CN112069886A - Transformer substation respirator state intelligent identification method and system - Google Patents
Transformer substation respirator state intelligent identification method and system Download PDFInfo
- Publication number
- CN112069886A CN112069886A CN202010756159.1A CN202010756159A CN112069886A CN 112069886 A CN112069886 A CN 112069886A CN 202010756159 A CN202010756159 A CN 202010756159A CN 112069886 A CN112069886 A CN 112069886A
- Authority
- CN
- China
- Prior art keywords
- respirator
- transformer substation
- judging
- area ratio
- denotes
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
-
- 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
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- 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
- G06V10/30—Noise filtering
-
- 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
- G06V10/56—Extraction of image or video features relating to colour
-
- 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
Abstract
The invention discloses a transformer substation respirator state intelligent identification method and a transformer substation respirator state intelligent identification system, wherein the method comprises the following steps: acquiring a transformer substation respirator image, and carrying out preprocessing operation on the acquired image to eliminate noise influence; optimizing a YOLO v3 deep neural network training model; identifying and extracting a respirator ROI area by using the trained model; and (3) performing HSV feature transformation on the ROI, analyzing the color component proportion in different states according to the inherent attribute of the color information of the respirator in an HSV color space, and comprehensively judging the running state of the respirator. The method and the system can automatically identify the operation state of the transformer substation respirator, reduce the workload of transformer substation operation and maintenance personnel, replace or partially replace the workload of inspection personnel, and prompt and provide warning information.
Description
Technical Field
The invention relates to the field of power system automation, in particular to a transformer substation respirator state intelligent identification method and system.
Background
With the continuous expansion of the power grid scale, the progress of the smart power grid is continuously accelerated, and the power grid operation and maintenance face the challenges of increasing devices, but the number of people is continuously reduced. The traditional transformer respirator lacks a monitoring means, the state of the transformer respirator is determined and the fault defect is found by completely depending on manual visual observation, and the respirator is the guarantee of stable work of the transformer and has important influence on the stable operation of electric power. The manual confirmation mode cannot meet the requirement of modern power development.
The common respirator mainly comprises a glass cover tank, an oil seal cup, a connecting pipe, an isolation net, an adsorbent (silica gel), a sealing ring and the like. The adsorbent (silica gel) is used as a drying agent, is blue under a normal state, turns pink or pink white after absorbing a tide device, and can generate brownish yellow, brownish black and the like under individual conditions. The color of the sorbent may be indicative of the operating state of the respirator. How in the actual work, the respirator state is discerned to timely high efficiency to change to the respirator that fails and put forward higher requirement to the personnel of patrolling and examining. The situation of untimely detection or missed detection may occur in a pure manual mode, and further damage is caused to the transformer.
Disclosure of Invention
The invention aims to provide a transformer substation respirator state intelligent identification method and system, which are used for solving at least one technical problem, can automatically identify the operation state of a transformer substation respirator, reduce the workload of transformer substation operation and maintenance personnel, replace or partially replace the workload of inspection personnel, and prompt and provide warning information.
In a first aspect, an embodiment of the present application provides a transformer substation respirator state intelligent identification method, including:
the method comprises the steps of collecting images of a respirator of the transformer substation, preprocessing the obtained images and eliminating noise influence.
Optimizing the YOLO v3 deep neural network training model.
The trained model is used to identify and extract the respirator ROI region.
And (3) performing HSV feature transformation on the ROI, analyzing the color component proportion in different states according to the inherent attribute of the color information of the respirator in an HSV color space, and comprehensively judging the running state of the respirator.
With reference to the first aspect, the present application provides a first possible implementation manner of the first aspect, where the acquiring a transformer substation respirator image, performing a preprocessing operation on the acquired image, and eliminating noise influence includes:
and acquiring transformer substation respirator images containing various positions, heights and angles by using the inspection robot.
And labeling the image by using a Yolo _ mark or a labelImg, and storing the image in a txt format.
The images of the breathing apparatuses in various states of the transformer substation are collected, and the mounting positions of the breathing apparatuses in the actual transformer substation are changeable and have various angles and heights. At present, intelligent transformer substations are equipped with intelligent inspection robots to acquire image information at a certain fixed position and fixed time. The data is derived from a large number of respirator images shot by the inspection robot of an intelligent substation within a period of time.
With reference to the first aspect, an embodiment of the present application provides a second possible implementation manner of the first aspect, where the optimizing a YOLO v3 deep neural network training model includes:
and keeping the first time and the second time of the residual blocks unchanged, modifying the residual blocks in the third time and the fourth time of downsampling from 8 to 3, and modifying the residual blocks in the fifth time of downsampling from 4 to 2.
The feature extraction layer in the detection head is kept as 1/3.
And (4) utilizing a network clipping method to perform clipping on the channel dimension.
And generating the anchors size of the respirator by using a k-Means clustering method and combining the actual targets of the respirator to replace the anchors in the configuration file.
Wherein, YOLO (You only look once) is a typical representation in a first-order detection model, and can well balance precision and speed in engineering application.
In combination with the first aspect, the present application provides a third possible implementation manner of the first aspect, where the identifying and extracting a respirator ROI region using a trained model includes:
and recognizing the image to be detected by using the trained model, and detecting the target of the respirator.
And intercepting a target ROI area of the respirator on the original image according to the position information output by the detection result.
The optimized YOLO neural network is used for training, the network input is an image to be detected, and the network output is confidence coefficient and position information of the respirator.
With reference to the first aspect, the present application provides a fourth possible implementation manner of the first aspect, where the analyzing, by using HSV feature transformation for the ROI area, color component proportions in different states according to an inherent attribute of color information of a ventilator in an HSV color space, and comprehensively determining an operating state of the ventilator includes:
and judging the three-color area ratio p, and if the three-color area ratio p is less than or equal to 20%, determining that the color of the respirator is abnormal and the silica gel needs to be replaced.
If the three-color area ratio p is larger than 20%, a median l is judged, and if the median l is smaller than or equal to 0, the respirator is determined to be not tightly sealed and the silica gel needs to be replaced.
If the median l is larger than 0, judging the ratio r of the monochromatic areas, and if the ratio r of the monochromatic areas is smaller than or equal to 33%, determining that the respirator needs to be replaced by silica gel.
And if the monochromatic area ratio r is larger than 33%, determining that the state of the respirator is normal.
Wherein the formula for judging the area ratio of three colors is
The formula for judging the median value is
The formula for judging the ratio of the monochromatic areas is
Where S denotes the number of pixels, subscript P denotes the pink component, subscript W denotes the pink-white component, subscript B denotes the blue component, subscript O denotes the total color component, and L denotes the number of rows.
And performing HSV (hue, saturation, value) feature transformation on the ROI, and judging the state of the respirator according to the distribution characteristics of the respirator in H, S, V three components under different states, wherein the judgment strategy mainly judges the state of the respirator according to the proportion and the position information of the blue component B, the pink component P and the pink-white component W.
In a second aspect, an embodiment of the present application further provides a transformer substation respirator state intelligent identification system, which is used for implementing the transformer substation respirator state intelligent identification method, and includes:
and the preprocessing module is used for acquiring the images of the respirator of the transformer substation, preprocessing the acquired images and eliminating noise influence.
And the model design module is used for optimizing the YOLO v3 deep neural network training model.
And the target extraction module is used for identifying and extracting a respirator ROI area by using the trained model.
And the state judgment module is used for performing HSV feature transformation on the ROI area, analyzing the color component proportion in different states according to the inherent attribute of the color information of the respirator in an HSV color space, and comprehensively judging the running state of the respirator.
With reference to the second aspect, embodiments of the present application provide a first possible implementation manner of the second aspect, where the preprocessing module includes:
and the acquisition unit is used for acquiring the transformer substation respirator images containing various positions, heights and angles by using the inspection robot.
And the labeling unit labels the image by using a Yolo _ mark or a labelImg and stores the image in a txt format.
With reference to the second aspect, embodiments of the present application provide a second possible implementation manner of the second aspect, where the model design module includes:
and the residual block optimization unit is used for keeping the first residual block and the second residual block unchanged, modifying the residual blocks in the third downsampling for the fourth time from 8 to 3, and modifying the residual blocks in the fifth downsampling from 4 to 2.
And the feature extraction layer optimization unit is used for keeping the feature extraction layer in the detection head as the original 1/3.
And the channel dimension optimization unit is used for cutting the channel dimension by using a network cutting method.
And the Anchors optimizing unit is used for generating the Anchors size of the respirator by utilizing a k-Means clustering method and combining the actual targets of the respirator to replace the Anchors in the configuration file.
With reference to the second aspect, an embodiment of the present application provides a third possible implementation manner of the second aspect, where the target extraction module includes:
and the recognition detection unit is used for recognizing the image to be detected by using the trained model to detect the target of the respirator.
And the extraction unit is used for intercepting the ROI area of the respirator target on the original image according to the position information output by the detection result.
With reference to the second aspect, an embodiment of the present application provides a fourth possible implementation manner of the second aspect, where the state determining module includes:
and the three-color area ratio p judging unit is used for judging the three-color area ratio p, and if the three-color area ratio p is less than or equal to 20%, the color of the respirator is determined to be abnormal, and the silica gel needs to be replaced.
And the median l judgment unit is used for judging the median l when the three-color area ratio p is greater than 20%, and if the median l is less than or equal to 0, determining that the respirator is not tightly sealed and the silica gel needs to be replaced.
And the monochromatic area ratio r judging unit is used for judging the monochromatic area ratio r when the median l is larger than 0, determining that the respirator needs to be replaced by silica gel if the monochromatic area ratio r is smaller than or equal to 33%, and determining that the state of the respirator is normal if the monochromatic area ratio r is larger than 33%.
Wherein the formula for judging the area ratio of three colors is
The formula for judging the median value is
The formula for judging the ratio of the monochromatic areas is
Where S denotes the number of pixels, subscript P denotes the pink component, subscript W denotes the pink-white component, subscript B denotes the blue component, subscript O denotes the total color component, and L denotes the number of rows.
The embodiment of the invention has the beneficial effects that:
according to the transformer substation respirator state intelligent identification method and system, a large amount of sample data and an optimized yolov3 network are used for obtaining model weight parameters, HSV space conversion is carried out through OpenCV, a judgment strategy is formulated on the basis of analyzing color information of different samples and used for respirator state identification, the operation state of a transformer substation respirator can be automatically identified, the workload of operation and maintenance personnel can be effectively reduced, the workload of inspection personnel is replaced or partially replaced, warning information is provided in a prompt mode, and missing detection and false detection of personnel are avoided.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
The transformer substation respirator state intelligent identification method and system of the present invention are further described in detail with reference to the accompanying drawings and the detailed description.
FIG. 1 is a general flowchart of the intelligent identification method for the respirator state of the transformer substation of the present invention;
FIG. 2 is a schematic block diagram of the intelligent transformer substation respirator state identification method of the invention;
FIG. 3 is a diagram of a normal state HSV distribution map in the intelligent identification method for the state of the respirator of the transformer substation;
FIG. 4 is a pink HSV distribution diagram in the intelligent identification method for the state of the transformer substation respirator;
fig. 5 is a white HSV distribution diagram in the intelligent identification method for the respirator state of the transformer substation.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein can be arranged and designed in a wide variety of different configurations.
Some embodiments of the invention are described in detail below with reference to the accompanying drawings. The embodiments described below and the features of the embodiments can be combined with each other without conflict.
Referring to fig. 1 to 5, a first embodiment of the present invention provides a transformer substation respirator state intelligent identification method, including:
and S1, acquiring the transformer substation respirator image, and carrying out preprocessing operation on the acquired image to eliminate noise influence.
S2, optimizing a YOLO v3 deep neural network training model.
And S3, identifying and extracting a respirator ROI area by using the trained model.
And S4, performing HSV feature transformation on the ROI, analyzing the color component proportion in different states according to the inherent attribute of the color information of the respirator in an HSV color space, and comprehensively judging the running state of the respirator.
With reference to the first aspect, the present application provides a first possible implementation manner of the first aspect, where the acquiring a transformer substation respirator image, performing a preprocessing operation on the acquired image, and eliminating noise influence includes:
and S11, acquiring transformer substation respirator images containing various positions, heights and angles by using the inspection robot.
S12, labeling the image by using a Yolo _ mark or a labelImg, and storing the image in a txt format.
The images of the breathing apparatuses in various states of the transformer substation are collected, and the mounting positions of the breathing apparatuses in the actual transformer substation are changeable and have various angles and heights. At present, intelligent transformer substations are equipped with intelligent inspection robots to acquire image information at a certain fixed position and fixed time. The data is derived from a large number of respirator images shot by the inspection robot of an intelligent substation within a period of time.
With reference to the first aspect, an embodiment of the present application provides a second possible implementation manner of the first aspect, where the optimizing a YOLO v3 deep neural network training model includes:
s21, the first and second residual blocks are kept unchanged, the number of residual blocks in the third and fourth downsampling is changed from 8 to 3, and the number of residual blocks in the fifth downsampling is changed from 4 to 2.
And S22, keeping the feature extraction layer in the detection head as the original 1/3.
And S23, cutting on the channel dimension by using a network cutting method.
And S24, generating the anchors size of the respirator by utilizing a k-Means clustering method and combining the actual targets of the respirator, and replacing the anchors in the configuration file.
Wherein the network is modified on the basis of Yolov3-spp for this particular goal of the respirator.
Wherein, YOLO (You only look once) is a typical representation in a first-order detection model, and can well balance precision and speed in engineering application.
In combination with the first aspect, the present application provides a third possible implementation manner of the first aspect, where the identifying and extracting a respirator ROI region using a trained model includes:
and S31, recognizing the image to be detected by using the trained model, and detecting the target of the respirator.
And S32, intercepting the ROI area of the respirator target on the original image according to the position information output by the detection result.
The optimized YOLO neural network is used for training, the network input is an image to be detected, and the network output is confidence coefficient and position information of the respirator.
With reference to the first aspect, the present application provides a fourth possible implementation manner of the first aspect, where the analyzing, by using HSV feature transformation for the ROI area, color component proportions in different states according to an inherent attribute of color information of a ventilator in an HSV color space, and comprehensively determining an operating state of the ventilator includes:
and S41, judging the three-color area ratio p, and if the three-color area ratio p is less than or equal to 20%, determining that the color of the respirator is abnormal and the silica gel needs to be replaced.
S42, if the three-color area ratio p is larger than 20%, judging a median l, and if the median l is smaller than or equal to 0, determining that the respirator is not tightly sealed and the silica gel needs to be replaced.
And S43, if the median l is larger than 0, judging the ratio r of the monochromatic areas, and if the ratio r of the monochromatic areas is smaller than or equal to 33%, determining that the respirator needs to be replaced by silica gel.
And S44, if the monochromatic area ratio r is larger than 33%, determining that the state of the respirator is normal.
Wherein the formula for judging the area ratio of three colors is
The formula for judging the median value is
The formula for judging the ratio of the monochromatic areas is
Where S denotes the number of pixels, subscript P denotes the pink component, subscript W denotes the pink-white component, subscript B denotes the blue component, subscript O denotes the total color component, and L denotes the number of rows.
And performing HSV (hue, saturation, value) feature transformation on the ROI, and judging the state of the respirator according to the distribution characteristics of the respirator in H, S, V three components under different states, wherein the judgment strategy mainly judges the state of the respirator according to the proportion and the position information of the blue component B, the pink component P and the pink-white component W.
Referring to fig. 1 to 5, a second embodiment of the present invention provides a transformer substation respirator state intelligent identification system, which is used for implementing the transformer substation respirator state intelligent identification method, and includes:
and the preprocessing module is used for acquiring the images of the respirator of the transformer substation, preprocessing the acquired images and eliminating noise influence.
And the model design module is used for optimizing the YOLO v3 deep neural network training model.
Wherein the network is modified on the basis of Yolov3-spp for this particular goal of the respirator.
And the target extraction module is used for identifying and extracting a respirator ROI area by using the trained model.
And the state judgment module is used for performing HSV feature transformation on the ROI area, analyzing the color component proportion in different states according to the inherent attribute of the color information of the respirator in an HSV color space, and comprehensively judging the running state of the respirator.
With reference to the second aspect, embodiments of the present application provide a first possible implementation manner of the second aspect, where the preprocessing module includes:
and the acquisition unit is used for acquiring the transformer substation respirator images containing various positions, heights and angles by using the inspection robot.
And the labeling unit labels the image by using a Yolo _ mark or a labelImg and stores the image in a txt format.
With reference to the second aspect, embodiments of the present application provide a second possible implementation manner of the second aspect, where the model design module includes:
and the residual block optimization unit is used for keeping the first residual block and the second residual block unchanged, modifying the residual blocks in the third downsampling for the fourth time from 8 to 3, and modifying the residual blocks in the fifth downsampling from 4 to 2.
And the feature extraction layer optimization unit is used for keeping the feature extraction layer in the detection head as the original 1/3.
And the channel dimension optimization unit is used for cutting the channel dimension by using a network cutting method.
And the Anchors optimizing unit is used for generating the Anchors size of the respirator by utilizing a k-Means clustering method and combining the actual targets of the respirator to replace the Anchors in the configuration file.
With reference to the second aspect, an embodiment of the present application provides a third possible implementation manner of the second aspect, where the target extraction module includes:
and the recognition detection unit is used for recognizing the image to be detected by using the trained model to detect the target of the respirator.
And the extraction unit is used for intercepting the ROI area of the respirator target on the original image according to the position information output by the detection result.
With reference to the second aspect, an embodiment of the present application provides a fourth possible implementation manner of the second aspect, where the state determining module includes:
and the three-color area ratio p judging unit is used for judging the three-color area ratio p, and if the three-color area ratio p is less than or equal to 20%, the color of the respirator is determined to be abnormal, and the silica gel needs to be replaced.
And the median l judgment unit is used for judging the median l when the three-color area ratio p is greater than 20%, and if the median l is less than or equal to 0, determining that the respirator is not tightly sealed and the silica gel needs to be replaced.
And the monochromatic area ratio r judging unit is used for judging the monochromatic area ratio r when the median l is larger than 0, determining that the respirator needs to be replaced by silica gel if the monochromatic area ratio r is smaller than or equal to 33%, and determining that the state of the respirator is normal if the monochromatic area ratio r is larger than 33%.
Wherein the formula for judging the area ratio of three colors is
The formula for judging the median value is
The formula for judging the ratio of the monochromatic areas is
Where S denotes the number of pixels, subscript P denotes the pink component, subscript W denotes the pink-white component, subscript B denotes the blue component, subscript O denotes the total color component, and L denotes the number of rows.
The embodiment of the invention aims to protect a transformer substation respirator state intelligent identification method and a transformer substation respirator state intelligent identification system, and the transformer substation respirator state intelligent identification method and the transformer substation respirator state intelligent identification system have the following effects:
the method comprises the steps of obtaining model weight parameters by utilizing a large amount of sample data and an optimized yolov3 network, carrying out HSV space conversion by utilizing OpenCV, making a judgment strategy on the basis of analyzing color information of different samples, identifying the state of a respirator, automatically identifying the running state of the respirator of the transformer substation, effectively lightening the workload of operation and maintenance personnel, replacing or partially replacing the workload of inspection personnel, prompting to provide warning information, and avoiding missing detection and false detection of personnel.
The computer program product of the method and the system for intelligently identifying the state of the transformer substation respirator provided by the embodiment of the application comprises a computer readable storage medium storing program codes, instructions included in the program codes can be used for executing the method in the previous method embodiment, and specific implementation can refer to the method embodiment, and is not described herein again.
In particular, the storage medium can be a general storage medium, such as a mobile disk, a hard disk, and the like, and when a computer program on the storage medium is executed, the method can be executed, so that the state of the transformer substation respirator can be intelligently identified.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present application, and are used for illustrating the technical solutions of the present application, but not limiting the same, and the scope of the present application is not limited thereto, and although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope disclosed in the present application; such modifications, changes or substitutions do not depart from the spirit and scope of the exemplary embodiments of the present application, and are intended to be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (10)
1. A transformer substation respirator state intelligent identification method is characterized by comprising the following steps:
acquiring a transformer substation respirator image, and carrying out preprocessing operation on the acquired image to eliminate noise influence;
optimizing a YOLO v3 deep neural network training model;
identifying and extracting a respirator ROI area by using the trained model;
and (3) performing HSV feature transformation on the ROI, analyzing the color component proportion in different states according to the inherent attribute of the color information of the respirator in an HSV color space, and comprehensively judging the running state of the respirator.
2. The transformer substation respirator state intelligent identification method according to claim 1, wherein the collecting transformer substation respirator images, performing preprocessing operation on the obtained images, and eliminating noise influence comprises:
acquiring transformer substation respirator images containing various positions, heights and angles by using an inspection robot;
and labeling the image by using a Yolo _ mark or a labelImg, and storing the image in a txt format.
3. The transformer substation respirator state intelligent recognition method of claim 1, wherein the optimized YOLO v3 deep neural network training model comprises:
keeping the first time and the second time of the residual blocks unchanged, modifying the residual blocks in the third time and the fourth time of downsampling from 8 to 3, and modifying the residual blocks in the fifth time of downsampling from 4 to 2;
keeping the feature extraction layer in the detection head as the original 1/3;
cutting on the channel dimension by using a network cutting method;
and generating the anchors size of the respirator by using a k-Means clustering method and combining the actual targets of the respirator to replace the anchors in the configuration file.
4. The substation respirator state intelligent identification method according to claim 1, wherein the identifying and extracting respirator ROI areas using the trained model comprises:
recognizing an image to be detected by using the trained model, and detecting a target of the respirator;
and intercepting a target ROI area of the respirator on the original image according to the position information output by the detection result.
5. The transformer substation respirator state intelligent identification method according to claim 1, wherein the ROI region is transformed by using HSV features, and color component specific gravity in different states is analyzed according to inherent properties of color information of a respirator in an HSV color space, so as to comprehensively judge the operating state of the respirator, and the method comprises the following steps:
judging the area ratio p of three colors, and if the area ratio p of three colors is less than or equal to 20%, determining that the color of the respirator is abnormal and the silica gel needs to be replaced;
if the area ratio p of the three colors is more than 20%, judging a median l, and if the median l is less than or equal to 0, determining that the respirator is not tightly sealed and the silica gel needs to be replaced;
if the median l is larger than 0, judging the ratio r of the monochromatic areas, and if the ratio r of the monochromatic areas is smaller than or equal to 33%, determining that the respirator needs to be replaced by silica gel;
if the monochromatic area ratio r is larger than 33%, determining that the state of the breathing machine is normal;
wherein the judgment is madeThe formula of the area ratio of three colors is
The formula for judging the median value is
The formula for judging the ratio of the monochromatic areas is
Where S denotes the number of pixels, subscript P denotes the pink component, subscript W denotes the pink-white component, subscript B denotes the blue component, subscript O denotes the total color component, and L denotes the number of rows.
6. A transformer substation respirator state intelligent identification system for realizing the transformer substation respirator state intelligent identification method of any one of claims 1 to 5, which is characterized by comprising the following steps:
the preprocessing module is used for acquiring images of the respirator of the transformer substation, preprocessing the acquired images and eliminating noise influence;
the model design module is used for optimizing a YOLO v3 deep neural network training model;
the target extraction module is used for identifying and extracting a respirator ROI area by using the trained model;
and the state judgment module is used for performing HSV feature transformation on the ROI area, analyzing the color component proportion in different states according to the inherent attribute of the color information of the respirator in an HSV color space, and comprehensively judging the running state of the respirator.
7. The substation respirator state intelligent recognition system of claim 6, wherein the preprocessing module comprises:
the acquisition unit is used for acquiring transformer substation respirator images containing various positions, heights and angles by using the inspection robot;
and the labeling unit labels the image by using a Yolo _ mark or a labelImg and stores the image in a txt format.
8. The substation respirator state intelligent recognition system of claim 6, wherein the model design module comprises:
a residual block optimization unit, configured to keep the first and second residual blocks unchanged, modify 8 residual blocks from 8 to 3 residual blocks when the third downsampling is performed for the fourth time, and modify 4 residual blocks from 2 residual blocks when the fifth downsampling is performed;
the characteristic extraction layer optimization unit is used for keeping the characteristic extraction layer in the detection head as 1/3;
the channel dimension optimizing unit is used for cutting the channel dimension by using a network cutting method;
and the Anchors optimizing unit is used for generating the Anchors size of the respirator by utilizing a k-Means clustering method and combining the actual targets of the respirator to replace the Anchors in the configuration file.
9. The substation respirator state intelligent recognition system of claim 6, wherein the target extraction module comprises:
the identification detection unit is used for identifying the image to be detected by using the trained model and detecting the target of the respirator;
and the extraction unit is used for intercepting the ROI area of the respirator target on the original image according to the position information output by the detection result.
10. The transformer substation respirator state intelligent recognition system of claim 6, wherein the state judgment module comprises:
the three-color area ratio p judging unit is used for judging the three-color area ratio p, and if the three-color area ratio p is less than or equal to 20%, the color of the respirator is determined to be abnormal, and the silica gel needs to be replaced;
the median l judging unit is used for judging the median l when the area ratio p of three colors is greater than 20%, and if the median l is less than or equal to 0, determining that the respirator is not tightly sealed and the silica gel needs to be replaced;
the monochromatic area ratio r judging unit is used for judging the monochromatic area ratio r when the median value l is larger than 0, determining that the respirator needs to be replaced by silica gel if the monochromatic area ratio r is smaller than or equal to 33%, and determining that the respirator is normal if the monochromatic area ratio r is larger than 33%;
wherein the formula for judging the area ratio of three colors is
The formula for judging the median value is
The formula for judging the ratio of the monochromatic areas is
Where S denotes the number of pixels, subscript P denotes the pink component, subscript W denotes the pink-white component, subscript B denotes the blue component, subscript O denotes the total color component, and L denotes the number of rows.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010756159.1A CN112069886A (en) | 2020-07-31 | 2020-07-31 | Transformer substation respirator state intelligent identification method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010756159.1A CN112069886A (en) | 2020-07-31 | 2020-07-31 | Transformer substation respirator state intelligent identification method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112069886A true CN112069886A (en) | 2020-12-11 |
Family
ID=73656740
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010756159.1A Pending CN112069886A (en) | 2020-07-31 | 2020-07-31 | Transformer substation respirator state intelligent identification method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112069886A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113807348A (en) * | 2021-08-24 | 2021-12-17 | 许继集团有限公司 | High-voltage cable target identification and positioning method and device |
| CN114037887A (en) * | 2021-09-26 | 2022-02-11 | 佛山中科云图智能科技有限公司 | State detection method and device for transformer respirator, medium and electronic equipment |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110707823A (en) * | 2019-11-08 | 2020-01-17 | 国网上海市电力公司 | Multi-dimensional intelligent remote inspection system for transformer substation |
| CN111046886A (en) * | 2019-12-12 | 2020-04-21 | 吉林大学 | Automatic identification method, device and equipment for number plate and computer readable storage medium |
| CN111062282A (en) * | 2019-12-05 | 2020-04-24 | 武汉科技大学 | Transformer substation pointer type instrument identification method based on improved YOLOV3 model |
| CN111080634A (en) * | 2019-12-23 | 2020-04-28 | 北京新松融通机器人科技有限公司 | Transformer appearance defect identification method based on inspection robot and Cascade RCNN algorithm |
| CN111104906A (en) * | 2019-12-19 | 2020-05-05 | 南京工程学院 | Transmission tower bird nest fault detection method based on YOLO |
| CN111210974A (en) * | 2020-02-10 | 2020-05-29 | 国网河南省电力公司检修公司 | Transformer respiratory system |
| CN111275124A (en) * | 2020-02-10 | 2020-06-12 | 国网河南省电力公司检修公司 | Transformer respirator detection method, classification model construction method, computer readable medium and transformer respirator monitoring system |
-
2020
- 2020-07-31 CN CN202010756159.1A patent/CN112069886A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110707823A (en) * | 2019-11-08 | 2020-01-17 | 国网上海市电力公司 | Multi-dimensional intelligent remote inspection system for transformer substation |
| CN111062282A (en) * | 2019-12-05 | 2020-04-24 | 武汉科技大学 | Transformer substation pointer type instrument identification method based on improved YOLOV3 model |
| CN111046886A (en) * | 2019-12-12 | 2020-04-21 | 吉林大学 | Automatic identification method, device and equipment for number plate and computer readable storage medium |
| CN111104906A (en) * | 2019-12-19 | 2020-05-05 | 南京工程学院 | Transmission tower bird nest fault detection method based on YOLO |
| CN111080634A (en) * | 2019-12-23 | 2020-04-28 | 北京新松融通机器人科技有限公司 | Transformer appearance defect identification method based on inspection robot and Cascade RCNN algorithm |
| CN111210974A (en) * | 2020-02-10 | 2020-05-29 | 国网河南省电力公司检修公司 | Transformer respiratory system |
| CN111275124A (en) * | 2020-02-10 | 2020-06-12 | 国网河南省电力公司检修公司 | Transformer respirator detection method, classification model construction method, computer readable medium and transformer respirator monitoring system |
Non-Patent Citations (2)
| Title |
|---|
| 李瑞生 等: "基于HSV特征变换与目标检测的变压器呼吸器缺陷智能识别方法", 《高电压技术》, 3 March 2020 (2020-03-03), pages 1 - 9 * |
| 石青 等: "《微小型仿生机器鼠设计与控制》", 31 December 2019, pages: 192 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113807348A (en) * | 2021-08-24 | 2021-12-17 | 许继集团有限公司 | High-voltage cable target identification and positioning method and device |
| CN114037887A (en) * | 2021-09-26 | 2022-02-11 | 佛山中科云图智能科技有限公司 | State detection method and device for transformer respirator, medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110070033B (en) | Method for detecting wearing state of safety helmet in dangerous working area in power field | |
| CN111259892B (en) | Inspection method, inspection device, inspection equipment and inspection medium for state of indicator lamp | |
| CN112069886A (en) | Transformer substation respirator state intelligent identification method and system | |
| CN111401314B (en) | Dressing information detection method, device, equipment and storage medium | |
| CN107784310A (en) | Status information of equipment acquisition methods, device, system, storage medium and robot | |
| CN107832770A (en) | A kind of equipment routing inspection method, apparatus, system, storage medium and crusing robot | |
| CN101625723A (en) | Rapid image-recognizing method of power line profile | |
| CN108416294A (en) | A kind of fan blade fault intelligent identification method based on deep learning | |
| CN111080634A (en) | Transformer appearance defect identification method based on inspection robot and Cascade RCNN algorithm | |
| CN111401419A (en) | Improved RetinaNet-based employee dressing specification detection method | |
| CN111080600A (en) | Fault identification method for split pin on spring supporting plate of railway wagon | |
| JP2016223815A (en) | Deterioration diagnostic system and deterioration diagnostic method | |
| CN107679495A (en) | A kind of detection method of transmission line of electricity periphery activity engineering truck | |
| CN113076683A (en) | Modeling method of convolutional neural network model for substation behavior monitoring | |
| CN106846304B (en) | Electrical equipment detection method and device based on infrared detection | |
| CN116543241B (en) | Detection method and device for leakage gas cloud, storage medium and electronic equipment | |
| CN111178424A (en) | Petrochemical production site safety compliance real-time detection system and method | |
| CN116740654B (en) | Substation operation prevention and control method based on image recognition technology | |
| CN116310274A (en) | State evaluation method for power transmission and transformation equipment | |
| CN112711793B (en) | Mountain cutting construction identification method and device | |
| CN111414825B (en) | Method for detecting wearing of safety helmet | |
| CN110969081B (en) | Power transmission line external force damage detection method based on KL divergence of multi-module division | |
| CN112211793A (en) | Wind turbine generator fault automatic identification method based on video image analysis | |
| Gao et al. | Design and implementation of fire detection system using new model mixing | |
| CN112905723A (en) | Analysis and visualization method for Micaps fourteenth type data |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201211 |