CN109389322A - The disconnected broken lot recognition methods of grounded-line based on target detection and long memory models in short-term - Google Patents
The disconnected broken lot recognition methods of grounded-line based on target detection and long memory models in short-term Download PDFInfo
- Publication number
- CN109389322A CN109389322A CN201811276351.XA CN201811276351A CN109389322A CN 109389322 A CN109389322 A CN 109389322A CN 201811276351 A CN201811276351 A CN 201811276351A CN 109389322 A CN109389322 A CN 109389322A
- Authority
- CN
- China
- Prior art keywords
- grounded
- line
- term
- short
- training
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- 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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
- G07C1/00—Registering, indicating or recording the time of events or elapsed time, e.g. time-recorders for work people
- G07C1/20—Checking timed patrols, e.g. of watchman
Abstract
The present invention proposes a kind of disconnected broken lot recognition methods of the grounded-line based on target detection and long memory models in short-term, comprising the following steps: step S1: collects the grounded-line failure picture in transmission line of electricity as initial training data；Step S2: pre-processing grounded-line data set, and by treated, data are marked；Grounded-line data set is divided into training set and test set later；Step S3: initializing the model of pre-training, enters data into training in Faster-Rcnn target detection model, terminates training after model convergence；Step S4: the data sectional for extracting grounded-line by step S3 is input to training in long memory models in short-term；Step S5: the variable and network structure that are served only for trained are rejected, retain the weight finally detected, pass through the model inspection failure after solidifying by the long memory models in short-term of solidification.This method can distinguish faulted line segment and normal line segment in conjunction with the grounded-line of segmentation, identify that stranded and broken lot accuracy rate will be promoted greatly in this way.
Description
Technical field
The invention belongs to electric inspection process fault identification fields, more particularly to a kind of target detection that is based on to combine long short-term memory
The grounded-line of model is stranded, broken lot knows method for distinguishing.
Background technique
Growing with national economy, power grid also becomes huger to adapt to its development, electrical grid transmission route
Also therefore become to become increasingly complex, manual inspection is no longer satisfied requirement instantly.Therefore unmanned plane electric inspection process is met the tendency of
And give birth to, electric inspection process becomes human inspection, manned helicopter routing inspection and unmanned plane inspection " association after the inspection that have passed through practice
With development ".Unmanned plane is flexible, at low cost in construction of line traction and line data-logging upper type, can not only find on route
Small size parts, such as insulator, Bird's Nest, but also it can be found that the manual inspections such as the split pin of fitting burn into and nut missing
The problem of cannot finding.
The fault detection that unmanned plane is applied to power components judges generally by direct labor after live shooting or will
Data band is diagnosed back into row.But unmanned plane line walking process can generate a large amount of picture, whether artificial judgment equipment breaks down
Necessarily inefficiency and not accurate enough.In the state that grounded-line is due to being chronically at exposure, in many places by stress
It will lead to its material under effect to become fragile, destroy to become easy to generate with influence grounded-lines such as lightning stroke flashovers along with external force and break
The failures such as stock, broken lot and crackle, these failures are serious after occurring to will cause line-outage contingency, influences the operation of power transmission.Cause
The fault identification of this grounded-line is at urgent problem.
Summary of the invention
It is an object of the invention to: solve that the existing grounded-line in complicated transmission of electricity environment is stranded or the fault point of broken lot
Identify not high problem, provide it is a kind of more general, to pinpoint in face of grounded-line fault point in complex environment
The technology that grounded-line based on the long memory models in short-term of target detection combination is stranded, broken lot identifies.
To achieve the above object, the present invention specifically uses following technical scheme:
It is a kind of to be broken broken lot recognition methods based on target detection with the grounded-line of long memory models in short-term, which is characterized in that including with
Lower step:
Step S1: the grounded-line failure picture in transmission line of electricity is collected as initial training data, generates grounded-line data set；
Step S2: pre-processing grounded-line data set, and by treated, data are marked；Later by grounded-line data set point
For training set and test set；
Step S3: initializing the model of pre-training, enters data into training, model in Faster-Rcnn target detection model
Training is terminated after convergence；
Step S4: the data sectional for extracting grounded-line by step S3 is input to training in long memory models in short-term；
Step S5: the variable and network structure that are served only for trained are rejected, retain the weight finally detected, pass through by curing model
Model inspection failure after solidification.
Preferably, in step sl, cubic spline interpolation scaled size is used to original image, in the initial training number
Grounded-line normal picture is mixed into.
Preferably, step S2 the following steps are included:
Step S21: in grounded-line data set, positive sample and negative sample are respectively defined as to grounded-line failure and normal picture
This, and positive sample is identical with the specific gravity of negative sample；
Step S22: removal is fuzzy, identifies unclear sample；
Step S23: sample is labeled；
Step S24: grounded-line data set is divided into training set and test set according to the ratio of 8:2.
Preferably, in step S23, sample is enhanced first, then enhanced sample is labeled.
Preferably, in step s3: target detection model uses transfer learning, and weight, which derives from, uses COCO data set
The model initialization of pre-training；In step s 4, the weight of long memory models in short-term uses Xavier algorithm initialization.
Preferably, in step s 4: to each image data for extracting grounded-line by step S3, using cubic spline
Interpolation is of the same size all inputs all, and is input in long memory models in short-term according to the long segmentation of fixed column.It is using
After cubic spline interpolation scaling pictures size, be input to before long memory models in short-term, image data is normalized and
The processing of enhancing.
Preferably, in step s 4: the data that grounded-line is extracted by step S3 are segmented input in the way of column
Pixel, the training data as long memory network in short-term.
Since in actual unmanned plane inspection image, grounded-line component is very small, then plus under accurately to orient
Fault point, the two difficulties make common target identification network that will be difficult to be competent at.The present invention and its preferred embodiment use
The method for first identifying fault inspecting again after grounded-line, be will test out grounded-line all in picture by one-time detection, improved
The recall rate of fault detection.Long memory network in short-term is input to after cutting out the grounded-line that one-time detection goes out in original image again,
The network can distinguish faulted line segment and normal line segment in conjunction with the grounded-line of segmentation, identify stranded and broken lot accuracy rate in this way
It will greatly be promoted.
Of the invention and its preferred embodiment is realized by combining target detection model and long memory network in short-term to complicated back
The grounded-line failure of scape is accurately positioned its fault point.By carrying out fault identification after first classifying again, processing is for failure in this way
Recall rate has obtained great promotion.In the training process we used multiple dimensioned training, energy when prediction ensure that
It can enough be navigated to for the grounded-line of different scale, simultaneously because the place that stranded and broken lot is all failure is all very small
, grounded-line is segmented by the long segmentation of memory network in short-term of fault detection to be detected, and models coupling is so the information of history can be found
It is different from the fault point of normal region, is greatly improved the IOU of prediction block.
Detailed description of the invention
The present invention is described in more detail with reference to the accompanying drawings and detailed description:
Fig. 1 is present invention method overall flow schematic diagram；
Fig. 2 is schematic network structure of the embodiment of the present invention；
Fig. 3 is in the embodiment of the present invention to first and second detection block diagram representation of grounded-line mark；
Fig. 4 is the Fault Model input stepwise schematic views in the embodiment of the present invention using long memory network in short-term.
Specific embodiment
For the feature and advantage of this patent can be clearer and more comprehensible, special embodiment below is described in detail below:
As shown in Figure 1 and Figure 2, the present embodiment method overall flow the following steps are included:
Step S1: the grounded-line failure picture in transmission line of electricity is collected as initial training data, generates grounded-line data set；
Step S2: pre-processing grounded-line data set, and by treated, data are marked；Later by grounded-line data set point
For training set and test set；
Step S3: initializing the model of pre-training, enters data into training, model in Faster-Rcnn target detection model
Training is terminated after convergence；Complete one-time detection；
Step S4: the data sectional for extracting grounded-line by step S3 is input to training in long memory models in short-term；It completes
Secondary detection；
Step S5: the variable and network structure that are served only for trained are rejected, retain the weight finally detected, pass through by curing model
Model inspection failure after solidification.
In step sl, cubic spline interpolation scaled size is used to original image, is mixed into and leads in initial training data
Ground wire normal picture.
Since the image of unmanned plane acquisition is used in this embodiment, size is larger, therefore is used uniformly three to it
Secondary spline interpolation scaled size.
In addition, should have the grounded-line of the different conditions of different scenes when choosing image, preferably in the data set
Various types of grounded-line ratio is identical, guarantees that model will not be partial to the grounded-line of a certain scene with this.In order in the first rank
The Model of Target Recognition of section can preferably find the position of grounded-line, and normal class appropriate should be mixed into data set leads ground
Line sample improves the discrimination of model.
Step S2 specifically includes the following steps:
Step S21: in grounded-line data set, positive sample and negative sample are respectively defined as to grounded-line failure and normal picture
This, and positive sample is identical with the specific gravity of negative sample；
Step S22: removal is fuzzy, identifies unclear sample；
Step S23: sample is enhanced (since component causes the location of in picture with different shooting angle first
The external influence such as color difference, the enhancing carried out to data can enable model to learn to these external factor brings
Error), then enhanced sample is labeled；As shown in figure 3, it is whole (as primary inspection to need to mark grounded-line when mark
Survey frame) and there is stranded, broken lot fault zone (as secondary detection frame), in order to facilitate the training of network, mark should be accorded with
The xml label file of standardization PASCAL VOC format.Wherein in xml label comprising component area detection block (xmin,
Ymin, xmax, ymax), four values are respectively the coordinate of one-time detection frame and the secondary detection frame upper left corner, the lower right corner.
Step S24: grounded-line data set is divided into training set and test set according to the ratio of 8:2.
Later, in order to preferably be fitted the characteristic of picture, operation sample can be normalized.
As shown in Fig. 2, in step s3: target detection model (one-time detection) uses transfer learning, and weight derives from
Use the model initialization of COCO data set pre-training.
In step s 4, the weight of long memory models (secondary detection) in short-term uses Xavier algorithm initialization.The algorithm
The range that initialization can be automatically determined according to the quantity for outputting and inputting neuron, doing so, which can help to reduce gradient disperse, asks
Topic, transmits signal in training process deeper.
Training dataset and corresponding label file overall model as shown in Figure 2 is input to after initialization to work as
In start to train, in the training process can with the effect picture of real time inspection training, when discovery model restrained may be selected it is manual
Deconditioning.
In step s 4: it is inconsistent in order to solve the grounded-line dimension of picture come out due to interception, it is mentioned to by step S3
The each image data for getting grounded-line, being of the same size all inputs all using cubic spline interpolation, (Reshape is grasped
Make), after size is fixed, press as shown in Figure 4 that picture is (long by pixel input fault identification network according to the long segmentation of fixed column
Short-term memory model) in, segments here can be adjusted according to real data.
A most preferred scheme, the picture inputted herein are also required to normalization and data enhancing, while by prediction block
Also it is normalized between 0-1, the purpose for the arrangement is that in order to preferably restrain model.
The present embodiment saves training pattern in the training process, every certain step number, and the model of preservation will be on test set
Its model performance is tested, observing and nursing performance is with the variation of step number, and when model performance is stablized in certain level, selection is protected at this time
Model file of the model deposited as mold curing.Specific method is to reject training process variable contained in model, is only protected
The constants such as the weight of propagated forward are stayed, model volume is reduced, facilitates model transplantations.
This patent is not limited to above-mentioned preferred forms, anyone can obtain other each under the enlightenment of this patent
The disconnected broken lot recognition methods of the grounded-line based on target detection and long memory models in short-term of kind form, it is all according to the present patent application patent
The equivalent changes and modifications that range is done should all belong to the covering scope of this patent.
Claims (8)
- The broken lot recognition methods 1. a kind of grounded-line based on target detection and long memory models in short-term is broken, which is characterized in that including Following steps:Step S1: the grounded-line failure picture in transmission line of electricity is collected as initial training data, generates grounded-line data set；Step S2: pre-processing grounded-line data set, and by treated, data are marked；Later by grounded-line data set point For training set and test set；Step S3: initializing the model of pre-training, enters data into training, model in Faster-Rcnn target detection model Training is terminated after convergence；Step S4: the data sectional for extracting grounded-line by step S3 is input to training in long memory models in short-term；Step S5: the long memory models in short-term of solidification reject the variable and network structure that are served only for trained, what reservation finally detected Weight passes through the model inspection failure after solidifying.
- The broken lot recognition methods 2. the grounded-line according to claim 1 based on target detection and long memory models in short-term is broken, It is characterized in that, in step sl, cubic spline interpolation scaled size is used to original image, in the initial training data It is mixed into grounded-line normal picture.
- The broken lot recognition methods 3. the grounded-line according to claim 2 based on target detection and long memory models in short-term is broken, It is characterized by:Step S2 the following steps are included:Step S21: in grounded-line data set, positive sample and negative sample are respectively defined as to grounded-line failure and normal picture This, and positive sample is identical with the specific gravity of negative sample；Step S22: removal is fuzzy, identifies unclear sample；Step S23: sample is labeled；Step S24: grounded-line data set is divided into training set and test set according to the ratio of 8:2.
- The broken lot recognition methods 4. the grounded-line according to claim 3 based on target detection and long memory models in short-term is broken, It is characterized by:In step S23, sample is enhanced first, then enhanced sample is labeled.
- The broken lot recognition methods 5. the grounded-line according to claim 1 based on target detection and long memory models in short-term is broken, It is characterized by:In step s3: target detection model uses transfer learning, and weight derives from the mould using COCO data set pre-training Type initialization；In step s 4, the weight of long memory models in short-term uses Xavier algorithm initialization.
- The broken lot recognition methods 6. the grounded-line according to claim 1 based on target detection and long memory models in short-term is broken, It is characterized by:In step s 4: to each image data for extracting grounded-line by step S3, making to own using cubic spline interpolation Input is all of the same size, and is input in long memory models in short-term according to the long segmentation of fixed column.
- The broken lot recognition methods 7. the grounded-line according to claim 6 based on target detection and long memory models in short-term is broken, It is characterized by: being input to before long memory models in short-term, after using cubic spline interpolation scaling pictures size to picture The processing that data are normalized and enhance.
- The broken lot recognition methods 8. the grounded-line according to claim 1 based on target detection and long memory models in short-term is broken, It is characterized by:In step s 4: the data that grounded-line is extracted by step S3 are segmented input pixel in the way of column, as The training data of long memory network in short-term.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811276351.XA CN109389322A (en) | 2018-10-30 | 2018-10-30 | The disconnected broken lot recognition methods of grounded-line based on target detection and long memory models in short-term |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811276351.XA CN109389322A (en) | 2018-10-30 | 2018-10-30 | The disconnected broken lot recognition methods of grounded-line based on target detection and long memory models in short-term |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109389322A true CN109389322A (en) | 2019-02-26 |
Family
ID=65427929
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811276351.XA Pending CN109389322A (en) | 2018-10-30 | 2018-10-30 | The disconnected broken lot recognition methods of grounded-line based on target detection and long memory models in short-term |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109389322A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110009620A (en) * | 2019-04-03 | 2019-07-12 | 国网四川省电力公司检修公司 | The method and system of power grid stockbridge damper intelligent recognition are carried out based on transfer learning algorithm |
| CN110059076A (en) * | 2019-04-19 | 2019-07-26 | 国网山西省电力公司电力科学研究院 | A kind of Mishap Database semi-automation method for building up of power transmission and transformation line equipment |
| CN111028377A (en) * | 2019-11-22 | 2020-04-17 | 深圳供电局有限公司 | Inspection system of power transmission line and transmission method of inspection data |
| CN111199213A (en) * | 2020-01-03 | 2020-05-26 | 云南电网有限责任公司电力科学研究院 | Equipment defect detection method and device for transformer substation |
| CN112365876A (en) * | 2020-11-27 | 2021-02-12 | 北京百度网讯科技有限公司 | Method, device and equipment for training speech synthesis model and storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140082725A1 (en) * | 2006-02-28 | 2014-03-20 | The Trustees Of Columbia University In The City Of New York | Systems, Methods, and Media for Outputting a Dataset Based Upon Anomaly Detection |
| CN107506768A (en) * | 2017-10-11 | 2017-12-22 | 电子科技大学 | A kind of stranded recognition methods of transmission line wire based on full convolutional neural networks |
| CN108010030A (en) * | 2018-01-24 | 2018-05-08 | 福州大学 | A kind of Aerial Images insulator real-time detection method based on deep learning |
| CN108037133A (en) * | 2017-12-27 | 2018-05-15 | 武汉市智勤创亿信息技术股份有限公司 | A kind of power equipments defect intelligent identification Method and its system based on unmanned plane inspection image |
| CN108120900A (en) * | 2017-12-22 | 2018-06-05 | 北京映翰通网络技术股份有限公司 | A kind of electrical power distribution network fault location method and system |
| CN108647655A (en) * | 2018-05-16 | 2018-10-12 | 北京工业大学 | Low latitude aerial images power line foreign matter detecting method based on light-duty convolutional neural networks |
-
2018
- 2018-10-30 CN CN201811276351.XA patent/CN109389322A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140082725A1 (en) * | 2006-02-28 | 2014-03-20 | The Trustees Of Columbia University In The City Of New York | Systems, Methods, and Media for Outputting a Dataset Based Upon Anomaly Detection |
| CN107506768A (en) * | 2017-10-11 | 2017-12-22 | 电子科技大学 | A kind of stranded recognition methods of transmission line wire based on full convolutional neural networks |
| CN108120900A (en) * | 2017-12-22 | 2018-06-05 | 北京映翰通网络技术股份有限公司 | A kind of electrical power distribution network fault location method and system |
| CN108037133A (en) * | 2017-12-27 | 2018-05-15 | 武汉市智勤创亿信息技术股份有限公司 | A kind of power equipments defect intelligent identification Method and its system based on unmanned plane inspection image |
| CN108010030A (en) * | 2018-01-24 | 2018-05-08 | 福州大学 | A kind of Aerial Images insulator real-time detection method based on deep learning |
| CN108647655A (en) * | 2018-05-16 | 2018-10-12 | 北京工业大学 | Low latitude aerial images power line foreign matter detecting method based on light-duty convolutional neural networks |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110009620A (en) * | 2019-04-03 | 2019-07-12 | 国网四川省电力公司检修公司 | The method and system of power grid stockbridge damper intelligent recognition are carried out based on transfer learning algorithm |
| CN110059076A (en) * | 2019-04-19 | 2019-07-26 | 国网山西省电力公司电力科学研究院 | A kind of Mishap Database semi-automation method for building up of power transmission and transformation line equipment |
| CN111028377A (en) * | 2019-11-22 | 2020-04-17 | 深圳供电局有限公司 | Inspection system of power transmission line and transmission method of inspection data |
| CN111028377B (en) * | 2019-11-22 | 2022-07-05 | 深圳供电局有限公司 | Inspection system of power transmission line and transmission method of inspection data |
| CN111199213A (en) * | 2020-01-03 | 2020-05-26 | 云南电网有限责任公司电力科学研究院 | Equipment defect detection method and device for transformer substation |
| CN111199213B (en) * | 2020-01-03 | 2023-09-26 | 云南电网有限责任公司电力科学研究院 | Method and device for detecting defects of equipment for transformer substation |
| CN112365876A (en) * | 2020-11-27 | 2021-02-12 | 北京百度网讯科技有限公司 | Method, device and equipment for training speech synthesis model and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109389322A (en) | The disconnected broken lot recognition methods of grounded-line based on target detection and long memory models in short-term | |
| CN110598736B (en) | Power equipment infrared image fault positioning, identifying and predicting method | |
| Wang et al. | Automatic fault diagnosis of infrared insulator images based on image instance segmentation and temperature analysis | |
| CN110033453A (en) | Based on the power transmission and transformation line insulator Aerial Images fault detection method for improving YOLOv3 | |
| CN107742093B (en) | Real-time detection method, server and system for infrared image power equipment components | |
| CN108010030A (en) | A kind of Aerial Images insulator real-time detection method based on deep learning | |
| CN108898077B (en) | Power equipment infrared chart identification method and power equipment infrared identification system | |
| CN109345529A (en) | Based on the secondary target detection network wire clamp of modified, grading ring fault recognition method | |
| CN108680833B (en) | Composite insulator defect detection system based on unmanned aerial vehicle | |
| CN113205039B (en) | Power equipment fault image recognition disaster investigation system and method based on multiple DCNN networks | |
| CN114494185B (en) | Electrical equipment fault detection method based on RGB-T multi-scale feature fusion | |
| Wang et al. | Research on image recognition of insulators based on YOLO algorithm | |
| CN116681979A (en) | Power equipment target detection method under complex environment | |
| Manninen et al. | Multi-stage deep learning networks for automated assessment of electricity transmission infrastructure using fly-by images | |
| CN116994161A (en) | Insulator defect detection method based on improved YOLOv5 | |
| CN110009620A (en) | The method and system of power grid stockbridge damper intelligent recognition are carried out based on transfer learning algorithm | |
| CN113361520B (en) | Transmission line equipment defect detection method based on sample offset network | |
| CN115147591A (en) | Transformer equipment infrared image voltage heating type defect diagnosis method and system | |
| CN113837178A (en) | Deep learning-based automatic positioning and unified segmentation method for meter of transformer substation | |
| CN112508905A (en) | Hardware rust image detection method and computer readable storage medium | |
| Bao et al. | Automatic identification and defect diagnosis of transmission line insulators based on YOLOv3 network | |
| Jalil et al. | To identify hot spots in power lines using infrared and visible sensors | |
| CN112329863B (en) | Method for identifying state of isolation switch in traction substation | |
| CN116912721B (en) | Power distribution network equipment body identification method and system based on monocular stereoscopic vision | |
| Jiafeng et al. | Image recognition and processing model of multi-circuit lines on the same tower based on artificial intelligence |
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: 20190226 |