CN108470161A - A kind of high voltage isolator state identification method based on target following - Google Patents
A kind of high voltage isolator state identification method based on target following Download PDFInfo
- Publication number
- CN108470161A CN108470161A CN201810199222.9A CN201810199222A CN108470161A CN 108470161 A CN108470161 A CN 108470161A CN 201810199222 A CN201810199222 A CN 201810199222A CN 108470161 A CN108470161 A CN 108470161A
- Authority
- CN
- China
- Prior art keywords
- switch
- state
- high voltage
- target
- candidate
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
- G06V20/42—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items of sport video content
-
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
- G06T7/251—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments involving 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
- G06T2207/10016—Video; Image sequence
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of high voltage isolator state identification method based on target following, target following technology is applied in the division state recognition of high voltage isolator, this method is using disconnecting switch division process video as object, by Object Detecting and Tracking in conjunction with to disconnecting switch position and movement locus captured and carry out the division state of detection and isolation switch, provide it is a kind of efficiently, strong robustness, using the wide method for detecting high voltage isolator state based on target tracking of degree.
Description
Technical field
The present invention relates to high voltage isolator state inspection field, more specifically to one kind based on target with
The high voltage isolator state identification method of track.
Background technology
High voltage isolator is mainly used for closing for high-tension line and cut-offs, and is to ensure that the important of safe operation of power system sets
It is standby.In high-tension line maintenance, it is necessary to confirm that disconnecting switch is off, to ensure the safety of line attendant.Though
The switch arm shaft of right existing disconnecting switch has the auxiliary switch positioning function of simple limit switch etc, but is based on machine
The positioning of tool principle, when machinery is worn, this positioning may result in larger deviation.Long-range monitoring means shortage causes
Line attendant must arrive disconnecting switch scene, and the state of switch is confirmed in a manner of range estimation, thus cause maintenance personnel
Waste of human resource, maintenance efficiency is low.
It, can not switching devices division although video monitoring system has been widely used in electric apparatus monitoring
State carries out automatic identification.The domestic research for this respect at present is mainly based upon image processing techniques, such as uses local edge
Edge detection algorithm, by carrying out curve fitting to the edge of detection switch key position in preset detection range, then
Inclination angle range under the curve slope being calculated and the normal division state of switch is compared, to determine the shape of disconnecting switch
State.The generalization ability of this algorithm is weaker, only has good effect to the switch of specific shape, and video camera is in real work
In the unstable situation such as be possible to occur shaking, how selecting the error range with reference to tilt angles and permission, to become one big
Difficult point.Also a kind of method based on template matches, by taking interlacing slightly to be matched every column data, on the basis of thick matched
Power equipment is accurately matched again by the correlation of calculation template and target area.The algorithm of this template matches is to figure
The clarity of picture is more demanding, and fixed template does not adapt to the dynamic characteristic rotated during switch division, and due to not having
There is the candidate region of anticipation, the method for template matches takes consumption and calculates very much, it is difficult to meet detection high voltage isolator shape in real time
The requirement of state.
The existing method based on image procossing has specific limit to the detection of equipment, position stability and working environment
System, needs to extract characteristic point from image manually, and the method based on image procossing is easily lost target, such algorithm more
Robustness and application range be not strong.In addition, the recognition methods based on still image cannot utilize the contextual information pair of video
Motion state residing for high-voltage switch gear is identified, and complete reliable information cannot be provided for related personnel.Therefore, at present there is an urgent need for
It is a kind of towards video processing dynamic identifying method high voltage isolator state is identified.
Invention content
The technical problem to be solved in the present invention is, provides a kind of high voltage isolator state recognition based on target following
Method.
The technical solution adopted by the present invention to solve the technical problems is：Construct a kind of high_voltage isolation based on target following
On off state recognition methods, includes the following steps：
Step S010, one switch detection model of training export the first of switch by detecting the dynamic of monitoring area in real time
Beginning position；
Switch initial position is input to switch tracking mould by step S020, the start-up trace algorithm when disconnecting link, which moves, to be opened and closed
In type, each frame of video is analyzed using motion model, next switch can for the information inference based on video frame before
The candidate region that can occur, and assign identical weight for candidate region；
Step S030 solves the sparse reconstruct of group, according to the reconstruct solved using display model for each candidate region
Error size is that weight is distributed in candidate region again, the position of switch is obtained according to the weighted sum of candidate position, and then opened
Close movement locus；
Step S040 goes out the division state of switch by the Distance Judgment between moving track calculation plug-in strip.
Preferably, in the step S010, one switch detection model of training includes the following steps：
Step S010-1, the positive negative sample of acquisition high voltage isolator switch, difference is in including switching and switching
The image block of the folding condition of degree is as positive sample, and the background segment not including switch is as negative sample；
Step S010-2, to each positive and negative sample extraction Gradient Features, dimension rotation invariant features and edge contour feature,
It is spliced into feature vector of the subscript range vectors as each positive negative sample, trains a position of the switch to examine using feature vector
Survey device.
Preferably, in the step S020, include the following steps：
Positive negative sample in step S010-1 is launched into one-dimensional vector by step S020-1 by row, and to the vector normalizing
Change makes its l2Norm is 1, and the corresponding classification of sample is constant, obtains dictionary atom di, then form dictionary matrix D=[d1,
d2,…,dm]；
Above-mentioned positive sample is clustered into G-1 classes by step S020-2 using K mean algorithms, and negative sample is individually for one kind, record
Lower diCorresponding cluster classification g, finally obtains G class samples, here classification g representative be meant that switch be in out, half opening and closing
Various different conditions；
Step S020-3 utilizes trained detector detection video first frame and initial several frames in step step S010-2
The position of middle disconnecting switch is used in combination box to mark out the position of switch, if in a certain frame video box and first frame box
Overlapping area is less than a threshold value, judges that switch is now in motion state, enables tracing algorithm；
Step S020-4, tracing algorithm is divided into two parts of motion model and display model, when enabling tracing algorithm, profit
With particle wave pattern may be in p are switched according in the location of switch of present frame and form prediction next frame video
Position is that weight w is distributed in the position of each particlei=1/p.
Preferably, in the step S030, include the following steps：
The image of candidate position is launched into one-dimensional vector by row, equally makees l by step S030-12Norm normalized,
It is Y={ y to obtain candidate target1,y2,…,yN}；All candidate targets are solvedIts
In, a is that the sparse coefficient of candidate target collection indicates,Indicate candidate target yiReconstructed error,agThe vector of the corresponding coefficient composition of atom to be grouped into g classifications in dictionary D atoms, λ is canonical term system
Number is to weigh the proportion of reconstructed error and regular terms in object function；
Step S030-2, step S030-1 are that entire candidate target collection calculates a rarefaction representation vector a, and a is substituted intoFor each candidate target yiReconstructed error is calculated, particle filter model prediction is redistributed according to reconstructed error size
Each candidate target weight wiSo that ownership heavy phase adds as 1, and the final position for tracking target is all candidate target positions
The weighted average set and；
Step S030-3 repeats the movement locus that step S020-4 is switched to step S030-2.
Preferably, in the step S040, on the basis of taking the central axes between plug-in strip, and the selected tracking of real-time tracking plug-in strip
Point position is then identified disconnecting switch state according to following rule if the distance of trace point to benchmark is s：
(1) disconnecting switch is in open state：S ＞ s0；
(2) disconnecting switch is in conjunction state：S ＜ s1；
(3) disconnecting switch is in by opening to conjunction state：st＞ st+Δt；
(4) disconnecting switch is in by being bonded to open state：st＜ st+Δt；
Wherein, s0,s1Respectively judge the threshold value of opening and closing state, st,st+ΔtThe respectively video frame of t and t+ time Δts
Distance of the middle trace point to benchmark.
Implement a kind of high voltage isolator state identification method based on target following of the present invention, has below beneficial to effect
Fruit：
1, target following technology is applied in the division state recognition of high voltage isolator by the present invention, with disconnecting switch point
Conjunction process video is object, by Object Detecting and Tracking in conjunction with to disconnecting switch position and movement locus is captured
The division state of detection and isolation switch, provide it is a kind of efficiently, strong robustness, using degree it is wide based on target tracking detection high pressure every
The method for leaving off status.
2, thinking of the present invention is simple, artificially need not carry out Accurate Model to switch shape, and it is specific outer to be not limited to certain
The switch of shape or model increases the range of model application.
3, strong robustness of the present invention, tracking target is not easy to lose, need not be to the working environment of video camera and the work of switch
Excessive restriction and hypothesis are carried out as state.
4, the division state that the present invention is not only able to disconnecting switch judges, moreover it is possible to be sentenced by the movement locus of switch
It disconnects Guan Shiyou and assigns to conjunction or by closing to the state divided.
Description of the drawings
Present invention will be further explained below with reference to the attached drawings and examples, in attached drawing：
Fig. 1 is high voltage isolator " dividing " " conjunction " status diagram；
Fig. 2 is the recognizer flow chart of high voltage isolator state；
Fig. 3 is masterplate update and target search schematic diagram；
Fig. 4 is that switch folding condition judges schematic diagram.
Specific implementation mode
For a clearer understanding of the technical characteristics, objects and effects of the present invention, now control attached drawing is described in detail
The specific implementation mode of the present invention.
The present invention provides a kind of high voltage isolator state identification method based on target following, as shown in Fig. 2, include with
Lower step：
1, acquisition switchs positive negative sample, that is, includes to switch and switch the image block for being in different degrees of folding condition
As positive sample, as shown in Figure 1, acquire simultaneously not include switch background segment as negative sample, by positive sample picture classification
Labeled as 1, negative sample picture classification is labeled as -1；
2, to the positive and negative sample extraction Gradient Features of each of step 1, dimension rotation invariant features and edge contour feature,
It is spliced into feature vector of the subscript range vectors as each positive negative sample, trains a position of the switch to examine using feature vector
Survey device, the initial position of detection switch；
3, the positive negative sample in step 1 is launched into one-dimensional vector by row, and its l is made to vector normalization2Norm is
1, the corresponding classification of sample is constant, obtains dictionary atom di, then form dictionary matrix D=[d1,d2,…,dm], build dictionary
It is to be able to generate the judgement template of dynamic change, adapts to the metamorphosis during switch rotary；
4, the positive sample in step 3 is clustered into G-1 classes using K mean algorithms, negative sample is individually for one kind, records di
Corresponding cluster classification g, finally obtains G class samples, classification g representatives here be meant that switch be in out, half opening and closing etc. it is each
Kind different conditions, it is therefore an objective to utilize the group structure feature of rarefaction representation；
5, the position of disconnecting switch in video first frame and initially several frames is detected using trained detector in step 2,
It is used in combination box to mark out the position of switch, if box and the box overlapping area of first frame are less than a threshold in a certain frame video
Value judges that switch is now in motion state, enables tracing algorithm；
6, tracing algorithm is divided into two parts of motion model and display model and utilizes particle wave when enabling tracing algorithm
Model switchs the p position that may be according in the location of switch of present frame and form prediction next frame video, is
The position distribution weight w of each particlei=1/p；
7, the image of candidate position is launched into one-dimensional vector by row, equally makees l2Norm normalized, obtains candidate
Target is Y={ y1,y2,…,yN}；
8, following problem is solved to all candidate targets in step 7：
Wherein, a is that the sparse coefficient of candidate target collection indicates,Indicate candidate target yiReconstructed error,agThe vector of the corresponding coefficient composition of atom to be grouped into g classifications in dictionary D atoms, λ is canonical term system
Number is to weigh the proportion of reconstructed error and regular terms in object function.Due to rarefaction representation coefficient with the variation of video frame and
Variation, this be equivalent to differentiate template be also between a using group sparse constraint an intuitivism apprehension be make coefficient class it is sparse and
It is not sparse in class, it is thus possible to preferably the form of switch to be estimated using information in class, as shown in Figure 3；
9, it can be that entire candidate target collection calculates a rarefaction representation vector a in step 8, a is substituted intoFor
Each candidate target yiReconstructed error is calculated, each candidate of particle filter model prediction is redistributed according to reconstructed error size
The weight w of targetiSo that ownership heavy phase adds as 1, and the final position for tracking target is that the weighting of all candidate target positions is flat
And；
10, the movement locus that step 6-9 is switched is repeated, the state of switch is judged according to track.Because camera is fixed
Immediately below switch, and positioned at the middle position of two shoe brake knives.On the basis of taking the central axes between plug-in strip, and real-time tracking
Plug-in strip selectes trace point position, if the distance of trace point to benchmark is s, then knows to disconnecting switch state according to following rule
Not：
(1) disconnecting switch is in open state：S ＞ s0；
(2) disconnecting switch is in conjunction state：S ＜ s1；
(3) disconnecting switch is in by opening to conjunction state：st＞ st+Δt；
(4) disconnecting switch is in by being bonded to open state：st＜ st+Δt；
Wherein, s0,s1Respectively judge the threshold value of opening and closing state, st,st+ΔtThe respectively video frame of t and t+ time Δts
Middle trace point to benchmark distance, as shown in Figure 4.
The embodiment of the present invention is described with above attached drawing, but the invention is not limited in above-mentioned specific
Embodiment, the above mentioned embodiment is only schematical, rather than restrictive, those skilled in the art
Under the inspiration of the present invention, without breaking away from the scope protected by the purposes and claims of the present invention, it can also make very much
Form, all of these belong to the protection of the present invention.
Claims (5)
1. a kind of high voltage isolator state identification method based on target following, which is characterized in that include the following steps：
Step S010, one switch detection model of training export the initial bit of switch by detecting the dynamic of monitoring area in real time
It sets；
Switch initial position is input in switch trace model by step S020, the start-up trace algorithm when disconnecting link, which moves, to be opened and closed,
Each frame of video is analyzed using motion model, next switch is likely to occur the information inference based on video frame before
Candidate region, and assign identical weight for candidate region；
Step S030 solves the sparse reconstruct of group, according to the reconstructed error solved using display model for each candidate region
Size is that weight is distributed in candidate region again, the position of switch is obtained according to the weighted sum of candidate position, and then obtain switch fortune
Dynamic rail mark；
Step S040 goes out the division state of switch by the Distance Judgment between moving track calculation plug-in strip.
2. a kind of high voltage isolator state identification method based on target following according to claim 1, feature exist
In in the step S010, one switch detection model of training includes the following steps：
Step S010-1, the positive negative sample of acquisition high voltage isolator switch, is in various degree including switching and switching
Folding condition image block as positive sample, not include switch background segment as negative sample；
Step S010-2, to each positive and negative sample extraction Gradient Features, dimension rotation invariant features and edge contour feature, splicing
Feature vector at a subscript range vectors as each positive negative sample, trains a position of the switch to detect using feature vector
Device.
3. a kind of high voltage isolator state identification method based on target following according to claim 2, feature exist
In in the step S020, including the following steps：
Positive negative sample in step S010-1 is launched into one-dimensional vector by row, and made to vector normalization by step S020-1
Its l2Norm is 1, and the corresponding classification of sample is constant, obtains dictionary atom di, then form dictionary matrix D=[d1,d2,…,dm]；
Above-mentioned positive sample is clustered into G-1 classes by step S020-2 using K mean algorithms, and negative sample is individually for one kind, records di
Corresponding cluster classification g, finally obtains G class samples, here classification g representative be meant that switch be in out, half opening and closing it is various
Different conditions；
Step S020-3 utilizes trained detector detection video first frame and initial a few frame intervals in step step S010-2
The position for leaving pass is used in combination box to mark out the position of switch, if box and the box of first frame overlap in a certain frame video
Area is less than a threshold value, judges that switch is now in motion state, enables tracing algorithm；
Step S020-4, tracing algorithm are divided into two parts of motion model and display model and utilize grain when enabling tracing algorithm
Wavelet model switchs the p position that may be according in the location of switch of present frame and form prediction next frame video
It sets, is that weight w is distributed in the position of each particlei=1/p.
4. a kind of high voltage isolator state identification method based on target following according to claim 3, feature exist
In in the step S030, including the following steps：
The image of candidate position is launched into one-dimensional vector by row, equally makees l by step S030-12Norm normalized, obtains
Candidate target is Y={ y1,y2,…,yN}；All candidate targets are solvedWherein, a is
The sparse coefficient expression of candidate target collection,Indicate candidate target yiReconstructed error,agFor
Be grouped into the vector of the atom corresponding coefficient composition of g classifications in dictionary D atoms, λ be regularization coefficient with weigh reconstructed error and
Proportion of the regular terms in object function；
Step S030-2, step S030-1 are that entire candidate target collection calculates a rarefaction representation vector a, and a is substituted intoFor each candidate target yiReconstructed error is calculated, particle filter model prediction is redistributed according to reconstructed error size
Each candidate target weight wiSo that ownership heavy phase adds as 1, and the final position for tracking target is all candidate target positions
The weighted average set and；
Step S030-3 repeats the movement locus that step S020-4 is switched to step S030-2.
5. a kind of high voltage isolator state identification method based on target following according to claim 4, feature exist
In, in the step S040, on the basis of taking the central axes between plug-in strip, and real-time tracking plug-in strip selectes trace point position, if with
The distance of track point to benchmark is s, then disconnecting switch state is identified according to following rule：
(1) disconnecting switch is in open state：S ＞ s0；
(2) disconnecting switch is in conjunction state：S ＜ s1；
(3) disconnecting switch is in by opening to conjunction state：st＞ st+Δt；
(4) disconnecting switch is in by being bonded to open state：st＜ st+Δt；
Wherein, s0,s1Respectively judge the threshold value of opening and closing state, st,st+ΔtRespectively in the video frame of t and t+ time Δts with
Distance of the track point to benchmark.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810199222.9A CN108470161A (en) | 2018-03-12 | 2018-03-12 | A kind of high voltage isolator state identification method based on target following |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810199222.9A CN108470161A (en) | 2018-03-12 | 2018-03-12 | A kind of high voltage isolator state identification method based on target following |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108470161A true CN108470161A (en) | 2018-08-31 |
Family
ID=63265100
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810199222.9A Pending CN108470161A (en) | 2018-03-12 | 2018-03-12 | A kind of high voltage isolator state identification method based on target following |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108470161A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814734A (en) * | 2020-07-24 | 2020-10-23 | 南方电网数字电网研究院有限公司 | Method for identifying state of knife switch |
| CN112712547A (en) * | 2020-12-25 | 2021-04-27 | 华雁智科(杭州)信息技术有限公司 | State detection method of isolating switch and establishment method of model |
| CN113866620A (en) * | 2021-09-26 | 2021-12-31 | 国网福建省电力有限公司福州供电公司 | Disconnecting link state detection device and method based on wireless photoelectric sensing |
| CN113936230A (en) * | 2021-11-02 | 2022-01-14 | 石家庄铁道大学 | Method for detecting safety distance of disconnecting switch leading-in wire based on unmanned aerial vehicle |
| CN118298201A (en) * | 2024-06-04 | 2024-07-05 | 合肥工业大学 | Equipment identification matching model training method, equipment identification matching method and equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103514600A (en) * | 2013-09-13 | 2014-01-15 | 西北工业大学 | Method for fast robustness tracking of infrared target based on sparse representation |
| CN106447696A (en) * | 2016-09-29 | 2017-02-22 | 郑州轻工业学院 | Bidirectional SIFT (scale invariant feature transformation) flow motion evaluation-based large-displacement target sparse tracking method |
| CN106683116A (en) * | 2016-08-31 | 2017-05-17 | 电子科技大学 | Particle filter integrated tracking method based on support vector machine |
-
2018
- 2018-03-12 CN CN201810199222.9A patent/CN108470161A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103514600A (en) * | 2013-09-13 | 2014-01-15 | 西北工业大学 | Method for fast robustness tracking of infrared target based on sparse representation |
| CN106683116A (en) * | 2016-08-31 | 2017-05-17 | 电子科技大学 | Particle filter integrated tracking method based on support vector machine |
| CN106447696A (en) * | 2016-09-29 | 2017-02-22 | 郑州轻工业学院 | Bidirectional SIFT (scale invariant feature transformation) flow motion evaluation-based large-displacement target sparse tracking method |
Non-Patent Citations (4)
| Title |
|---|
| JINGYI WANG,ET AL.: "Recognition of high voltage isolating switch"s states based on object tracking", 《2017 4TH INTERNATIONAL CONFERENCE ON SYSTEMS AND INFORMATICS（ICSAI）》 * |
| 薛斌等: "融合局部加权余弦与稀疏表示的目标跟踪算法", 《电讯技术》 * |
| 郁道银等: "基于随机投影和稀疏表示的跟踪算法", 《电子与信息学报》 * |
| 钟伟: "基于稀疏表示的目标跟踪算法", 《中国优秀博硕士学位论文全文数据库(硕士) 信息科技辑（月刊）》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814734A (en) * | 2020-07-24 | 2020-10-23 | 南方电网数字电网研究院有限公司 | Method for identifying state of knife switch |
| CN111814734B (en) * | 2020-07-24 | 2024-01-26 | 南方电网数字电网研究院有限公司 | Method for identifying state of disconnecting link |
| CN112712547A (en) * | 2020-12-25 | 2021-04-27 | 华雁智科(杭州)信息技术有限公司 | State detection method of isolating switch and establishment method of model |
| CN112712547B (en) * | 2020-12-25 | 2024-06-04 | 华雁智科(杭州)信息技术有限公司 | State detection method and model building method of isolating switch |
| CN113866620A (en) * | 2021-09-26 | 2021-12-31 | 国网福建省电力有限公司福州供电公司 | Disconnecting link state detection device and method based on wireless photoelectric sensing |
| CN113936230A (en) * | 2021-11-02 | 2022-01-14 | 石家庄铁道大学 | Method for detecting safety distance of disconnecting switch leading-in wire based on unmanned aerial vehicle |
| CN118298201A (en) * | 2024-06-04 | 2024-07-05 | 合肥工业大学 | Equipment identification matching model training method, equipment identification matching method and equipment |
| CN118298201B (en) * | 2024-06-04 | 2024-09-03 | 合肥工业大学 | Equipment identification matching model training method, equipment identification matching method and equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108470161A (en) | A kind of high voltage isolator state identification method based on target following | |
| CN113034548B (en) | Multi-target tracking method and system suitable for embedded terminal | |
| Miao et al. | Insulator detection in aerial images for transmission line inspection using single shot multibox detector | |
| Wang et al. | Insulator identification from aerial images using support vector machine with background suppression | |
| CN111179249A (en) | Power equipment detection method and device based on deep convolutional neural network | |
| Qu et al. | Moving vehicle detection with convolutional networks in UAV videos | |
| CN109298785A (en) | A kind of man-machine joint control system and method for monitoring device | |
| CN109241829A (en) | The Activity recognition method and device of convolutional neural networks is paid attention to based on space-time | |
| CN103164858A (en) | Adhered crowd segmenting and tracking methods based on superpixel and graph model | |
| CN104281857B (en) | A kind of isolator detecting method projected based on orthogonal direction | |
| Prasad et al. | Review on Machine Vision based Insulator Inspection Systems for Power Distribution System. | |
| CN109934849B (en) | Online multi-target tracking method based on trajectory metric learning | |
| CN112985263A (en) | Method, device and equipment for detecting geometrical parameters of bow net | |
| CN117235583B (en) | Monitoring method and system for GIS breaker actuating mechanism | |
| CN113936210A (en) | Anti-collision method for tower crane | |
| Niu et al. | Electrical equipment identification method with synthetic data using edge-oriented generative adversarial network | |
| Manninen et al. | Multi-stage deep learning networks for automated assessment of electricity transmission infrastructure using fly-by images | |
| Lu et al. | Thermal Fault Diagnosis of Electrical Equipment in Substations Based on Image Fusion. | |
| CN108960241A (en) | Insulator based on statistical shape model falls to go here and there detection algorithm | |
| CN110852203B (en) | Multi-factor suspicious person identification method based on video feature learning | |
| CN103400382A (en) | Abnormal panel detection algorithm based on ATM (Automatic Teller Machine) scene | |
| Wang et al. | Recognition of high voltage isolating switch's states based on object tracking | |
| CN113780224A (en) | Transformer substation unmanned inspection method and system | |
| Nan et al. | Safety Helmet Detection Dynamic Model Based on the Critical Area Attention Mechanism | |
| Gandhamal et al. | Evaluation of background subtraction algorithms for object extraction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180831 |
|
| RJ01 | Rejection of invention patent application after publication |