CN111080634A - Transformer appearance defect identification method based on inspection robot and Cascade RCNN algorithm - Google Patents
Transformer appearance defect identification method based on inspection robot and Cascade RCNN algorithm Download PDFInfo
- Publication number
- CN111080634A CN111080634A CN201911338035.5A CN201911338035A CN111080634A CN 111080634 A CN111080634 A CN 111080634A CN 201911338035 A CN201911338035 A CN 201911338035A CN 111080634 A CN111080634 A CN 111080634A
- Authority
- CN
- China
- Prior art keywords
- image
- transformer
- model
- inspection robot
- inspection
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
- G01N21/9515—Objects of complex shape, e.g. examined with use of a surface follower device
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8883—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges involving the calculation of gauges, generating models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
Abstract
The invention discloses a transformer appearance defect identification method based on a patrol robot and a Cascade RCNN algorithm, belonging to the field of image identification and comprising the following steps: s1: image acquisition, the operation and maintenance personnel and the engineer through in the power station adopt tools such as camera, handheld terminal, camera to carry out image acquisition to the appearance defect condition of transformer, S2: making an image book set, randomly dividing a training set and a testing set according to a certain proportion according to the acquired image of the transformer and the marked xml file, and respectively training a model and verifying the accuracy of the model, S3: training a Cascade RCNN model by using training set data, S4: the model is deployed on a centralized control platform of the inspection robot, and the high-definition camera acquires the appearance image of the transformer equipment and is used for inspecting the appearance state of the equipment in the transformer area. The invention verifies the feasibility of identifying the appearance defects of the transformer equipment based on the transformer substation inspection robot and the deep learning technology, and promotes the ground application of the artificial intelligence technology in the transformer substation.
Description
Technical Field
The invention relates to the field of image recognition, in particular to a transformer appearance defect recognition method based on a patrol robot and a Cascade RCNN algorithm.
Background
The intelligent inspection robot for the transformer substation has the vision capability, acquires the appearance image of equipment in a substation area of the transformer substation through a visible light imaging device arranged on the head, automatically tracks and monitors the equipment with the defects in real time, and an operation and maintenance person selects corresponding defective equipment through a remote client, sets a defect tracking task and selects a corresponding period to track and repeatedly inspect; or the robot is controlled to monitor the defect equipment at fixed points all day, so that the data of the defect equipment can be acquired in real time, and the workload of personnel is reduced.
The transformer substation inspection operation and maintenance mode is changed by replacing a robot with a robot, a large amount of data can be generated in the inspection process of the robot, the difference from the traditional inspection data is that the traditional inspection data is structured data such as voltage and current, and the machine inspection data is data such as images and videos, which is a challenge faced by a power grid in the data processing process, the traditional image processing and mode identification method is utilized to obtain good identification precision on meter reading and oil level state reading, but the intelligent identification capability of the appearance defect of equipment is still deficient, and the main reasons are as follows:
(1) because the transformer substation is of a complex structure, the number of devices in a factory area is large, so that background features in an image are difficult to distinguish;
(2) the method has the following problems that similar contour interference exists, such as the identification of a circular meter, but a large number of circular areas which are not meters exist in the background, and false identification is easily caused;
(3) the traditional algorithm has poor applicability to ambient light, especially in the case of low illumination or backlight;
(4) the expression capability of the artificial design characteristics to the target is insufficient, and the detection accuracy is low.
Disclosure of Invention
The invention aims to provide a transformer appearance defect identification method based on an inspection robot and a Cascade RCNN algorithm, so as to solve the problems in the background technology.
In order to achieve the purpose, the invention provides the following technical scheme: the transformer appearance defect identification method based on the inspection robot and the Cascade RCNN algorithm comprises the following steps:
s1: collecting an image;
s2: making an image album, comprising the steps of:
s2-1: labeling a defect image:
s2-2: data cleaning;
s2-3: data enhancement comprising the steps of:
s2-3-1: inputting an image img;
s2-3-2: setting an upper limit clipping area max;
s2-3-3: comparing the size of the image to be cut with the size of the upper limit cutting area, wherein the size comprises a width value and a height value;
s2-3-4: when the size of the image to be cut is larger than the upper limit cutting area, the image needs to be cut, and the object in the xml tag file of the image is analyzed firstly;
s2-3-5: finding a minimum bounding rectangle that encloses all objects;
s2-3-6: cutting out a minimum circumscribed rectangular area from an original image;
s2-3-7: calculating the coordinates of each object in the minimum circumscribed rectangle in the newly-cut image, and writing the coordinates into an xml tag of the newly-cut image;
s2-3-8: completing data enhancement;
s2-4: respectively zooming the original image and the enhanced image;
s2-5: combining the scaled original image and the scaled cutting image to produce an image sample set, wherein the image sample set comprises a training set and a testing set and is used for the following steps of model training and model testing;
s3: training a Cascade RCNN model by using training set data, and testing the model precision, wherein the method comprises the following steps:
s3-1: identifying the appearance defects of the transformer by selecting a Cascade RCNN target detection algorithm;
s3-2: utilizing the training sample;
s3-3: training to obtain a model;
s3-4: testing the precision of the model;
s3-5: judging whether the model meets the requirements of inspection operation and maintenance;
s4: deploying the model on a centralized control platform of the inspection robot, and comprising the following steps:
s4-1: the inspection robot takes a picture by visible light;
s4-2: deploying a server side;
s4-3: loading a model;
s4-4: picture reasoning;
s4-5: a defect identification result;
s4-6: the web service is sent to the centralized control platform;
s4-7: and (5) displaying the inspection result.
Preferably, in S1, the operation and maintenance personnel and engineers in the power station use tools such as a camera, a handheld terminal, and a camera to capture images of the appearance defects of the transformer.
Preferably, in S1, the image acquisition is mainly directed to the following common species defect types:
(1) oil stains leak from the surface of the component;
(2) metal corrosion on the surface of the part;
(3) changing color of the silica gel of the respirator;
(4) abnormal oil seal of the oil level of the respirator;
(5) the main transformer area foreign matter comprises suspended matter and a bird nest.
Preferably, in S2, the data enhancement is performed by image cropping.
Preferably, in S3, the Cascade RCNN algorithm belongs to a three-stage Cascade detection algorithm, and is more suitable for the cases where the target size difference is large and there is a small target, such as oil leakage in the appearance defect of the transformer and discoloration of the silica gel of the respirator, which belong to two defect types with large size difference.
Preferably, in S3, if the model does not meet the inspection operation and maintenance requirements, if there is a high missed inspection rate and a high false inspection rate, the hyper-parameters of the model need to be optimized and adjusted again, and the model needs to be trained again, and when the model indexes can meet the national grid operation and maintenance requirements, the model can be output and deployed to the remote server of the inspection robot.
Preferably, in S4, the intelligent inspection robot employs a visible light inspection function, that is, the high-definition camera collects an appearance image of the transformer device in the daytime, and the intelligent inspection robot is mainly used for inspecting the appearance state of the device in the transformer area.
Preferably, in S4, the identification result of the appearance defect of the transformer may be stored as inspection history data and a report is generated, and the operation and maintenance staff can grasp the operation status of the defect device at the operation and maintenance master station according to the defect report automatically generated and uploaded by the robot, and check the device status and report the scheduling processing according to the alarm information of the inspection robot.
Compared with the prior art, the invention has the beneficial effects that: the development of artificial intelligence and deep learning technology provides a new mode for detecting and identifying the appearance defects of the transformer substation equipment, a deep learning model can be established based on a large amount of data by using an image identification algorithm of deep learning, the extraction capability of a network on the high-level characteristics of images can be improved by using massive parameters in the network, the adaptability to the changeful target equipment of the transformer substation is strong, the image identification technology of the appearance defects of the transformer substation equipment has good application value and application prospect, and the transformer operation and maintenance inspection efficiency and the intelligentization level can be remarkably improved.
Drawings
FIG. 1 is an overall flow chart of the present invention;
FIG. 2 is a flow chart of a transformer appearance defect image acquisition and identification method based on Cascade RCNN algorithm of the invention;
FIG. 3 is a flow chart of data enhancement in the image cropping mode of the present invention;
fig. 4 is a flow chart showing the appearance defect image recognition and inspection result of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1 to 4, the present invention provides a technical solution: the transformer appearance defect identification method based on the inspection robot and the Cascade RCNN algorithm specifically comprises the following steps as shown in FIG. 1:
s1: collecting an image;
the method comprises the following steps that operation and maintenance personnel and engineers in a power station adopt tools such as a camera, a handheld terminal and a camera to acquire images of appearance defect conditions of a transformer;
image acquisition is mainly aimed at the following common species defect types:
(1) oil stains leak from the surface of the component;
(2) metal corrosion on the surface of the part;
(3) changing color of the silica gel of the respirator;
(4) abnormal oil seal of the oil level of the respirator;
(5) foreign matter in main transformer area, including suspended matter and bird nest
S2: the method for creating the image book set, as shown in fig. 2, comprises the following steps:
s2-1: labeling a defect image:
because the defect form on the transformer is complex, the labeling quality of a defect image sample directly influences the image identification effect, a clear and identifiable image with a correct focal length is selected for labeling, an RGB three-channel color mode is preferably adopted for the image, a labelImg labeling tool is selected for labeling the image, and an xml-format labeling file can be obtained;
s2-2: data cleaning;
the data cleaning aims at eliminating images which do not meet the requirements of model training, and the main contents comprise: converting the image extension JPG into a JPG format; rejecting incomplete defect labels in the image and labels with main defect features blocked; labels of the defective target boundary frames exceeding the image boundary are removed;
s2-3: data enhancement, the purpose of which is to expand a data set, the main contents of which include: the transformer image does not meet the up-down inversion invariance, and the image data is inverted left and right; the method for cutting the appearance defect image of the transformer specifically shown in fig. 3 comprises the following steps:
s2-3-1: inputting an image img;
inputting a transformer appearance defect image with an marked xml file;
s2-3-2: setting an upper limit clipping area max;
setting the upper limit size of a cutting area, wherein the width max _ w of the upper limit cutting area max is set to be 800, and the height max _ h is set to be 600;
s2-3-3: comparing the size of the image to be cut with the size of the upper limit cutting area, wherein the size comprises a width value and a height value;
when the following conditions are satisfied: when the img _ w < max _ w & & img _ h < max _ h condition, namely the image to be cut is smaller than the width and the height of the upper limit cutting area, cutting is abandoned and quitting is performed; if the condition is not met, executing the next step;
s2-3-4: firstly, analyzing an object in an xml tag file of an image;
analyzing the bndbox value in the defect target object in the image annotation xml file, namely determining the position of each object in the image in the annotation file;
s2-3-5: finding a minimum bounding rectangle that encloses all objects;
according to the positions of the objects in the image, in the bndbox, the minimum bounding rectangle which can surround all the objects is obtained and used as the image clipping area of the next step;
s2-3-6: cutting a minimum circumscribed rectangular area from an original image;
according to the determined image clipping area, clipping a small image in the area from an original image, and ensuring that the clipped outer contour does not exceed the boundary of the image;
s2-3-7: calculating the coordinates of each object in the minimum circumscribed rectangle in the newly-cut image, and writing the coordinates into an xml tag of the newly-cut image;
recalculating the bndbox position of each object in the newly-cropped image in the obtained newly-cropped image, calculating the bndbox coordinate position of each object in the newly-cropped image, and writing the bndbox coordinate position into an xml annotation file of the newly-cropped image;
s2-3-8: completing data enhancement;
s2-4: respectively zooming the original image and the enhanced image;
the image scaling aims at scaling the image within a set size on the premise of not setting the aspect ratio of the image in the sample set so as to improve the training speed of the model and ensure that the condition of super GPU video memory does not occur in the training process;
s2-5: combining the scaled original image and the scaled cutting image to produce an image sample set, wherein the image sample set comprises a training set and a testing set and is used for the following steps of model training and model testing;
taking the new cut image as a sample after data enhancement, combining the sample with the original image sample into a sample set, and making a finished image sample set, wherein the finished image sample set comprises a training set and a testing set;
s3: training a Cascade RCNN model by using training set data, and testing the model precision, wherein the method comprises the following steps:
s3-1: identifying the appearance defects of the transformer by selecting a Cascade RCNN target detection algorithm;
the Cascade RCNN algorithm belongs to a three-stage Cascade detection algorithm, is more suitable for the conditions that the target size difference is large and small targets exist, such as oil leakage in the appearance defect of a transformer and color change of silica gel of a respirator, belonging to two defect types with large size difference, and can improve the recall rate of the defect target by adopting the algorithm;
s3-2: utilizing the training sample;
in the embodiment of the invention, 8 GPU cards are adopted for model training, the type of the GPU card is NVIDIA RTX 2080Ti, and each GPU card has 12G video memory, so that the model convergence speed can be greatly improved;
s3-3: training to obtain a model;
drawing a loss curve of the model in a training stage, continuously reducing the learning rate, stopping model training when the loss value is not converged any more, and obtaining a Cascade RCNN target detection model capable of identifying 5 types of defects of the appearance of the transformer;
s3-4: testing the precision of the model;
evaluating the precision of the model by using the test sample set, and outputting the precision value, the recall rate and the ap value of each type of defect type to measure the detection precision and the recall condition of the model on the appearance defects of the 5 types of transformers;
s3-5: judging whether the model meets the requirements of inspection operation and maintenance;
evaluating the precision of the model by using the test sample set, and outputting the precision value, the recall rate and the ap value of each type of defect type to measure the detection precision and the recall condition of the model on the appearance defects of the 5 types of transformers;
if the model can not meet the requirements of inspection operation and maintenance, if high undetected rate and false undetected rate exist, the hyper-parameters of the model need to be optimized and adjusted again, and the model needs to be trained again;
when the model indexes can meet the national network operation and maintenance requirements, the model can be output and deployed to a remote server of the inspection robot;
s4: the model is deployed on a centralized control platform of the inspection robot, and specifically as shown in fig. 4, the method comprises the following steps:
s4-1: the inspection robot takes a picture by visible light;
the intelligent inspection robot of the transformer substation adopts a visible light inspection function, namely, an appearance image of transformer equipment is acquired by a high-definition camera in the daytime, and the intelligent inspection robot is mainly used for inspecting the appearance state of the equipment in a transformer area, such as oil stain leakage on the surface of a part, metal corrosion on the surface of the part, color change of silica gel of a respirator, abnormal oil seal of the oil level of the respirator and foreign matters in a main transformer area, including suspended objects, bird nests and the like;
s4-2: deploying a server side;
the transformer appearance defect identification method is deployed to a server side of the national grid operation and maintenance department, and the advantage of deploying the algorithm model at the server side is that the strong model loading and reasoning capacity of the algorithm model can be utilized;
s4-3: loading a model;
loading an algorithm model at a server side, and loading binary weight parameters of the model;
s4-4: picture reasoning;
when transformer appearance images shot by the inspection robot are acquired, real-time algorithm reasoning is carried out on each frame of image;
s4-5: a defect identification result;
if the transformer appearance image shot by the inspection robot has appearance defects, outputting the types of the defects and the coordinate positions of the defects in the image;
s4-6: the web service is sent to the centralized control platform;
the identification result of the appearance defect of the transformer is sent to a centralized control platform of an operation and maintenance master station computer in real time in a web service mode;
s4-7: the inspection result is displayed, the identification result of the appearance defect of the transformer is displayed on the interface of the centralized control platform, according to the serious program of the appearance defect of the transformer, the alarm is given out according to the area of leaked oil stains, the corrosion degree of metal corrosion, the abnormal degree of color change of silica gel of the respirator and the like, the identification result of the appearance defect of the transformer can be stored as inspection historical data and a report form is generated, operation and maintenance personnel can master the operation condition of the defect equipment at an operation and maintenance main station according to the defect report form automatically generated and uploaded by the robot, and the state of the equipment is checked and reported and scheduled according to the alarm information of the inspection robot.
According to the invention, through the development of artificial intelligence and deep learning technology, a new mode is provided for the detection and identification of the appearance defects of the substation equipment, a deep learning model can be established based on a large amount of data by using an image identification algorithm of deep learning, the extraction capability of the network on the high-level characteristics of the image can be improved by using massive parameters in the network, the adaptability to the target equipment with the changeability of the substation is stronger, the image identification technology of the appearance defects of the substation equipment has good application value and application prospect, and the substation operation and maintenance inspection efficiency and the intelligentization level can be remarkably improved.
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (8)
1. The transformer appearance defect identification method based on the inspection robot and the Cascade RCNN algorithm is characterized by comprising the following steps of:
s1: collecting an image;
s2: making an image album, comprising the steps of:
s2-1: labeling a defect image:
s2-2: data cleaning;
s2-3: data enhancement comprising the steps of:
s2-3-1: inputting an image img;
s2-3-2: setting an upper limit clipping area max;
s2-3-3: comparing the size of the image to be cut with the size of the upper limit cutting area, wherein the size comprises a width value and a height value;
s2-3-4: when the size of the image to be cut is larger than the upper limit cutting area, the image needs to be cut, and the object in the xml tag file of the image is analyzed firstly;
s2-3-5: finding a minimum bounding rectangle that encloses all objects;
s2-3-6: cutting out a minimum circumscribed rectangular area from an original image;
s2-3-7: calculating the coordinates of each object in the minimum circumscribed rectangle in the newly-cut image, and writing the coordinates into an xml tag of the newly-cut image;
s2-3-8: completing data enhancement;
s2-4: respectively zooming the original image and the enhanced image;
s2-5: combining the scaled original image and the scaled cutting image to produce an image sample set, wherein the image sample set comprises a training set and a testing set and is used for the following steps of model training and model testing;
s3: training a Cascade RCNN model by using training set data, and testing the model precision, wherein the method comprises the following steps:
s3-1: identifying the appearance defects of the transformer by selecting a Cascade RCNN target detection algorithm;
s3-2: utilizing the training sample;
s3-3: training to obtain a model;
s3-4: testing the precision of the model;
s3-5: judging whether the model meets the requirements of inspection operation and maintenance;
s4: deploying the model on a centralized control platform of the inspection robot, and comprising the following steps:
s4-1: the inspection robot takes a picture by visible light;
s4-2: deploying a server side;
s4-3: loading a model;
s4-4: picture reasoning;
s4-5: a defect identification result;
s4-6: the web service is sent to the centralized control platform;
s4-7: and (5) displaying the inspection result.
2. The transformer appearance defect identification method based on the inspection robot and the Cascade RCNN algorithm according to claim 1, characterized in that: in S1, the appearance defect condition of the transformer is captured by operation and maintenance personnel and engineers in the power station using tools such as cameras, handheld terminals, cameras, etc.
3. The transformer appearance defect identification method based on the inspection robot and the Cascade RCNN algorithm according to claim 1, characterized in that: in S1, the image acquisition is mainly directed to the following common species defect types:
(1) oil stains leak from the surface of the component;
(2) metal corrosion on the surface of the part;
(3) changing color of the silica gel of the respirator;
(4) abnormal oil seal of the oil level of the respirator;
(5) the main transformer area foreign matter comprises suspended matter and a bird nest.
4. The transformer appearance defect identification method based on the inspection robot and the Cascade RCNN algorithm according to claim 1, characterized in that: in S2, data enhancement is performed by image cropping.
5. The transformer appearance defect identification method based on the inspection robot and the Cascade RCNN algorithm according to claim 1, characterized in that: in S3, the Cascade RCNN algorithm belongs to a three-stage Cascade detection algorithm, and is more suitable for the conditions that the target size difference is large and small targets exist, such as oil leakage in the appearance defect of a transformer and color change of silica gel of a respirator, which belong to two defect types with large size difference.
6. The transformer appearance defect identification method based on the inspection robot and the Cascade RCNN algorithm according to claim 1, characterized in that: in S3, if the model can not meet the inspection operation and maintenance requirements, if the model has high missed inspection rate and false inspection rate, the hyper-parameters of the model need to be optimized and adjusted again, and the model needs to be trained again, and when the model indexes can meet the national network operation and maintenance requirements, the model can be output and deployed on a remote server of the inspection robot.
7. The transformer appearance defect identification method based on the inspection robot and the Cascade RCNN algorithm according to claim 1, characterized in that: in S4, the intelligent inspection robot adopts a visible light inspection function, namely, the high-definition camera collects the appearance image of the transformer equipment in the daytime, and the intelligent inspection robot is mainly used for inspecting the appearance state of the equipment in the transformer area.
8. The inspection robot and Cascade RCNN algorithm-based transformer appearance defect identification method according to claim 7, wherein the transformer appearance defect identification method comprises the following steps: in S4, the recognition result of the appearance defect of the transformer may be saved as inspection history data and a report may be generated, and the operation and maintenance staff may grasp the operation status of the defect device at the operation and maintenance master station according to the defect report automatically generated and uploaded by the robot, and check the device status in time and report the scheduling processing according to the alarm information of the inspection robot.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911338035.5A CN111080634A (en) | 2019-12-23 | 2019-12-23 | Transformer appearance defect identification method based on inspection robot and Cascade RCNN algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911338035.5A CN111080634A (en) | 2019-12-23 | 2019-12-23 | Transformer appearance defect identification method based on inspection robot and Cascade RCNN algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111080634A true CN111080634A (en) | 2020-04-28 |
Family
ID=70316722
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911338035.5A Pending CN111080634A (en) | 2019-12-23 | 2019-12-23 | Transformer appearance defect identification method based on inspection robot and Cascade RCNN algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111080634A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112036463A (en) * | 2020-08-26 | 2020-12-04 | 国家电网有限公司 | Power equipment defect detection and identification method based on deep learning |
| CN112069886A (en) * | 2020-07-31 | 2020-12-11 | 许继集团有限公司 | Transformer substation respirator state intelligent identification method and system |
| CN112183509A (en) * | 2020-12-01 | 2021-01-05 | 广州市玄武无线科技股份有限公司 | Warehouse auditing method and system based on target detection |
| CN112229845A (en) * | 2020-10-12 | 2021-01-15 | 国网河南省电力公司濮阳供电公司 | Unmanned aerial vehicle high-precision winding tower intelligent inspection method based on visual navigation technology |
| CN112380907A (en) * | 2020-10-20 | 2021-02-19 | 广东电网有限责任公司中山供电局 | Intelligent image identification method for defects of power transformation equipment and management system thereof |
| CN112383698A (en) * | 2020-10-09 | 2021-02-19 | 西安星闪数创智能科技有限公司 | Transformer platform district terminal intelligent monitoring system |
| CN113065608A (en) * | 2021-04-22 | 2021-07-02 | 深圳华瑞通科技有限公司 | Intelligent troubleshooting system and method based on multiple image recognition |
| CN113191362A (en) * | 2021-04-26 | 2021-07-30 | 南瑞集团有限公司 | Transformer equipment oil leakage defect detection device and method |
| CN113283541A (en) * | 2021-06-15 | 2021-08-20 | 无锡锤头鲨智能科技有限公司 | Automatic floor sorting method |
| CN113421263A (en) * | 2021-08-24 | 2021-09-21 | 深圳市信润富联数字科技有限公司 | Part defect detection method, device, medium and computer program product |
| CN115239734A (en) * | 2022-09-23 | 2022-10-25 | 成都数之联科技股份有限公司 | Model training method, device, storage medium, equipment and computer program product |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107944453A (en) * | 2017-12-05 | 2018-04-20 | 江苏省电力试验研究院有限公司 | Based on Hu not bushing detection methods of bending moment and support vector machines |
| CN109344905A (en) * | 2018-10-22 | 2019-02-15 | 王子蕴 | A kind of transmission facility automatic fault recognition methods based on integrated study |
| CN110097053A (en) * | 2019-04-24 | 2019-08-06 | 上海电力学院 | A kind of power equipment appearance defect inspection method based on improvement Faster-RCNN |
| CN110555513A (en) * | 2019-08-16 | 2019-12-10 | 国电南瑞科技股份有限公司 | Deep learning-based power equipment defect integrated diagnosis method |
-
2019
- 2019-12-23 CN CN201911338035.5A patent/CN111080634A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107944453A (en) * | 2017-12-05 | 2018-04-20 | 江苏省电力试验研究院有限公司 | Based on Hu not bushing detection methods of bending moment and support vector machines |
| CN109344905A (en) * | 2018-10-22 | 2019-02-15 | 王子蕴 | A kind of transmission facility automatic fault recognition methods based on integrated study |
| CN110097053A (en) * | 2019-04-24 | 2019-08-06 | 上海电力学院 | A kind of power equipment appearance defect inspection method based on improvement Faster-RCNN |
| CN110555513A (en) * | 2019-08-16 | 2019-12-10 | 国电南瑞科技股份有限公司 | Deep learning-based power equipment defect integrated diagnosis method |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112069886A (en) * | 2020-07-31 | 2020-12-11 | 许继集团有限公司 | Transformer substation respirator state intelligent identification method and system |
| CN112036463A (en) * | 2020-08-26 | 2020-12-04 | 国家电网有限公司 | Power equipment defect detection and identification method based on deep learning |
| CN112383698B (en) * | 2020-10-09 | 2022-05-06 | 西安星闪世图科技有限公司 | Transformer platform district terminal intelligent monitoring system |
| CN112383698A (en) * | 2020-10-09 | 2021-02-19 | 西安星闪数创智能科技有限公司 | Transformer platform district terminal intelligent monitoring system |
| CN112229845A (en) * | 2020-10-12 | 2021-01-15 | 国网河南省电力公司濮阳供电公司 | Unmanned aerial vehicle high-precision winding tower intelligent inspection method based on visual navigation technology |
| CN112380907A (en) * | 2020-10-20 | 2021-02-19 | 广东电网有限责任公司中山供电局 | Intelligent image identification method for defects of power transformation equipment and management system thereof |
| CN112183509A (en) * | 2020-12-01 | 2021-01-05 | 广州市玄武无线科技股份有限公司 | Warehouse auditing method and system based on target detection |
| CN113065608A (en) * | 2021-04-22 | 2021-07-02 | 深圳华瑞通科技有限公司 | Intelligent troubleshooting system and method based on multiple image recognition |
| CN113191362A (en) * | 2021-04-26 | 2021-07-30 | 南瑞集团有限公司 | Transformer equipment oil leakage defect detection device and method |
| CN113191362B (en) * | 2021-04-26 | 2022-07-22 | 南瑞集团有限公司 | Transformer equipment oil leakage defect detection device and method |
| CN113283541A (en) * | 2021-06-15 | 2021-08-20 | 无锡锤头鲨智能科技有限公司 | Automatic floor sorting method |
| CN113283541B (en) * | 2021-06-15 | 2022-07-22 | 无锡锤头鲨智能科技有限公司 | Automatic floor sorting method |
| CN113421263A (en) * | 2021-08-24 | 2021-09-21 | 深圳市信润富联数字科技有限公司 | Part defect detection method, device, medium and computer program product |
| CN113421263B (en) * | 2021-08-24 | 2021-11-30 | 深圳市信润富联数字科技有限公司 | Part defect detection method, device, medium and computer program product |
| CN115239734A (en) * | 2022-09-23 | 2022-10-25 | 成都数之联科技股份有限公司 | Model training method, device, storage medium, equipment and computer program product |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111080634A (en) | Transformer appearance defect identification method based on inspection robot and Cascade RCNN algorithm | |
| CN112633535A (en) | Photovoltaic power station intelligent inspection method and system based on unmanned aerial vehicle image | |
| CN109308447A (en) | The method of equipment operating parameter and operating status is automatically extracted in remote monitoriong of electric power | |
| CN109858367B (en) | Visual automatic detection method and system for worker through supporting unsafe behaviors | |
| CN110186375A (en) | Intelligent high-speed rail white body assemble welding feature detection device and detection method | |
| CN112560816A (en) | Equipment indicator lamp identification method and system based on YOLOv4 | |
| CN112837290A (en) | Crack image automatic identification method based on seed filling algorithm | |
| CN113837991A (en) | Image processing method, device, equipment and storage medium | |
| CN111127445A (en) | Distribution network line high-temperature area detection method and system based on deep learning | |
| CN104535589A (en) | Online detection method and device for low-voltage current mutual inductor | |
| CN109840905A (en) | Power equipment rusty stain detection method and system | |
| CN113988573A (en) | Risk judgment method, system and medium for routing inspection unmanned aerial vehicle based on power system | |
| CN111561967A (en) | Real-time online detection method and system for pantograph-catenary operation state | |
| CN112766795A (en) | Pipe gallery intelligent information management method and system based on cloud processing | |
| CN111626104B (en) | Cable hidden trouble point detection method and device based on unmanned aerial vehicle infrared thermal image | |
| CN116229052B (en) | Method for detecting state change of substation equipment based on twin network | |
| CN113469938A (en) | Pipe gallery video analysis method and system based on embedded front-end processing server | |
| CN112508911A (en) | Rail joint touch net suspension support component crack detection system based on inspection robot and detection method thereof | |
| CN111948213A (en) | Lipstick defect detection device based on attention capsule network and detection method thereof | |
| CN111028250A (en) | Real-time intelligent cloth inspecting method and system | |
| CN116231504A (en) | Remote intelligent inspection method, device and system for booster station | |
| CN112784914B (en) | Pipe gallery video intelligent attribute detection method and system based on cloud processing | |
| CN116308185A (en) | Visual progress input and early warning method for constructional engineering | |
| CN112861681B (en) | Pipe gallery video intelligent analysis method and system based on cloud processing | |
| CN115689206A (en) | Intelligent monitoring method for transformer substation infrastructure progress based on deep learning |
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: 20200428 |