CN112966552B - Routine inspection method and system based on intelligent identification - Google Patents
Routine inspection method and system based on intelligent identification Download PDFInfo
- Publication number
- CN112966552B CN112966552B CN202110128417.6A CN202110128417A CN112966552B CN 112966552 B CN112966552 B CN 112966552B CN 202110128417 A CN202110128417 A CN 202110128417A CN 112966552 B CN112966552 B CN 112966552B
- Authority
- CN
- China
- Prior art keywords
- monitoring
- patrol
- image
- neural network
- network model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000007689 inspection Methods 0.000 title claims abstract description 106
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000012544 monitoring process Methods 0.000 claims abstract description 280
- 230000005540 biological transmission Effects 0.000 claims abstract description 96
- 238000003062 neural network model Methods 0.000 claims abstract description 72
- 238000009434 installation Methods 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 239000013589 supplement Substances 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 claims description 2
- 230000000007 visual effect Effects 0.000 description 4
- 101100494773 Caenorhabditis elegans ctl-2 gene Proteins 0.000 description 3
- 101100112369 Fasciola hepatica Cat-1 gene Proteins 0.000 description 3
- 101100005271 Neurospora crassa (strain ATCC 24698 / 74-OR23-1A / CBS 708.71 / DSM 1257 / FGSC 987) cat-1 gene Proteins 0.000 description 3
- 238000010191 image analysis Methods 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- HBBGRARXTFLTSG-UHFFFAOYSA-N Lithium ion Chemical compound [Li+] HBBGRARXTFLTSG-UHFFFAOYSA-N 0.000 description 2
- QVFWZNCVPCJQOP-UHFFFAOYSA-N chloralodol Chemical compound CC(O)(C)CC(C)OC(O)C(Cl)(Cl)Cl QVFWZNCVPCJQOP-UHFFFAOYSA-N 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 229910001416 lithium ion Inorganic materials 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- JBQRDRKWCBEQKP-UHFFFAOYSA-N 2-amino-5-iodo-6-phenyl-1h-pyrimidin-4-one Chemical compound N1C(N)=NC(=O)C(I)=C1C1=CC=CC=C1 JBQRDRKWCBEQKP-UHFFFAOYSA-N 0.000 description 1
- 238000012935 Averaging Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/20—Administration of product repair or maintenance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Molecular Biology (AREA)
- Tourism & Hospitality (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Business, Economics & Management (AREA)
- Biophysics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Multimedia (AREA)
- Alarm Systems (AREA)
- Image Analysis (AREA)
Abstract
The application provides a routine inspection method and a system based on intelligent identification. The method comprises the following steps: the patrol device collects monitoring images related to patrol scenes in the transformer substation based on the received patrol instructions sent by the monitoring server; the patrol device sends the monitoring image to a transmission gateway; and the transmission gateway inputs the received monitoring image sent by the inspection device into the lightweight image recognition neural network model so as to recognize the potential safety hazard target in the monitoring image. The transmission gateway sends the monitoring image output by the lightweight image recognition neural network model to a monitoring server; and the monitoring server determines the monitoring image with the potential safety hazard target, and sends the identification code and the alarm information corresponding to the inspection device to the monitoring terminal corresponding to the inspection device based on the monitoring image with the potential safety hazard target. The mode with image recognition replaces the manual work and patrols and examines the discernment trouble, uses manpower sparingly, can improve greatly and patrol and examine the frequency, improves the trouble recognition rate.
Description
Technical Field
The application relates to the technical field of substation patrol, in particular to a routine patrol method and system based on intelligent identification.
Background
In the transformer substation, various equipment are installed in different regions, and in order to ensure the fault-free operation of all the equipment, the equipment needs to be manually and periodically inspected, and the operation state of the equipment and key readings on reading equipment, the states of switch indicator lamps and the like need to be checked every time. The running state of the equipment of the whole station is judged by manually checking the performance state of the equipment on the spot, the polling period is 2-3 times per week, and the polling time needs about one day each time. The manual inspection has the problems of occupying more personnel and untimely inspection. Some transformer substations have installed the robot of patrolling and examining for coping with these problems, but the robot of patrolling and examining is expensive, and because space restriction, the robot of patrolling and examining can't reach because many places are indoor, still need the manual work to patrol and examine.
In addition, most of the existing image analysis technologies rely on a back-end image analysis server, the front-end equipment uploads the pictures to the server after the pictures are taken, and then the server analyzes the images and pushes alarms. This method will generate a large delay, and needs to deploy a special server, which is costly in investment and high in traffic cost. Therefore, the methods can not ensure lower polling cost, higher polling efficiency and higher identification accuracy.
Disclosure of Invention
Aiming at the problems in the prior art, the embodiment of the application provides a routine inspection method and a routine inspection system based on intelligent identification, and solves the technical problems that the conventional method for routine inspection of a transformer substation is high in cost, low in inspection efficiency and low in fault identification accuracy.
In one aspect, an embodiment of the present application provides a routine inspection method based on intelligent identification, and the method includes: the monitoring server trains a lightweight image recognition neural network model and deploys the lightweight image recognition neural network model to a transmission gateway; the monitoring server sends a patrol instruction to the patrol device in a narrow-band transmission mode so as to awaken the patrol devices installed at a plurality of preset installation positions in the transformer substation; the awakened patrol device collects monitoring images related to patrol scenes in the transformer substation based on the patrol instruction; the default state of the patrol device is a dormant state; the patrol device sends the monitoring image to a transmission gateway in a broadband transmission mode; wherein the transmission gateway is installed in the substation; and the transmission gateway preprocesses the received monitoring image sent by the inspection device and inputs the preprocessed monitoring image into the lightweight image recognition neural network model so as to recognize the potential safety hazard target in the monitoring image.
According to the embodiment of the application, the small patrol device is installed at each position of the transformer substation, the monitoring images of each corner of the transformer substation are collected, the image recognition mode is achieved, and the manpower input of transformer substation patrol is reduced. The comprehensiveness and the safety of the substation patrol are improved. The light-weight front-end neural network model is adopted to identify the image, so that the cost can be saved, the pressure of a server can be relieved, and the image identification precision and the identification speed can be improved. In addition, the patrol device is awakened to shoot through the narrow-band transmission patrol instruction, the patrol device is in a dormant state at ordinary times, the standby power consumption of the patrol device is low, and only microampere standby current exists. Therefore, the patrol device is more power-saving, the battery is smaller, the patrol device can be smaller, and the cost is lower.
In one embodiment, the training of the lightweight image recognition neural network model by the monitoring server specifically includes: the monitoring server obtains a reference image recognition neural network model through a FasterRCNN algorithm, wherein a Res101 algorithm is adopted in an image feature extraction method in the FasterRCNN algorithm; the monitoring server trains and adjusts the reference image recognition neural network model, and performs model conversion on the trained and adjusted reference image recognition neural network model to obtain the lightweight image recognition neural network model, wherein the lightweight neural network model is an offline neural network model.
In one embodiment, after the transmission gateway preprocesses the monitoring image received from the patrol device and inputs the preprocessed monitoring image into the lightweight image recognition neural network model to identify the potential safety hazard target in the monitoring image, the method further includes: the transmission gateway sends the monitoring image output by the lightweight image recognition neural network model to a monitoring server; the monitoring server determines the monitoring image with the potential safety hazard target, and determines an identity identification code corresponding to the first inspection device based on the monitoring image with the potential safety hazard target; the first inspection device is used for shooting the monitoring image of the target with the potential safety hazard; the monitoring server determines a monitoring terminal corresponding to the first inspection device based on the identity identification code corresponding to the first inspection device; and the monitoring server sends the identification code and the alarm information corresponding to the first patrol device to the monitoring terminal corresponding to the first patrol device.
In one embodiment, after the monitoring server sends the id code and the alarm information corresponding to the first patrol device to the monitoring terminal corresponding to the first patrol device, the method further includes: and the monitoring terminal receives monitoring videos which are collected by the first inspection device and are related to the inspection scene in the transformer substation based on the identification code corresponding to the first inspection device.
According to the embodiment of the application, the monitoring server sends the alarm information to the monitoring terminal, so that inspection personnel can check the monitoring picture shot by the inspection device for detecting the potential safety hazard target according to the alarm information, confirm the reason of the potential safety hazard, or inspect on site, and greatly improve inspection efficiency.
In one embodiment, before the patrol apparatus sends the monitoring image to a transmission gateway, the method further includes: the inspection device sends the collected monitoring images related to the inspection scene in the transformer substation to a monitoring server based on a preset rule; and the monitoring server inputs the received monitoring image sent by the inspection device into a depth image recognition neural network model so as to recognize a potential safety hazard target in the monitoring image.
According to the embodiment of the application, the back-end neural network model is installed on the monitoring server, so that the model can be flexibly selected according to the requirement of a user on the identification precision, and the use experience of the user is improved. And the monitoring image can be directly sent to the monitoring server for identification through other transmission modes such as 4G and the like under the condition that the individual patrol device is not covered by the network such as wifi and the like, so that the condition of missing report is avoided.
In one embodiment, after the monitoring server determines the monitoring image with the potential safety hazard target and determines the identification code corresponding to the first inspection device based on the monitoring image with the potential safety hazard target, the method further includes: the monitoring server generates alarm information corresponding to the first inspection device based on the identity identification code corresponding to the first inspection device; and the monitoring server writes the alarm information corresponding to the first inspection device into an alarm report.
In one embodiment, after the transmission gateway sends the monitoring image output by the lightweight image recognition neural network model to a monitoring server, the method further includes: the monitoring server classifies the monitoring images output by the lightweight image recognition neural network model based on a preset classification rule to obtain a first class image and a second class image; the first category images comprise monitoring images without potential safety hazard targets, and the second category images comprise monitoring images with potential safety hazard targets; and visually displaying the second category of images.
According to the embodiment of the application, the monitoring server is used for classifying and displaying the identified monitoring images, so that the monitoring scene conditions of each inspection device can be clearly presented to the working personnel.
In one embodiment, before the patrol device after being woken up collects a monitoring image related to a patrol scene in the substation based on the patrol instruction, the method further includes: the monitoring server sends a plurality of inspection instructions to the transmission gateway; the transmission gateway sends the plurality of inspection instructions to the inspection device at preset time intervals; or the transmission gateway sends the plurality of patrol instructions to the patrol device according to a preset sending sequence.
In the embodiment of the application, the monitoring server sends the configuration table recording the patrol command and the command sending rule to the transmission gateway, and the transmission gateway sends the command to the plurality of patrol devices according to the rule according to the content in the configuration table. Thousands of patrol devices are managed through one gateway, so that management, maintenance and construction of the patrol devices are facilitated.
In one embodiment, the method further comprises: the patrol device receives a working instruction from the monitoring server; wherein the work order at least comprises any one or more of the following items: and turning on a light supplement lamp instruction, a video recording instruction and a data transmission instruction. The patrol device determines the current electric quantity and refuses to execute the working instruction under the condition that the current electric quantity is lower than a preset threshold value; and the patrol device sends the current electric quantity to a monitoring server.
On the other hand, the embodiment of the present application provides a routine patrol system based on intelligent identification, and the system includes: the monitoring server is used for training the lightweight image recognition neural network model and deploying the lightweight image recognition neural network model to a transmission gateway; the monitoring server is also used for sending a patrol instruction to the patrol device in a narrow-band transmission mode; awakening a plurality of inspection devices arranged at preset installation positions in the transformer substation; the inspection device is used for acquiring monitoring images related to inspection scenes in the transformer substation based on the inspection instructions and sending the monitoring images to the transmission gateway in a broadband transmission mode; and the transmission gateway is used for preprocessing the received monitoring image sent by the patrol device and inputting the preprocessed monitoring image into the lightweight image recognition neural network model so as to recognize a potential safety hazard target in the monitoring image.
According to the embodiment of the application, the inspection devices with low cost are installed at all positions of the transformer substation, the scene images in the transformer substation are shot and the images are identified, and the inspection robot which is dozens of millions at a glance has great advantages in cost. The patrol device is extremely low in power consumption and small in size, and is easy to install through the support, so that light-weight installation can be realized; the device is used for remotely shooting images through the inspection device, hidden dangers in the transformer substation are remotely inspected through image recognition, and the manpower input of transformer substation inspection can be reduced. And the patrol device equipment can be almost installed in all places in the transformer substation, can cover the whole station, realizes the whole scene patrol, and solves the problem that manpower and robots can not patrol places which can not be reached.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
fig. 1 is a flowchart of a routine patrol method based on intelligent recognition according to an embodiment of the present disclosure;
fig. 2 is a flowchart of a method for controlling a patrol device by a monitoring server according to an embodiment of the present disclosure;
FIG. 3 is a flow chart of another intelligent identification based routine patrol method provided by an embodiment of the present application;
fig. 4 is a schematic diagram of a routine patrol system based on intelligent identification according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the technical solutions of the present application will be described in detail and completely with reference to the following specific embodiments of the present application and the accompanying drawings. It should be apparent that the described embodiments are only a few embodiments of the present application, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Due to the complex environment in the transformer substation, a plurality of scenes needing monitoring are provided, and common visualization equipment is difficult to apply to the transformer substation. The existing transformer substation mainly has some equipment parameter sensors used, the number of visual equipment is less, some large transformer substations use inspection robots, but the inspection robots are troublesome to install and debug, and are often months after installation and debugging are finished, and the operation and maintenance are complex. The conventional patrol method based on image analysis at present depends on a back-end server, and has time delay and high flow cost.
In order to solve the above problems, embodiments of the present application provide a routine inspection method and system based on intelligent identification.
The technical solutions proposed in the embodiments of the present application are described in detail below with reference to the accompanying drawings.
Fig. 1 is a flowchart of a routine patrol method based on intelligent recognition according to an embodiment of the present application, and as shown in fig. 1, the method may include steps S101 to S106:
s101, the monitoring server trains a lightweight image recognition neural network model, and deploys the lightweight image recognition neural network model to a transmission gateway.
And the monitoring server obtains a reference image recognition neural network model through a FasterRCNN algorithm, wherein an image feature extraction method in the FasterRCNN algorithm adopts a Res101 algorithm, then the reference image recognition neural network model is trained and tuned, and model conversion is carried out on the trained and tuned reference image recognition neural network model to obtain a light-weight image recognition neural network model.
Specifically, the image feature extraction algorithm in the FasterRCNN algorithm is replaced by the Res101 algorithm for image feature extraction, so that the FasterRCNN algorithm and the Res101 algorithm are fused to obtain a reference image recognition neural network model. And performing parameter optimization on the reference image recognition neural network model to obtain the reference image recognition neural network model with the recognition precision meeting the requirement. The reference image recognition neural network model is then converted into an offline neural network model, i.e., a lightweight image recognition neural network model, by an ATC (approach sensor Compiler) tool.
Further, the obtained lightweight image recognition neural network model is trained through the training samples, and the trained lightweight image recognition neural network model is issued to the transmission gateway.
S102, the patrol device collects monitoring images related to patrol scenes in the transformer substation based on the patrol instructions received from the monitoring server and sends the monitoring images to the transmission gateway.
The monitoring server controls the plurality of inspection devices to acquire monitoring images of all positions of the transformer substation through one transmission gateway, and transmits the acquired monitoring images to the transmission gateway under the default condition. Specifically, fig. 2 is a flowchart of a method for controlling a patrol device by a monitoring server according to an embodiment of the present application, and as shown in fig. 2, S102 may specifically include steps S201 to S203:
s201, the monitoring server sends a patrol instruction and an instruction sending rule to the transmission gateway.
Specifically, the monitoring server sends a sending rule of the patrol instruction to the transmission gateway, where the sending rule of the patrol instruction may include: the system comprises a time interval for transmitting a patrol command to the patrol device by the transmission gateway, a sequence for transmitting the patrol command to the patrol device by the transmission gateway, an identity identification code of the patrol device for transmitting the patrol command by the transmission gateway and the like. The sequence of sending the patrol instruction to the patrol device by the transmission gateway may include the following cases: the transmission gateway sends patrol instructions to all patrol devices in a broadcast mode; and the transmission gateway sequentially sends patrol instructions and the like to the patrol device.
In one embodiment, each patrol device has a unique ID, i.e., an identification code, and the monitoring server can manage the patrol device through the ID. The identity ID may be a Mobile Equipment Identifier MEID (Mobile Equipment Identifier MEID) of the patrol apparatus.
In one embodiment, the patrol device receives a work order from the monitoring server; wherein the work order at least comprises any one or more of the following items: and turning on a light supplement lamp instruction, a video recording instruction and a data transmission instruction. The patrol device determines the current residual capacity and refuses to execute the working instruction under the condition that the current residual capacity is lower than a preset threshold value; and sending the current electric quantity to the monitoring server.
In the embodiment of the application, one transmission gateway is connected with all patrol devices installed in one transformer substation in a WIFI or 4G/5G mode, and manages all patrol devices. The transmission gateway is installed to serve as a medium between the monitoring server and the inspection device, so that the pressure of the monitoring server is reduced, and the installation and maintenance are convenient.
And S202, after receiving the instruction sending rule and the patrol instruction sent by the monitoring server, the transmission gateway sends the patrol instruction to the connected patrol device according to the instruction sending rule.
Specifically, the transmission gateway sends out the patrol instruction in a Lora transmission mode according to an instruction sending rule sent by the monitoring server. Because the data volume of the patrol command is very small, and a lot of flow cost can be wasted by adopting a broadband technology, the patrol command is sent by adopting a Lora narrowband transmission mode, and the running cost of the device is saved.
In one embodiment, if the instruction issue rule is: and transmitting the patrol command to all patrol devices every other day in a simultaneous transmission sequence. And the transmission gateway sends the patrol command to all patrol devices according to the sending rule, and continuously sends the patrol command at intervals of one day until a new command sending rule is received.
S203, the patrol device receives the patrol instruction sent by the transmission gateway and collects the monitoring image of the patrol scene according to the content of the patrol instruction.
Specifically, after the patrol device receives the patrol instruction, the patrol instruction content is read, the monitoring image of the substation patrol scene is shot according to the patrol instruction content, and the monitoring image is sent to the transmission gateway in a WIFI or 4G/5G mode.
In one embodiment, the patrol device is in a dormant state when not receiving the patrol instruction, the patrol device only has microampere standby current, and the power consumption is extremely low, so that a non-rechargeable lithium-ion battery can be adopted as a battery of the patrol device, the battery can be mounted at one time to support the working time of the patrol device for several years, the cost is saved, the size of the patrol device can be reduced, and the patrol device can be conveniently mounted at a visual dead corner of a transformer substation. After the patrol device receives the patrol command sent by the transmission gateway, the patrol device is waken from the dormant state, and photographing is started according to the content of the patrol command. For example, the contents of the patrol instruction are: the patrol device shoots patrol scene images once every ten minutes for ten times. The patrol device takes a monitoring image every ten minutes from the receipt of the instruction, sends the taken monitoring image to the transmission gateway ten times later, and then enters the sleep state again.
S103, after the patrol device sends the monitoring image to the transmission gateway, the transmission gateway inputs the received monitoring image sent by the patrol device into the lightweight image recognition neural network model so as to recognize the potential safety hazard target in the monitoring image, and sends the monitoring image output by the lightweight image recognition neural network model to the monitoring server.
Specifically, when the transmission gateway identifies an image through the lightweight image identification neural network model, the received monitoring image is first preprocessed through the AIPP operator, such as processing of changing the image size, color gamut conversion (converting the image format), and averaging/multiplying factor (changing the image pixels). And then inputting the preprocessed monitoring image into a lightweight image recognition neural network model for target recognition, and outputting the recognized monitoring image. In the identified monitoring images, the identification frames are arranged on the monitoring images with the potential safety hazard targets to frame out the potential safety hazard targets, and the identification frames are not arranged on the monitoring images without the potential safety hazard targets. And the transmission gateway sends the output monitoring image to a monitoring server in a 4G or 5G mode.
In one embodiment, the lightweight image recognition neural network model deployed on the transmission gateway can be customized according to the requirements of a user, and if the user needs to recognize which kind of hidden danger target or objects, samples of the hidden danger targets are collected for model training, and the trained model is deployed in the transmission gateway.
It should be noted that, according to the routine patrol method based on intelligent identification provided in the embodiment of the present application, in a default case, the patrol apparatus sends the monitoring image to the transmission gateway, and the monitoring image is identified by the transmission gateway and then sent to the monitoring server. However, in the case where the user has a special need for the recognition model, the patrol device may directly transmit the monitoring image to the monitoring server.
In one embodiment, if the user needs to identify the monitoring image through the depth image recognition neural network model on the monitoring server to obtain a better image recognition effect, the patrol device may be used to directly send the monitoring image to the monitoring server, and the monitoring server performs the image recognition. Specifically, as shown in fig. 3, the method may include steps S301 to S303:
s301, the monitoring server sends a preset sending rule and the IP address of the monitoring server to the patrol device through the transmission gateway.
Specifically, the monitoring server sends a preset sending rule corresponding to the monitoring image to the patrol device based on the user requirement, and the content of the preset sending rule is as follows: and sending the monitoring image to a monitoring server. And attaches the IP address of the monitoring server so that the patrol apparatus transmits the monitoring image.
And S302, after receiving the preset sending rule, the patrol device sends the monitoring image to the monitoring server through a transmission mode of 4G CAT1 based on the preset sending rule and the IP address of the monitoring server.
The transmission cost can be reduced by the transmission mode of 4G CAT1, and due to the fact that a plurality of servers need to be configured for building the neural network model for recognizing the depth images at the rear end, the generated flow cost is very high, and therefore the cost can be saved for users by the transmission mode.
And S303, the monitoring server inputs the received monitoring image sent by the inspection device into the depth image recognition neural network model so as to recognize the potential safety hazard target in the monitoring image.
In an embodiment, a depth image recognition neural network model is deployed on a monitoring server, and compared with a lightweight image recognition neural network model deployed on a transmission gateway, the depth image recognition neural network model on the monitoring server is higher in recognition accuracy, but due to the fact that the cost of setting up the server is high and the traffic cost is high, the cost of image recognition is increased, the time for recognizing an image is increased due to a complex model, and the timeliness is poor, so that the method is used as a second choice of the embodiment of the application, and the image is recognized by the lightweight image recognition neural network model on the transmission gateway by default. Unless the user has a requirement on the recognition accuracy, the depth image recognition neural network model on the monitoring server is selected to recognize the monitoring image.
And the monitoring server inputs the received monitoring image into the depth image recognition neural network model and outputs the recognized monitoring image, and in the recognized monitoring image, the monitoring image with the potential safety hazard target marks the potential safety hazard target in a recognition frame mode.
And S104, after receiving the monitoring image output by the lightweight image recognition neural network model and sent by the transmission gateway, the monitoring server determines the monitoring image of the target with the potential safety hazard, and determines the identification code corresponding to the first inspection device based on the monitoring image of the target with the potential safety hazard.
Specifically, the monitoring server receives a monitoring image of the monitoring image output by the lightweight image recognition neural network model sent by the transmission gateway, or after the monitoring server recognizes the monitoring image, firstly detects whether a potential safety hazard target is recognized on the monitoring image, namely whether a recognition frame is on the monitoring image. If the potential safety hazard target is identified on the monitored image, determining an identity identification code corresponding to the first inspection device, namely, determining the identity ID of the inspection device recorded on the monitored image, wherein the identity ID represents the identity identification code of the inspection device for shooting the monitored image with the potential safety hazard target.
And S105, the monitoring server generates alarm information of the first patrol device based on the identity identification code corresponding to the first patrol device, determines a monitoring terminal corresponding to the first patrol device, and sends the identity identification code corresponding to the first patrol device and the alarm information to the monitoring terminal corresponding to the first patrol device.
Specifically, the monitoring server confirms the inspection device which takes the monitoring image through the identity ID detected on the monitoring image with the potential safety hazard target, and generates the alarm information of the inspection device. The alarm information may include the following contents: and (4) a potential safety hazard reminding message and what specific targets the detected potential safety hazard targets are. For example, the hidden trouble targets may be people, equipment operating states, obstacles, and the like.
Further, the alarm information and the identity ID of the inspection device are sent to the monitoring terminal corresponding to the inspection device, so that inspection personnel can confirm and check the reason of the hidden danger.
In one embodiment, the routine patrol system of the embodiment of the application comprises a plurality of monitoring terminals, wherein the monitoring terminals comprise mobile phones, computers and other terminals of patrol personnel. Each monitoring terminal is responsible for receiving the alarm information of a plurality of inspection devices, so that the inspection device can be confirmed to be responsible for by which monitoring terminal by identifying the identity ID on the potential safety hazard target, and after confirmation, the monitoring server sends the alarm information to the monitoring terminal.
For example, the monitoring server detects that the identity ID of the inspection device that captures the monitoring image is 001 on a monitoring image that identifies the potential safety hazard target, then finds that the 001 inspection device is in charge of the second monitoring terminal, and sends the identity ID of the 001 inspection device and the alarm information to the second monitoring terminal.
In one embodiment, the monitoring server arranges the alarm conditions of all the patrol devices into an alarm condition report, and updates the alarm condition report after a new patrol device gives an alarm, so that a user can download and view the alarm conditions of the patrol devices.
Further, the monitoring server classifies the received monitoring images based on the scene or the alarm condition, and performs visual display so that the user can view the monitoring images.
In an embodiment of the application, the monitoring server classifies the received monitoring images in a preset manner to obtain a first category monitoring image (a monitoring image with a potential safety hazard target) and a second category monitoring image (a monitoring image with a potential safety hazard target), and then visually displays the first category monitoring image. The second category of monitoring images can also be displayed visually. It should be noted that the visual display here may be realized by the monitoring server itself, or may be displayed by a display device externally connected to the monitoring server, which is not limited in the embodiment of the present application.
S106, the monitoring terminal receives monitoring videos which are collected by the first patrol device and related to patrol scenes in the transformer substation based on the identity identification codes corresponding to the first patrol device.
Specifically, after the monitoring terminal receives the alarm information and the identity ID corresponding to the inspection device, an alarm frame is popped up to prompt inspection personnel to check. The monitoring terminal responds to the operation that the patrol inspection personnel clicks the alarm box, the interface of the patrol inspection device is searched in the patrol inspection device which is responsible for the monitoring terminal according to the identity ID, the monitoring video of the patrol inspection device is called out by calling the interface, and the patrol inspection personnel can check the video to find the source of the hidden danger.
The method embodiment provided by the embodiment of the application also provides a routine inspection system based on intelligent identification based on the same inventive concept.
Fig. 4 is a schematic diagram of a routine patrol system based on intelligent recognition according to an embodiment of the present application, and as shown in fig. 4, the routine patrol system 400 based on intelligent recognition includes: a patrol device 410, a transmission gateway 420, a monitoring server 430, and a monitoring terminal 440.
Specifically, the patrol device 410 is configured to capture a scene image of the substation according to an instruction of the monitoring server 430. The patrol devices 410 are installed at various positions of the substation so that the patrol devices 410 can photograph images of the substation without dead angles.
The transmission gateway 420 is used for receiving the instructions of the monitoring server 430 and distributing the instructions to the plurality of patrol apparatuses 410. And is further configured to receive the monitoring image uploaded by the inspection device 410, and identify the monitoring image through the lightweight image being recognized by the neural network model, so as to determine the monitoring image with the potential safety hazard target.
The monitoring server 430 is configured to send a monitoring instruction to the inspection device 410, control the inspection device 410 to work, summarize alarm conditions of the inspection device 410, and send the summarized alarm conditions to the corresponding monitoring terminal 440, and is further configured to receive the identified monitoring images uploaded by the transmission gateway 420, and perform classified display on the monitoring images.
The monitoring terminals 440, each of which is responsible for a plurality of inspection devices in this embodiment of the application. After receiving the alarm information sent by the monitoring server 430, the monitoring terminal 440 calls the monitoring video of the inspection device 410 where the alarm occurs, so that the inspection personnel can check what potential safety hazard occurs in the substation and judge whether to eliminate the potential safety hazard on site. The monitoring terminal 440 can be a mobile phone WeChat client and a computer management interface of a user, so that inspection personnel can conveniently check images and alarms uploaded by equipment in real time and process faults at any time without being limited by regions.
In one embodiment, the patrol device 410 may transmit the monitoring image to the transmission gateway 420 in a WIFI or 4G/5G manner, or may transmit the monitoring image to the monitoring server 430 in a 4G CAT1 manner.
In one embodiment, the patrol device 410 can be a low-power-consumption photographing device, the size is small, the device can be magnetically attracted, the device can be pasted, screws, various modes such as binding bands and the like can be conveniently installed at any corner in a transformer substation, unpacking and installation are realized for no more than 3 minutes, ultralow self-discharge lithium-ion battery power supply is adopted, the standby power consumption is less than 60uA/3.6V, the service life of a single 14000mah battery for 5 years can be prolonged, the trouble that monitoring equipment needs to be frequently charged or the power supply is changed is reduced, the lightweight patrol is truly realized, and the use convenience is improved. The transmission gateway 420 can manage more than 6 ten thousand inspection devices, the monitoring points of a general transformer substation are not more than 3000 point positions, and a single transmission gateway completely meets the management and monitoring of all the monitoring points in one transformer substation, so that the monitoring without dead angles in a whole scene is realized.
According to the routine patrol method and system based on intelligent identification, the patrol device which is small and exquisite and low in power consumption is installed at each position in the transformer substation, the scene images of each position of the transformer substation are collected through the patrol device, and the images are identified through the lightweight neural network model, so that the identification efficiency is improved, and the time-consuming problem caused by image transmission is reduced. The image recognition method is used for replacing manual inspection, so that the inspection efficiency and the inspection safety of the transformer substation are improved, the inspection cost of the transformer substation is saved, and the method has important positive significance.
The above description is only a preferred embodiment of the present application and is not intended to limit the present application. It should be noted that various modifications and changes may be made to the present application by those of ordinary skill in the art. Any modification, equivalent replacement, improvement and the like made without departing from the principle of the present application shall be included in the scope of the claims of the present application.
Claims (8)
1. A routine patrol method based on intelligent identification is characterized by comprising the following steps:
the monitoring server trains a lightweight image recognition neural network model and deploys the lightweight image recognition neural network model to a transmission gateway;
the monitoring server sends a patrol instruction to the patrol device in a narrow-band transmission mode so as to awaken the patrol devices installed at a plurality of preset installation positions in the transformer substation;
the awakened patrol device collects monitoring images related to patrol scenes in the transformer substation based on the patrol instruction; the default state of the patrol device is a dormant state;
the patrol device sends the monitoring image to a transmission gateway in a broadband transmission mode; wherein the transmission gateway is installed in the substation;
the transmission gateway preprocesses the received monitoring image sent by the inspection device, and inputs the preprocessed monitoring image into the lightweight image recognition neural network model so as to recognize the potential safety hazard target in the monitoring image;
the transmission gateway sends the monitoring image output by the lightweight image recognition neural network model to a monitoring server;
the monitoring server determines the monitoring image with the potential safety hazard target, and determines an identification code corresponding to the first inspection device based on the monitoring image with the potential safety hazard target; the first inspection device is used for shooting the monitoring image of the target with the potential safety hazard;
the monitoring server determines a monitoring terminal corresponding to the first inspection device based on the identity identification code corresponding to the first inspection device;
the monitoring server sends the identity identification code and the alarm information corresponding to the first inspection device to a monitoring terminal corresponding to the first inspection device;
and the monitoring terminal receives monitoring videos which are collected by the first patrol device and are related to patrol scenes in the transformer substation based on the identity identification codes corresponding to the first patrol device.
2. The routine patrol method based on intelligent recognition according to claim 1, wherein the training of the lightweight image recognition neural network model by the monitoring server specifically comprises:
the monitoring server obtains a reference image recognition neural network model through a FasterRCNN algorithm, wherein a Res101 algorithm is adopted in an image feature extraction method in the FasterRCNN algorithm;
the monitoring server trains and adjusts the reference image recognition neural network model, and performs model conversion on the trained and adjusted reference image recognition neural network model to obtain the lightweight image recognition neural network model, wherein the lightweight image recognition neural network model is an offline neural network model.
3. The routine patrol method based on intelligent identification as claimed in claim 1, wherein before the patrol device sends the monitoring image to the transmission gateway through a broadband transmission mode, the method further comprises:
the inspection device sends the collected monitoring images related to the inspection scene in the transformer substation to a monitoring server based on a preset rule;
and the monitoring server inputs the received monitoring image sent by the inspection device into a depth image recognition neural network model so as to recognize a potential safety hazard target existing in the monitoring image.
4. The routine patrol method based on intelligent identification as claimed in claim 1, wherein after the monitoring server determines the monitored images with potential safety targets and determines the corresponding identification codes of the first patrol devices based on the monitored images with potential safety targets, the method further comprises:
the monitoring server generates alarm information corresponding to the first inspection device based on the identity identification code corresponding to the first inspection device;
and the monitoring server writes the alarm information corresponding to the first inspection device into an alarm report.
5. The routine patrol method based on intelligent recognition of claim 1, wherein after the transmission gateway sends the monitoring image output by the lightweight image recognition neural network model to a monitoring server, the method further comprises:
the monitoring server classifies the monitoring images output by the lightweight image recognition neural network model based on a preset classification rule to obtain a first class image and a second class image; the first category images comprise monitoring images without potential safety hazard targets, and the second category images comprise monitoring images with potential safety hazard targets;
and visually displaying the second category of images.
6. The routine patrol method based on intelligent recognition of claim 1, wherein before the patrol device after being awakened collects monitoring images related to patrol scenes in the substation based on the patrol instruction, the method further comprises:
the monitoring server sends a plurality of inspection instructions to the transmission gateway;
the transmission gateway sends the plurality of patrol instructions to the patrol device at preset time intervals; alternatively, the first and second electrodes may be,
and the transmission gateway sends the plurality of patrol instructions to the patrol device according to a preset sending sequence.
7. The routine patrol method based on intelligent recognition is characterized by further comprising the following steps of:
the patrol device receives a working instruction from the monitoring server; wherein the work order at least comprises any one or more of the following items: turning on a light supplement lamp instruction, a video recording instruction and a data transmission instruction;
the patrol device determines the current electric quantity and refuses to execute the working instruction under the condition that the current electric quantity is lower than a preset threshold value;
and the patrol device sends the current electric quantity to a monitoring server.
8. A routine patrol system based on intelligent recognition, the system comprising:
the monitoring server is used for training the lightweight image recognition neural network model and deploying the lightweight image recognition neural network model to a transmission gateway;
the monitoring server is also used for sending a patrol instruction to the patrol device in a narrow-band transmission mode; awakening a plurality of inspection devices arranged at preset installation positions in the transformer substation;
the patrol device is used for acquiring monitoring images related to patrol scenes in the transformer substation based on the patrol instruction and sending the monitoring images to the transmission gateway in a broadband transmission mode;
the transmission gateway is used for preprocessing the received monitoring image sent by the inspection device and inputting the preprocessed monitoring image into the lightweight image recognition neural network model so as to recognize a potential safety hazard target in the monitoring image;
the transmission gateway is further used for sending the monitoring image output by the lightweight image recognition neural network model to a monitoring server so that the monitoring server can determine the monitoring image with the potential safety hazard target, and an identity identification code corresponding to the first inspection device is determined based on the monitoring image with the potential safety hazard target; the first inspection device is used for shooting the monitoring image of the target with the potential safety hazard; determining a monitoring terminal corresponding to the first inspection device based on the identity identification code corresponding to the first inspection device; sending the identity identification code and the alarm information corresponding to the first inspection device to a monitoring terminal corresponding to the first inspection device;
and the monitoring terminal is used for receiving monitoring videos which are collected by the first patrol device and are related to patrol scenes in the transformer substation based on the identity identification codes corresponding to the first patrol device.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110128417.6A CN112966552B (en) | 2021-01-29 | 2021-01-29 | Routine inspection method and system based on intelligent identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110128417.6A CN112966552B (en) | 2021-01-29 | 2021-01-29 | Routine inspection method and system based on intelligent identification |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112966552A CN112966552A (en) | 2021-06-15 |
| CN112966552B true CN112966552B (en) | 2022-10-21 |
Family
ID=76272579
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110128417.6A Active CN112966552B (en) | 2021-01-29 | 2021-01-29 | Routine inspection method and system based on intelligent identification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112966552B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113538171A (en) * | 2021-09-14 | 2021-10-22 | 南方电网调峰调频发电有限公司 | Power station safety monitoring system |
| CN114004519A (en) * | 2021-11-05 | 2022-02-01 | 国网山东省电力公司青岛供电公司 | Power distribution visualization system and method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101917061A (en) * | 2010-07-14 | 2010-12-15 | 山东电力集团公司泰安供电公司 | Automatic inspection method and device for substation |
| CN205566523U (en) * | 2016-04-14 | 2016-09-07 | 国网山东省电力公司无棣县供电公司 | Transformer substation of power supply enterprise, transmission line system of patrolling and examining |
| CN107018364A (en) * | 2017-03-31 | 2017-08-04 | 中国南方电网有限责任公司超高压输电公司曲靖局 | A kind of centralized control system and method for substation inspection |
| CN111860170A (en) * | 2020-06-19 | 2020-10-30 | 包头钢铁（集团）有限责任公司 | Safety protective clothing wearing detection system based on 5G and deep learning target detection |
| CN112149522A (en) * | 2020-09-04 | 2020-12-29 | 国网天津市电力公司 | Intelligent visual external-damage-prevention monitoring system and method for cable channel |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107092251B (en) * | 2016-12-22 | 2019-11-15 | 合肥工业大学 | The automatic station-keeping system and method for the unmanned inspection car of railway based on image recognition |
| CN210605473U (en) * | 2019-11-12 | 2020-05-22 | 北京华能新锐控制技术有限公司 | Power plant inspection system |
| CN112131936B (en) * | 2020-08-13 | 2023-07-21 | 华瑞新智科技（北京）有限公司 | Inspection robot image recognition method and inspection robot |
-
2021
- 2021-01-29 CN CN202110128417.6A patent/CN112966552B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101917061A (en) * | 2010-07-14 | 2010-12-15 | 山东电力集团公司泰安供电公司 | Automatic inspection method and device for substation |
| CN205566523U (en) * | 2016-04-14 | 2016-09-07 | 国网山东省电力公司无棣县供电公司 | Transformer substation of power supply enterprise, transmission line system of patrolling and examining |
| CN107018364A (en) * | 2017-03-31 | 2017-08-04 | 中国南方电网有限责任公司超高压输电公司曲靖局 | A kind of centralized control system and method for substation inspection |
| CN111860170A (en) * | 2020-06-19 | 2020-10-30 | 包头钢铁（集团）有限责任公司 | Safety protective clothing wearing detection system based on 5G and deep learning target detection |
| CN112149522A (en) * | 2020-09-04 | 2020-12-29 | 国网天津市电力公司 | Intelligent visual external-damage-prevention monitoring system and method for cable channel |
Non-Patent Citations (1)
| Title |
|---|
| "变电站巡检机器人视频监控***设计与实现";李梁;《中国优秀博硕士学位论文全文数据库（硕士） 工程科技Ⅱ辑》;20150715(第07期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112966552A (en) | 2021-06-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104581087B (en) | A kind of power transmission line intelligent remote monitoring system based on Internet of Things | |
| CN204465736U (en) | A kind of power transmission line intelligent long distance control system based on Internet of Things | |
| CN112966552B (en) | Routine inspection method and system based on intelligent identification | |
| CN104581054A (en) | Electric transmission line inspecting method and system based on videos | |
| CN102616617B (en) | Elevator monitoring method | |
| CN103152601A (en) | Intelligent failure-reporting camera and network management client system thereof | |
| CN110807460B (en) | Transformer substation intelligent patrol system based on image recognition and application method thereof | |
| CN211509189U (en) | Electric power operation field supervision equipment and system | |
| CN105139471A (en) | Inspection method for electrical equipment | |
| CN106294602B (en) | Multimedia information visualization system | |
| CN112688434A (en) | Monitoring and early warning method and device for power transmission and distribution line, computer equipment and medium | |
| CN107613344B (en) | Instant feedback system for display terminal playing state and implementation method | |
| CN108833574A (en) | A kind of transmission line long-distance monitoring system suitable for Internet of Things | |
| CN116111721A (en) | Smart grid power equipment safety monitoring system and method based on big data | |
| CN114445782A (en) | Power transmission line image acquisition system based on edge AI and Beidou short messages | |
| CN108847985A (en) | Power transmission line intelligent remote monitoring system suitable for Internet of Things | |
| CN112001810A (en) | Intelligent forestry patrolling system and method based on machine vision | |
| CN113858194A (en) | A other robot that stands for job site auxiliary management | |
| CN105811578A (en) | Power transmission line monitoring platform, power source monitoring algorithm thereof and image early warning algorithm of power transmission line monitoring platform | |
| CN108663970B (en) | Electric power operation safety monitoring system | |
| CN115393340A (en) | AI vision product quality detection system based on 5G algorithm | |
| CN104767969A (en) | Intelligent operation safety monitoring system and method for railway crossing | |
| CN112272285B (en) | Intelligent operation and maintenance system, software equipment and device | |
| CN210428558U (en) | Intelligent auxiliary system for power optical cable inspection line | |
| CN209896769U (en) | Intelligent integrated non-contact online monitoring device for transformer |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |