Overall contact network identification method

Abstract

The present invention provides an overall contact network identification method, provides an identification trigger device for identifying the contact network of a railway system and a collection camera connected to the identification trigger device, including the following steps: step S10, installing the identification trigger device to a position directly opposite to the contact network; step S20, using the identification trigger device to collect image information near the contact network in real time; step S30, the identification trigger device analyzes the image information and detects the target features, and after identifying the target features, sends a working signal to the collection camera; step S40, the collection camera works to collect high-definition images. Compared with the related art, the overall contact network identification method provided by the present invention has high triggering accuracy, low probability of false touch, low requirements for installation space, and can simultaneously ensure the triggering effect during the day and at night.

Description

Technical Field

[0001] The present invention relates to the technical field of insurance case processing, and in particular to an appraisal system and method for investigating and assessing losses. Background Art

[0002] With the rapid development of railway infrastructure construction and people's high requirements for railway transportation safety and operation quality, relevant railway departments need to conduct real-time inspection and timely maintenance of railway infrastructure within the skylight to ensure the safety of train operation.

[0003] Railway infrastructure inspection is an important guarantee for transportation safety, including various professional inspection contents, such as track inspection, track flaw detection, track geometry parameters, contact network geometry parameters detection, contact suspension support device detection, etc. The overall contact network suspension state detection and monitoring device is mainly composed of a trigger module, an acquisition control module, and an operation control module. Therefore, triggering is the key process of this type of non-contact detection.

[0004] However, see Figure 1 , Figure 1 It is a schematic diagram of the contact network identification method of the prior art. As shown in the figure, the prior art usually adopts a method in which a high-frequency camera and a laser work together, and mainly adopts the following three methods for triggering: 1) manual triggering, judging the target by the human eye, and pressing the trigger button to trigger the acquisition camera to shoot; 2) automatic laser triggering, using a high-frequency camera to capture the bright spot formed by the laser directly hitting the target, and automatically triggering the acquisition camera to shoot by judging the bright spot in the camera picture; 3) feature recognition automatic triggering, through the recognition camera to perform high-frequency image acquisition, and then convert the acquired image, extract the target features (mainly positioning the wire clamp), and then identify the target and trigger.

[0005] Therefore, the trigger modules in the prior art also usually have the following defects: 1) The triggering accuracy is not high. Whether it is laser triggering or feature recognition, the target image is optically converted. If the target is deformed or the installation position changes, it will cause missed triggering; 2) The probability of false triggering is high. The same as the previous principle, the tunnel wall during laser triggering or the anchor section during feature recognition will form features consistent with the target, resulting in false triggering; 3) The installation space requirements are high. In laser triggering, since the target needs to be directly approached at close range, and the acquisition camera needs a large object distance to capture a complete image, the installation space is large; 4) There is a large difference in the triggering effect between daytime and nighttime. The light spot formed by the laser on the target at night is more obvious, while under strong light during the day, the light spot formed by the laser cannot be recognized and does not meet the triggering requirements.

[0006] Therefore, it is necessary to provide an overall contact network identification method with high triggering accuracy, low probability of false triggering, low installation space requirements, and the ability to ensure triggering effects both during the day and at night to solve the above technical problems. Summary of the invention

[0007] In view of the shortcomings of the prior art, the purpose of the present invention is to provide an overall contact network identification method with high triggering accuracy, low probability of false triggering, low requirement for installation space, and the ability to ensure triggering effects both during the day and at night.

[0008] The technical solution adopted by the present invention to solve the technical problem is:

[0009] A method for identifying an entire contact network, providing an identification trigger device for identifying a contact network of a railway system and a collection camera connected to the identification trigger device, comprising the following steps:

[0010] Step S10, installing the identification trigger device to a position facing the contact network;

[0011] Step S20, using the recognition trigger device to collect image information near the contact network in real time;

[0012] Step S30, the recognition trigger device analyzes the image information and detects the target features, and sends a working signal to the acquisition camera after identifying the target features;

[0013] Step S40: the acquisition camera operates to acquire high-definition images.

[0014] Preferably, the identification trigger device includes a shell, a starlight camera and an industrial camera installed in the shell, and in the shell, the starlight camera and the industrial camera are installed in parallel.

[0015] Preferably, the recognition range of the starlight camera covers the entire contact network, and the recognition range of the industrial camera falls within the recognition range of the starlight camera.

[0016] Preferably, the industrial camera is connected to the starlight camera; the overall contact network identification method further includes a step between step S10 and step S20.

[0017] Step S15, the starlight camera works to shoot a first calibration image, and sends the first calibration image to the industrial camera, the industrial camera works to shoot a second calibration image, the industrial camera compares the first calibration image with the second calibration image, obtains its own shooting direction, and adjusts the focus so that the recognition range of the industrial camera falls within the recognition range of the starlight camera.

[0018] Preferably, the shooting range of the industrial camera is located in the middle of the shooting range of the starlight camera.

[0019] Preferably, the industrial camera has a built-in target retrieval module and an image learning module connected to the target retrieval module, and the target retrieval module and the image learning module work together to realize position judgment and identification interval judgment of target features.

[0020] Preferably, the connection between the identification trigger device and the acquisition camera is an electrical connection or a communication connection.

[0021] Preferably, the target feature is a horizontal arm insulator, an inclined arm insulator, or a positioning clamp of the contact network.

[0022] To sum up, compared with the prior art, the overall contact network identification method provided by the present invention has high triggering accuracy. Through deep learning of the detection target samples, the intelligent algorithm has high recognition accuracy. The same target recognition time interval judgment strategy is used to further improve the recognition accuracy, and basically no false triggering will occur; the probability of missed triggering is low, and through direct recognition of the target image, the recognition rate is higher than that of feature recognition after traditional image processing. When the portable detection device runs at a low speed (5KM / H), basically no missed triggering will occur; the installation space requirement is low, and the starlight camera for recognition image acquisition is installed in parallel with the industrial camera for detection image acquisition. There is no need to approach the recognition target, and there is no installation space requirement; the triggering effect is consistent during the day and at night, and the starlight camera for recognition image acquisition can automatically adjust the exposure, and can form a clearer image at night without fill light for recognition by the intelligent algorithm, and the recognition result is consistent with the image collected during the day. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A schematic diagram of a method for realizing contact network identification in the prior art;

[0024] Figure 2 A horizontal schematic diagram of an identification trigger device in the overall contact network identification method provided by the present invention;

[0025] Figure 3 A longitudinal schematic diagram of an identification trigger device in the overall contact network identification method provided by the present invention;

[0026] Figure 4 The present invention provides a flow chart of the overall contact network identification method. DETAILED DESCRIPTION

[0027] The present invention is described in detail below in conjunction with the accompanying drawings and embodiments. The following experimental examples and embodiments are used to further illustrate but not limit the present invention.

[0028] Please refer to Figures 2 to 4The present invention provides a method for identifying an entire overhead contact network, and provides an identification trigger device 10 for identifying an overhead contact network 200 of a railway system and a collection camera 20 connected to the identification trigger device 10. The connection between the identification trigger device 10 and the collection camera 20 is an electrical connection or a communication connection.

[0029] The identification trigger device 10 includes a housing 11, a starlight camera 12 and an industrial camera 13 installed in the housing 11, and the starlight camera 12 is installed in parallel with the industrial camera in the housing 11. This configuration breaks through the limitation of the traditional method that high-frequency cameras and lasers and other components need to be installed close to the identification target, reduces the space requirements for the installation of the identification trigger 10, and is also more convenient for transportation.

[0030] The recognition range of the starlight camera 12 covers the entire overhead line 200, and the recognition range of the industrial camera 13 falls within the recognition range of the starlight camera 12. Figure 1 The contact network 200 includes a pillar 201, and a horizontal arm 202 and an oblique arm 203 fixed to the pillar 201. The recognition range of the starlight camera 12 should cover the pillar 201, the horizontal arm 202 and the oblique arm 203.

[0031] The target feature of the industrial camera 13 can be set as an insulator fixed on the horizontal arm 202 and / or the oblique arm 203. Of course, the target feature of the industrial camera 13 can also be the horizontal arm 202 and / or the oblique arm 203 itself, a positioning clamp or other identifiable targets. In this way, the trigger accuracy can be effectively improved by directly judging the position of the insulator and the recognition interval. At the same time, by directly identifying the target feature, the recognition rate is higher than the method in the prior art that requires image processing first. Under low-speed (5KM / H) operation, there is basically no missed triggering.

[0032] Specifically, in order to achieve the above objectives, the industrial camera 13 has a built-in target retrieval module and an image learning module connected to the target retrieval module. The target detection module 131 and the image learning module work together to realize the position judgment and identification interval judgment of the target feature.

[0033] More specifically, in this embodiment, the target retrieval module uses the V5 version of YOLO to obtain the fastest target retrieval speed; the image learning module uses tensorRT to improve the running speed of the model, and the two cooperate to improve the trigger accuracy. Of course, this is not a limitation of the present invention. In other embodiments of the present invention, the target retrieval module and the image learning module can also be any other software application that meets the conditions.

[0034] Preferably, in order to further ensure the recognition effect of the industrial camera 13 , in this embodiment, the shooting range of the industrial camera 13 is set to be located in the middle part of the shooting range of the starlight camera 11 .

[0035] The specific steps of the overall contact network identification method are as follows:

[0036] Step S10, installing the identification trigger device 10 to a position facing the contact network 200;

[0037] In this step, there is no need to deliberately control the distance between the identification trigger device 10 and the contact network 200, but the identification trigger device 10 needs to be completely fixed to ensure that it does not shake randomly.

[0038] Step S15, the starlight camera 12 works to shoot a first calibration image, and sends the first calibration image to the industrial camera 13, the industrial camera 13 works to shoot a second calibration image, the industrial camera 13 compares the first calibration image with the second calibration image, obtains its own shooting direction, and adjusts the focus so that the recognition range of the industrial camera 13 falls within the recognition range of the starlight camera 12;

[0039] It should be noted that this step is a calibration step. The starlight camera 12 and the industrial camera 13 work simultaneously, and the calibration of the industrial camera 13 is carried out by comparing the first calibration image with the second calibration image, adapting to the real-time changing on-site environment, to achieve automatic calibration with higher accuracy. At the same time, it is also faster than the manual proofreading method in the prior art.

[0040] Step S20, using the recognition trigger device to collect image information near the contact network in real time;

[0041] Step S30, the recognition trigger device analyzes the image information and detects the target features, and sends a working signal to the acquisition camera after identifying the target features;

[0042] Step S40: the acquisition camera operates to acquire high-definition images.

[0043] At the same time, in order to ensure that the recognition trigger device 10 has the same function during the day and at night, the starlight camera 12 and the industrial camera 13 need to have the function of automatically adjusting the exposure to ensure that clearer image information can be formed at night without fill light for subsequent recognition.

[0044] Compared with the prior art, the overall contact network identification method provided by the present invention has high triggering accuracy. Through deep learning of the detection target samples, the intelligent algorithm has high recognition accuracy. The same target recognition time interval judgment strategy is used to further improve the recognition accuracy, and basically no false triggering will occur; the probability of missed triggering is low. Through direct recognition of the target image, the recognition rate is higher than that of feature recognition after traditional image processing. When the portable detection device runs at a low speed (5KM / H), basically no missed triggering will occur; the installation space requirement is low, and the starlight camera for recognition image acquisition is installed in parallel with the industrial camera for detection image acquisition. There is no need to approach the recognition target, and there is no installation space requirement; the triggering effect is consistent during the day and at night, and the starlight camera for recognition image acquisition can automatically adjust the exposure. Even at night without fill light, a clearer image can be formed for recognition by the intelligent algorithm, and the recognition result is consistent with the image collected during the day.

[0045] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A method for identifying an entire contact network, providing an identification trigger device for identifying a contact network of a railway system and a collection camera connected to the identification trigger device, characterized in that: The steps include: Step S10, installing the identification trigger device to a position facing the contact network; Step S20, using the recognition trigger device to collect image information near the contact network in real time; Step S30, the recognition trigger device analyzes the image information and detects the target features, and sends a working signal to the acquisition camera after identifying the target features; Step S40, the acquisition camera works to acquire high-definition images, the recognition trigger device includes a shell, a starlight camera and an industrial camera installed in the shell, and in the shell, the starlight camera for recognition image acquisition and the industrial camera for detection image acquisition are installed in parallel, the starlight camera for recognition image acquisition can automatically adjust the exposure, and can form a clearer image at night without fill light for intelligent algorithm recognition, and the recognition result is consistent with the image collected during the day, the recognition range of the starlight camera covers the entire contact network, and the recognition range of the industrial camera falls within the recognition range of the starlight camera; The industrial camera has a built-in target retrieval module and an image learning module connected to the target retrieval module, and the target retrieval module and the image learning module work together to realize position judgment and identification interval judgment of the target feature; the industrial camera is connected to the starlight camera; the overall contact network identification method also includes a step S15 arranged between the step S10 and the step S20, the starlight camera works to shoot a first calibration image, and sends the first calibration image to the industrial camera, the industrial camera works to shoot a second calibration image, the industrial camera compares the first calibration image with the second calibration image, obtains its own shooting direction, and adjusts the focus so that the recognition range of the industrial camera falls within the recognition range of the starlight camera, and the shooting range of the industrial camera is located in the middle part of the shooting range of the starlight camera.

2. The overall contact network identification method according to claim 1, characterized in that: The connection between the identification trigger device and the acquisition camera is an electrical connection or a communication connection.

3. The overall contact network identification method according to claim 1, characterized in that: The target feature is a horizontal arm insulator, an inclined arm insulator, or a positioning wire clamp of the contact network.

Images