CN115171091A - Meter identification method for substation inspection - Google Patents
Meter identification method for substation inspection Download PDFInfo
- Publication number
- CN115171091A CN115171091A CN202211081495.6A CN202211081495A CN115171091A CN 115171091 A CN115171091 A CN 115171091A CN 202211081495 A CN202211081495 A CN 202211081495A CN 115171091 A CN115171091 A CN 115171091A
- Authority
- CN
- China
- Prior art keywords
- meter
- image
- point
- pointer
- template
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000007689 inspection Methods 0.000 title claims abstract description 29
- 230000007246 mechanism Effects 0.000 claims abstract description 17
- 239000013598 vector Substances 0.000 claims abstract description 9
- 238000003062 neural network model Methods 0.000 claims abstract description 4
- 238000004364 calculation method Methods 0.000 claims description 28
- 238000013528 artificial neural network Methods 0.000 claims description 18
- 238000001514 detection method Methods 0.000 claims description 16
- 238000004458 analytical method Methods 0.000 claims description 7
- 230000003044 adaptive effect Effects 0.000 claims description 5
- 230000002776 aggregation Effects 0.000 claims description 3
- 238000004220 aggregation Methods 0.000 claims description 3
- 238000003384 imaging method Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000006116 polymerization reaction Methods 0.000 claims description 3
- 238000007619 statistical method Methods 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 3
- 125000004122 cyclic group Chemical group 0.000 claims 1
- 238000012423 maintenance Methods 0.000 abstract description 4
- 238000007405 data analysis Methods 0.000 abstract description 3
- 238000013135 deep learning Methods 0.000 abstract 1
- 239000011159 matrix material Substances 0.000 description 9
- 238000012549 training Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000009466 transformation Effects 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
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
-
- 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
- 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
- G06V10/225—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 based on a marking or identifier characterising the area
-
- 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
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
-
- 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
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
-
- 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
- G06V2201/02—Recognising information on displays, dials, clocks
Abstract
The invention discloses a meter identification method for substation inspection, which comprises the following steps: s1, acquiring a meter image; s2, detecting characteristic points of the meter image, and matching the characteristic points with the meter template image; s3, calculating the pointer direction of the meter; and S4, analyzing the pointer reading and outputting in a structured mode. The intelligent meter identification method adopted by the invention can realize automatic and accurate meter reading identification in daily power grid inspection work, record statistics and data analysis are simple and convenient, the operation and maintenance labor cost of manual meter reading can be reduced, the efficiency and the reliability are effectively improved, meanwhile, a multi-layer perceptron and a graph neural network model are constructed by utilizing a deep learning method, the expression information of high-dimensional vectors of characteristic points is extracted, a self-attention mechanism and a cross-attention mechanism are introduced, the internal and external relations and the corresponding relations of the characteristic points can be better learned, and the deep information expression capability of the characteristic points is enhanced.
Description
Technical Field
The invention relates to the technical field of meter identification, in particular to a meter identification method for substation inspection.
Background
At present, in order to ensure that various devices of a transformer substation normally operate and monitor index values and states of various devices of the transformer substation, daily routing inspection of the transformer substation is very important work, and reading and copying of meter devices are an important part of the work. The reading and copying work of the traditional manual meter is operated by special operation and maintenance personnel, so that the traditional manual meter is complex in work, easy to make mistakes, incapable of real-time monitoring, difficult to count, unsafe and the like.
With the rapid development of digital intelligence in various industries, artificial intelligence technology is very colorful in various scene fields, and the application of artificial intelligence technology to the reading and recording technology of meter pointer equipment is still imperfect, so that a safer and more reliable method for accurately and rapidly identifying the meter reading of the transformer substation is required.
An effective solution to the problems in the related art has not been proposed yet.
Disclosure of Invention
Aiming at the problems in the related art, the invention provides a meter identification method for substation inspection, which aims to overcome the technical problems in the prior art.
Therefore, the invention adopts the following specific technical scheme:
a meter identification method for substation routing inspection comprises the following steps:
s1, acquiring a meter image;
s2, detecting characteristic points of the meter image, and matching the characteristic points with the meter template image;
s3, calculating the pointer direction of the meter;
and S4, analyzing the pointer reading and outputting in a structured mode.
Further, the step of acquiring the meter image further comprises the steps of:
s11, configuring and connecting front-end shooting equipment, and remotely controlling the shooting equipment;
s12, calling equipment capable of shooting the meter, focusing on a meter dial by changing the angle and the focal length of the equipment, configuring shooting point positions capable of imaging clearly and recording;
s13, configuring the meter type of the shooting point;
s14, calling a shooting device for shooting the point location, and acquiring an initial meter image of the point location in a picture capturing or video stream pulling and analyzing mode;
and S15, decoding and preprocessing the initial meter image, acquiring the meter image, and sending the meter image to a downstream module for analysis and identification.
Further, the detecting the feature points of the meter image and matching with the meter template image further comprises the following steps:
s21, acquiring a meter image and a meter type through a meter shooting point, and retrieving a corresponding meter template image from a meter template image library based on the mark type;
s22, sending the meter image and the meter template image into a graph neural network feature matching module, and calculating feature points and corresponding relations of the meter image and the meter template image;
s23, matching and aligning the meter image with the meter template image, and outputting the aligned meter image;
s24, obtaining the coordinate area position of each component of the meter image from the meter template configuration;
each assembly of the meter image comprises a dial, scales, a center and a pointer.
Further, the step of sending the meter image and the meter template image into the graph neural network feature matching module and calculating the feature points and the corresponding relationship of the two images further comprises the following steps:
s221, respectively collecting images X of the meter 1 Meter-mixing templateImage X 2 Inputting the input size W multiplied by H matched with the convolutional neural network by Resize, and after normalization, inputting the input size W multiplied by H into a convolutional neural network model;
s222, extracting the image X collected by the meter A Characteristic point of
And descriptor
；
S223, extracting a meter template image X B Characteristic point of
And descriptor
；
S224, constructing a multilayer perceptron of an encoder, and performing
And
and embedding the high-dimensional characteristics into a multilayer perceptron to learn the high-dimensional characteristics, and updating the high-dimensional characteristic vectors obtained by learning to the original descriptor.
Further, the learned formula is as follows:
wherein, the symbol
Indicating assignment operation, left side
For descriptors updated after an assignment operation, right side
In order to update the descriptor before the update,
is a multi-layer perceptron encoder,
are characteristic points.
Further, the multilayer perceptron of the encoder is constructed by
And
the method for learning the high-dimensional features by embedding the high-dimensional features into a multilayer perceptron and updating the learned high-dimensional feature vectors into the original descriptor further comprises the following steps:
s2241, introducing a multi-head attention mechanism and a graph neural network structure;
s2242, input pair through control graph neural network
And
respectively realizing a self-attention mechanism and a cross-attention mechanism;
s2243, through cycle alternate input, learning and updating image X A Is described in
Wherein, the first and the second end of the pipe are connected with each other,
two nodes both correspond to the edge formed by the descriptors in the same image;
Further, the formula of the cyclically alternating input is as follows:
wherein the content of the first and second substances,
the representation is based on
A constructed graph neural network;
and
are respectively an image X A To go to
Is characterized by
To go to
And
calculating an output intermediate result by the layer;
represents polymerization
Descriptors of feature points.
Further, the meter pointer direction calculation further includes the following steps:
s31, based on the meter template image, a pointer position area and a dial plate central point of the aligned meter image can be obtained;
s32, taking the pointer position area as a binary template;
s33, acquiring a pointer position area image by taking the binary template as a template;
and S34, calculating a binary image of the pointer position area image by an adaptive binarization method.
Further, the calculation of the self-adaptive binarization method comprises indirect calculation and direct calculation;
wherein the indirect calculation comprises:
performing linear detection in the binary image area to find a pointer line segment of the meter;
detecting the longest line segment by adopting an EDLines linear detection method;
after the pointer line segment is obtained, calculating the intersection point of the straight line and the scale mark region by constructing an equation set;
the direct calculation comprises the following steps:
knowing the central point of the dial, depicting an extension line from the central point of the dial to each pixel point of the binary image;
calculating a projection point of each pixel point of the dial plate central point pointing to the binary image falling to the scale mark area;
and (4) obtaining the intersection point of the scale mark area pointed by the meter pointer through the projection point aggregation analysis.
Further, the analyzing the pointer reading and structuring the output further comprises the steps of:
s41, acquiring a scale mark area, scale point coordinates and numerical values configured by the meter template graph;
s42, performing approximate interpolation calculation on the intersection points and the configured scale points to obtain approximate readings of the pointer of the meter pointing to the intersection points;
and S43, outputting the reading result to the system in a structuralized mode, and recording the subsequent statistical analysis.
The beneficial effects of the invention are as follows:
1. the intelligent meter identification method adopted by the invention can realize automatic and accurate meter reading identification in the routine power grid inspection work, record statistics and data analysis are simply and conveniently carried out, the operation and maintenance labor cost of manual meter reading can be reduced, and the efficiency and the reliability can be effectively improved.
2. Compared with the traditional feature matching methods such as SIFT and ORB, the method can extract the expression information of the feature point high-dimensional vectors, introduces a self-attention mechanism and a cross-attention mechanism, can better learn the relation and the corresponding relation between the inside and the outside of the feature point, and further enhances the deep information expression capability of the feature point.
3. The invention constructs a deep feature detection matching model for continuous learning, can construct a training data set aiming at a meter equipment image of a power grid inspection scene, enables the model to adapt to a real scene for learning, and continuously improves the accuracy of the model in the scene, which is also a characteristic that the traditional feature matching method cannot achieve.
4. Compared with a Hough Line or LSD linear detection method adopted by the traditional method, the pointer direction calculation method of the meter equipment adopted by the invention not only adopts a new EDlines linear detection method, but also introduces a method for densely calculating and analyzing projection points, and has better robustness in an actual scene compared with the linear detection method.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
Fig. 1 is a flowchart of a meter identification method for substation inspection according to an embodiment of the present invention;
fig. 2 is a functional flow chart of a meter image feature point detection and matching module in a meter identification method for substation inspection according to an embodiment of the invention;
FIG. 3 is a functional flow chart of a graph neural network model feature point detection and matching module in the meter identification method for substation inspection according to the embodiment of the invention;
fig. 4 is a flow chart of calculation of pointer pointing of meter equipment in a meter identification method for substation inspection according to an embodiment of the present invention.
Detailed Description
For further explanation of the various embodiments, the drawings which form a part of the disclosure and which are incorporated in and constitute a part of this specification, illustrate embodiments and, together with the description, serve to explain the principles of operation of the embodiments, and to enable others of ordinary skill in the art to understand the various embodiments and advantages of the invention, and, by reference to these figures, reference is made to the accompanying drawings, which are not to scale and wherein like reference numerals generally refer to like elements.
According to the embodiment of the invention, a meter identification method for substation inspection is provided.
Referring to the drawings and the detailed description, the invention is further described, as shown in fig. 1 to 4, according to the meter identification method for substation inspection of the embodiment of the invention, the meter identification method includes the following steps:
s1, acquiring a meter image;
in one embodiment, said acquiring a meter image further comprises the steps of:
s11, configuring and connecting front-end shooting equipment, and remotely controlling the shooting equipment;
s12, calling equipment capable of shooting the meter, focusing on a meter dial by changing the angle and the focal length of the equipment, configuring shooting point positions capable of imaging clearly and recording;
s13, configuring the meter type of the shooting point;
s14, calling a shooting device for shooting a point location, and acquiring an initial meter image of the point location in a picture capturing or video stream pulling analysis mode;
and S15, decoding and preprocessing the initial meter image, acquiring the meter image, and sending the meter image to a downstream module for analysis and identification.
S2, detecting characteristic points of the meter image, and matching the characteristic points with the meter template image;
in one embodiment, the detecting the feature points of the meter image and matching with the meter template image further comprises the following steps:
s21, acquiring a meter image and a meter type through a meter shooting point, and retrieving a corresponding meter template image from a meter template image library based on the mark type;
s22, sending the meter image and the meter template image into a graph neural network feature matching module, and calculating feature points and corresponding relations of the meter image and the meter template image;
s23, matching and aligning the meter image with the meter template image, and outputting the aligned meter image;
s24, obtaining the coordinate area position of each component of the meter image from the meter template configuration;
each assembly of the meter image comprises a dial, scales, a center and a pointer.
In one embodiment, the step of sending the meter image and the meter template image to the graph neural network feature matching module, and calculating the feature points and the corresponding relationship between the two further includes the following steps:
s221, respectively collecting images X of the meter 1 And Meter template image X 2 Inputting the input dimension W multiplied by H matched with the Resize convolution neural network, and after normalization, inputting the input dimension W multiplied by H into a convolution neural network model;
s222, extracting the acquired image X of the meter A Characteristic point of
And descriptor
；
S223, extracting a meter template image X B Characteristic point of
And descriptor
；
S224, constructing a multilayer perceptron of an encoder, and enabling the multilayer perceptron to be a multi-layer perceptron
And
and embedding the high-dimensional features into a multilayer perceptron to learn, and updating the high-dimensional feature vectors obtained by learning to the original descriptor.
In one embodiment, the learned formula is as follows:
wherein the symbols
Indicating assignment operation, left side
For descriptors updated after an assignment operation, right side
In order to update the descriptor before the update,
is a multi-layer perceptron encoder,
are characteristic points.
In one embodiment, the construction of the encoder multi-layer perceptron is to
And
embedding the high-dimensional characteristics into a multi-layer perceptron to learn, and updating the high-dimensional characteristic vectors obtained by learning toThe original descriptor further comprises the following steps:
s2241, introducing a multi-head attention mechanism and a graph neural network structure;
s2242, input pair through control graph neural network
And
respectively realizing a self-attention mechanism and a cross-attention mechanism;
s2243, through cycle alternate input, learning and updating image X A A descriptor of (1);
wherein, the first and the second end of the pipe are connected with each other,
two nodes both correspond to the edge formed by the descriptors in the same image;
In one embodiment, the formula for the cyclically alternating input is as follows:
wherein, the first and the second end of the pipe are connected with each other,
the representation is based on
A constructed graph neural network;
and
are respectively provided withAs an image X A Go to the first
A characteristic point is
To go to
And
calculating an output intermediate result by the layer;
represents polymerization
A descriptor of the feature points.
In specific application, after L (multiple) operations, the matching descriptor is finally obtained through linear projection
Can be calculated by the same principle
；
Where W and b are the weight and offset of the linear transformation, respectively, y A And y B GNN in image X, respectively A And image X B The final result obtained by the above calculation, F A And F B Respectively, are matching descriptors for subsequent feature point matching calculations.
The score matrix is then constructed
And an allocation matrix
For the allocation matrix
The meaning of which is to be understood as image X A M feature points and image X B The soft adjacent matrixes of the N characteristic points can finally obtain the matching relation of all the characteristic points; to calculate
First, a score matrix is calculated
Let us order
Represents a matching descriptor F A To (1) a
Element and F B For a general graph matching process, the most transmission principle can be used for reference, and the distribution matrix can be calculated by maximizing the cost function
：
Wherein the maximum cost function calculates the distribution matrix
The calculation formula of (a) is as follows:
and restrain
Each row and each column of the array are summed to be 1, and the acquisition of a 'soft' adjacency matrix is realized;
for a model inference, the assignment matrix can be used to this step
Calculating image X A And image X B Can match the image X with the image X A By image X B Converting the template to realize the alignment correction of the template and the template;
however, if in the model training phase, the distribution matrix can be minimized based on the calibrated True value (GT for short)
Negative log-likelihood function of (2), loss function of computational model
Carrying out model training through gradient back propagation, and updating each parameter weight of the model;
by constructing a learnable feature matching model, not only can the high-dimensional expression of feature points be learnt through a multilayer perceptron, and a self-attention mechanism and a cross-attention mechanism are introduced by taking a graph neural network as a carrier, the intrinsic and extrinsic deep-level expression capability of features can be better learnt, but also a data set can be constructed by collecting meter equipment of an actual power grid inspection scene as an object, model training and learning of a real scene can be carried out, and continuous promotion and optimization can be realized in the scene.
S3, calculating the pointer direction of the meter;
in one embodiment, the meter pointer pointing calculation further comprises the steps of:
s31, acquiring a pointer position area and a dial plate central point of the aligned meter image based on the meter template image;
s32, constructing a binary template (M) by taking the pointer position area (R) R ）；
S33, acquiring pointer position area images by taking the binary template as the template
；
And S34, calculating a binary image of the pointer position area image by an adaptive binarization method.
In one embodiment, the adaptive binarization method calculation comprises indirect calculation and direct calculation;
wherein the indirect calculation comprises:
performing linear detection in the binary image region to find a pointer line segment of the meter;
detecting the longest line segment by adopting an EDLines linear detection method;
after the pointer line segment is obtained, calculating the intersection point of the straight line and the scale mark region by constructing an equation set;
the direct calculation comprises the following steps:
knowing the central point of the dial, depicting an extension line from the central point of the dial to each pixel point of the binary image;
calculating a projection point of each pixel point of the dial plate central point pointing to the binary image falling to the scale mark area;
and (4) obtaining the intersection point of the scale mark area pointed by the pointer of the meter through the projection point aggregation analysis.
S4, analyzing the pointer reading and outputting in a structured mode;
in one embodiment, said analyzing the pointer reading and structuring the output further comprises the steps of:
s41, acquiring a scale mark area, scale point coordinates and numerical values configured by the meter template graph;
s42, performing approximate interpolation calculation on the intersection points and the configured scale points to obtain approximate readings of the pointer of the meter pointing to the intersection points;
and S43, outputting the reading result to the system in a structuralized mode, and recording the subsequent statistical analysis.
In summary, by means of the technical scheme of the invention, the intelligent meter identification method adopted by the invention can realize automatic and accurate meter reading identification in daily power grid inspection work, and record statistics and data analysis are simply and conveniently carried out. The operation and maintenance labor cost of manual meter reading can be reduced, and meanwhile, the efficiency and the reliability can be effectively improved. Compared with the traditional feature matching methods such as SIFT, ORB and the like, the method can extract the expression information of the high-dimensional vectors of the feature points, introduce a self-attention mechanism and a cross-attention mechanism, better learn the internal and external relations and the corresponding relations of the feature points, and further enhance the deep information expression capability of the feature points. The invention constructs a deep feature detection matching model for continuous learning, can construct a training data set aiming at a meter equipment image of a power grid inspection scene, enables the model to adapt to a real scene for learning, and continuously improves the accuracy of the model in the scene, which is also a characteristic that the traditional feature matching method cannot achieve. Compared with a Hough Line or LSD linear detection method adopted by the traditional method, the pointer direction calculation method of the meter equipment adopted by the invention not only adopts a new EDlines linear detection method, but also introduces a method for densely calculating and analyzing projection points, and has better robustness in an actual scene compared with the linear detection method.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and should not be taken as limiting the scope of the present invention, which is intended to cover any modifications, equivalents, improvements, etc. within the spirit and scope of the present invention.
Claims (10)
1. A meter identification method for substation inspection is characterized by comprising the following steps:
s1, collecting a meter image;
s2, detecting characteristic points of the meter image, and matching the characteristic points with the meter template image;
s3, calculating the pointer direction of the meter;
and S4, analyzing the pointer reading and outputting in a structured mode.
2. The meter identification method for substation inspection according to claim 1, wherein the step of acquiring the meter image further comprises the steps of:
s11, configuring and connecting front-end shooting equipment, and remotely controlling the shooting equipment;
s12, calling equipment capable of shooting the meter, focusing on a meter dial by changing the angle and the focal length of the equipment, configuring shooting point positions capable of imaging clearly and recording;
s13, configuring the meter type of the shooting point;
and S14, calling a shooting device for shooting the point location, and acquiring an initial meter image of the point location in a picture capturing or video stream pulling analysis mode.
3. The meter identification method for substation routing inspection according to claim 1, wherein the step of detecting the feature points of the meter image and matching the feature points with the meter template image further comprises the steps of:
s21, acquiring a meter image and a meter type through a meter shooting point, and retrieving a corresponding meter template image from a meter template image library based on the mark type;
s22, sending the meter image and the meter template image into a graph neural network feature matching module, and calculating feature points and corresponding relations of the meter image and the meter template image;
s23, matching and aligning the meter image with the meter template image, and outputting the aligned meter image;
s24, obtaining the coordinate area position of each component of the meter image from the meter template configuration;
each assembly of the meter image comprises a dial, scales, a center and a pointer.
4. The meter identification method for substation inspection according to claim 3, wherein the step of sending the meter image and the meter template image to the neural network feature matching module and calculating feature points and corresponding relations between the meter image and the meter template image further comprises the following steps:
s221, respectively collecting images X by the meter 1 And Meter template image X 2 Inputting the input dimension W multiplied by H matched with the Resize convolution neural network, and after normalization, inputting the input dimension W multiplied by H into a convolution neural network model;
s222, extracting the acquired image X of the meter A Characteristic point of
And the descriptor
；
S223, extracting a meter template image X B Characteristic point of
And the descriptor
；
S224, constructing a multilayer perceptron of an encoder, and enabling the multilayer perceptron to be a multi-layer perceptron
And
and embedding the high-dimensional characteristics into a multilayer perceptron to learn the high-dimensional characteristics, and updating the high-dimensional characteristic vectors obtained by learning to the original descriptor.
5. The meter identification method for substation inspection according to claim 4, wherein the learning formula is as follows:
wherein, the symbol
Indicating assignment operation, left side
For descriptors updated after an assignment operation, right side
To updateThe description of the preceding paragraphs should be read as follows,
is a multi-layer perceptron encoder,
are characteristic points.
6. The substation inspection tour-oriented meter identification method according to claim 4, wherein the encoder multilayer perceptron is constructed to identify the encoder multilayer perceptron
And
the method for learning the high-dimensional features by embedding the high-dimensional features into a multilayer perceptron and updating the learned high-dimensional feature vectors into the original descriptor further comprises the following steps:
s2241, introducing a multi-head attention mechanism and a graph neural network structure;
s2242, input pair through control graph neural network
And
respectively realizing a self-attention mechanism and a cross-attention mechanism;
s2243, through cycle alternate input, learning and updating image X A The descriptor of (1);
wherein the content of the first and second substances,
two nodes both correspond to the edge formed by the descriptors in the same image;
7. The meter identification method for substation inspection according to claim 6, wherein the formula of cyclic alternate input is as follows:
wherein the content of the first and second substances,
the representation is based on
A constructed graph neural network;
and
are respectively an image X A Go to the first
A characteristic point is
Go to the first
And
the intermediate result of the layer calculation output;
represents polymerization
A descriptor of the feature points.
8. The meter identification method for substation routing inspection according to claim 1, wherein the meter pointer direction calculation further comprises the following steps:
s31, based on the meter template image, a pointer position area and a dial plate central point of the aligned meter image can be obtained;
s32, taking the pointer position area as a binary template;
s33, acquiring a pointer position area image by taking the binary template as a template;
and S34, calculating a binary image of the pointer position area image by an adaptive binarization method.
9. The meter identification method for substation inspection according to claim 8, wherein the adaptive binarization method calculation comprises indirect calculation and direct calculation;
wherein the indirect calculation comprises:
performing linear detection in the binary image region to find a pointer line segment of the meter;
detecting the longest line segment by adopting an EDLines linear detection method;
after the pointer line segment is obtained, calculating the intersection point of the straight line and the scale mark region by constructing an equation set;
the direct calculation includes the steps of:
knowing the center point of the dial, depicting an extension line from the center point of the dial to each pixel point of the binary image;
calculating a projection point of each pixel point of the dial plate central point pointing to the binary image falling to the scale mark area;
and (4) obtaining the intersection point of the scale mark area pointed by the meter pointer through the projection point aggregation analysis.
10. The substation inspection tour-oriented meter identification method according to claim 1, wherein the analyzing and structuring the pointer reading output further comprises the steps of:
s41, acquiring scale mark areas, scale point coordinates and numerical values configured by the meter template drawing;
s42, performing approximate interpolation calculation on the intersection points and the configured scale points to obtain approximate readings of the pointer of the meter pointing to the intersection points;
and S43, outputting the reading result to the system in a structuralized mode, and recording the subsequent statistical analysis.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211081495.6A CN115171091A (en) | 2022-09-06 | 2022-09-06 | Meter identification method for substation inspection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211081495.6A CN115171091A (en) | 2022-09-06 | 2022-09-06 | Meter identification method for substation inspection |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115171091A true CN115171091A (en) | 2022-10-11 |
Family
ID=83482138
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211081495.6A Pending CN115171091A (en) | 2022-09-06 | 2022-09-06 | Meter identification method for substation inspection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115171091A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117114657A (en) * | 2023-10-23 | 2023-11-24 | 国网江西省电力有限公司超高压分公司 | Fault information early warning system and method based on power equipment inspection knowledge graph |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109508709A (en) * | 2018-09-29 | 2019-03-22 | 中国南方电网有限责任公司超高压输电公司检修试验中心 | A kind of simple pointer meter reading method based on machine vision |
| CN110826549A (en) * | 2019-11-04 | 2020-02-21 | 山东欧玛嘉宝电气科技有限公司 | Inspection robot instrument image identification method and system based on computer vision |
| CN113536895A (en) * | 2021-05-28 | 2021-10-22 | 立得空间信息技术股份有限公司 | Disc pointer meter identification method |
| CN114743189A (en) * | 2022-04-19 | 2022-07-12 | 中国联合网络通信集团有限公司 | Pointer instrument reading identification method and device, electronic equipment and storage medium |
-
2022
- 2022-09-06 CN CN202211081495.6A patent/CN115171091A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109508709A (en) * | 2018-09-29 | 2019-03-22 | 中国南方电网有限责任公司超高压输电公司检修试验中心 | A kind of simple pointer meter reading method based on machine vision |
| CN110826549A (en) * | 2019-11-04 | 2020-02-21 | 山东欧玛嘉宝电气科技有限公司 | Inspection robot instrument image identification method and system based on computer vision |
| CN113536895A (en) * | 2021-05-28 | 2021-10-22 | 立得空间信息技术股份有限公司 | Disc pointer meter identification method |
| CN114743189A (en) * | 2022-04-19 | 2022-07-12 | 中国联合网络通信集团有限公司 | Pointer instrument reading identification method and device, electronic equipment and storage medium |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117114657A (en) * | 2023-10-23 | 2023-11-24 | 国网江西省电力有限公司超高压分公司 | Fault information early warning system and method based on power equipment inspection knowledge graph |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109949316B (en) | Power grid equipment image weak supervision example segmentation method based on RGB-T fusion | |
| US20240153272A1 (en) | Method for detecting urine formed elements based on deep learning and context relation | |
| CN110659637A (en) | Electric energy meter number and label automatic identification method combining deep neural network and SIFT features | |
| CN115171091A (en) | Meter identification method for substation inspection | |
| CN112712052A (en) | Method for detecting and identifying weak target in airport panoramic video | |
| CN114882039A (en) | PCB defect identification method applied to automatic PCB sorting process | |
| CN115457556A (en) | Reading method for disc pointer type instrument of nuclear power plant | |
| CN113313176A (en) | Point cloud analysis method based on dynamic graph convolution neural network | |
| CN116205905B (en) | Power distribution network construction safety and quality image detection method and system based on mobile terminal | |
| Ni et al. | Multi-meter intelligent detection and recognition method under complex background | |
| CN114937153B (en) | Visual characteristic processing system and method based on neural network in weak texture environment | |
| CN115661446A (en) | Pointer instrument indication automatic reading system and method based on deep learning | |
| CN116188755A (en) | Instrument angle correction and reading recognition device based on deep learning | |
| CN112766305B (en) | Visual SLAM closed loop detection method based on end-to-end measurement network | |
| Zeng et al. | Evaluation of physical electrical experiment operation process based on YOLOv5 and ResNeXt cascade networks | |
| CN113936019A (en) | Method for estimating field crop yield based on convolutional neural network technology | |
| CN111414844B (en) | Container number identification method based on convolutional neural network | |
| CN115082770B (en) | Image center line structure extraction method based on machine learning | |
| CN113888741B (en) | Method for correcting rotating image of instrument in power distribution room | |
| CN117809169B (en) | Small-sample underwater sonar image classification method and model building method thereof | |
| CN113128460B (en) | Knowledge distillation-based multi-resolution pedestrian re-identification method | |
| CN117079178A (en) | Learning process sequence data imaging method based on convolutional neural network | |
| Huang et al. | Exploiting Memory-based Cross-Image Contexts for Salient Object Detection in Optical Remote Sensing Images | |
| CN115953727A (en) | Floc settling rate detection method and system, electronic equipment and medium | |
| Dai et al. | Intelligent Quality Inspection of Secondary Wiring based on Rotating Object Detection and Pairwise Matching Strategy |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20221011 |
|
| RJ01 | Rejection of invention patent application after publication |