CN108960241A - Insulator based on statistical shape model falls to go here and there detection algorithm - Google Patents
Insulator based on statistical shape model falls to go here and there detection algorithm Download PDFInfo
- Publication number
- CN108960241A CN108960241A CN201810836240.3A CN201810836240A CN108960241A CN 108960241 A CN108960241 A CN 108960241A CN 201810836240 A CN201810836240 A CN 201810836240A CN 108960241 A CN108960241 A CN 108960241A
- Authority
- CN
- China
- Prior art keywords
- insulator
- shape
- model
- statistical
- here
- 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
- 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
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
Fall to go here and there detection algorithm the present invention provides a kind of insulator based on statistical shape model comprising: input picture is pre-processed；Interested target area is extracted, insulation sub-information should be included in the region；Son is described using GLOH to be accurately positioned insulator, and advantage of the statistical shape model in terms of image segmentation is recycled to carry out Accurate Segmentation insulator；The position for the insulator being partitioned into is analyzed, to achieve the purpose that detect whether insulator falls string.
Description
Technical field
The present invention relates to electric system, pattern-recognition and the classification and detection of classify field, especially specific objective.
Background technique
In the power system, insulator is a kind of special insulation control, for preventing the charging member shape of transmission line of electricity
At Grounding.It is chronically at due to insulator in severe natural environment, so often will appear " falling string " failure, this event
Barrier can make power grid parallel off, lead to large-area power-cuts, cause high risks (bibliography " Wang Yin to the safety and stability of electric system
It is vertical, Yan Bin, the detection and positioning [J] computer engineering and design of insulator " falling string " defect of view-based access control model, 2014,5
(2): 583-587. ").Dependent on manually carrying out the mode of inspection and maintenance not only low efficiency, at high cost to transmission line of electricity, and
And there is very big safety issues.With the development of intellectual technology, using computer vision and image processing techniques to exhausted
Edge state detect, automatic identification defects of insulator, has in terms of the safe operation for ensureing electric system very heavy
The effect wanted.
In the recent period, researchers propose the algorithm of many recognition detection insulators.Document (" Wang Yinli, Yan Bin, based on view
The detection of insulator " falling string " defect of feel and positioning [J] computer engineering and design, 2014,5 (2): 583-587. ") it mentions
A kind of insulator of view-based access control model falls to go here and there defect recognition algorithm out, the color space algorithm combination Lab, " maximum between-cluster variance
Method " and " area morphology " carry out coarse segmentation to insulation subgraph, and according to the insulation subgraph of acquisition, founding mathematical models lead to
The methods of center position coordinates and the area coordinate for crossing mathematical derivation calculating insulator judge whether insulator has string defect.Text
Offer (" Tong Weiguo, the transmission line of electricity identification based on Aerial Images study [D] North China Electric Power University with condition detection method,
2011. ") by extracting the features such as invariant moment features, color characteristic, shape feature and wavelet coefficient of insulator in Aerial Images
It carries out, proposes a kind of insulator recognizer of combination multi-characteristicquantity quantity, common failure is diagnosed.Document (" Zheng Tao,
Taking photo by plane based on PCNN insulate subgraph segmentation and Position Research [D] the Maritime Affairs University Of Dalian, 2011. ") utilize image
Partitioning algorithm simultaneously combines Pulse Coupled Neural Network (PCNN), and to taking photo by plane, the subgraph that insulate is divided, to the figure after segmentation
As carrying out Generalized Hough Transform, and then identify insulator.Document (" Tong Weiguo, in great into equal insulation neural network based
Sub- fault diagnosis [J] Computer Simulation, 2013,30 (9): 310-313. ") by full skirt ratio, smoothness and being connected to
It is extracted than these three features, nerve net is established using the improved algorithm that adaptation rate is combined with additional momentum
Network identifies the filth of insulator, falls string and three kinds of failures of crackle.Document (" Li Hong, the insulator state based on rarefaction representation
Recognition methods study [D] North China Electric Power University, 2016. ") by Hough transform detection straight line determine insulator in the picture
Position, then using SVM classifier by the result of Primary Location carry out classification realize insulator positioning, pass through extract crackle
The super complete dictionary that rarefaction representation classifier is established with the feature vector for falling insulator string carries out fault identification to insulator, but
This method is higher to image quality requirements.
Influence due to insulator image data acquisition vulnerable to many factors such as sensor, shooting environmental, shooting angle,
There are the complicated cases such as dimension rotation, scaling, illumination variation for image obtained, under these conditions, the above method there is
It positions, extract the problems such as not accurate enough.
Summary of the invention
Fall to go here and there detection algorithm the present invention provides a kind of insulator based on statistical shape model comprising: input is schemed
As being pre-processed；Interested target area is extracted, insulation sub-information should be included in the region；Son is described come smart using GLOH
It determines position insulator, advantage of the statistical shape model in terms of image segmentation is recycled to carry out Accurate Segmentation insulator；To being partitioned into
The position of insulator analyzed, to achieve the purpose that detect whether insulator falls string.
Before carrying out image processing and analysis, need to carry out the image comprising insulator necessary pretreatment to reach
Reduce the influence of the factors such as noise.Median filtering is carried out to image first, median filtering can overcome edge blurry and noise etc.
Problem, can preferably retention insulator marginal information；Enhance the contrast of image, using histogram equalization with convenient
The positioning of insulator.Then bianry image is obtained to realize the separation of foreground and background using thresholding method.
Image is detected using Hessian-Affine Region detection algorithms to obtain characteristic area, in detection part
After characteristic point, characteristic point is described using GLOH operator.The extracted characteristics of image of GLOH operator, to dimension rotation, contracting
Put, illumination variation is able to maintain invariance, have stronger anti-noise ability.
Utilize K-Means algorithm training visual dictionary: K central point of random initializtion first；Secondly the distance that each sample point arrives K central point respectively is calculated,And each sample mark is classified as that one kind nearest from it；According to the classification of current sample,
It recalculates, update central point；Terminate if central point does not update, otherwise continues each sample point and arrive K central point respectively
Distance is iterated.
The visual dictionary of the insulator of generation can indicate are as follows:
(1)
WhereinIndicate the histogram vectors of k-th of vision word, it is the average value of all feature vectors in kth class.For matching threshold, obtained by training sample.
By matching, the feature of most of nonisulated sons is excluded, the feature of all insulators is left.By characteristic matching
Afterwards, the position of target is it will be apparent that defining ballot matrix according to the local feature region of acquisition:
(2)
WhereinIndicate the scale of support area.Ballot matrix is handled using threshold value, the area of insulator can be obtained
Domain.
After region detection, usePass through points distribution models (point distribution
Model, PDM) shape described.But often there is the differences in size, position and direction for the shape sample in different images
It is different, need to analyze progress shape alignment by broad sense Pu Shi, process is the difference minimized between each shape and average shape
Away fromAs target alignment shape.
Then a statistical model is obtained to the shape progress principal component analysis after alignment to describe shape.Finally, using
During model guidance segmentation, by adjusting aspect of model parameter, iterative model be allowed to constantly with object matching and reach point
Cut the purpose of target.
After the complete image for obtaining single insulator, the number of statistics insulation, it is assumed that have a insulator, be denoted asThe distance between, and calculate adjacent insulatorIndicate theA insulation
Son is the distance between to the insulation subcenter.It willCarry out statistic of classification, it is assumed that be divided into class, it is contemplated that the error of calculating is pressed
Classify according to following methods, that is, takes lesser numberIf, then it is assumed thatWithIt is same
Class, will
(3)
As the actual distance between insulator, wherein,Indicate element in set D
Number.It enables, whereinExpression and immediate integer, if, then it is assumed that insulatorWith
Between there is string.And falls insulator string number and be.If from this method description as can be seen that fall to go here and there phenomenon it is very serious,
Calculating inaccuracy then, the judgment method may be not suitable for, need empirically to be manually set at this time.But such case is not
It is common, even if there are be easy to judgement and exist to fall to go here and there phenomenon.
Detailed description of the invention
Fig. 1 is insulator localization method；
Fig. 2 is that the insulator based on statistical shape model falls to go here and there detection algorithm flow chart.
Specific embodiment
According to Fig. 1, method shown in Fig. 2 and flow chart, before carrying out image processing and analysis, need to comprising insulation
The image of son carries out necessary pretreatment to reach the influence for reducing the factors such as noise.Median filtering is carried out to image first, in
Value filtering can overcome the problems such as edge blurry and noise, can preferably retention insulator marginal information；Utilize histogram
It equalizes to enhance the contrast of image, to facilitate the positioning of insulator.Then using thresholding method obtain bianry image with
Realize the separation of foreground and background.
Image is detected using Hessian-Affine Region detection algorithms to obtain characteristic area, in detection part
After characteristic point, characteristic point is described using GLOH operator.The extracted characteristics of image of GLOH operator, to dimension rotation, contracting
Put, illumination variation is able to maintain invariance, have stronger anti-noise ability.
Utilize K-Means algorithm training visual dictionary: K central point of random initializtion first；Secondly the distance that each sample point arrives K central point respectively is calculated,And each sample mark is classified as that one kind nearest from it；According to the classification of current sample,
It recalculates, update central point；Terminate if central point does not update, otherwise continues each sample point and arrive K central point respectively
Distance is iterated.
The visual dictionary of the insulator of generation can indicate are as follows:
(1)
WhereinIndicate the histogram vectors of k-th of vision word, it is the average value of all feature vectors in kth class.For matching threshold, obtained by training sample.
By matching, the feature of most of nonisulated sons is excluded, the feature of all insulators is left.By characteristic matching
Afterwards, the position of target is it will be apparent that defining ballot matrix according to the local feature region of acquisition:
(2)
WhereinIndicate the scale of support area.Ballot matrix is handled using threshold value, the area of insulator can be obtained
Domain.
After region detection, usePass through points distribution models (point distribution
Model, PDM) shape described.But often there is the differences in size, position and direction for the shape sample in different images
It is different, need to analyze progress shape alignment by broad sense Pu Shi, process is the difference minimized between each shape and average shape
Away fromAs target alignment shape.
Then a statistical model is obtained to the shape progress principal component analysis after alignment to describe shape.Finally, using
During model guidance segmentation, by adjusting aspect of model parameter, iterative model be allowed to constantly with object matching and reach point
Cut the purpose of target.
After the complete image for obtaining single insulator, the number of statistics insulation, it is assumed that have a insulator, be denoted asThe distance between, and calculate adjacent insulator,Indicate theA insulator
The distance between to the insulation subcenter.It willCarry out statistic of classification, it is assumed that be divided into class, it is contemplated that the error of calculating, according to
Following methods are classified, that is, take lesser numberIf, then it is assumed thatWithIt is same
Class, will
(3)
As the actual distance between insulator, wherein,Indicate element in set D
Number.It enables, whereinExpression and immediate integer, if, then it is assumed that insulatorWith
Between there is string.And falls insulator string number and be.If from this method description as can be seen that fall to go here and there phenomenon it is very serious,
ThenCalculating inaccuracy, which may be not suitable for, need empirically to be manually set at this time.But such case is simultaneously
It is rare, even if there are be easy to judgement and exist to fall to go here and there phenomenon.
Claims (5)
1. the insulator based on statistical shape model falls to go here and there detection algorithm, it is characterised in that:
It is different from existing insulator and falls crosstalk detecting method, the accurate segmentation of insulator is carried out using statistical shape model；
It is used in feature description: (1) gradient orientation histogram (Gradient Location-Orientation
Histogram, GLOH) feature description is carried out, have and invariance is able to maintain to dimension rotation, scaling, illumination variation, and
There is stronger anti-noise ability；(2) K-Means clustering algorithm training visual dictionary, excludes the feature of nonisulated son, retention insulator
Feature achievees the purpose that detect insulator in image.
2. the insulator based on statistical shape model as described in claim 1 falls to go here and there detection algorithm, which is characterized in that described
Method uses:
Statistical shape model carries out shape alignment by broad sense Pu Shi analysis method, and all shapes in training set are aligned in system
Under one coordinate system, the gap between each shape and average shape is minimizedAs target alignment shape；It is right
Shape after alignment carries out principal component analysis, after statistical analysis, obtains a statistical model to describe shape；Use model
Guidance segmentation during, by adjusting aspect of model parameter, iterative model be allowed to constantly with object matching and reach segmentation mesh
Target purpose.
3. the insulator based on statistical shape model as described in claim 1 falls to go here and there detection algorithm, which is characterized in that described
Method uses:
GLOH description divides feature vertex neighborhood using log-polar, that is, establishes the nesting that three radiuses are gradually increased
Border circular areas (radius size is respectively 6,11,15) simultaneously carries out the division of 8 angle directions in two border circular areas of outermost, this
Sample just forms 17 nonoverlapping subregions, and GLOH description establishes the gradient orientation histogram of 16 columns to each subregion, this
The sub- dimension of description constructed by sample is 17 × 16=272 dimensions.
4. most representational 128 dimension element is selected using PCA algorithm to form final GLOH description.
5. the insulator based on statistical shape model as described in claim 1 falls to go here and there detection algorithm, which is characterized in that described
Method uses:
K-Means cluster is a kind of unsupervised learning method, is widely used, is an iterative algorithm, learning objective is
The quadratic sum of all samples to corresponding cluster centre is minimized, to make the similar sample generated as close possible to different
Class sample is as separated as possible, can be indicated using the visual dictionary of the insulator of K-Means clustering algorithm training are as follows:
(1).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810836240.3A CN108960241A (en) | 2018-07-26 | 2018-07-26 | Insulator based on statistical shape model falls to go here and there detection algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810836240.3A CN108960241A (en) | 2018-07-26 | 2018-07-26 | Insulator based on statistical shape model falls to go here and there detection algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108960241A true CN108960241A (en) | 2018-12-07 |
Family
ID=64464874
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810836240.3A Pending CN108960241A (en) | 2018-07-26 | 2018-07-26 | Insulator based on statistical shape model falls to go here and there detection algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108960241A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113538382A (en) * | 2021-07-19 | 2021-10-22 | 安徽炬视科技有限公司 | Insulator detection algorithm based on non-deep network semantic segmentation |
| CN116596908A (en) * | 2023-05-30 | 2023-08-15 | 南京亦鑫同电气有限责任公司 | Wire and cable safety state assessment method and system based on data processing |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103529362A (en) * | 2013-10-28 | 2014-01-22 | 国家电网公司 | Perception based insulator recognition and defect diagnosis method |
| CN104483326A (en) * | 2014-12-19 | 2015-04-01 | 长春工程学院 | High-voltage wire insulator defect detection method and high-voltage wire insulator defect detection system based on deep belief network |
| US20150221079A1 (en) * | 2014-01-31 | 2015-08-06 | Pictometry International Corp. | Augmented Three Dimensional Point Collection of Vertical Structures |
| CN106408025A (en) * | 2016-09-20 | 2017-02-15 | 西安工程大学 | Classification and recognition method of aerial image insulators based on image processing |
| CN106570853A (en) * | 2015-10-08 | 2017-04-19 | 上海深邃智能科技有限公司 | Shape and color integration insulator identification and defect detection method |
| CN106778734A (en) * | 2016-11-10 | 2017-05-31 | 华北电力大学（保定） | A kind of insulator based on rarefaction representation falls to go here and there defect inspection method |
-
2018
- 2018-07-26 CN CN201810836240.3A patent/CN108960241A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103529362A (en) * | 2013-10-28 | 2014-01-22 | 国家电网公司 | Perception based insulator recognition and defect diagnosis method |
| US20150221079A1 (en) * | 2014-01-31 | 2015-08-06 | Pictometry International Corp. | Augmented Three Dimensional Point Collection of Vertical Structures |
| CN104483326A (en) * | 2014-12-19 | 2015-04-01 | 长春工程学院 | High-voltage wire insulator defect detection method and high-voltage wire insulator defect detection system based on deep belief network |
| CN106570853A (en) * | 2015-10-08 | 2017-04-19 | 上海深邃智能科技有限公司 | Shape and color integration insulator identification and defect detection method |
| CN106408025A (en) * | 2016-09-20 | 2017-02-15 | 西安工程大学 | Classification and recognition method of aerial image insulators based on image processing |
| CN106778734A (en) * | 2016-11-10 | 2017-05-31 | 华北电力大学（保定） | A kind of insulator based on rarefaction representation falls to go here and there defect inspection method |
Non-Patent Citations (1)
| Title |
|---|
| 汪晓 等: "基于统计形状模型的绝缘子 "掉串"检测算法", 《计算机测量与控制》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113538382A (en) * | 2021-07-19 | 2021-10-22 | 安徽炬视科技有限公司 | Insulator detection algorithm based on non-deep network semantic segmentation |
| CN113538382B (en) * | 2021-07-19 | 2023-11-14 | 安徽炬视科技有限公司 | Insulator detection algorithm based on non-deep network semantic segmentation |
| CN116596908A (en) * | 2023-05-30 | 2023-08-15 | 南京亦鑫同电气有限责任公司 | Wire and cable safety state assessment method and system based on data processing |
| CN116596908B (en) * | 2023-05-30 | 2024-02-06 | 南京亦鑫同电气有限责任公司 | Wire and cable safety state assessment method and system based on data processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109118479B (en) | Capsule network-based insulator defect identification and positioning device and method | |
| CN110543837B (en) | Visible light airport airplane detection method based on potential target point | |
| CN111428748B (en) | HOG feature and SVM-based infrared image insulator identification detection method | |
| Oberweger et al. | Visual recognition and fault detection for power line insulators | |
| Gleason et al. | Vehicle detection from aerial imagery | |
| CN107610114B (en) | optical satellite remote sensing image cloud and snow fog detection method based on support vector machine | |
| Liao et al. | A robust insulator detection algorithm based on local features and spatial orders for aerial images | |
| Chen et al. | Vehicle detection in high-resolution aerial images via sparse representation and superpixels | |
| Wang et al. | Insulator identification from aerial images using support vector machine with background suppression | |
| CN112183667B (en) | Insulator fault detection method in cooperation with deep learning | |
| CN106610969A (en) | Multimodal information-based video content auditing system and method | |
| CN113205063A (en) | Visual identification and positioning method for defects of power transmission conductor | |
| CN105678760A (en) | Method for recognizing insulator image on the basis of Canny edge detection algorithm | |
| CN111814686A (en) | Vision-based power transmission line identification and foreign matter invasion online detection method | |
| CN111080691A (en) | Infrared hot spot detection method and device for photovoltaic module | |
| CN110751619A (en) | Insulator defect detection method | |
| CN105303162B (en) | A kind of Aerial Images insulator recognition methods based on target proposed algorithm | |
| CN108073940B (en) | Method for detecting 3D target example object in unstructured environment | |
| CN111460881A (en) | Traffic sign countermeasure sample detection method and classification device based on neighbor discrimination | |
| Wang et al. | Railway insulator detection based on adaptive cascaded convolutional neural network | |
| CN113447771A (en) | Partial discharge pattern recognition method based on SIFT-LDA characteristics | |
| CN108960241A (en) | Insulator based on statistical shape model falls to go here and there detection algorithm | |
| CN117197700B (en) | Intelligent unmanned inspection contact net defect identification system | |
| CN111507249A (en) | Transformer substation nest identification method based on target detection | |
| Liu et al. | A robust algorithm for text detection in color images |
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: 20181207 |