CN107358259A - Covering ice for insulator detection method based on GLOH descriptions and GVF Snake models - Google Patents

Covering ice for insulator detection method based on GLOH descriptions and GVF Snake models Download PDF

Info

Publication number
CN107358259A
CN107358259A CN201710568690.4A CN201710568690A CN107358259A CN 107358259 A CN107358259 A CN 107358259A CN 201710568690 A CN201710568690 A CN 201710568690A CN 107358259 A CN107358259 A CN 107358259A
Authority
CN
China
Prior art keywords
mrow
insulator
msub
class
characteristic
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
Application number
CN201710568690.4A
Other languages
Chinese (zh)
Other versions
CN107358259B (en
Inventor
赵常威
程登峰
刘安迪
季坤
操松元
严波
李森林
陈江
何凯
陈忠
杨为
杨海涛
邓倩倩
张国宝
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Anhui Nari Jiyuan Power Grid Technology Co Ltd
State Grid Corp of China SGCC
Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd
Anhui University
Original Assignee
Anhui Nari Jiyuan Power Grid Technology Co Ltd
State Grid Corp of China SGCC
Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd
Anhui University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Anhui Nari Jiyuan Power Grid Technology Co Ltd, State Grid Corp of China SGCC, Electric Power Research Institute of State Grid Anhui Electric Power Co Ltd, Anhui University filed Critical Anhui Nari Jiyuan Power Grid Technology Co Ltd
Priority to CN201710568690.4A priority Critical patent/CN107358259B/en
Publication of CN107358259A publication Critical patent/CN107358259A/en
Application granted granted Critical
Publication of CN107358259B publication Critical patent/CN107358259B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • 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
    • G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
    • 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
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a kind of covering ice for insulator detection method based on GLOH descriptions and GVF Snake, including：1 pair of input picture pre-processes；2 extractions target area interested, region planted agent include insulation sub-information；3 describe son to position insulator using GLOH；4 accurately detect insulator contour using GVF Snake models；5 pairs of insulator contours are analyzed, and determine that insulator whether there is icing situation by calculating maximum radial distance, and the thickness of icing is calculated in the case of icing.The defects of present invention can overcome traditional method to be difficult to intactly separate insulator with complex background, the accuracy of covering ice for insulator situation judgement can be effectively improved.

Description

Covering ice for insulator detection method based on GLOH descriptions and GVF-Snake models
Technical field
The present invention relates to power system, pattern-recognition and the classification and detection of classify field, especially specific objective.
Background technology
In power system, covering ice for insulator is always a very big threat for power grid security.Particularly in cold Winter, covering ice for insulator situation occurs often, and icing flashover is that larger harm occurs in a kind of transmission line of electricity often, right The safe operation of the insulating properties and system of circuit has very big potential safety hazard, thus carries out detection to covering ice for insulator and have Important meaning.With the development of machine vision technique and image processing techniques, using vision and image processing techniques to insulation Sub- state detect, automatic identification covering ice for insulator situation, has very in terms of the safe operation of power system is ensured Important effect.
In the recent period, researchers propose the algorithm of many recognition detection covering ice for insulator situations.Document (" is opened into Sheng Ge Covering ice for insulator automatic identifications [J] of the such as Hao, Jiang Xiuchen based on image processing techniques, east china electric power, 2009,37 (1):146- 149. ") covering ice for insulator feature is analyzed, it is special to the situation of covering ice for insulator and insulation spirte using image processing techniques Sign amount analyzed and identified, obtains insulator contour image the methods of application image smoothing processing and threshold transformation, and calculate The parameters such as ice covering thickness, but used threshold transformation method is difficult to the interference that excludes contextual factor completely, affects insulation The extraction of sub- profile；Document (" the electric transmission line isolator ice covering thickness image recognition algorithms such as Zhang Ye, Feng Ling, Mu Jingyu [J], Automation of Electric Systems, 2016,40 (21):195-202. ") realize determining for insulator in the picture by template matching technique Position, then by technologies such as image preprocessing, image segmentation, edge extractings, insulator edge before and after icing is extracted in image, Then contrast district number of pixels determines whether icing, but insulator shape is changeable in the case of icing, used template Matching is difficult to provide accurate template, can reduce the precision of insulator positioning；Document (" Lin Yang, Xiaolan Jiang, Yanpeng Hao etc.Recognition of natural ice types on in-service glass insulators based on texture feature descriptor[J],IEEE Transactions on Dielectrics and Electrical Insulation,2017,24(1):535-542. ") calculated based on textural characteristics description Son establishes the ice kind identification method of glass insulator, and this method utilizes consistent local binary patterns and improved consistent local two Value pattern is used for the textural characteristics for extracting six type icing images, is calculated by the coefficient correlation of Texture similarity and is covered to identify Ice type, but this method is merely able to identify six type icing, and insulator is not implemented and is automatically positioned；Document (" Wang little Peng, recklessly Woods is built, grandson just newly waits application images edge detection method on-line monitoring electric power line ice-covering thickness research [J] High-Voltage Electrical Appliances, 2009,45 (6):69-74. ") based on the method for detecting image edge of small echo come extract profile carry out ice covering thickness measurement, and Experimental verification is carried out in artificial-climate laboratory is simulated, but this method is imaged by a technical grade on shaft tower Machine gathers image, can not carry out icing judgement to any input picture, universality is poor；(" Yang Hao, fearless are based on three-dimensional to document Covering ice for insulator image monitoring [J] Electric Power Automation Equipments of reconstruction, 2013,33 (2):92-98. ") pass through the three-dimensional of image Rebuild to carry out covering ice for insulator on-line monitoring, using computer binocular vision technology, diverse location is placed on using two Camera acquisition insulation subgraph, the three-dimensional point cloud model that insulator is carried out by the parallax between image is rebuild, so as to calculate To the thickness of icing, but this method requires installed video camera to quasi-insulator, and needs to carry out the demarcation of video camera.
In conventional images monitoring method, need the position of pre-set insulator mostly, or simply to laboratory this Covering ice for insulator situation under kind of ecotopia is identified, apart from practical application, also there is a certain distance.Due to figure As the complexity obtained, insulator usually combines together with background, and traditional method is difficult by its complete separation.
The content of the invention
The present invention is the defects of overcoming traditional method to be difficult to intactly separate insulator with complex background, there is provided a kind of Covering ice for insulator detection algorithm based on GLOH descriptions and GVF-Snake models, to can accurately be automatically positioned insulator, So as to effectively improve the accuracy of covering ice for insulator situation judgement.
The present invention is that technical scheme is used by solving technical problem：
The present invention it is a kind of based on GLOH description son and GVF-Snake models covering ice for insulator detection method the characteristics of be by Following steps are carried out：
Some width insulation subgraph in step 1, collection transmission line of electricity, and in some width insulation subgraphs progress The pretreatment of value filtering and histogram equalization, obtain some pretreated insulation subgraphs；
Step 2, using Hessian-Affine Region detection algorithms to described some pretreated insulation subgraphs Detected, obtain corresponding characteristic area；Assuming that n characteristic point is included in any one characteristic area；
Step 3, using GLOH description son the n characteristic point is described, obtain n characteristic vector；J-th of spy of note It is T to levy vectorj, 1≤j≤n；
Step 4, the visual dictionary using machine learning method training insulator：
K step 4.1, random initializtion central point Represent i-th of central point, i=1, 2,…,K；
Step 4.2, calculate j-th of characteristic vector TjTo i-th of central pointDistanceSo as to obtain jth Individual characteristic vector TjTo the distance { d of K central point1,d2,…,di,…,dK}；From the distance { d1,d2,…,di,…,dKIn Choose minimum value and by j-th of characteristic vector TjIt is referred in the class corresponding to minimum value；So as to which n characteristic vector be referred to In corresponding class, and obtain K class；
Step 4.3, the central point for calculating the K classAnd with original K central pointIt is compared, if all same, performs step 4.4；Otherwise, by the center for K class being calculated PointIt is assigned to the K central pointAfter being updated, return to step 4.2；
Step 4.4, the central point for calculating the K classIn each class average value, put down Mean vector Represent the average value of i-th of class；
Step 4.5, using formula (1) obtain the visual dictionary L of insulator：
In formula (1), μiFor the matching threshold of i-th of class, matching threshold μiFor all characteristic vectors in i-th of class and i-th The average value of classBetween difference maximum norm；
Step 5, by dictionary pattern matching, eliminate nonisulated subcharacter
Step 5.1, the width obtained on transmission line of electricity insulate subgraph as test image, and according to step 1 to step 3 Handled, obtain m characteristic vector of the test image, be designated as { T(1),T(2),…,T(k),…,T(m)}；T(k)Represent institute State the characteristic vector of k-th of characteristic point in the characteristic area of test image, 1≤k≤m；
Step 5.2, initialization k=1；
Step 5.3, initialization i=1；
Step 5.4, by k-th of characteristic vector T(k)With the average value of i-th of classThe difference that is obtained after subtracting each other with i-th The matching threshold μ of classiCompare, if difference is less than or equal to matching threshold μi, then it represents that k-th of characteristic vector T(k)Belong to i-th Class, perform step 5.5；Otherwise, k-th of characteristic vector T is represented(k)I-th of class is not belonging to, performs step 5.6；
Step 5.5, judge whether k ＞ m set up, if so, then represent that test image has matched with the visual dictionary L Into the characteristic vector after being updated is as insulation subcharacter；Otherwise, after k+1 being assigned into k, return to step 5.3；
Step 5.6, judge whether i ＞ K set up, if so, then represent k-th of characteristic vector T(k)It is not belonging to any one Class, and k-th of characteristic vector T is deleted from m characteristic vector(k)Afterwards, k+1 is assigned to k, performs step 5.3；Otherwise, by i+ After 1 is assigned to i, return to step 5.4；
Step 6, the ballot using any one pixel (x, y) in the characteristic area of formula (2) the calculating test image Value V (x, y)：
In formula (2), xhAnd yhRepresent h-th of insulation subcharacter institute in the test image in the characteristic vector after renewal The abscissa and ordinate of corresponding pixel, 1≤h≤H, H represent the sum of insulation subcharacter；σ represents the chi of characteristic area Degree；
Step 7, judge whether V (x, y) is more than or equal to set threshold value, if being more than or equal to, then it represents that pixel (x, y) The region belonged to where insulator, and retain pixel (x, y)；Otherwise, it is the area where nonisulated son to represent pixel (x, y) The pixel in domain；
Step 8, repeat step 6- steps 7, so as to complete the inspection of all pixels point in the characteristic area of the test image Survey, and obtain the subregion that insulate；
Step 9, using GVF-Snake models to it is described insulation subregion carry out contour detecting, obtain insulator contour；
Step 10, icing judge
Step 10.1, maximum radial distance in the insulator contour is calculated, and be designated as d；
Step 10.2, judge | d-d ' | whether ＞ δ set up, if so, then represent that insulator has icing, and perform step 10.3；Otherwise, the non-icing of insulator is represented；Threshold value set by δ；Radial direction when d ' is insulator non-icing in the picture away from From；
Step 10.3, using formula (3) obtain the thickness D of icing on insulator：
D=0.5 (d-d') d "/d ' (3)
In formula (3), d " represents the actual radial distance of insulator.
Compared with existing technology, beneficial effects of the present invention are embodied in：
1st, the inventive method pre-processes to input picture first；Secondly target area interested, the region are extracted Planted agent includes insulation sub-information；Then machine learning method combination GLOH describes son to train visual dictionary, so as to position insulation Son, recycles GVF-Snake models accurately to detect insulator contour；Finally insulator contour is analyzed, passes through calculating Maximum radial distance whether there is icing situation to determine insulator, and the thickness of icing, Neng Gouzhun are calculated in the case of icing Insulator really is automatically positioned, so as to effectively increase the accuracy of covering ice for insulator situation judgement；
2nd, the present invention is used and image is detected to obtain characteristic area using Hessian-Affine Region detection algorithms Domain, affine invariants are obtained by way of multiple dimensioned iteration, so as to effectively detect affine covariant region；
3rd, characteristics of image of the present invention given by using GLOH description, can change to scaling, rotation, illumination etc. Maintain the invariance, anti-noise ability is strong, can describe the subcharacter that insulate well；
4th, the present invention trains visual dictionary using machine learning method, and model training link is simple, while sample is more Sample also improves the adaptability to different scenes, by dictionary pattern matching, so as to improve the precision of insulator positioning；
5th, the present invention extracts the profile of insulator using GVF-Snake models, takes into account the continuity, smooth of contour curve Property, while initial value to contour curve and insensitive, concave edge circle can be converged to, ensure that the accurate of insulator contour extraction Property.
Brief description of the drawings
Fig. 1 is the inventive method flow chart；
Embodiment
As shown in figure 1, in the present embodiment, a kind of covering ice for insulator based on GLOH descriptions and GVF-Snake models is examined Survey method is that input picture is pre-processed；Secondly target area interested is extracted, region planted agent believes comprising insulator Breath；Then machine learning method combination GLOH describes son to train visual dictionary, so as to position insulator, recycles GVF Snake models accurately detect insulator contour；Finally insulator contour is analyzed, by calculate maximum radial distance come Determine that insulator whether there is icing situation, and the thickness of icing is calculated in the case of icing, specifically enter as follows OK：
Some width insulation subgraph in step 1, collection transmission line of electricity, and intermediate value filter is carried out to some width insulation subgraph The pretreatment of ripple and histogram equalization, the problems such as medium filtering can remove edge blurry and noise, can preferably will be absolutely The marginal information of edge remains；Strengthen the contrast of image using histogram equalization, to facilitate the positioning of insulator And obtain some pretreated insulation subgraphs；
Step 2, using Hessian-Affine Region detection algorithms it is pretreated to some insulation subgraph carry out Detection, obtains corresponding characteristic area；And characteristic point is detected in characteristic area, it is assumed that n is included in any one characteristic area Individual characteristic point；
Step 3, using GLOH description son n characteristic point is described, obtain n characteristic vector；Remember j-th of feature to Measure as Tj, 1≤j≤n；
GLOH is that a kind of description is proposed on the basis of SIFT description, its purpose in order to strengthen its robustness and Independence, divided using the space of log-polar to calculate histogram of gradients and reduce description using Principal Component Analysis Algorithm The dimension of son, specific implementation may be referred to Mikolajczyk in IEEE mode identification in 2005 and the phase of artificial intelligence transactions the 10th The article delivered.
Step 4, using the K-means clustering algorithms in machine learning method train insulator visual dictionary：
Characteristic vector set corresponding to K step 4.1, random initializtion central point Represent I-th of central point, i=1,2 ..., K；
Step 4.2, calculate j-th of characteristic vector TjTo i-th of central pointDistanceRepresent vector 2 norms, so as to obtain j-th of characteristic vector TjTo the distance { d of K central point1,d2,…,di,…,dK}；From distance { d1, d2,…,di,…,dKIn choose minimum value and by j-th of characteristic vector TjIt is referred in the class corresponding to minimum value；So as to by n Individual characteristic vector is referred in corresponding class, and obtains K class；
Step 4.3, the central point for calculating K class of gainedAnd with original K central pointIt is compared, if all same, performs step 4.4；Otherwise, by the center for K class being calculated PointIt is assigned to K central pointAfter being updated, return to step 4.2；
Step 4.4, the central point for calculating K classIn each class average value, obtain average value Vector Represent the average value of i-th of class；
Step 4.5, using formula (1) obtain the visual dictionary L of insulator：
In formula (1), μiFor the matching threshold of i-th of class, matching threshold μiFor all characteristic vectors in i-th of class and i-th The average value T of classiBetween difference maximum norm；
Step 5, by dictionary pattern matching, eliminate nonisulated subcharacter
Step 5.1, the width obtained on transmission line of electricity insulate subgraph as test image, and according to step 1 to step 3 Handled, obtain m characteristic vector of test image, be designated as { T(1),T(2),…,T(k),…,T(m)}；T(k)Represent test chart The characteristic vector of k-th of characteristic point in the characteristic area of picture, 1≤k≤m；
Step 5.2, initialization k=1；
Step 5.3, initialization i=1；
Step 5.4, by k-th of characteristic vector T(k)With the average value of i-th of classThe difference that is obtained after subtracting each other with i-th The matching threshold μ of classiCompare, if difference is less than or equal to matching threshold μi, then it represents that k-th of characteristic vector T(k)Belong to i-th Class, perform step 5.5；Otherwise, k-th of characteristic vector T is represented(k)I-th of class is not belonging to, performs step 5.6；
Step 5.5, judge whether k ＞ m set up, if so, then represent that test image matches completion with visual dictionary L, obtains Characteristic vector after to renewal is as insulation subcharacter；Otherwise, after k+1 being assigned into k, return to step 5.3；
Step 5.6, judge whether i ＞ K set up, if so, then represent k-th of characteristic vector T(k)It is not belonging to any one Class, and k-th of characteristic vector T is deleted from m characteristic vector(k)Afterwards, k+1 is assigned to k, performs step 5.3；Otherwise, by i+ After 1 is assigned to i, return to step 5.4；
Step 6, the ballot value V using any one pixel (x, y) in the characteristic area of formula (2) calculating test image (x,y)：
In formula (2), xhAnd yhRepresent that h-th of insulation subcharacter is corresponding in test image in the characteristic vector after renewal Pixel abscissa and ordinate, 1≤h≤H, H represent insulation subcharacter sum；σ represents the yardstick of characteristic area；
(x, y) is any point, by carrying out distance relation function V (x, y) calculating with characteristic point on insulator, is judged Whether it belongs to insulation subregion.
Step 7, judge whether V (x, y) is more than or equal to set threshold value, threshold value can take max { V (xh,yh),1≤h≤ H }, if being more than or equal to, then it represents that the region that pixel (x, y) belongs to where insulator, and retain pixel (x, y)；Otherwise, table Show pixel of the pixel (x, y) for the region where nonisulated son；
Step 8, repeat step 6- steps 7, so as to complete the detection of all pixels point in the characteristic area of test image, and Obtain the subregion that insulate；
Step 9, using GVF-Snake models contour detecting is carried out to the insulation subregion obtained by step 8, obtained absolutely Edge profile；
GVF-Snake models：Assuming that f (x, y) is image I (x, y) contour images, noteIt is f (x, y) ladder Field is spent, will in GVF-Snake modelsSpread to image border, so as to form diffusion gradient vector flow field GVF (x, y) =[u (x, y), v (x, y)], meet the minimum solution of following energy function：
Wherein μ is adjustment parameter, and u (x, y), v (x, y) are respectively the size of GVF both horizontally and vertically.
Under the principle that energy function minimizes, GVF (x, y) can pass through corresponding Euler equation solutions.
To ask above formula to obtain GVF (x, y), u, v are regarded as function on time t, i.e.,
According to formula (3), (4) grey iterative generation GVF (x, y), initial value typically takes the gradient of image.
Step 10, icing judge
The direct performance of step 10.1, icing on image is exactly that can increase the radial distance of insulator, is insulated obtaining After the contour images of substring, in order to judge icing situation, it is necessary to calculate the radial distance of insulator in image, directly from image In the insulator maximum radial distance measured refer to distance in image in units of single pixel, calculate in insulator contour most Big radial distance, and it is designated as d；
Step 10.2, judge | d-d ' | whether ＞ δ set up, if so, then represent that insulator has icing, and perform step 10.3；Otherwise, the non-icing of insulator is represented；δ avoids calculation error and set small threshold；When d ' is insulator non-icing Radial distance in the picture；
Step 10.3, foundation survey parameter, and the thickness D of icing on insulator is obtained using formula (3)：
D=0.5 (d-d') d "/d ' (5)
In formula (5), d " represents the actual radial distance of insulator.

Claims (1)

  1. A kind of 1. covering ice for insulator detection method based on GLOH descriptions and GVF-Snake models, it is characterized in that by following step It is rapid to carry out：
    Some width insulation subgraph in step 1, collection transmission line of electricity, and intermediate value filter is carried out to some width insulation subgraphs The pretreatment of ripple and histogram equalization, obtain some pretreated insulation subgraphs；
    Step 2, using Hessian-Affine Region detection algorithms it is pretreated to described some insulation subgraph carry out Detection, obtains corresponding characteristic area；Assuming that n characteristic point is included in any one characteristic area；
    Step 3, using GLOH description son the n characteristic point is described, obtain n characteristic vector；Remember j-th of feature to Measure as Tj, 1≤j≤n；
    Step 4, the visual dictionary using machine learning method training insulator：
    K step 4.1, random initializtion central point Represent i-th of central point, i=1,2 ..., K；
    Step 4.2, calculate j-th of characteristic vector TjTo i-th of central pointDistanceIt is special so as to obtain j-th Levy vector TjTo the distance { d of K central point1,d2,…,di,…,dK}；From the distance { d1,d2,…,di,…,dKIn choose Minimum value and by j-th of characteristic vector TjIt is referred in the class corresponding to minimum value；So as to which n characteristic vector be referred to accordingly Class in, and obtain K class；
    Step 4.3, the central point for calculating the K classAnd with original K central point It is compared, if all same, performs step 4.4；Otherwise, by the central point for K class being calculated It is assigned to the K central pointAfter being updated, return to step 4.2；
    Step 4.4, the central point for calculating the K classIn each class average value, obtain average value Vector Represent the average value of i-th of class；
    Step 4.5, using formula (1) obtain the visual dictionary L of insulator：
    <mrow> <mi>L</mi> <mo>=</mo> <mo>{</mo> <mrow> <mo>(</mo> <mover> <msub> <mi>T</mi> <mn>1</mn> </msub> <mo>&amp;OverBar;</mo> </mover> <mo>,</mo> <msub> <mi>&amp;mu;</mi> <mn>1</mn> </msub> <mo>)</mo> </mrow> <mo>,</mo> <mrow> <mo>(</mo> <mover> <msub> <mi>T</mi> <mn>2</mn> </msub> <mo>&amp;OverBar;</mo> </mover> <mo>,</mo> <msub> <mi>&amp;mu;</mi> <mn>2</mn> </msub> <mo>)</mo> </mrow> <mo>,</mo> <mo>...</mo> <mo>,</mo> <mrow> <mo>(</mo> <mover> <msub> <mi>T</mi> <mi>i</mi> </msub> <mo>&amp;OverBar;</mo> </mover> <mo>,</mo> <msub> <mi>&amp;mu;</mi> <mi>i</mi> </msub> <mo>)</mo> </mrow> <mo>,</mo> <mo>...</mo> <mo>,</mo> <mrow> <mo>(</mo> <mover> <msub> <mi>T</mi> <mi>K</mi> </msub> <mo>&amp;OverBar;</mo> </mover> <mo>,</mo> <msub> <mi>&amp;mu;</mi> <mi>K</mi> </msub> <mo>)</mo> </mrow> <mo>}</mo> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow>
    In formula (1), μiFor the matching threshold of i-th of class, matching threshold μiFor all characteristic vectors in i-th of class and i-th class Average valueBetween difference maximum norm；
    Step 5, by dictionary pattern matching, eliminate nonisulated subcharacter
    Step 5.1, the width insulation subgraph obtained on transmission line of electricity are carried out as test image, and according to step 1 to step 3 Processing, obtains m characteristic vector of the test image, is designated as { T(1),T(2),…,T(k),…,T(m)}；T(k)Represent the survey Attempt the characteristic vector of k-th of characteristic point in the characteristic area of picture, 1≤k≤m；
    Step 5.2, initialization k=1；
    Step 5.3, initialization i=1；
    Step 5.4, by k-th of characteristic vector T(k)With the average value of i-th of classOf the difference obtained after subtracting each other and i-th of class With threshold value μiCompare, if difference is less than or equal to matching threshold μi, then it represents that k-th of characteristic vector T(k)Belong to i-th of class, perform Step 5.5；Otherwise, k-th of characteristic vector T is represented(k)I-th of class is not belonging to, performs step 5.6；
    Step 5.5, judge whether k ＞ m set up, if so, then represent that test image matches completion with the visual dictionary L, obtains Characteristic vector after to renewal is as insulation subcharacter；Otherwise, after k+1 being assigned into k, return to step 5.3；
    Step 5.6, judge whether i ＞ K set up, if so, then represent k-th of characteristic vector T(k)Any one class is not belonging to, and K-th of characteristic vector T is deleted from m characteristic vector(k)Afterwards, k+1 is assigned to k, performs step 5.3；Otherwise, by i+1 assignment After i, return to step 5.4；
    Step 6, the ballot value V using any one pixel (x, y) in the characteristic area of formula (2) the calculating test image (x,y)：
    <mrow> <mi>V</mi> <mrow> <mo>(</mo> <mi>x</mi> <mo>,</mo> <mi>y</mi> <mo>)</mo> </mrow> <mo>=</mo> <munderover> <mo>&amp;Sigma;</mo> <mi>h</mi> <mi>H</mi> </munderover> <mfrac> <mn>1</mn> <mrow> <mn>2</mn> <msup> <mi>&amp;pi;&amp;sigma;</mi> <mn>2</mn> </msup> </mrow> </mfrac> <mi>exp</mi> <mrow> <mo>(</mo> <mo>-</mo> <mfrac> <mrow> <msup> <mrow> <mo>(</mo> <mi>x</mi> <mo>-</mo> <msub> <mi>x</mi> <mi>h</mi> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> <mo>+</mo> <msup> <mrow> <mo>(</mo> <mi>y</mi> <mo>-</mo> <msub> <mi>y</mi> <mi>h</mi> </msub> <mo>)</mo> </mrow> <mn>2</mn> </msup> </mrow> <mrow> <mn>2</mn> <msup> <mi>&amp;sigma;</mi> <mn>2</mn> </msup> </mrow> </mfrac> <mo>)</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>2</mn> <mo>)</mo> </mrow> </mrow>
    In formula (2), xhAnd yhRepresent that h-th of insulation subcharacter is corresponding in the test image in the characteristic vector after renewal Pixel abscissa and ordinate, 1≤h≤H, H represent insulation subcharacter sum；σ represents the yardstick of characteristic area；
    Step 7, judge whether V (x, y) is more than or equal to set threshold value, if being more than or equal to, then it represents that pixel (x, y) belongs to Region where insulator, and retain pixel (x, y)；Otherwise, it is the region where nonisulated son to represent pixel (x, y) Pixel；
    Step 8, repeat step 6- steps 7, so as to complete the detection of all pixels point in the characteristic area of the test image, and Obtain the subregion that insulate；
    Step 9, using GVF-Snake models to it is described insulation subregion carry out contour detecting, obtain insulator contour；
    Step 10, icing judge
    Step 10.1, maximum radial distance in the insulator contour is calculated, and be designated as d；
    Step 10.2, judge | d-d ' | whether ＞ δ set up, if so, then represent that insulator has icing, and perform step 10.3； Otherwise, the non-icing of insulator is represented；Threshold value set by δ；Radial distance when d ' is insulator non-icing in the picture；
    Step 10.3, using formula (3) obtain the thickness D of icing on insulator：
    D=0.5 (d-d') d "/d ' (3)
    In formula (3), d " represents the actual radial distance of insulator.
CN201710568690.4A 2017-07-13 2017-07-13 Covering ice for insulator detection method based on GLOH description son and GVF-Snake model Active CN107358259B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710568690.4A CN107358259B (en) 2017-07-13 2017-07-13 Covering ice for insulator detection method based on GLOH description son and GVF-Snake model

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710568690.4A CN107358259B (en) 2017-07-13 2017-07-13 Covering ice for insulator detection method based on GLOH description son and GVF-Snake model

Publications (2)

Publication Number Publication Date
CN107358259A true CN107358259A (en) 2017-11-17
CN107358259B CN107358259B (en) 2019-09-06

Family

ID=60293532

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710568690.4A Active CN107358259B (en) 2017-07-13 2017-07-13 Covering ice for insulator detection method based on GLOH description son and GVF-Snake model

Country Status (1)

Country Link
CN (1) CN107358259B (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108846398A (en) * 2018-06-26 2018-11-20 北京天工矩阵信息技术有限公司 CAD knows drawing method and device
CN111079852A (en) * 2019-12-27 2020-04-28 国网福建省电力有限公司宁德供电公司 Method for detecting icing thickness of insulator of power transmission line
CN111964865A (en) * 2020-08-12 2020-11-20 中国商用飞机有限责任公司 Method for calculating ice roughness of sand paper for airplane icing wind tunnel test
CN112150495A (en) * 2020-09-24 2020-12-29 西安工程大学 Insulator ice flashover risk prediction method based on deduction rule
CN112988079A (en) * 2021-05-07 2021-06-18 成都奥伦达科技有限公司 Management method and system for ultra-mass point clouds
US11093744B1 (en) * 2017-09-21 2021-08-17 Electric Power Science & Research Institute Of State Grid Tianjin Electric Power Company Method and device for determining types of ice-and-snow cover
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
CN113850977A (en) * 2021-09-17 2021-12-28 国网江苏省电力有限公司盐城供电分公司 Power transmission line icing detection early warning method
CN116191680A (en) * 2023-04-25 2023-05-30 丰德信电力发展股份有限公司 Monitoring management system applied to photovoltaic power generation

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102346015A (en) * 2011-06-24 2012-02-08 西安工程大学 Method for measuring icing thickness of insulator of power transmission line based on video difference analysis
CN102679935A (en) * 2012-03-02 2012-09-19 凯里供电局 System and method for calculating icing thickness of power transmission line
CN103196380A (en) * 2013-03-20 2013-07-10 北京国网富达科技发展有限责任公司 Electric transmission line image acquisition device and electric transmission line icing thickness detection system
CN103400140A (en) * 2013-07-24 2013-11-20 华南理工大学 Method for processing ice coated on insulator on basis of improved image method
CN103852018A (en) * 2012-12-03 2014-06-11 西安元朔科技有限公司 Electric transmission line icing thickness measuring algorithm based on image processing

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102346015A (en) * 2011-06-24 2012-02-08 西安工程大学 Method for measuring icing thickness of insulator of power transmission line based on video difference analysis
CN102679935A (en) * 2012-03-02 2012-09-19 凯里供电局 System and method for calculating icing thickness of power transmission line
CN103852018A (en) * 2012-12-03 2014-06-11 西安元朔科技有限公司 Electric transmission line icing thickness measuring algorithm based on image processing
CN103196380A (en) * 2013-03-20 2013-07-10 北京国网富达科技发展有限责任公司 Electric transmission line image acquisition device and electric transmission line icing thickness detection system
CN103400140A (en) * 2013-07-24 2013-11-20 华南理工大学 Method for processing ice coated on insulator on basis of improved image method

Non-Patent Citations (8)

* Cited by examiner, † Cited by third party
Title
LIN YANG等: "Recognition of Natural Ice Types on In-service Glass Insulators Based on Texture Feature Descriptor", 《IEEE TRANSACTIONS ON DIELECTRICS AND ELECTRICAL INSULATION》 *
张成等: "基于图像处理技术的绝缘子覆冰自动识别", 《华东电力》 *
张烨等: "输电线路绝缘子覆冰厚度图像识别算法", 《电力系统自动化》 *
朱述龙等: "运用GVF Snake算法提取水域的不规则边界", 《遥感学报》 *
杨浩等: "基于三维重建的绝缘子覆冰图像监测", 《电子自动化设备》 *
王小朋等: "应用图像边缘监测方法在线监测输电线路覆冰厚度研究", 《高压电器》 *
郜晓晶等: "基于GLOH算子和局部特征融合的人脸识别", 《计算机应用与软件》 *
韩海梅等: "基于K-means 和GVF Snake 模型的纤维图像分割", 《东华大学学报（自然科学版）》 *

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11093744B1 (en) * 2017-09-21 2021-08-17 Electric Power Science & Research Institute Of State Grid Tianjin Electric Power Company Method and device for determining types of ice-and-snow cover
CN108846398A (en) * 2018-06-26 2018-11-20 北京天工矩阵信息技术有限公司 CAD knows drawing method and device
CN111079852A (en) * 2019-12-27 2020-04-28 国网福建省电力有限公司宁德供电公司 Method for detecting icing thickness of insulator of power transmission line
CN111079852B (en) * 2019-12-27 2023-04-11 国网福建省电力有限公司宁德供电公司 Method for detecting icing thickness of insulator of power transmission line
CN111964865A (en) * 2020-08-12 2020-11-20 中国商用飞机有限责任公司 Method for calculating ice roughness of sand paper for airplane icing wind tunnel test
CN111964865B (en) * 2020-08-12 2021-04-27 中国商用飞机有限责任公司 Method for calculating ice roughness of sand paper for airplane icing wind tunnel test
CN112150495A (en) * 2020-09-24 2020-12-29 西安工程大学 Insulator ice flashover risk prediction method based on deduction rule
CN112150495B (en) * 2020-09-24 2023-03-14 西安工程大学 Insulator ice flashover risk prediction method based on deduction rule
CN112988079A (en) * 2021-05-07 2021-06-18 成都奥伦达科技有限公司 Management method and system for ultra-mass point clouds
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
CN113850977A (en) * 2021-09-17 2021-12-28 国网江苏省电力有限公司盐城供电分公司 Power transmission line icing detection early warning method
CN116191680A (en) * 2023-04-25 2023-05-30 丰德信电力发展股份有限公司 Monitoring management system applied to photovoltaic power generation
CN116191680B (en) * 2023-04-25 2023-07-18 丰德信电力发展股份有限公司 Monitoring management system applied to photovoltaic power generation

Also Published As

Publication number Publication date
CN107358259B (en) 2019-09-06

Similar Documents

Publication Publication Date Title
CN107358259B (en) Covering ice for insulator detection method based on GLOH description son and GVF-Snake model
CN109118479B (en) Capsule network-based insulator defect identification and positioning device and method
CN109389180A (en) A power equipment image-recognizing method and inspection robot based on deep learning
Wang et al. Detection of insulator defects with improved ResNeSt and region proposal network
Zhang et al. InsuDet: A fault detection method for insulators of overhead transmission lines using convolutional neural networks
CN103279765B (en) Steel wire rope surface damage detection method based on images match
CN112200178B (en) Transformer substation insulator infrared image detection method based on artificial intelligence
CN108564565A (en) A kind of power equipment infrared image multi-target orientation method based on deep learning
CN109359697A (en) Graph image recognition methods and inspection system used in a kind of power equipment inspection
CN102519846B (en) Hyperspectrum-based composite insulator hydrophobicity detection method
CN111044570A (en) Defect identification and early warning method and device for power equipment and computer equipment
CN106557740B (en) The recognition methods of oil depot target in a kind of remote sensing images
CN104281838A (en) Lug piece breakage detection method for high-speed rail overhead line system supporting device based on HOG features and two-dimensional Gabor wavelet transformation
CN107833211A (en) Zero resistance insulator automatic testing method and device based on infrared image
Wang et al. Railway insulator detection based on adaptive cascaded convolutional neural network
Li et al. PEDNet: A lightweight detection network of power equipment in infrared image based on YOLOv4-Tiny
Zhang et al. Detection and condition assessment of icicle bridging for suspension glass insulator by image analysis
Lu et al. Thermal Fault Diagnosis of Electrical Equipment in Substations Based on Image Fusion.
Qiu et al. A lightweight yolov4-edam model for accurate and real-time detection of foreign objects suspended on power lines
CN112419243B (en) Power distribution room equipment fault identification method based on infrared image analysis
CN106355187A (en) Application of visual information to electrical equipment monitoring
CN109543546A (en) The gait age estimation method returned based on the distribution of depth sequence
Niu et al. Electrical equipment identification method with synthetic data using edge-oriented generative adversarial network
CN110033082B (en) Method for identifying deep learning model in AI (Artificial intelligence) equipment
Wang et al. Substation Equipment Defect Detection based on Temporal-spatial Similarity Calculation

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