the still image data G1, and the relaxation determination can be performed promptly. Also, by comparing the reference video data D2 with the video data D1 obtained during the inspection by the learning AI, abnormalities within the shooting angle of the imaging means 1 can be detected as defects. Therefore, it can also contribute to finding defects that were overlooked in the conventional inspection method.

[0081] The description of the above-described embodiments is an example of the present invention, and the present invention is not limited to the above-described embodiments. Therefore, even if it is other than the above-described embodiments, various changes can be made according to the design and the like as long as it does not deviate from the technical idea according to the present invention. Further, the above-described various effects are merely an enumeration of the preferable effects resulting from the present invention, and the effects according to the present invention are not limited to those described in the present embodiment.

Explanation of Signs

[0082] M Fastening tool relaxation determination system 1 Imaging means 2 Information processing terminal 21 Fastening tool detection means 22 Relaxation determination means V Rail inspection device 3 Cart 4 Electric motor 5 Control unit 6 Lighting unit 7 Terminal protection unit 8 Seat unit P Rail inspection method S1 AI learning process S2 Rail inspection process TO Rail TR Rail VN Fastening tool D1 Video data D2 Reference video data G1 Still image data G2 Reference still image data A Relaxation information B Defect information

Claims

【Claim 1】 An in-rail inspection device (V) comprising a vehicle that travels automatically on the above-described track equipped with a fastening relaxation determination system for a fastener that fixes a rail constituting a train running track, The in-rail inspection device, On the base plate (32) laid on the base frame (31) made of metal angle materials or pipe materials, a vehicle body part (3) composed of a base frame (31), a base plate (32) laid on the base frame 31, axles (33) provided at the front and rear of the base frame, wheels (34) connected to both ends of the two axles, strut parts (361, 361) extending upward and rearward from the vicinity of the left and right wheels on the front side, a laying part (362) laid on the upper part of the strut parts, and a handle part (37) for performing operation provided on the laying part, An electric motor (4) for driving the wheels, a control unit (5) for controlling the electric motor, an information processing terminal (2) that functions as a fastening detection means and a relaxation determination means described later, and a seat part (8) for an operator to board are provided, Rail imaging means (11) fixed respectively in front of and above the left and right two wheels constituting the front wheels of the vehicle body part, capable of photographing the fastening members and the periphery of the track to be inspected from above the rail as moving image data as the vehicle moves forward, and a rail illumination (61) arranged above the rail imaging means and illuminating downward, Forward imaging means (12) fixed to the laying part with the lens facing the forward direction, capable of photographing the front of the travel as moving image data, and a forward illumination (62) arranged beside the forward imaging means and illuminating the front, Composed of the terminal device held by the operator, and The relaxation determination system, The rail imaging means and the forward imaging means, Fastening detection means capable of executing a fastening extraction process that uses an AI (Artificial Intelligence) that has learned to detect the fastener based on reference moving image data photographed in the same angle of view as the moving image data in advance, detects the fastener from the moving image data photographed by the imaging means, and generates still image data recording the fastener, Relaxation determination means for comparing the still image data with reference still image data photographed in a state where the fastening by the fastener is in good condition, determining the relaxation state of the fastener, generating relaxation information (A) when relaxation is recognized, and transmitting it to the terminal device to notify the operator. Moreover, the fastener detection means also uses an AI that has been trained to detect differences by comparing the video data with the reference video data, and detects defects that occur in the vicinity of the track determined by comparison with the fastener and the reference video data and in the front part of the travel from the video data obtained by the track imaging means and the front imaging means, and further executes a defect detection process of transmitting the defect information (B) to the terminal device and notifying the worker, and the track inspection device is characterized in that it can perform the above.

Images