CN116363573A - Transformer substation equipment state anomaly identification method and system - Google Patents

Transformer substation equipment state anomaly identification method and system Download PDF

Info

Publication number
CN116363573A
CN116363573A CN202310080950.9A CN202310080950A CN116363573A CN 116363573 A CN116363573 A CN 116363573A CN 202310080950 A CN202310080950 A CN 202310080950A CN 116363573 A CN116363573 A CN 116363573A
Authority
CN
China
Prior art keywords
image
preset
positioning label
substation equipment
offset
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.)
Pending
Application number
CN202310080950.9A
Other languages
Chinese (zh)
Inventor
岳绍龙
王飞
李睿
胡志坤
朱言庆
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhiyang Innovation Technology Co Ltd
Original Assignee
Zhiyang Innovation Technology Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhiyang Innovation Technology Co Ltd filed Critical Zhiyang Innovation Technology Co Ltd
Priority to CN202310080950.9A priority Critical patent/CN116363573A/en
Publication of CN116363573A publication Critical patent/CN116363573A/en
Pending legal-status Critical Current

Links

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
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
    • G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
    • G06T7/337—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods involving reference images or patches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/24—Aligning, centring, orientation detection or correction of the image
    • G06V10/245—Aligning, centring, orientation detection or correction of the image by locating a pattern; Special marks for positioning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
    • H02J13/00—Circuit arrangements for providing remote indication of network conditions, e.g. an instantaneous record of the open or closed condition of each circuitbreaker in the network; Circuit arrangements for providing remote control of switching means in a power distribution network, e.g. switching in and out of current consumers by using a pulse code signal carried by the network
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00—Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Databases & Information Systems (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Power Engineering (AREA)
  • Image Analysis (AREA)

Abstract

The invention provides a method and a system for identifying abnormal state of substation equipment, wherein the method comprises the steps of sticking a positioning label in a point position picture of a preset camera for identifying the substation equipment, and selecting a target to be identified in a frame; storing the preset bit image into a preset bit image template according to the point position ID; after the positioning label area is determined, calculating the offset of the second coordinate position of the positioning label in the inspection image and the first coordinate position of the positioning label in the preset image template; resetting the preset point position if the offset exceeds the set threshold value; if the correction is performed within the correctable range, the target to be identified is located in the frame selection area; cutting the inspection image according to the frame selection coordinates, and inputting the cut inspection image into a classification model to identify the current state of the target to be identified. Based on the method, a substation equipment state abnormality identification system is also provided. The invention builds the image recognition network, and realizes the state recognition of the substation equipment by classifying and learning the image data.

Description

Transformer substation equipment state anomaly identification method and system
Technical Field
The invention belongs to the technical field of substation equipment state monitoring, and particularly relates to a substation equipment state anomaly identification method and system.
Background
With the technical progress, the original monitoring and operation and maintenance modes of various industries cannot meet the requirement of intellectualization. Whether it is a conventional monitoring mode of an automation type or a video monitoring mode mainly of security protection, capability upgrading is urgently needed to adapt to higher requirements of industry clients.
In the power industry, conventional power stations face a large amount of repetitive work of manual inspection and manual rewinding, so that an intelligent inspection scheme for equipment in the stations is needed. The current intelligent monitoring inspection scheme or a target detection mode is adopted, but the mode is inevitably free from missed detection and can not alarm the state change of equipment. In addition, in actual inspection, the equipment such as an inspection ball machine and a track robot can have the problem of inaccurate positioning due to errors in hardware, so that an object to be identified deviates from a framed area, and an identification result is error.
Disclosure of Invention
In order to solve the technical problems, the invention provides a substation equipment state abnormality identification method and system. Setting a specific positioning label, calculating offset to carry out coordinate correction by positioning the positions of the positioning labels in the template image and the inspection image, and carrying out offset supplement calculation by adopting an image key point pair registration mode under the condition of failure detection of the positioning label so as to further ensure correction effect, thereby fully ensuring the operation stability of an algorithm.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
a transformer substation equipment state anomaly identification method comprises the following steps:
pasting a positioning label in a point position picture of a preset image acquisition module for identifying substation equipment, and selecting a target to be identified; storing the preset bit image into a preset bit image template according to the point positions I D;
after the positioning label area is determined, calculating the offset of the second coordinate position of the positioning label in the inspection image and the first coordinate position of the positioning label in the preset image template; if the offset exceeds the set offset threshold, resetting the preset point position; if the offset is in the correctable range, correcting to enable the target to be identified to be located in the frame selection area;
cutting the inspection image according to the frame selection coordinates, and inputting the cut image into a pre-trained classification model to identify the current state of the target to be identified.
Further, the positioning tag is used for identifying substation equipment in the point position picture through preset colors or preset shapes; the substation equipment comprises substation primary equipment and substation secondary equipment.
Further, after the positioning label is attached, the method further comprises: and selecting the target to be identified in the preset point position in a frame mode, and storing the frame coordinate, the point position I D and the target normal value in a database.
Further, the method further comprises the step of comparing the current state of the target to be identified with a target normal value stored in a database, and if the current state is inconsistent with the target normal value, carrying out abnormal warning on the current point position.
Further, the process of determining the location tag area includes:
detecting a second coordinate position of a positioning label in a patrol image, converting the patrol image into an HSV color gamut, and acquiring a region mask with preset color according to a preset color HSV value of the positioning label;
performing bit-wise operation on the region mask with the preset color and the inspection image to obtain the content of a mask region;
and (3) carrying out Canny edge detection on the inspection image to obtain a contour, drawing a contour line, and obtaining the number of contour corner points, wherein the area with the number of corner points being the number of positioning labels is the positioning label area.
Further, after determining the positioning tag area, the method further includes: if the positioning label cannot be positioned, respectively extracting a first characteristic point of the preset image template and a second characteristic point of the inspection image, and calculating offset by matching the first characteristic point and the second characteristic point; if the positioning label can be positioned, directly calculating the offset of the second coordinate position of the positioning label in the inspection image and the first coordinate position of the positioning label in the preset image template.
Further, the process of calculating the offset by matching the first feature point and the second feature point includes:
extracting a first characteristic point position and a second characteristic point position by adopting a SIFT algorithm; then determining a first characteristic point direction and a second characteristic point direction;
and finding out matched characteristic point pairs in the preset bitmap image template and the inspection image by matching the descriptors of the first characteristic points and the descriptors of the second characteristic points, and calculating the offset.
Further, the classification model is a lightweight image classification network; the lightweight image classification network utilizes depth separable convolution to extract image features, and disassembles the convolution into depth convolution and point convolution;
after multi-layer convolution and pooling operations, the extracted high-dimensional features output prediction categories through a Softmax function.
Further, the Softmax function has the expression:
Figure BDA0004067407490000031
wherein y represents the class of the current sample, Z i The characteristic value belonging to the i-th class extracted by the network is represented, and C represents the total classification number.
The invention also provides a substation equipment state abnormality identification system, which comprises a manufacturing module, a correction module and an identification module;
the manufacturing module is used for pasting a positioning label in the point position picture of the preset image acquisition module for identifying substation equipment, and selecting a target to be identified; storing the preset bit image into a preset bit image template according to the point position ID;
the correction module is used for determining a positioning label area and calculating the offset of the second coordinate position of the positioning label in the inspection image and the first coordinate position of the positioning label in the preset image template; if the offset exceeds the set offset threshold, resetting the preset point position; if the offset is in the correctable range, correcting to enable the target to be identified to be located in the frame selection area;
the identification module is used for cutting the inspection image according to the frame selection coordinates, and the cut image is input into a pre-trained classification model to identify the current state of the target to be identified.
The effects provided in the summary of the invention are merely effects of embodiments, not all effects of the invention, and one of the above technical solutions has the following advantages or beneficial effects:
the invention provides a method and a system for identifying abnormal state of substation equipment, wherein the method comprises the steps of sticking a positioning label in a point position picture of a preset image acquisition module for identifying the substation equipment, and selecting a target to be identified in a frame; storing the preset bit image into a preset bit image template according to the point position ID; after the positioning label area is determined, calculating the offset of the second coordinate position of the positioning label in the inspection image and the first coordinate position of the positioning label in the preset image template; if the offset exceeds the set offset threshold, resetting the preset point position; if the offset is in the correctable range, correcting to enable the target to be identified to be located in the frame selection area; cutting the inspection image according to the frame selection coordinates, and inputting the cut image into a pre-trained classification model to identify the current state of the target to be identified. Based on a substation equipment state abnormality identification method, a substation equipment state abnormality identification system is also provided. The invention provides a substation equipment state identification method, which utilizes a deep learning technology to build a lightweight image identification network, and has stable state identification capability for primary equipment and secondary equipment of a substation through classifying and learning image data.
The invention adopts a method of framing the target area in advance, and aims at solving the problems that the target detection can only identify the equipment state and can not judge the equipment deflection alarm by comparing the identification result with the normal value stored in the database to judge whether the current point position has the abnormal alarm when a plurality of scenes of identifying the targets exist in the inspection image, so that the omission is avoided and the normal state of each target is supported.
Because the stepping motor of the cradle head has deviation in the operation process, the position of the target to be identified in the image is deviated, the cut image cannot contain the complete target to be identified, and finally, the state identification model is inaccurate in prediction. According to the method, the specific positioning label is adhered, the positions of the positioning labels in the template image and the inspection image are positioned, the offset is calculated for coordinate correction, in order to further ensure the correction effect, the offset is calculated in a complementary mode of registering image key points under the condition that the positioning label fails to be detected, and the operation stability of an algorithm is fully ensured.
The substation equipment state anomaly identification method combined with image offset correction provided by the invention systematically solves the key problem of substation equipment inspection and anomaly alarm. The invention combines the digital image processing technology and the lightweight deep learning network, has small model volume and high operation speed, and has innovation.
Drawings
Fig. 1 is a flowchart of a method for identifying abnormal states of substation equipment according to embodiment 1 of the present invention;
FIG. 2 is a diagram showing a device status detection model in embodiment 1 of the present invention;
FIG. 3 is a schematic diagram of the deep split convolution operation in embodiment 1 of the present invention;
FIG. 4 is a diagram showing a process of detecting a positioning tag according to embodiment 1 of the present invention;
fig. 5 is a flowchart of a substation equipment status anomaly identification system according to embodiment 2 of the present invention.
Detailed Description
In order to clearly illustrate the technical features of the present solution, the present invention will be described in detail below with reference to the following detailed description and the accompanying drawings. The following disclosure provides many different embodiments, or examples, for implementing different structures of the invention. In order to simplify the present disclosure, components and arrangements of specific examples are described below. Furthermore, the present invention may repeat reference numerals and/or letters in the various examples. This repetition is for the purpose of simplicity and clarity and does not in itself dictate a relationship between the various embodiments and/or configurations discussed. It should be noted that the components illustrated in the figures are not necessarily drawn to scale. Descriptions of well-known components and processing techniques and processes are omitted so as to not unnecessarily obscure the present invention.
Example 1
The embodiment 1 of the invention provides a method for identifying abnormal state of transformer substation equipment, which aims at the difficult problem encountered in abnormal state identification alarm of the transformer substation intelligent patrol, on one hand, adopts a deep learning image identification algorithm to accurately identify the states of primary equipment and secondary equipment, compares the states with normal values stored in a database to judge whether the alarm is generated, on the other hand, provides a method for correcting the offset aiming at the point offset problem in the patrol process, ensures the quality of an input image of an equipment state identification model, designs and adheres a specific positioning label, detects the position of the positioning label to calculate the offset, simultaneously aims at ensuring the stability of correction, assists the point position of the positioning label which fails to detect or cannot adhere the positioning label in a key point to calculate the offset, and finally adjusts the position of a target frame in the database according to the offset to ensure that the cut image contains a complete target to be identified.
The embodiment 1 of the invention provides a substation equipment state abnormality identification method; in order to solve the problem of image offset, the quality of the input image of the state identification model is ensured, a positioning label is stuck in a preset point position picture, a target to be identified in the preset point position is selected in a frame mode, frame selection coordinates, point position id and normal values of the target are stored in a database, and meanwhile, the preset point position image is stored as a template according to point positions i d. Detecting the coordinate position of a positioning label in a patrol image by utilizing a digital image processing technology, calculating the coordinate position offset of the positioning label in a template image, and if the positioning label fails to be detected or the positioning label is not configured in the scene, respectively extracting characteristic point pairs of the template image and the patrol image, matching the characteristic point pairs, calculating the offset of the characteristic point pairs, correcting frame selection coordinates stored in a database according to the obtained offset, and always ensuring that a target is positioned in a frame selection area; the device state is identified by building a lightweight image identification network.
Fig. 1 is a flowchart of a method for identifying abnormal states of substation equipment according to embodiment 1 of the present invention;
in step S0, the process flow starts.
In step S1, preset points of an image acquisition module are set, in this application the image acquisition module employs a camera,
in step S2, a specific positioning label is manufactured, and the positioning label is stuck on the preset point position picture. In the application, the positioning tag identifies substation equipment in the point position picture through a preset color or a preset shape; the substation equipment comprises substation primary equipment and substation secondary equipment. Triangle with side length of 3CM is adopted in the method and the device in the transformer station to be well distinguished.
In step S3, a target to be identified in the preset point location is selected in a frame, and the frame coordinates, the point location id and the normal value of the target are stored in a database.
In step S4, the preset bit image is stored as a preset bit image template by the dot bit ID.
In step S5, a positioning tag area is determined, as shown in fig. 4, which is a diagram of a positioning tag detection process in embodiment 1 of the present invention;
detecting a second coordinate position of the positioning label in the inspection image, converting the inspection image into an HSV color gamut, and acquiring a region mask with preset color according to the HSV value of the preset color of the positioning label;
carrying out bit-wise operation on the regional mask with the preset color and the inspection image to obtain the content of the mask region;
and (3) carrying out Canny edge detection on the inspection image to obtain a contour, drawing a contour line, and obtaining the number of contour corner points, wherein the area with the number of corner points being the number of positioning labels is the positioning label area. The area with the number of corner points being 3 is the positioning label area.
In step S6, it is determined whether the positioning tag cannot be positioned, if so, step S7 is executed, and if so, step S8 is executed.
In step S7, first feature points of the preset bit image template and second feature points of the patrol image are extracted. The specific process comprises the following steps: extracting a first characteristic point position and a second characteristic point position by adopting a SIFT algorithm; then determining a first characteristic point direction and a second characteristic point direction;
and finding out matched characteristic point pairs in the preset bitmap image template and the inspection image by matching the descriptors of the first characteristic points and the descriptors of the second characteristic points, and calculating the offset.
Extracting characteristic points of a template image and a patrol image respectively, taking a SIFT characteristic extraction algorithm as an example, firstly generating a Gaussian differential image pyramid by continuously shrinking and Gaussian filtering an original image, then screening by using a Hessian matrix by detecting extreme points of images between different layers, and obtaining the positions of the characteristic points according to the following formula, wherein D is xx ,D xy And D yy The difference between the corresponding positions of the neighborhood of the candidate points is obtained.
Figure BDA0004067407490000071
After the positions of the feature points are obtained, the directions of the feature points need to be obtained, wherein L is the scale space value where the key points are located. And collecting the gradient and direction characteristics of pixels in a 3 sigma neighborhood window of the image with respect to the key points detected in the Gaussian differential pyramid, and carrying out statistics. The direction with the highest amplitude is taken as the main direction, and the direction exceeding 80 percent of the peak value is taken as the auxiliary direction.
Figure BDA0004067407490000072
Figure BDA0004067407490000073
And matching the characteristic point pairs. The feature point neighborhood is divided into a plurality of blocks, and gradient direction histograms of eight directions are calculated. There are 16 regions, so 16×8=128-dimensional data is generated. And finding out matched characteristic point pairs in the two images by matching the template image with the patrol image characteristic point descriptors, and calculating the offset.
In step S8, after determining the positioning tag region, an offset between the second coordinate position of the positioning tag in the inspection image and the first coordinate position of the positioning tag in the preset image template is calculated.
And calculating the offset of the second coordinate position of the positioning label in the inspection image and the first coordinate position of the positioning label in the preset image template.
In step S9, it is determined whether the offset exceeds the threshold, and if the offset does not exceed the threshold, step S10 is performed. Otherwise, step S11 is performed.
In step S10, the offset is within the correctable range, and the frame selection coordinates stored in the database are corrected according to the obtained offset, so that the target is always ensured to be in the frame selection area.
And constructing and training a lightweight image classification network, reducing model quantity by adopting deep separation convolution, and acquiring scene data to train an image classification model. The depth convolution uses only one convolution kernel per channel, so the number of output channels after the convolution operation for a single channel is also 1. FIG. 2 is a diagram showing a device status detection model in embodiment 1 of the present invention; FIG. 3 is a schematic diagram of the deep split convolution operation in embodiment 1 of the present invention; and then the N feature images are spliced in sequence to obtain an output feature image with the channel N. The convolution kernel size of the point convolution is 1×1×n, and the feature map of the previous step is weighted and combined in the depth direction to generate a new feature map. After multi-layer convolution and pooling operations, the extracted high-dimensional features are output through Softmax.
Figure BDA0004067407490000081
Wherein y represents the class of the current sample, Z i The characteristic value belonging to the i-th class extracted by the network is represented, and C represents the total classification number. The magnitude of the error between the true classification result and the predicted classification result can be measured by the loss function, and then the network weight is optimized and corrected based on the error.
In step S11, the preset point location is considered to be out of the correction range, and the reset of the preset point location is prompted.
In step S12, after correction, target clipping is performed.
In step S13, the cropped image is input into the trained classification model to identify the current state of the object.
In step S14, the current state is compared with the normal state values stored in the database;
in step S15, if they are not identical, an abnormality warning is performed on the point location.
In step S16, the flow ends.
According to the substation equipment state identification method provided by the embodiment 1 of the invention, a lightweight image identification network is built by using a deep learning technology, and the primary equipment and the secondary equipment of the substation are provided with stable state identification capability through classifying and learning of image data.
According to the substation equipment state identification method provided by the embodiment 1 of the invention, a method of framing target areas in advance is adopted, so that the condition that a plurality of identification targets exist in a patrol image is avoided, the record of the normal state of each target is supported while the omission is avoided, whether the current point position has abnormal alarms or not is judged by comparing the identification result with the normal value stored in the database, and the problem that the target detection can only identify the equipment state and cannot judge the equipment deflection alarms is solved.
Because the stepping motor of the cradle head has deviation in the operation process, the position of the target to be identified in the image is deviated, the cut image cannot contain the complete target to be identified, and finally, the state identification model is inaccurate in prediction. According to the substation equipment state identification method provided by the embodiment 1 of the invention, the position of the positioning label in the positioning template image and the position of the positioning label in the inspection image are stuck with a specific positioning label, the offset is calculated for coordinate correction, and in order to further ensure the correction effect, the offset is calculated in a registration mode by adopting image key points under the condition that the positioning label fails to be detected, so that the operation stability of an algorithm is fully ensured.
The substation equipment state identification method combined with image offset correction provided by the embodiment 1 of the invention systematically solves the key problems of inspection and abnormal alarm of the substation equipment. The invention combines the digital image processing technology and the lightweight deep learning network, has small model volume and high operation speed, and has innovation.
Example 2
Based on the method for identifying the state of the substation equipment according to embodiment 1 of the present invention, embodiment 2 of the present invention further provides a system for identifying the state of the substation equipment, and fig. 5 is a flowchart of a system for identifying the abnormal state of the substation equipment according to embodiment 2 of the present invention. The system comprises a manufacturing module, a correction module and an identification module;
the manufacturing module is used for pasting a positioning label in the point position picture of the preset image acquisition module for identifying substation equipment, and selecting a target to be identified; storing the preset bit image into a preset bit image template according to the point position ID;
the correction module is used for determining a positioning label area and calculating the offset of the second coordinate position of the positioning label in the inspection image and the first coordinate position of the positioning label in the preset image template; if the offset exceeds the set offset threshold, resetting the preset point position; if the offset is in the correctable range, correcting to enable the target to be identified to be located in the frame selection area;
the identification module is used for cutting the inspection image according to the frame selection coordinates, and the cut image is input into a pre-trained classification model to identify the current state of the target to be identified.
In the manufacturing module, the image acquisition module adopts a camera.
The positioning tag identifies substation equipment in the point position picture through a preset color or a preset shape; the substation equipment comprises substation primary equipment and substation secondary equipment. Triangle with side length of 3CM is adopted in the method and the device in the transformer station to be well distinguished.
The method further comprises the following steps of: and selecting the target to be identified in the preset point position in a frame mode, and storing the frame coordinate, the point position I D and the target normal value in a database.
The process implemented by the correction module comprises the following steps: determining a positioning label area, detecting a second coordinate position of the positioning label in the inspection image, converting the inspection image into an HSV color gamut, and acquiring an area mask with a preset color according to the HSV value of the preset color of the positioning label; carrying out bit-wise operation on the regional mask with the preset color and the inspection image to obtain the content of the mask region; and (3) carrying out Canny edge detection on the inspection image to obtain a contour, drawing a contour line, and obtaining the number of contour corner points, wherein the area with the number of corner points being the number of positioning labels is the positioning label area. The area with the number of corner points being 3 is the positioning label area.
If the positioning label cannot be positioned, respectively extracting a first characteristic point of the preset image template and a second characteristic point of the inspection image, and calculating offset by matching the first characteristic point and the second characteristic point; if the positioning label can be positioned, directly calculating the offset of the second coordinate position of the positioning label in the inspection image and the first coordinate position of the positioning label in the preset image template.
And extracting a first characteristic point of the preset bit image template and a second characteristic point of the inspection image. The specific process comprises the following steps: extracting a first characteristic point position and a second characteristic point position by adopting a SIFT algorithm; then determining a first characteristic point direction and a second characteristic point direction;
and finding out matched characteristic point pairs in the preset bitmap image template and the inspection image by matching the descriptors of the first characteristic points and the descriptors of the second characteristic points, and calculating the offset.
Extracting characteristic points of a template image and a patrol image respectively, taking a SIFT characteristic extraction algorithm as an example, firstly generating a Gaussian differential image pyramid by continuously shrinking and Gaussian filtering an original image, then screening by using a Hessian matrix by detecting extreme points of images between different layers, and obtaining the positions of the characteristic points according to the following formula, wherein D is xx ,D xy And D yy The difference between the corresponding positions of the neighborhood of the candidate points is obtained.
Figure BDA0004067407490000101
After the positions of the feature points are obtained, the directions of the feature points need to be obtained, wherein L is the scale space value where the key points are located. And collecting the gradient and direction characteristics of pixels in a 3 sigma neighborhood window of the image with respect to the key points detected in the Gaussian differential pyramid, and carrying out statistics. The direction with the highest amplitude is taken as the main direction, and the direction exceeding 80 percent of the peak value is taken as the auxiliary direction.
Figure BDA0004067407490000102
Figure BDA0004067407490000103
And matching the characteristic point pairs. The feature point neighborhood is divided into a plurality of blocks, and gradient direction histograms of eight directions are calculated. There are 16 regions, so 16×8=128-dimensional data is generated. And finding out matched characteristic point pairs in the two images by matching the template image with the patrol image characteristic point descriptors, and calculating the offset.
And after the positioning label area is determined, calculating the offset of the second coordinate position of the positioning label in the inspection image and the first coordinate position of the positioning label in the preset image template.
And calculating the offset of the second coordinate position of the positioning label in the inspection image and the first coordinate position of the positioning label in the preset image template.
If the offset exceeds the set offset threshold, resetting the preset point position;
if the offset is in the correctable range, correcting to enable the target to be identified to be located in the frame selection area; and correcting the frame selection coordinates stored in the database according to the obtained offset within the correctable range, and always ensuring that the target is in the frame selection area.
The identification module is implemented by the following steps: and constructing and training a lightweight image classification network, reducing model quantity by adopting deep separation convolution, and acquiring scene data to train an image classification model. The depth convolution uses only one convolution kernel per channel, so the number of output channels after the convolution operation for a single channel is also 1. FIG. 2 is a diagram showing a device status detection model in embodiment 1 of the present invention; FIG. 3 is a schematic diagram of the deep split convolution operation in embodiment 1 of the present invention; and then the N feature images are spliced in sequence to obtain an output feature image with the channel N. The convolution kernel size of the point convolution is 1×1×n, and the feature map of the previous step is weighted and combined in the depth direction to generate a new feature map. After multi-layer convolution and pooling operations, the extracted high-dimensional features are output through Softmax.
Figure BDA0004067407490000111
Wherein y represents the class of the current sample, Z i The characteristic value belonging to the i-th class extracted by the network is represented, and C represents the total classification number. The error between the true classification result and the predicted classification result can be measured by the loss functionThe magnitude and then the network weights are optimized and corrected based on the error.
After correction, the target crop is performed. The cropped image is input into the trained classification model to identify the current state of the object. Comparing the current state with a normal state value stored in a database; if the two points are inconsistent, abnormal alarm is carried out on the points.
According to the substation equipment state identification system provided by the embodiment 2 of the invention, a lightweight image identification network is built by using a deep learning technology, and the primary equipment and the secondary equipment of the substation are provided with stable state identification capability through classifying and learning of image data.
According to the substation equipment state identification system provided by the embodiment 2 of the invention, a method of framing target areas in advance is adopted, so that the situation that a plurality of identification targets exist in a patrol image is avoided, the record of the normal state of each target is supported while the omission is avoided, whether the current point position has abnormal alarms or not is judged by comparing the identification result with the normal value stored in the database, and the problem that the target detection can only identify the equipment state and cannot judge the equipment deflection alarms is solved.
Because the stepping motor of the cradle head has deviation in the operation process, the position of the target to be identified in the image is deviated, the cut image cannot contain the complete target to be identified, and finally, the state identification model is inaccurate in prediction. According to the substation equipment state identification system provided by the embodiment 2 of the invention, the position of the positioning label in the positioning template image and the position of the positioning label in the inspection image are stuck with a specific positioning label, the offset is calculated for coordinate correction, and in order to further ensure the correction effect, the offset is calculated in a registration mode by adopting image key points under the condition that the positioning label fails to be detected, so that the running stability of an algorithm is fully ensured.
The substation equipment state identification system provided by the embodiment 2 of the invention provides the substation equipment state abnormality identification method combined with image offset correction, and the key problems of inspection and abnormal alarm of the substation equipment are systematically solved. The invention combines the digital image processing technology and the lightweight deep learning network, has small model volume and high operation speed, and has innovation.
The description of the relevant parts in the substation equipment state identification system provided in embodiment 2 of the present application may refer to the detailed description of the corresponding parts in the substation equipment state identification method provided in embodiment 1 of the present application, which is not repeated herein.
It is noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements is inherent to. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element. In addition, the parts of the above technical solutions provided in the embodiments of the present application, which are consistent with the implementation principles of the corresponding technical solutions in the prior art, are not described in detail, so that redundant descriptions are avoided.
While the specific embodiments of the present invention have been described above with reference to the drawings, the scope of the present invention is not limited thereto. Other modifications and variations to the present invention will be apparent to those of skill in the art upon review of the foregoing description. It is not necessary here nor is it exhaustive of all embodiments. On the basis of the technical scheme of the invention, various modifications or variations which can be made by the person skilled in the art without the need of creative efforts are still within the protection scope of the invention.

Claims (10)

1. The method for identifying the abnormal state of the substation equipment is characterized by comprising the following steps of:
pasting a positioning label in a point position picture of a preset image acquisition module for identifying substation equipment, and selecting a target to be identified; storing the preset bit image into a preset bit image template according to the point position ID;
after the positioning label area is determined, calculating the offset of the second coordinate position of the positioning label in the inspection image and the first coordinate position of the positioning label in the preset image template; if the offset exceeds the set offset threshold, resetting the preset point position; if the offset is in the correctable range, correcting to enable the target to be identified to be located in the frame selection area;
cutting the inspection image according to the frame selection coordinates, and inputting the cut image into a pre-trained classification model to identify the current state of the target to be identified.
2. The method for identifying abnormal state of substation equipment according to claim 1, wherein the positioning tag is used for identifying the substation equipment in the point location picture through a preset color or a preset shape; the substation equipment comprises substation primary equipment and substation secondary equipment.
3. The substation equipment state anomaly identification method according to claim 2, further comprising, after attaching the positioning label: and selecting the target to be identified in the preset point location by a frame, and storing the frame coordinate, the point location ID and the target normal value in a database.
4. A substation equipment state anomaly identification method according to claim 3, further comprising comparing the current state of the object to be identified with the normal values of the objects stored in the database, and if they are inconsistent, performing anomaly warning on the current point location.
5. The substation equipment state anomaly identification method according to claim 2, wherein the process of determining the location tag area comprises:
detecting a second coordinate position of a positioning label in a patrol image, converting the patrol image into an HSV color gamut, and acquiring a region mask with preset color according to a preset color HSV value of the positioning label;
performing bit-wise operation on the region mask with the preset color and the inspection image to obtain the content of a mask region;
and (3) carrying out Canny edge detection on the inspection image to obtain a contour, drawing a contour line, and obtaining the number of contour corner points, wherein the area with the number of corner points being the number of positioning labels is the positioning label area.
6. The method for identifying abnormal state of substation equipment according to claim 1, wherein after determining the positioning tag area, further comprises: if the positioning label cannot be positioned, respectively extracting a first characteristic point of the preset image template and a second characteristic point of the inspection image, and calculating offset by matching the first characteristic point and the second characteristic point; if the positioning label can be positioned, directly calculating the offset of the second coordinate position of the positioning label in the inspection image and the first coordinate position of the positioning label in the preset image template.
7. The substation equipment state anomaly identification method according to claim 6, wherein the process of calculating the offset by matching the first feature point and the second feature point comprises:
extracting a first characteristic point position and a second characteristic point position by adopting a SIFT algorithm; then determining a first characteristic point direction and a second characteristic point direction;
and finding out matched characteristic point pairs in the preset bitmap image template and the inspection image by matching the descriptors of the first characteristic points and the descriptors of the second characteristic points, and calculating the offset.
8. The substation equipment state anomaly identification method according to claim 1, wherein the classification model is a lightweight image classification network; the lightweight image classification network utilizes depth separable convolution to extract image features, and disassembles the convolution into depth convolution and point convolution;
after multi-layer convolution and pooling operations, the extracted high-dimensional features output prediction categories through a Softmax function.
9. The substation equipment state anomaly identification method according to claim 8, wherein the Softmax function has the expression:
Figure FDA0004067407480000021
wherein y represents the class of the current sample, Z i The characteristic value belonging to the i-th class extracted by the network is represented, and C represents the total classification number.
10. The system for identifying the abnormal state of the substation equipment is characterized by comprising a manufacturing module, a correcting module and an identifying module;
the manufacturing module is used for pasting a positioning label in the point position picture of the preset image acquisition module for identifying substation equipment, and selecting a target to be identified; storing the preset bit image into a preset bit image template according to the point position ID;
the correction module is used for determining a positioning label area and calculating the offset of the second coordinate position of the positioning label in the inspection image and the first coordinate position of the positioning label in the preset image template; if the offset exceeds the set offset threshold, resetting the preset point position; if the offset is in the correctable range, correcting to enable the target to be identified to be located in the frame selection area;
the identification module is used for cutting the inspection image according to the frame selection coordinates, and the cut image is input into a pre-trained classification model to identify the current state of the target to be identified.
CN202310080950.9A 2023-01-31 2023-01-31 Transformer substation equipment state anomaly identification method and system Pending CN116363573A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310080950.9A CN116363573A (en) 2023-01-31 2023-01-31 Transformer substation equipment state anomaly identification method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310080950.9A CN116363573A (en) 2023-01-31 2023-01-31 Transformer substation equipment state anomaly identification method and system

Publications (1)

Publication Number Publication Date
CN116363573A true CN116363573A (en) 2023-06-30

Family

ID=86928891

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310080950.9A Pending CN116363573A (en) 2023-01-31 2023-01-31 Transformer substation equipment state anomaly identification method and system

Country Status (1)

Country Link
CN (1) CN116363573A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117036665A (en) * 2023-09-04 2023-11-10 南京航空航天大学 Knob switch state identification method based on twin neural network
CN117201122A (en) * 2023-09-11 2023-12-08 大连理工大学 Unsupervised attribute network anomaly detection method and system based on view level graph comparison learning

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117036665A (en) * 2023-09-04 2023-11-10 南京航空航天大学 Knob switch state identification method based on twin neural network
CN117036665B (en) * 2023-09-04 2024-03-08 南京航空航天大学 Knob switch state identification method based on twin neural network
CN117201122A (en) * 2023-09-11 2023-12-08 大连理工大学 Unsupervised attribute network anomaly detection method and system based on view level graph comparison learning

Similar Documents

Publication Publication Date Title
CN111310645B (en) Method, device, equipment and storage medium for warning overflow bin of goods accumulation
CN116363573A (en) Transformer substation equipment state anomaly identification method and system
CN109190636B (en) Remote sensing image ship target information extraction method
CN113077453B (en) Circuit board component defect detection method based on deep learning
CN112837315B (en) Deep learning-based transmission line insulator defect detection method
CN113034624B (en) Temperature early warning image identification method, system, equipment and storage medium based on thermochromic adhesive
CN112446370A (en) Method for recognizing text information of nameplate of power equipment
CN110827432A (en) Class attendance checking method and system based on face recognition
CN112419260A (en) PCB character area defect detection method
CN111652846A (en) Semiconductor defect identification method based on characteristic pyramid convolution neural network
CN115019294A (en) Pointer instrument reading identification method and system
CN109615610B (en) Medical band-aid flaw detection method based on YOLO v2-tiny
CN117114420B (en) Image recognition-based industrial and trade safety accident risk management and control system and method
CN112260402B (en) Monitoring method for state of intelligent substation inspection robot based on video monitoring
CN116704490B (en) License plate recognition method, license plate recognition device and computer equipment
CN114187247A (en) Ampoule bottle printing character defect detection method based on image registration
CN114219753A (en) Power equipment surface defect detection method based on deep learning and terminal
CN111832497B (en) Text detection post-processing method based on geometric features
CN113408630A (en) Transformer substation indicator lamp state identification method
CN117372956A (en) Method and device for detecting state of substation screen cabinet equipment
Steiger et al. Automatic detection of transmission towers
CN204613128U (en) Circuit board detection device
CN116664540A (en) Rubber sealing ring surface defect detection method based on Gaussian line detection
CN116523885A (en) PCB defect detection method based on multi-scale fusion and deep learning
CN112380970B (en) Video target detection method based on local area search

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