CN107944396A - A kind of disconnecting link state identification method based on improvement deep learning - Google Patents
A kind of disconnecting link state identification method based on improvement deep learning Download PDFInfo
- Publication number
- CN107944396A CN107944396A CN201711207726.2A CN201711207726A CN107944396A CN 107944396 A CN107944396 A CN 107944396A CN 201711207726 A CN201711207726 A CN 201711207726A CN 107944396 A CN107944396 A CN 107944396A
- Authority
- CN
- China
- Prior art keywords
- disconnecting link
- insulator
- deep learning
- link state
- identification method
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/34—Smoothing or thinning of the pattern; Morphological operations; Skeletonisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Abstract
The invention discloses a kind of based on the disconnecting link state identification method for improving deep learning, comprise the following steps：Obtain training pattern；The probability for obtaining prediction block is predicted to input picture by training pattern；Using sliding window policy selection candidate region and obtain label；Candidate region is carried out to delete choosing acquisition candidate rectangle frame；Fitting a straight line is carried out to candidate rectangle frame and obtains exact rectangular frame；Disconnecting link in exact rectangular frame and insulator state are judged, complete the identification of disconnecting link state.The present invention obtains training pattern using the convolutional neural networks of the pond stragetic innovation based on spatial weighting on image set；Then the potential site of insulator and disconnecting link is detected by training pattern, the closure of a variety of disconnecting links or off-state is identified according to the connectedness with insulator, the position of insulator and disconnecting link can be precisely located, significantly improve the precision of disconnecting link state recognition.
Description
Technical field
The present invention relates to pattern-recognition and the classification and detection of classify field, especially specific objective, and in particular to a kind of
Based on the disconnecting link state identification method for improving deep learning.
Background technology
Substation's Real-time Monitor Technique obtains swift and violent development in recent years, wherein the electric device maintenance based on image procossing
Become the hot spot of research with identification technology.Due to image captured in reality usually contain many other targets rather than just
In itself, while captured image background is also complex for power equipment target interested, such as different illumination conditions, shooting
Angle etc., this causes same target that different patterns is presented in different images.
Traditional power equipment (such as insulator and disconnecting link) recognition methods relies primarily on color of object feature and geometric properties,
These methods are often influenced by factors such as brightness change and complex backgrounds, cause these methods often to possess poor extensive energy
Power.Such as document (" Zhang Geng, Zhang Dahua, Li Dan, et al..The automatic identification
method ofswitch state[J].International Journal ofSimulation: Systems,Science
and Technology,2016,17(25):P21.1-21.4. a kind of simple and quick disconnecting link state identification method ") is proposed, should
Method carries out Hough transform to determine the state of disconnecting link directly on disconnecting link boundary image.Document (" Lin H, Zhang W, et
al.A condition monitoring algorithmbased on image geometric analysis for
substation switch[C]//International Conference on Intelligent Computing and
Internet ofThings.Harbin:IEEE,2015:72-76. ") propose a kind of real-time knife based on disconnecting link geological information
Lock condition monitoring algorithm, the algorithm mainly obtain tool arm straight line using Hough transform, and straight using the cosine law two tool arms of calculating
Angle between line judges the state of disconnecting link.Document (" the various features extractive technique of Zhao Junmei, Zhang Liping insulation subgraph
Research [J] electrical measurements and instrument, 2013,50 (12):37-41. ") propose insulation subgraph various features extractive technique, bag
The textural characteristics using gray level co-occurrence matrixes extraction image are included, invariant moment features is extracted using bianry image, utilizes the several of image
What feature extraction boundary profile, information is provided for isolator detecting and identification.(" Chen Anwei, happy complete bright, Zhang Zongyi, waits to document
The recognition methods of transformer substation switch status image [J] Automation of Electric Systems based on robot, 2012,36 (6):101-
105. ") robot of a kind of energy automatic identification disconnecting link position and state is developed, this method utilizes SIFT algorithms to carry out template matches
To find disconnecting link position, Hough transform extraction straight line is then carried out to judge disconnecting link state.Document (" Yuan Jinsha, Cui Kebin, Lee
The identification of treasured tree insulator video images of the based on ASIFT algorithms and positioning [J] electrical measurements and instrument, 2015,52 (7):106-
112. ") matching for the sub-pictures that insulate in transmission line of electricity video and standard gallery is then realized using ASIFT algorithms, so identify and
Insulator in positioning video.
Document (" transformer substation switch equipment detection and state recognition of Shao Jianxiong, Yan Yunfeng, the Qi Donglian based on Hough forest
[J] Automation of Electric Systems, 2016,40 (11):115-120. ") point out that the above method is also poor, anti-dry there is stability
Disturb the problems such as indifferent.In order to improve positioning and accuracy of identification, in recent years many researchers using the method for machine learning come
Carry out the positioning and state recognition of insulator or disconnecting link.Such as document, (" Shao Jianxiong, Yan Yunfeng, Qi Donglian are based on Hough forest
Transformer substation switch equipment detects and state recognition [J] Automation of Electric Systems, 2016,40 (11):115-120. ") propose
A kind of disconnecting link state identification method based on Hough forest, disconnecting link disconnection and closure state model are carried out using Hough forest
Practise, train the model that disconnecting link disconnects and closes two states, disconnecting link positioning and state are carried out to input picture using the model
Identification.(" high-strength, sun is military, insulator breakdown recognizer [J] electricity of the Li Qian based on sparse difference depth belief network for document
Survey and instrument, 2016,53 (1):19-25. ") by the gray feature matrix conversion of image it is difference representing matrix and to its average
Change, normalization and rarefaction, are then trained obtained difference characteristic using DBN networks, reach identification insulator breakdown
Purpose.Document (" LiuY, Yong J, Liu L, et al.The method ofinsulator recognition based
on deep learning[C]//International Conference onApplied Robotics forthe Power
Industry.Jinan,China,2016:1-5. ") propose one six layers of convolutional neural networks (Convolutinal
Neural Networks, CNNs) the training detection model on insulation subgraph, it is successfully accurately fixed using the training pattern
Position insulator.Document (" Chen Tuo, Zhang Guoyue, Qi Donglian.A Recognition Method of
Smart Substation Switchgear State Based on Fully Convolutional Networks[C]
.Proceedings ofthe 35th Chinese Control Conference,2016,Chengdu,China,9894-
9897. ") a kind of disconnecting link for being based on full convolutional neural networks (Fully Convolutional Networks, FCNs) is proposed
State identification method, this method carries out insulator and disconnecting link target segmentation positioning using FCNs, and utilizes insulator and disconnecting link
Between geometrical relationship identify the state of disconnecting link target.But the problem of this kind of method is interfered with each other there are more disconnecting link targets,
Accuracy of identification is still waited to further improve.
The content of the invention
The present invention is to overcome above-mentioned the deficiencies in the prior art part, there is provided is provided a kind of based on the knife for improving deep learning
Lock state identification method, to which the influence of brightness change and complex background to disconnecting link state recognition can be overcome, so as to can essence
Really the position of ground positioning insulator and disconnecting link, improves the precision of disconnecting link state recognition.
It is of the invention to be for technical scheme applied to solve the technical problem：
A kind of disconnecting link state identification method based on improvement deep learning, comprises the following steps：
Obtain training pattern；
The probability for obtaining prediction block is predicted to input picture by training pattern；
Using sliding window policy selection candidate region and obtain label；
Candidate region is carried out to delete choosing acquisition candidate rectangle frame；
Fitting a straight line is carried out to candidate rectangle frame and obtains exact rectangular frame；
Disconnecting link in exact rectangular frame and insulator state are judged, complete the identification of disconnecting link state.
Further, the acquisition training pattern is specially：Using the convolution of the pond stragetic innovation based on spatial weighting
Neutral net obtains training pattern on image set：Specifically include following steps：
Training set is used as by the use of image set of the camera device shooting comprising insulator, disconnecting link and different background；
Smooth operation is carried out to image using Gaussian kernel wave filter；
Strengthen the contrast of image using histogram equalization；
Image border is obtained using Canny operators, the enclosed region of maximum area is chosen in image border as mesh
Mark region；
Using Gaussian Profile come the weights of random initializtion CNNs, and the CNNs for building in training image collection 6 layers changes
For training network model.
Further, the convolutional neural networks of the pond stragetic innovation based on spatial weighting are expressed as follows：
Convolution Feature Mapping can be expressed as the tensor of H × W × D dimensions, and wherein H and W represent the size of convolution Feature Mapping
(length and width), D represent the quantity of Feature Mapping；Spatial weighting mask can be expressed as the Chi Huatong of K H × W size
Road, makes PkRepresent the pond feature of k-th of space mask, then it can be expressed as：
Wherein xiRepresent i-th of local feature of Feature Mapping,It is the coefficient of correspondence (weight) of k-th of space mask；
It can be seen that PkDimension be 1 × 1 × D, the pond feature for multiple pond passages that can connect can form whole pond
The graphical representation of layer, i.e.,:
P=[P1 T,P2 T,…,Pk T,…,PK T]T；
P is the output of pond layer, is K × D character representations of input picture.
Further, described image collection includes amount of images more than or equal to 3000.
Further, it is described to be predicted the probability for obtaining prediction block to input picture by training pattern, specifically include：
Input picture is predicted using training pattern, prediction block is obtained as classifier using the full contiguous functions of Softmax
Probability.
Further, it is described to utilize sliding window policy selection candidate region and obtain label, specifically include：Use sliding window
Mouthful strategy selects candidate region, designs the sliding windows of multiple and different scales to obtain insulator and disconnecting link target area, when
Label is obtained when insulator and disconnecting link target area are as input；
Further, the quantity of the sliding window is more than or equal to 10.
Further, it is described that candidate region is carried out to delete choosing acquisition candidate rectangle frame；Specifically include：Pressed down using non-maximum
Preparation method excludes suspected target region, only leaves the candidate rectangle frame of some different classes of very close real goals, it is excluded
According to being to remove the suspected target region with smaller IoU (intersection ofunion) value, wherein IoU detects for algorithm
Region (DR：Detection Region) and real estate (GT:GroundTruth coincidence factor between)：
Further, the disconnecting link and insulator state in exact rectangular frame judges, completes disconnecting link state
Identification, is specially：If a disconnecting link region and two insulator regional connectivities, then it is assumed that the disconnecting link is in closure state, no
Then it is off.
Compared with existing technology, beneficial effects of the present invention are embodied in：
The present invention first pre-processes input picture, secondly using the volume of the pond stragetic innovation based on spatial weighting
Product neutral net obtains training pattern on image set；Then the potential position of insulator and disconnecting link is detected by training pattern
Put, accurate insulator and disconnecting link position are obtained using non-maxima suppression and Algorithm of fitting a straight line；According to the company with insulator
The general character identifies the closure of a variety of disconnecting links or off-state, the position of insulator and disconnecting link can be precisely located, significantly improves
The precision of disconnecting link state recognition；The improvement of spatial weighting is carried out to pondization operation in CNNs can further improve depth characteristic table
The robustness and validity shown；Strengthen the contrast of image using histogram equalization, eliminate reflected light to insulator and knife
The influence of lock positioning；Image border is obtained using Canny operators, the enclosed region that maximum area is chosen in image border is made
For target area interested, the precision that insulator and disconnecting link position is improved；Trained mould is trained using deep learning method
Type detects the potential site of insulator and disconnecting link, and the diversity of sample also improves the adaptability to different scenes；Utilize cunning
Dynamic window policy selects candidate region, designs the sliding window of different scale to obtain insulation sub-goal, can overcome due to
The problem of difference of shooting angle, insulator shows different size in different images；Intended using non-maxima suppression and straight line
Hop algorithm, being capable of accurately insulator and disconnecting link position come the insulator obtained and disconnecting link position.
Brief description of the drawings
Fig. 1 is proposed by the present invention a kind of based on the disconnecting link state identification method flow chart for improving deep learning；
Fig. 2 is the method flow diagram for obtaining training pattern proposed by the present invention；
Fig. 3 is disconnecting link state recognition phase flow figure proposed by the present invention.
Embodiment
The specific embodiment of the invention is described with reference to the accompanying drawings and embodiments：
As shown in Figure 1, it is a kind of disconnecting link state identification method structure chart based on improvement deep learning proposed by the present invention；
A kind of disconnecting link state identification method based on improvement deep learning, comprises the following steps：
Step 101, training pattern is obtained；
Step 102, the probability for obtaining prediction block is predicted to input picture by training pattern；
Step 103, using sliding window policy selection candidate region and label is obtained；
Step 104, candidate region is carried out deleting choosing acquisition candidate rectangle frame；
Step 105, fitting a straight line is carried out to candidate rectangle frame and obtains exact rectangular frame；
Step 106, the disconnecting link in exact rectangular frame and insulator state are judged, completes the identification of disconnecting link state.
The present invention first pre-processes input picture, using the convolution god of the pond stragetic innovation based on spatial weighting
Training pattern is obtained on image set through network；Then the potential site of insulator and disconnecting link is detected by training pattern, profit
Accurate insulator and disconnecting link position are obtained with non-maxima suppression and Algorithm of fitting a straight line；Come according to the connectedness of insulator
Identify the closure or off-state of a variety of disconnecting links.The position of insulator and disconnecting link can be precisely located in the present invention, significantly improve
The precision of disconnecting link state recognition.
Fig. 2 is the method flow diagram for obtaining training pattern proposed by the present invention；
In a step 101, obtaining training pattern is specially：Using the convolution god of the pond stragetic innovation based on spatial weighting
Training pattern is obtained on image set through network：Specifically include following steps：
Step 201, it is used as training set by the use of image set of the camera device shooting comprising insulator, disconnecting link and different background；
Step 202, smooth operation is carried out to image using Gaussian kernel wave filter；
Step 203, the contrast of image is strengthened using histogram equalization；
Step 204, image border is obtained using Canny operators, the closed area of maximum area is chosen in image border
Domain is as target area；
Step 205, using Gaussian Profile come the weights of random initializtion CNNs, and 6 layers are built in training image collection
CNNs carrys out repetitive exercise network model.
The improvement that the present invention carries out pondization operation in CNNs spatial weighting can further improve depth characteristic expression
Robustness and validity；Strengthen the contrast of image using histogram equalization, to eliminate reflected light to insulator and disconnecting link
The influence of positioning；Image border is obtained using Canny operators, the enclosed region conduct of maximum area is chosen in image border
Target area interested, improves the precision of insulator and disconnecting link positioning；Training pattern is trained using deep learning method
To detect the potential site of insulator and disconnecting link, image set is used as training sample, the diversity of sample using more than 3000
Improve the adaptability to different scenes；Candidate region is selected using sliding window strategy, designs the cunning of 10 different scales
Dynamic window obtains insulation sub-goal, can overcome difference due to shooting angle, and insulator shows difference in different images
The problem of size.
It is disconnecting link state recognition phase flow figure proposed by the present invention referring to Fig. 3.
As shown in figure 3, disconnecting link state recognition process is to carry out convolutional neural networks to shooting image to obtain training pattern, so
The target of insulator and disconnecting link is obtained by sliding window strategy afterwards, obtains suspected target candidate region, then by non-very big
Value suppression method handles suspected target candidate region to obtain close target area, then obtains accurate mesh by fitting a straight line
Region is marked, disconnecting link state recognition result is finally obtained according to disconnecting link condition discrimination condition.
It is provided by the invention a kind of based on the disconnecting link state identification method for improving deep learning, input picture is located in advance
Reason, training pattern is obtained using the convolutional neural networks of the pond stragetic innovation based on spatial weighting on image set；Then lead to
Training pattern is crossed to detect the potential site of insulator and disconnecting link, is obtained using non-maxima suppression and Algorithm of fitting a straight line accurate
Insulator and disconnecting link position；The closure of a variety of disconnecting links or off-state are identified according to the connectedness with insulator.Specifically press
Following steps carry out：
The convolutional neural networks of pond stragetic innovation based on spatial weighting are expressed as follows：
The tensor of H × W × D dimensions can be expressed as convolution Feature Mapping, wherein H and W represent convolution Feature Mapping
Size (length and width), D represent the quantity of Feature Mapping.Spatial weighting mask can be expressed as the pond of K H × W size
Passage, makes PkRepresent the pond feature of k-th of space mask, then it can be expressed as：
Wherein xiRepresent i-th of local feature of Feature Mapping,It is the coefficient of correspondence (weight) of k-th of space mask.
It can be seen that PkDimension be 1 × 1 × D, the pond feature for multiple pond passages that can connect can form the image of whole pond layer
Represent, i.e.,:
P=[P1 T,P2 T,…,Pk T,…,PK T]T (2)
P is the output of pond layer, is K × D character representations of input picture.
Training mould is obtained on image set using the convolutional Neural net of the pond stragetic innovation based on spatial weighting
Type：
Manually shot by the use of video camera and obtain the 3000 width images comprising insulator, disconnecting link and different background as training
Collection；
Smooth operation is carried out to image using Gaussian kernel wave filter to reduce the influence of noise；
Strengthen the contrast of image using histogram equalization, to eliminate the shadow that reflected light positions insulator and disconnecting link
Ring；
Carry out the weights of random initializtion CNNs using Gaussian Profile (average 0, standard deviation 0.01), and in training image
The CNNs for building 6 layers on collection carrys out repetitive exercise network model, uses ReLU (Rectified Linear Unit) activation primitive generations
For traditional sigmoid activation primitives, its mathematic(al) representation is f (x)=max (0, x), the purpose is to avoid diffusion with
Improve the convergence rate of stochastic gradient descent method；
Input picture is predicted using trained CNNs models, using the full contiguous functions of Softmax as classification
Machine obtains the probability of prediction block；
Candidate region is selected with sliding window strategy, designs the sliding windows of 10 different scales to obtain insulation specific item
Mark, when whole target (insulator and disconnecting link) region can obtain its label as when inputting；
Most of suspected target regions are excluded using non-maxima suppression method, some is only left and different classes of connects very much
The candidate rectangle frame of nearly real goal, it is to remove to have smaller IoU (intersection ofunion) value that it, which excludes foundation,
Suspicious region, wherein IoU are algorithm detection zone (DR：Detection Region) and real estate (GT:GroundTruth)
Between coincidence factor：
Fitting a straight line is carried out to resulting candidate rectangle frame to obtain including the accurate square of real insulator/disconnecting link target
Shape frame；
If a disconnecting link region and two insulator regional connectivities, then it is assumed that the disconnecting link is in closure state, is otherwise
It is off.
The preferred embodiment for the present invention is explained in detail above in conjunction with attached drawing, but the invention is not restricted to above-mentioned implementation
Mode, within the knowledge of a person skilled in the art, can also be on the premise of present inventive concept not be departed from
Make a variety of changes, these changes are related to correlation technique well-known to those skilled in the art, these both fall within patent of the present invention
Protection domain.
Many other changes and remodeling can be made by not departing from the spirit and scope of the present invention.It should be appreciated that the present invention is not
It is limited to specific embodiment, the scope of the present invention is defined by the following claims.
Claims (9)
- It is 1. a kind of based on the disconnecting link state identification method for improving deep learning, it is characterised in that to comprise the following steps：Obtain training pattern；The probability for obtaining prediction block is predicted to input picture by training pattern；Using sliding window policy selection candidate region and obtain label；Candidate region is carried out to delete choosing acquisition candidate rectangle frame；Fitting a straight line is carried out to candidate rectangle frame and obtains exact rectangular frame；Disconnecting link in exact rectangular frame and insulator state are judged, complete the identification of disconnecting link state.
- It is 2. according to claim 1 a kind of based on the disconnecting link state identification method for improving deep learning, it is characterised in that institute Stating acquisition training pattern is specially：Obtained using the convolutional neural networks of the pond stragetic innovation based on spatial weighting on image set Obtain training pattern：Specifically include following steps：Training set is used as by the use of image set of the camera device shooting comprising insulator, disconnecting link and different background；Smooth operation is carried out to image using Gaussian kernel wave filter；Strengthen the contrast of image using histogram equalization；Image border is obtained using Canny operators, the enclosed region of maximum area is chosen in image border as target area Domain；Using Gaussian Profile come the weights of random initializtion CNNs, and the CNNs of 6 layers of structure carrys out iteration instruction in training image collection Practice network model.
- It is 3. according to claim 2 a kind of based on the disconnecting link state identification method for improving deep learning, it is characterised in that institute The convolutional neural networks for stating the pond stragetic innovation based on spatial weighting are expressed as follows：Convolution Feature Mapping can be expressed as the tensor of H × W × D dimensions, and wherein H and W represent the size (length of convolution Feature Mapping And width), D represents the quantity of Feature Mapping；Spatial weighting mask can be expressed as the pond passage of K H × W size, make Pk Represent the pond feature of k-th of space mask, then it can be expressed as：Wherein xiRepresent i-th of local feature of Feature Mapping,It is the coefficient of correspondence (weight) of k-th of space mask；It can be seen that PkDimension be 1 × 1 × D, the pond feature for multiple pond passages that can connect can form the figure of whole pond layer As representing, i.e.,:P is the output of pond layer, is K × D character representations of input picture.
- It is 4. according to claim 2 a kind of based on the disconnecting link state identification method for improving deep learning, it is characterised in that institute State image set and include amount of images more than or equal to 3000.
- It is 5. according to claim 1 a kind of based on the disconnecting link state identification method for improving deep learning, it is characterised in that institute The probability for being predicted by training pattern to input picture and obtaining prediction block is stated, is specifically included：Using training pattern to input Image is predicted, and the probability of prediction block is obtained as classifier using the full contiguous functions of Softmax.
- It is 6. according to claim 1 a kind of based on the disconnecting link state identification method for improving deep learning, it is characterised in that institute State using sliding window policy selection candidate region and obtain label, specifically include：Candidate regions are selected with sliding window strategy Domain, designs the sliding windows of multiple and different scales to obtain insulator and disconnecting link target area, when insulator and disconnecting link target area Label is obtained when domain is as input；
- It is 7. according to claim 6 a kind of based on the disconnecting link state identification method for improving deep learning, it is characterised in that institute The quantity for stating sliding window is more than or equal to 10.
- It is 8. according to claim 1 a kind of based on the disconnecting link state identification method for improving deep learning, it is characterised in that institute State and candidate region is carried out to delete choosing acquisition candidate rectangle frame；Specifically include：Suspected target area is excluded using non-maxima suppression method Domain, only leaves the candidate rectangle frame of some different classes of very close real goals, it, which excludes foundation, is removed with smaller The suspected target region of IoU (intersection of union) value, wherein IoU are algorithm detection zone (DR：Detection Region) with real estate (GT:Ground Truth) between coincidence factor：
- It is 9. according to claim 1 a kind of based on the disconnecting link state identification method for improving deep learning, it is characterised in that institute State and the disconnecting link in exact rectangular frame and insulator state are judged, complete the identification of disconnecting link state, be specially：If one Disconnecting link region and two insulator regional connectivities, then it is assumed that the disconnecting link is in closure state, is otherwise off.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711207726.2A CN107944396B (en) | 2017-11-27 | 2017-11-27 | Knife switch state identification method based on improved deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711207726.2A CN107944396B (en) | 2017-11-27 | 2017-11-27 | Knife switch state identification method based on improved deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107944396A true CN107944396A (en) | 2018-04-20 |
| CN107944396B CN107944396B (en) | 2021-12-28 |
Family
ID=61949086
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711207726.2A Active CN107944396B (en) | 2017-11-27 | 2017-11-27 | Knife switch state identification method based on improved deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107944396B (en) |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109472268A (en) * | 2018-09-07 | 2019-03-15 | 广东电网有限责任公司 | A kind of protection pressing plate state identification method based on convolutional neural networks |
| CN109712118A (en) * | 2018-12-11 | 2019-05-03 | 武汉三江中电科技有限责任公司 | A kind of substation isolating-switch detection recognition method based on Mask RCNN |
| CN109741398A (en) * | 2018-12-28 | 2019-05-10 | 同方威视技术股份有限公司 | Image-recognizing method, device and computer storage medium |
| CN109753500A (en) * | 2018-12-21 | 2019-05-14 | 广州供电局有限公司 | Disconnecting link abnormal state data identification method, device and computer equipment |
| CN109753902A (en) * | 2018-12-21 | 2019-05-14 | 广西电网有限责任公司来宾供电局 | Power distribution cabinet square pressing plate state identification method based on the processing of hough changing image |
| CN110221200A (en) * | 2019-06-10 | 2019-09-10 | 河北工业大学 | A kind of omnipotent breaker attachment method for diagnosing faults based on deep learning |
| CN110533090A (en) * | 2019-08-21 | 2019-12-03 | 国网江苏省电力有限公司电力科学研究院 | Switch tool condition detection method and device |
| CN110728171A (en) * | 2019-08-19 | 2020-01-24 | 广东电网有限责任公司 | Fine detection and identification method for state of disconnecting link |
| CN111046943A (en) * | 2019-12-09 | 2020-04-21 | 国网智能科技股份有限公司 | Method and system for automatically identifying state of isolation switch of transformer substation |
| CN111199227A (en) * | 2019-12-20 | 2020-05-26 | 广西柳州联耕科技有限公司 | High-precision image identification method |
| CN111209958A (en) * | 2020-01-03 | 2020-05-29 | 南方电网科学研究院有限责任公司 | Transformer substation equipment detection method and device based on deep learning |
| CN111488990A (en) * | 2020-04-17 | 2020-08-04 | 苏州浪潮智能科技有限公司 | Model clipping method, device, equipment and medium based on performance perception |
| CN111754465A (en) * | 2020-06-04 | 2020-10-09 | 四川大学 | Insulator positioning and string drop detection method |
| CN111784677A (en) * | 2020-07-03 | 2020-10-16 | 安徽炬视科技有限公司 | Single-camera knife switch arm opening angle identification method based on scene analysis |
| CN111898481A (en) * | 2020-07-14 | 2020-11-06 | 济南信通达电气科技有限公司 | State identification method and device for pointer type opening and closing indicator |
| CN111950606A (en) * | 2020-07-28 | 2020-11-17 | 北京恒通智控机器人科技有限公司 | Disconnecting link state identification method, device, equipment and storage medium |
| CN111950369A (en) * | 2020-07-10 | 2020-11-17 | 济南信通达电气科技有限公司 | Transformer substation pointer type disconnecting link action degree identification method and device |
| CN112036402A (en) * | 2020-08-26 | 2020-12-04 | 济南信通达电气科技有限公司 | Split type disconnecting link state identification method and device |
| CN112418328A (en) * | 2020-11-25 | 2021-02-26 | 哈尔滨市科佳通用机电股份有限公司 | Deep learning-based drain outlet cover plate non-closing in-place fault detection method |
| CN113569819A (en) * | 2021-09-23 | 2021-10-29 | 泰豪软件股份有限公司 | Method and system for identifying opening and closing states of disconnecting link, storage medium and computer equipment |
| CN113569736A (en) * | 2021-07-28 | 2021-10-29 | 南方电网数字电网研究院有限公司 | Disconnecting link state identification method and device, computer equipment and storage medium |
| CN112036402B (en) * | 2020-08-26 | 2024-04-26 | 济南信通达电气科技有限公司 | Split type disconnecting link state identification method and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160018519A1 (en) * | 2014-07-18 | 2016-01-21 | Siemens Aktiengesellschaft | High frequency acoustic spectrum imaging method and device |
| CN106022345A (en) * | 2016-04-13 | 2016-10-12 | 杭州远鉴信息科技有限公司 | State identification method for high-voltage disconnecting switch based on Hough forest |
| CN106504233A (en) * | 2016-10-18 | 2017-03-15 | 国网山东省电力公司电力科学研究院 | Image electric power widget recognition methodss and system are patrolled and examined based on the unmanned plane of Faster R CNN |
| CN107257161A (en) * | 2017-06-20 | 2017-10-17 | 安徽南瑞继远电网技术有限公司 | A kind of transformer station's disconnecting link remote control auxiliary check method and system based on state recognition algorithm |
-
2017
- 2017-11-27 CN CN201711207726.2A patent/CN107944396B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160018519A1 (en) * | 2014-07-18 | 2016-01-21 | Siemens Aktiengesellschaft | High frequency acoustic spectrum imaging method and device |
| CN106022345A (en) * | 2016-04-13 | 2016-10-12 | 杭州远鉴信息科技有限公司 | State identification method for high-voltage disconnecting switch based on Hough forest |
| CN106504233A (en) * | 2016-10-18 | 2017-03-15 | 国网山东省电力公司电力科学研究院 | Image electric power widget recognition methodss and system are patrolled and examined based on the unmanned plane of Faster R CNN |
| CN107257161A (en) * | 2017-06-20 | 2017-10-17 | 安徽南瑞继远电网技术有限公司 | A kind of transformer station's disconnecting link remote control auxiliary check method and system based on state recognition algorithm |
Cited By (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109472268A (en) * | 2018-09-07 | 2019-03-15 | 广东电网有限责任公司 | A kind of protection pressing plate state identification method based on convolutional neural networks |
| CN109712118A (en) * | 2018-12-11 | 2019-05-03 | 武汉三江中电科技有限责任公司 | A kind of substation isolating-switch detection recognition method based on Mask RCNN |
| CN109753500A (en) * | 2018-12-21 | 2019-05-14 | 广州供电局有限公司 | Disconnecting link abnormal state data identification method, device and computer equipment |
| CN109753902A (en) * | 2018-12-21 | 2019-05-14 | 广西电网有限责任公司来宾供电局 | Power distribution cabinet square pressing plate state identification method based on the processing of hough changing image |
| CN109753500B (en) * | 2018-12-21 | 2021-02-02 | 广东电网有限责任公司广州供电局 | Method and device for identifying abnormal data of disconnecting link state and computer equipment |
| CN109753902B (en) * | 2018-12-21 | 2021-10-29 | 广西电网有限责任公司来宾供电局 | Hough transformation image processing-based power distribution cabinet square pressure plate state identification method |
| CN109741398A (en) * | 2018-12-28 | 2019-05-10 | 同方威视技术股份有限公司 | Image-recognizing method, device and computer storage medium |
| CN110221200B (en) * | 2019-06-10 | 2021-06-29 | 河北工业大学 | Universal circuit breaker accessory fault diagnosis method based on deep learning |
| CN110221200A (en) * | 2019-06-10 | 2019-09-10 | 河北工业大学 | A kind of omnipotent breaker attachment method for diagnosing faults based on deep learning |
| CN110728171A (en) * | 2019-08-19 | 2020-01-24 | 广东电网有限责任公司 | Fine detection and identification method for state of disconnecting link |
| CN110728171B (en) * | 2019-08-19 | 2023-01-20 | 广东电网有限责任公司 | Fine detection and identification method for state of disconnecting link |
| CN110533090A (en) * | 2019-08-21 | 2019-12-03 | 国网江苏省电力有限公司电力科学研究院 | Switch tool condition detection method and device |
| CN110533090B (en) * | 2019-08-21 | 2022-07-08 | 国网江苏省电力有限公司电力科学研究院 | Method and device for detecting state of switch knife switch |
| CN111046943A (en) * | 2019-12-09 | 2020-04-21 | 国网智能科技股份有限公司 | Method and system for automatically identifying state of isolation switch of transformer substation |
| CN111199227A (en) * | 2019-12-20 | 2020-05-26 | 广西柳州联耕科技有限公司 | High-precision image identification method |
| CN111209958A (en) * | 2020-01-03 | 2020-05-29 | 南方电网科学研究院有限责任公司 | Transformer substation equipment detection method and device based on deep learning |
| CN111209958B (en) * | 2020-01-03 | 2023-11-14 | 南方电网科学研究院有限责任公司 | Substation equipment detection method and device based on deep learning |
| CN111488990B (en) * | 2020-04-17 | 2022-11-29 | 苏州浪潮智能科技有限公司 | Model clipping method, device, equipment and medium based on performance perception |
| CN111488990A (en) * | 2020-04-17 | 2020-08-04 | 苏州浪潮智能科技有限公司 | Model clipping method, device, equipment and medium based on performance perception |
| CN111754465B (en) * | 2020-06-04 | 2023-06-09 | 四川大学 | Insulator positioning and string dropping detection method |
| CN111754465A (en) * | 2020-06-04 | 2020-10-09 | 四川大学 | Insulator positioning and string drop detection method |
| CN111784677A (en) * | 2020-07-03 | 2020-10-16 | 安徽炬视科技有限公司 | Single-camera knife switch arm opening angle identification method based on scene analysis |
| CN111950369A (en) * | 2020-07-10 | 2020-11-17 | 济南信通达电气科技有限公司 | Transformer substation pointer type disconnecting link action degree identification method and device |
| CN111898481A (en) * | 2020-07-14 | 2020-11-06 | 济南信通达电气科技有限公司 | State identification method and device for pointer type opening and closing indicator |
| CN111950606A (en) * | 2020-07-28 | 2020-11-17 | 北京恒通智控机器人科技有限公司 | Disconnecting link state identification method, device, equipment and storage medium |
| CN111950606B (en) * | 2020-07-28 | 2023-11-07 | 北京恒通智控机器人科技有限公司 | Knife switch state identification method, device, equipment and storage medium |
| CN112036402A (en) * | 2020-08-26 | 2020-12-04 | 济南信通达电气科技有限公司 | Split type disconnecting link state identification method and device |
| CN112036402B (en) * | 2020-08-26 | 2024-04-26 | 济南信通达电气科技有限公司 | Split type disconnecting link state identification method and device |
| CN112418328A (en) * | 2020-11-25 | 2021-02-26 | 哈尔滨市科佳通用机电股份有限公司 | Deep learning-based drain outlet cover plate non-closing in-place fault detection method |
| CN113569736A (en) * | 2021-07-28 | 2021-10-29 | 南方电网数字电网研究院有限公司 | Disconnecting link state identification method and device, computer equipment and storage medium |
| CN113569819B (en) * | 2021-09-23 | 2021-12-24 | 泰豪软件股份有限公司 | Method and system for identifying opening and closing states of disconnecting link, storage medium and computer equipment |
| CN113569819A (en) * | 2021-09-23 | 2021-10-29 | 泰豪软件股份有限公司 | Method and system for identifying opening and closing states of disconnecting link, storage medium and computer equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107944396B (en) | 2021-12-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107944396A (en) | A kind of disconnecting link state identification method based on improvement deep learning | |
| Zhou et al. | Safety helmet detection based on YOLOv5 | |
| CN106127204B (en) | A kind of multi-direction meter reading Region detection algorithms of full convolutional neural networks | |
| CN112380952B (en) | Power equipment infrared image real-time detection and identification method based on artificial intelligence | |
| CN105069472B (en) | A kind of vehicle checking method adaptive based on convolutional neural networks | |
| CN108564565A (en) | A kind of power equipment infrared image multi-target orientation method based on deep learning | |
| CN109508710A (en) | Based on the unmanned vehicle night-environment cognitive method for improving YOLOv3 network | |
| CN106971152A (en) | A kind of method of Bird's Nest in detection transmission line of electricity based on Aerial Images | |
| CN109492581A (en) | A kind of human motion recognition method based on TP-STG frame | |
| CN108764186A (en) | Personage based on rotation deep learning blocks profile testing method | |
| CN108197587A (en) | A kind of method that multi-modal recognition of face is carried out by face depth prediction | |
| CN109285139A (en) | A kind of x-ray imaging weld inspection method based on deep learning | |
| CN109635875A (en) | A kind of end-to-end network interface detection method based on deep learning | |
| CN110148162A (en) | A kind of heterologous image matching method based on composition operators | |
| CN109255776A (en) | A kind of transmission line of electricity split pin defect automatic identifying method | |
| CN107133569A (en) | The many granularity mask methods of monitor video based on extensive Multi-label learning | |
| CN106557740B (en) | The recognition methods of oil depot target in a kind of remote sensing images | |
| CN103390164A (en) | Object detection method based on depth image and implementing device thereof | |
| CN107545571A (en) | A kind of image detecting method and device | |
| CN110298265A (en) | Specific objective detection method in a kind of elevator based on YOLO neural network | |
| CN108229524A (en) | A kind of chimney and condensing tower detection method based on remote sensing images | |
| CN107194938A (en) | Image outline detection method based on depth convolutional neural networks | |
| CN108257148A (en) | The target of special object suggests window generation method and its application in target following | |
| CN108596121A (en) | A kind of face critical point detection method based on context and structural modeling | |
| CN116205905B (en) | Power distribution network construction safety and quality image detection method and system based on mobile terminal |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |