Method and device for detecting defects of overhead wire suspension parts

Abstract

The present application provides a method and device for detecting defects in a contact network suspension part, the method comprising: recording the position and number of poles between adjacent stations, where the poles contain contact network suspension parts; when a radar installed on the top of a vehicle detects a first pole, obtaining first image data containing the first pole through an industrial camera; determining whether the contact network suspension part at the first pole has defects according to the first image data; when the contact network suspension part at the first pole has defects, extracting a first feature vector corresponding to the first pole according to the first image, determining an identifier corresponding to the first pole according to the first feature vector through a trained neural network model, and then determining a first predicted position of the first pole. The method solves the problem that effective defect detection of contact network suspension parts in the field of rail transit in the related art can be performed, and high-definition imaging and defect detection of contact network suspension parts can be achieved, and the defect position can be effectively located.

Description

Technical Field

[0001] The present application relates to the technical field of catenary detection, and in particular to a method and device for detecting defects in catenary suspension components. Background Art

[0002] The overhead contact system refers to the entire power supply system that transmits the electric energy from the traction substation to the vehicle through the pantograph of the electric passenger car. It is the only equipment in the traction power supply system without backup. The status of the overhead contact suspension parts directly determines the operation of the urban rail transit system. In the early days, manual inspections were used to check the damage, defects, and loss of suspension parts caused by vibration, impact, and friction between the vehicle and the overhead contact system. Later, the overhead contact inspection devices were installed on track inspection vehicles and electric passenger cars to replace the human eye for inspection. The solutions were all subject to the influence of camera pixels and frame rates, and it was impossible to image the overhead contact network in high definition, and it was impossible to accurately locate the fault.

[0003] There is currently no effective solution to the problem in related technologies that effective defect detection of contact network suspension parts in the rail transit field cannot be carried out. Summary of the invention

[0004] The embodiments of the present application provide a method and device for detecting defects in a contact network suspension component, so as to at least solve the problem in the related art that effective defect detection of contact network suspension components in the field of rail transit cannot be performed.

[0005] According to an embodiment of the present application, a method for detecting defects in a contact network suspension component is provided, comprising: recording the positions and number of markers between adjacent sites, wherein each of the markers corresponds to an identifier, and the identifier is used to uniquely identify the marker, wherein the marker contains a contact network suspension component; when a radar installed on the top of the vehicle detects a first marker, acquiring first image data containing the first marker through an industrial camera; determining whether the contact network suspension component at the first marker has a defect based on the first image data; when the contact network suspension component at the first marker has a defect, extracting a first feature vector corresponding to the first marker based on the first image, determining the identifier corresponding to the first marker based on the first feature vector through a trained neural network model, and then determining a first predicted position of the first marker.

[0006] Optionally, the method also includes: acquiring in real time the running speed of the vehicle, the stations passed and the timestamp during the operation of the vehicle; determining the second predicted position of the first benchmark according to the timestamp of taking the first image, the running speed of the vehicle and the stations passed; and correcting the first predicted position by the second predicted position.

[0007] Optionally, recording the position and quantity of the benchmarks between adjacent sites includes: recording the position and quantity of at least one of the following workpieces between adjacent sites: an anchor segment joint, a center anchor segment, a line branch, and a segment insulator.

[0008] Optionally, before acquiring the first image data containing the first benchmark through an industrial camera, the method further includes: scanning the position of the anchor segment joint through a radar, wherein each anchor segment joint corresponds to a different preset height; when the radar scans the anchor segment joint at the preset height, it is determined that the benchmark is detected.

[0009] Optionally, the radar includes two radars, which are arranged along the direction in which the vehicle moves, and four industrial cameras and four LED high-frequency flash lights are symmetrically arranged on both sides of the radar.

[0010] According to an embodiment of the present application, a contact network suspension component defect detection device is also provided, including: a recording module, configured to record the position and number of markers between adjacent sites, wherein each of the markers corresponds to an identifier, and the identifier is used to uniquely identify the marker, and the marker contains a contact network suspension component; a first acquisition module, configured to acquire first image data containing the first marker through an industrial camera when a radar installed on the top of the vehicle detects a first marker; a first determination module, configured to determine whether the contact network suspension component at the first marker has a defect based on the first image data; a second determination module, configured to extract a first feature vector corresponding to the first marker based on the first image when a defect exists in the contact network suspension component at the first marker, determine the identifier corresponding to the first marker through a trained neural network model based on the first feature vector, and then determine the first predicted position of the first marker.

[0011] Optionally, the device also includes: a second acquisition module, configured to acquire the vehicle's running speed, passed stations and timestamp in real time during the operation of the vehicle; a third determination module, configured to determine the second predicted position of the first benchmark based on the timestamp of taking the first image, the vehicle's running speed and passed stations; and a correction module, configured to correct the first predicted position using the second predicted position.

[0012] Optionally, the device also includes: a scanning module, configured to scan the position of the anchor segment joint through radar, wherein each anchor segment joint corresponds to a different preset height; and a fourth determination module, configured to determine that the benchmark is detected when the radar scans the anchor segment joint at the preset height.

[0013] Through the embodiment of the present application, a method for detecting defects in a contact network suspension is provided, including: recording the position and number of poles between adjacent stations, where the poles contain contact network suspensions; when a radar installed on the top of a vehicle detects a first pole, obtaining first image data containing the first pole through an industrial camera; determining whether the contact network suspension at the first pole has defects based on the first image data; when the contact network suspension at the first pole has defects, extracting a first feature vector corresponding to the first pole based on the first image, determining an identifier corresponding to the first pole through a trained neural network model based on the first feature vector, and then determining a first predicted position of the first pole. The problem that effective defect detection of contact network suspensions in the field of rail transit in the related art cannot be performed is solved, and high-definition imaging and defect detection of contact network suspensions can be achieved, and the defect position can be effectively located. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0015] Figure 1 is a flow chart of an optional method for detecting defects of a contact network suspension component according to an embodiment of the present application;

[0016] Figure 2 is a flow chart of another optional method for detecting defects of a contact network suspension component according to an embodiment of the present application;

[0017] Figure 3 It is a schematic diagram of the viewing angle of radar and industrial camera;

[0018] Figure 4 yes Figure 3 Top view of the .

[0019] Description of Reference Numerals

[0020] 1. Radar; 2. Industrial camera; 3. LED high-frequency flash; A. Radar perspective; B. Camera longitudinal perspective; C. Camera lateral perspective. DETAILED DESCRIPTION

[0021] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0022] It should be noted that the terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0023] An embodiment of the present application provides a method for detecting defects in a contact network suspension component. Figure 1 is a flow chart of an optional method for detecting defects of a contact network suspension member according to an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0024] Step S102, recording the position and number of poles between adjacent sites, wherein each pole corresponds to an identifier, the identifier is used to uniquely identify the pole, and the pole contains a contact network suspension member;

[0025] Step S104, when the radar installed on the top of the vehicle detects the first benchmark, first image data containing the first benchmark is acquired by the industrial camera;

[0026] Step S106, determining whether there is a defect in the overhead line suspension component at the first benchmark according to the first image data;

[0027] Step S108, when there is a defect in the contact network suspension component at the first benchmark, extract the first feature vector corresponding to the first benchmark according to the first image, determine the identifier corresponding to the first benchmark through the trained neural network model according to the first feature vector, and then determine the first predicted position of the first benchmark.

[0028] Through the embodiment of the present application, a method for detecting defects in a contact network suspension is provided, including: recording the position and number of poles between adjacent stations, where the poles contain contact network suspensions; when a radar installed on the top of a vehicle detects a first pole, obtaining first image data containing the first pole through an industrial camera; determining whether the contact network suspension at the first pole has defects based on the first image data; when the contact network suspension at the first pole has defects, extracting a first feature vector corresponding to the first pole based on the first image, determining an identifier corresponding to the first pole through a trained neural network model based on the first feature vector, and then determining a first predicted position of the first pole. The problem that effective defect detection of contact network suspensions in the field of rail transit in the related art cannot be performed is solved, and high-definition imaging and defect detection of contact network suspensions can be achieved, and the defect position can be effectively located.

[0029] It should be noted that the neural network model can be a convolutional neural network trained using different benchmark images and corresponding logos as training data, and the logo corresponding to the benchmark in the image can be determined by feature recognition.

[0030] Optionally, the method also includes: acquiring in real time the running speed of the vehicle, the stations passed and the timestamp during the operation of the vehicle; determining the second predicted position of the first benchmark according to the timestamp of taking the first image, the running speed of the vehicle and the stations passed; and correcting the first predicted position by the second predicted position.

[0031] Optionally, recording the position and quantity of the benchmarks between adjacent sites includes: recording the position and quantity of at least one of the following workpieces between adjacent sites: anchor segment joints, center anchor segments, line branches, and segment insulators.

[0032] Optionally, before acquiring the first image data containing the first benchmark through an industrial camera, the method further includes: scanning the position of the anchor segment joint through a radar, wherein each anchor segment joint corresponds to a different preset height; when the radar scans the anchor segment joint at the preset height, it is determined that the benchmark is detected.

[0033] When the radar detects the anchor joint or any other benchmark, the corresponding benchmark identification can be determined according to the height of the anchor joint or the height of the benchmark, and the identification can be recorded. When it is determined that there is a defect in the contact network suspension at a certain benchmark, the benchmark position can be determined directly according to the corresponding identification recorded at that time.

[0034] Figure 2 is a flow chart of another optional method for detecting defects of contact network suspension parts according to an embodiment of the present application, such as Figure 2 As shown, after the laser radar scans to a specific height of the suspension mounting base, it triggers the industrial camera to take pictures, triggers the fill light (LED high-frequency flash) to strobe, and records the anchor segment number (corresponding logo) at the same time to facilitate anchor segment joint positioning. The captured photos are then used to perform high-definition imaging of the suspension parts, and the high-definition images are input into a trained image recognition neural network model for defect recognition, or defect recognition is performed directly manually.

[0035] Optionally, the radar includes two radars, which are arranged along the direction in which the vehicle moves, and four industrial cameras and four LED high-frequency flash lights are symmetrically arranged on both sides of the radar.

[0036] Figure 3 This is a schematic diagram of the viewing angle of radar and industrial camera. Figure 4 yes Figure 3 A top view of the Figure 2 and Figure 3 As shown, each detection module can include two radars, four high-definition industrial cameras and four LED high-frequency flashes. Of course, one radar, two industrial cameras and two LED high-frequency flashes, or three radars, six industrial cameras and six LED high-frequency flashes can be set accordingly according to practice. The embodiment of the present application does not limit the number of radars, industrial cameras and LED high-frequency flashes in the module.

[0037] According to an embodiment of the present application, a contact network suspension component defect detection device is also provided, including: a recording module, configured to record the position and number of markers between adjacent sites, wherein each of the markers corresponds to an identifier, and the identifier is used to uniquely identify the marker, and the marker contains a contact network suspension component; a first acquisition module, configured to acquire first image data containing the first marker through an industrial camera when a radar installed on the top of the vehicle detects a first marker; a first determination module, configured to determine whether the contact network suspension component at the first marker has a defect based on the first image data; a second determination module, configured to extract a first feature vector corresponding to the first marker based on the first image when a defect exists in the contact network suspension component at the first marker, determine the identifier corresponding to the first marker through a trained neural network model based on the first feature vector, and then determine the first predicted position of the first marker.

[0038] Optionally, the device also includes: a second acquisition module, configured to acquire the vehicle's running speed, passed stations and timestamp in real time during the operation of the vehicle; a third determination module, configured to determine the second predicted position of the first benchmark based on the timestamp of taking the first image, the vehicle's running speed and passed stations; and a correction module, configured to correct the first predicted position using the second predicted position.

[0039] Optionally, the device also includes: a scanning module, configured to scan the position of the anchor segment joint through radar, wherein each anchor segment joint corresponds to a different preset height; and a fourth determination module, configured to determine that the benchmark is detected when the radar scans the anchor segment joint at the preset height.

[0040] The rigid overhead wire suspension is photographed using the four industrial cameras of the suspension detection module. In dark conditions, high-brightness LED fill lights are used for illumination compensation to obtain high-definition images of the suspension components. The suspension and the pole are detected using a laser radar. When the radar waveform detects the pole, the industrial camera is triggered to image the object under test with the fill light. The overhead wire suspension area is then identified using an image recognition algorithm, and the boundary is identified and fitted. Finally, pattern recognition is used to determine whether there is an abnormality in the key position of the overhead wire suspension. This can effectively distinguish whether there is damage to the insulator, tilt, insufficient positioning slope, loose nuts, or other parts that are damaged.

[0041] Through the method of the embodiment of the present application, 25-megapixel high-definition imaging of the contact network can be achieved; the anchor segment joint positioning can be achieved with a positioning accuracy of up to ±8m; and real-time online monitoring of the contact network suspension parts can be achieved.

[0042] Optionally, in this embodiment, a person of ordinary skill in the art may understand that all or part of the steps in the various methods of the above embodiments may be completed by instructing hardware related to the terminal device through a program, and the program may be stored in a computer-readable storage medium, and the storage medium may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc.

[0043] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for detecting defects in a contact network suspension component, characterized in that: include: Recording the position and number of the poles between adjacent sites, wherein each pole corresponds to an identifier, the identifier is used to uniquely identify the pole, and the pole contains a contact network suspension member; When the radar installed on the top of the vehicle detects a first marker, first image data containing the first marker is acquired by an industrial camera; determining whether a contact network suspension component at the first pole has a defect according to the first image data; When there is a defect in the overhead line suspension component at the first benchmark, extracting a first feature vector corresponding to the first benchmark according to the first image, determining an identifier corresponding to the first benchmark through a trained neural network model according to the first feature vector, and then determining a first predicted position of the first benchmark; The recording of the positions and numbers of the poles between adjacent sites includes: Record the location and quantity of at least one of the following parts between adjacent stations: anchor section joints, center anchor sections, line branches, and segment insulators; Before acquiring the first image data including the first benchmark through the industrial camera, the method further includes: Scanning the position of the anchor segment joints by radar, wherein each anchor segment joint corresponds to a different preset height; When the radar scans the anchor section joint at the preset height, determining that the pole is detected; The method further comprises: Acquiring the running speed, passing stations and timestamp of the vehicle in real time during the running of the vehicle; Determining a second predicted position of the first benchmark according to a timestamp of capturing the first image, a running speed of the vehicle, and a passed station; The first predicted position is corrected by the second predicted position.

2. The method according to claim 1, characterized in that The radar includes two radars, which are arranged along the direction in which the vehicle moves. Four industrial cameras and four LED high-frequency flash lights are symmetrically arranged on both sides of the radar.

3. A contact network suspension defect detection device, characterized in that: include: A recording module, configured to record the position and number of the poles between adjacent sites, wherein each of the poles corresponds to an identifier, the identifier is used to uniquely identify the pole, and the pole contains a contact network suspension member; The recording of the location and number of poles between adjacent sites includes: Record the location and quantity of at least one of the following parts between adjacent stations: anchor section joints, center anchor sections, line branches, and segment insulators; A first acquisition module is configured to acquire first image data containing the first benchmark through an industrial camera when the radar installed on the top of the vehicle detects the first benchmark; a first determination module, configured to determine whether a contact network suspension component at the first pole has a defect according to the first image data; a second determination module configured to, when there is a defect in the overhead wire suspension component at the first benchmark, extract a first feature vector corresponding to the first benchmark according to the first image, determine an identifier corresponding to the first benchmark through a trained neural network model according to the first feature vector, and then determine a first predicted position of the first benchmark; A scanning module, configured to scan the position of the anchor segment joint by radar, wherein each of the anchor segment joints corresponds to a different preset height; A fourth determination module is configured to determine that the benchmark is detected when the radar scans the anchor segment joint at the preset height; A second acquisition module is configured to acquire the running speed, the passed stations and the timestamp of the vehicle in real time during the running of the vehicle; a third determination module, configured to determine a second predicted position of the first benchmark according to a timestamp of capturing the first image, a running speed of the vehicle, and a passed station; A correction module is configured to correct the first predicted position by using the second predicted position.

Images