CN106127756B - A kind of insulator recognition detection method based on multicharacteristic information integration technology - Google Patents

A kind of insulator recognition detection method based on multicharacteristic information integration technology Download PDF

Info

Publication number
CN106127756B
CN106127756B CN201610452634.XA CN201610452634A CN106127756B CN 106127756 B CN106127756 B CN 106127756B CN 201610452634 A CN201610452634 A CN 201610452634A CN 106127756 B CN106127756 B CN 106127756B
Authority
CN
China
Prior art keywords
image
feature
formula
insulator
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.)
Active
Application number
CN201610452634.XA
Other languages
Chinese (zh)
Other versions
CN106127756A (en
Inventor
黄新波
邢晓强
朱永灿
纪超
张烨
李菊清
刘新慧
张慧莹
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.)
XI'AN JIN POWER ELECTRICAL Co.,Ltd.
Original Assignee
Xian Polytechnic 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 Xian Polytechnic University filed Critical Xian Polytechnic University
Priority to CN201610452634.XA priority Critical patent/CN106127756B/en
Publication of CN106127756A publication Critical patent/CN106127756A/en
Application granted granted Critical
Publication of CN106127756B publication Critical patent/CN106127756B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G06V10/40—Extraction of image or video features
    • G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
    • G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20212—Image combination
    • G06T2207/20221—Image fusion; Image merging
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)

Abstract

The insulator recognition detection method based on multicharacteristic information integration technology that the invention discloses a kind of, it is specifically implemented according to the following steps: step 1, utilizing the picture signal of insulator on the transmission line of electricity for installing the focusing video camera acquisition with holder on steel tower at the scene or iron；Step 2, image is collected to step 1 to pre-process；Step 3, the texture feature vector of insulator is extracted from pretreated image；Step 4, insulator color feature vector to be identified is extracted, unequal interval quantization is carried out to the HSV color space of insulation subgraph collected in step 1；Step 5, insulator shape feature is extracted using based on Hu square algorithm in region description method；Step 6,3 kinds of features that insulator is extracted in step 3, step 4, step 5 are merged, method of the invention can simple, reliable, fast and automatically detect the operating status of insulator, thus electric power system fault caused by preventing because of Insulator detection.

Description

A kind of insulator recognition detection method based on multicharacteristic information integration technology
Technical field
The invention belongs to electric system to monitor field on-line, be related to a kind of insulator based on multicharacteristic information integration technology Recognition detection method.
Background technique
With the proposition that global energy internet theory is conceived, using extra-high voltage grid as bulk transmission grid, with powerful intelligence Power grid is to rely on, and real-time monitoring, security maintenance and the operational management for reinforcing transmission line of electricity are the most important things.And in high voltage transmission line Lu Zhong, defective insulation are that the main reason for accident occur, become electric power system fault rate because Insulator detection causes accident at present First.Insulator is run outdoors for a long time, and the factors such as various dunghill interference, easily cause interior insulator to split in atmosphere Line, surface fracture, dielectric strength reduce and the failures such as pollution flashover.When in insulator chain there are when low value or zero resistance insulator, it is powerful Lightning current and power frequency continued flow flowed through from the porcelain body on zero resistance insulator head, easily cause zero resistance insulator and cross thermal spalling.If The operating status of grasp insulator early, it will the failure of many electric system is reduced or avoided.Traditional inspection insulator The method of operating status is periodical power failure or electrification artificial detection, these operations not only need to step on bar to be detected piecewise, but also high-altitude Operation big, risk height affected by environment, working efficiency are lower.Therefore, how simple, reliable, quickly automatic without stepping on bar The operating status of monitoring insulator is to reduce human input, is eliminated safe hidden trouble, it is ensured that safe operation of power system is to be explored Important technology problem.
Summary of the invention
The insulator recognition detection method based on multicharacteristic information integration technology that the object of the present invention is to provide a kind of, can Simply, reliably, fast and automatically detect insulator operating status, thus prevention the electric system caused by Insulator detection therefore Barrier.
The technical scheme adopted by the invention is that a kind of insulator recognition detection side based on multicharacteristic information integration technology Method is specifically implemented according to the following steps:
Step 1, using on the transmission line of electricity for installing the focusing video camera acquisition with holder on steel tower at the scene or iron The picture signal of insulator；
Step 2, it collects image to step 1 to pre-process, specific step are as follows: utilize optimal entropic threshold method (OET) Image segmentation is carried out, by analyzing the entropy of image grey level histogram, finds optimal threshold；
Step 3, the texture feature vector that insulator is extracted in pretreated image was carried out from step 2；
Step 4, insulator color feature vector to be identified is extracted, to the HSV color of insulation subgraph collected in step 1 Color space carries out unequal interval quantization；
Step 5, insulator shape eigenvectors to be identified are extracted, the shape feature of insulator is unrelated with external environment, It is that the external most stable information of object and image are most intuitive, most directly visualizes, using based on Hu in region description method Square algorithm extracts insulator shape feature；
Step 6,3 kinds of features that insulator is extracted in step 3, step 4, step 5 are merged, three kinds of single features When fusion, each characteristic quantity is normalized before similarity measurement, adjust and scheme in target image to be identified and database As the weight between three features.Its weight is determined with analytic hierarchy process (AHP), calculates consistency ratio CR, verifies judgment matrix three Whether the weight of Fusion Features meets the requirements.
The features of the present invention also characterized in that
Step 2 is specifically implemented according to the following steps:
Step 2.1, the tonal range of acquired image in step 1 is denoted as { 0,1,2 ... ... L-1 }；
Step 2.2, if the region that pixel of the gray level lower than t is constituted is target area A, the then table of target area A entropy Up to formula are as follows:
P in formulaiIndicate that gray level is the probability that i occurs, i=0,1,2 ... ... t；PtIndicate that gray level is the general of t appearance Rate；
Step 2.3, if the region that pixel of the gray level higher than t is constituted is background area B, the then table of background area B entropy Up to formula are as follows:
I=t+1, t+2, t+3 ... .L-1；
P in formulaiIndicate that gray level is the probability that i occurs, i=0,1,2 ... ... t；PtIndicate that gray level is the general of t appearance Rate；
Step 2.4, max-thresholds are calculated,
It is defined according to entropy function,
In formulaI=0 therein, 1,2 ... ... t；I=0,1 therein, 2 ... ... L-1；
Then, when entropy function C (t) is maximum,
Calculate t=argmax { C (t) } (2-4)
Mean that the max-thresholds of gray scale t.
Step 3 is specifically implemented according to the following steps:
Step 3.1, using the statistical method in texture characteristic extracting method to 3 color spaces of former RGB color image Mean value, variance, the degree of bias, kurtosis, energy, entropy parameter of totally 6 color space features as textural characteristics are calculated separately,
Specific formula for calculation is as follows:
(1) mean value
(2) variance
(3) degree of bias
(4) kurtosis
(5) energy
(6) entropy
Wherein g (i) is the gray value of i-gray level, pgIt (i) is the probability of i-th of gray value；
Former RGB color image is just divided into 18 feature vectors by this 6 features in total.The single features different to image into Normalization inside feature is carried out when row feature extraction, is carried out this 18 feature vectors inside feature using Gaussian normalization method [- 1,1] are normalized to, other element value is reduced and the distribution of the element value after normalization is had an impact.Feature vector is returned One changes FiCalculation formula it is as follows:
Wherein μiIndicate the mean value of target image characteristics vector to be identified, σiIndicate target image characteristics vector to be identified Standard deviation, fiFor 18 one dimensional histograms initial characteristic values and fi∈[f1,f2,f3,.......f18]；I=1,2,3 ... ... 18；
Step 3.2, by the texture feature vector extracted in step 3.1 and the line in established sample database in advance Reason feature vector compares matching, calculates its similarity, method particularly includes:
The matched similarity of textural characteristics in target image to be identified and characteristic image library is calculated using chessboard distance formula Distance, the normalization characteristic vector of two images are that 1 18 dimensional vector may be expressed as:
Fxi[Fx1,Fx2,Fx3,......Fx18]；
Fyi[Fy1,Fy2,Fy3,......Fy18]；
Then, it is calculated according to chessboard distance formula:
Wherein S1Indicate the matched similarity distance of textural characteristics in target image to be identified and characteristic image library, FxiAnd Fyi Indicate the normalization characteristic vector of two images, S1Value is bigger, and the similarity of two images is higher.
Step 4 is specifically implemented according to the following steps:
Step 4.1, the space tone H is divided into 8 parts, saturation degree S and the space brightness V are divided into 3 parts, the tone H of image Range is [0,360 °], and the range of saturation degree S and brightness V are [0,1]；
Step 4.2, color quantization is carried out according to the different range of H, S, V, specific as follows shown:
3 color components after quantization are synthesized one-dimensional characteristic vector:
I=HQsQv+SQv+V (4-4)
I=HQsQvQ in+SQ formulas=3 be the quantization series of component S；Qv=3 be component V
Quantization series when 3-4 may be expressed as:
I=9H+3S+V (4-5)
The subspace that HSV space after quantization resolves into obtains 72 one dimensional histograms；
Step 4.3, the feature vector Gaussian normalization of the color histogram after quantifying in step 4.2 is subtracted to [- 1,1] Few other element value has an impact the distribution of the element value after normalization, the normalized F of feature vectorlCalculation formula is such as Under:
Wherein μlIndicate the mean value of target image characteristics vector to be identified, σlIndicate target image characteristics vector to be identified Standard deviation, flFor 72 one dimensional histograms initial characteristic values and fl∈[f1,f2,f3,.......f72]；L=1,2,3 ... ... 72,
Step 4.4, target image to be identified is calculated using chessboard distance formula to match with color characteristic in characteristic image library Similarity distance.The normalization characteristic of two images may be expressed as: to for 2 72 dimensional vectors
Fpl[Fp1,Fp2,Fp3,......Fp72]；Fql[Fq1,Fq2,Fq3,......Fq72]；
Wherein chessboard distance calculation formula:
Wherein S2Indicate the matched similarity distance of textural characteristics in target image to be identified and characteristic image library, FplAnd Fql Indicate the normalization characteristic vector of two images, S2Value is maximum, and the similarity of two images is higher.
Step 5 is specifically implemented according to the following steps:
Step 5.1, the p+q rank square of acquired image f (x, y) is defined are as follows:
Then, the central moment of p+q rank square are as follows:
In formulaIndicate the center of gravity of image-region；
Step 5.2, the central moment in (5-2) formula can be returned to obtain image property unrelated with scaling itself One changes, and the central moment after normalization indicates are as follows:
In formulaP+q=2,3,4 ... ... normalized centers are away from the translation to object, scaling
It is remained unchanged with rotation；
Step 5.3, image f (x, y) second order and third central moment are calculated using using formula (5-2), obtained
Wherein second-order moment around mean μ02And μ20Respectively indicate the inertia around the vertically and horizontally axis by gray scale mass center Square, third central moment μ03And μ30Asymmetric degree of the mensurable institute's analyzed area of amplitude to vertically and horizontally axis；
7 invariant moments are constructed, translation, scaling and invariable rotary can be kept under conditions of consecutive image.Described 7 are not Bending moment is defined respectively as:
Central moment expression formula after second order normalized:
Central moment expression formula after the normalization of three ranks:
φ1=η20+η02 (5-8)
φ2=(η20+η02)2+4η2 11 (5-9)
φ3=(η30-3η12)2+(3η21-η03)2 (5-10)
φ4=(η30+η12)2+(η21+η03)2 (5-11)
φ5=(η30-3η12)(η30+η12)[(η20+η12)2-3(η21+η03)2]+(3η21+η03)(η21+η03)[3(η30+η12)2- (η21+η03)] (5-12)
φ6=(η20-η02)[(η30+η12)2-(η21+η03)2]+4η11(η30+η12)(η21+η03) (5-13)
φ7=(3 η21-η03)(η30+η12)[(η30+η12)2-3(η21+η03)2]+(3η21-η03)(η21+η03)[3(η30+η12)2- (η21+η03)2] (5-14)
φ therein1~φ7For the calculation formula of 7 invariant moments, each η value indicates the center after second order, the normalization of three ranks Square；
Step 5.4,7 shape spies of insulator to be identified are extracted by Hu square algorithm (5-8)~(5-14) calculation formula Vector is levied, then to each feature vector Gaussian normalization to [- 1,1].By shape in target image to be identified and characteristic image library Feature normalization is that 27 dimensional vectors may be expressed as:
Fsi[Fs1,Fs2,Fs3,......Fs7]；Fti[Ft1,Ft2,Ft3,......Ft7]；
Equally, using chessboard distance formula calculate target image to be identified in characteristic image library shape feature it is matched Similarity distance S3:
S3Value is maximum, and the similarity of two images is higher；
Step 5.5, the color, shape and textural characteristics of the insulator extracted are subjected to Fusion Features.If three kinds of features The similarity (distance) of Vector Fusion be D, three kinds of single features vectors (feature of color, shape and texture) similarity (away from From) S1,S2,S3Corresponding weight is respectively w1,w2,w3, then the similarity (distance) of three kinds of Fusion Features vectors be,
D=s1w1+s2w2+s3w3 (5-16)
∑ w in formulai=1；I=1,2,3.
Step 6 is specifically implemented according to the following steps:
Step 6.1, if texture, color and the shape eigenvectors of a target image i to be identified are [Oi1,Oi2,Oi3], N number of image in recognition target image i and database is treated by chessboard distance and carries out similarity calculation, is calculated distance value and is remembered respectively For Di1,Di2,......DiN；DiNIt is also three characteristic distances of any one image in target image to be identified and data picture library:
DiN=[d1,iN,d2,iN,d3,iN] (6-1)
Step 6.2, the mean μ of 3N distance value is calculatedDAnd standard deviation sigmaD, specific formula for calculation are as follows:
F thereina、fb、fcRespectively represent texture, color and shape eigenvectors；
Step 6.3, N width image (N takes any positive integer) similarity distance D in recognition target image and feature database is treatedi1, Di2,......DiNIt is normalized, keeps three kinds of different characteristic vectors roughly the same on similarity calculation, DiNIt is worth normalizing [0,1] section, expression formula are largely fallen in after change are as follows:
Step 6.4, the weight of three Fusion Features is determined using analytic hierarchy process (AHP).With 1-9 scaling law to Fusion Features after Three feature description indexes importance of insulation subgraph are judged, export three using the judgment matrix A that policymaker provides The weight s of featurei, then the consistency met with consistency ratio CR test and judge matrix, three features are exported to verify Weighted value accuracy；
Step 6.5, consistency ratio CR is calculated, and whether is met unanimously using consistency ratio CR test and judge matrix Property, and then determine the weighted value of Fusion Features.
Step 6.4 is specifically implemented according to the following steps:
Step 6.4.1, according to 1-9 scaling law to three feature weight s1,s2,s3Importance proportion quotiety two-by-two carries out Assignment forms judgment matrix:
A1=(apq)n×nWherein apqProportion quotiety for pth feature with respect to q feature importance,
Step 6.4.2 takes width insulation subgraph, in hierarchy Model, only takes first layer to this insulation subgraph It is secondary, and this level can be divided into three classifications, i.e. texture (C), color (S) and shape (T), then corresponding judgment matrix indicates Are as follows:
It is thereinRespectively color and shape feature proportion quotiety, texture and color characteristic ratio Scale, texture and shape feature proportion quotiety.Generally according to the interpretation of result of three features of user's priori knowledge and extraction, shape Other two aspect ratios of aspect ratio are great,
When starting, one initial value of scale value is provided, that is, is setJudge one Cause sex rate CR.
Step 6.4.3, to A1In every row element multiplication and open cube and obtain vector Yi=(y1,y2,y3), calculation formula is such as Under:
To YiIt is normalized to obtain normalized weight vector si=(s1,s2,s3)
Step 6.4.4, to A1In every column element sum to obtain Zq=(Z1,Z2,Z3)
Step 6.4.5 calculates judgment matrix A1Maximum value characteristic value
Step 6.5 is specifically implemented according to the following steps:
Step 6.5.1 calculates judgment matrix A1Coincident indicator CI,
Wherein n is judgment matrix A1Order, the Maximum characteristic root λ of the smaller judgment matrix of CImaxMore meet completely the same Property, CI is bigger, and the degree for illustrating judgment matrix deviation crash consistency is bigger.
Judgment matrix A is calculated according to 6-101Other two characteristic value is denoted as respectively: λ2、λ3These three maximum eigenvalue are sought Average valueCalculation formula is as follows:
Step 6.5.2 calculates Aver-age Random Consistency Index RI according to formula (9-10)
Step 6.5.3 calculates consistency ratio CR using calculated CI and RI:
The judgment matrix A as CR < 0.11With satisfactory consistency, i.e. weight vectors siMeet matrix equation: A1si=λmaxsi There is maximum eigenvalue λmax, so judgment matrix A1The weighted value of three Fusion Features meets the requirements, the power used when calculating at this time The weighted value used when weight values are image co-registration in this method exports fused insulator identification image；
As CR >=0.1, judgment matrix A1Without satisfactory consistency, then goes to step 6.4 and re-start assignment decision It is right that person needs againThree scales carry out assignment according to 1-9 scaling law, with assignment when preceding primary calculating It is adjusted after comparison, constructs new judgment matrix A1, then calculate separately weight vector si, judgment matrix maximum feature λmaxAnd one Cause property index CI, finally calculates consistency ratio CR, until CR < 0.1 is set up.
The invention has the advantages that equipment used in the implementation present invention is less compared with existing system, structure letter Single, low in cost, this method can make full use of multiple features fusion thought in image processing techniques, by real-time interception in video flowing Insulator target image carries out three Fusion Features match cognizations with insulator feature picture library, can real time on-line monitoring insulator Operating status.Data are remotely handled convenient for monitoring center and global prison is carried out to the operating condition of the insulator of electric power networks Control, is advantageously implemented the security monitoring of automation.
Detailed description of the invention
Fig. 1 is a kind of flow chart of insulator recognition detection method based on multicharacteristic information integration technology of the invention.
Specific embodiment
The following describes the present invention in detail with reference to the accompanying drawings and specific embodiments.
The insulator recognition detection method based on multicharacteristic information integration technology that the present invention provides a kind of, as shown in Figure 1, It is specifically implemented according to the following steps:
Step 1, using on the transmission line of electricity for installing the focusing video camera acquisition with holder on steel tower at the scene or iron The picture signal of insulator；
Step 2, it collects image to step 1 to pre-process, specific step are as follows: utilize optimal entropic threshold method (OET) Image segmentation is carried out, by the entropy of analysis image grey level histogram, finds optimal threshold, specific step is:
Step 2.1, the tonal range of acquired image in step 1 is denoted as { 0,1,2 ... ... L-1 }.
Step 2.2, if the region that pixel of the gray level lower than t is constituted is target area A, the then table of target area A entropy Up to formula are as follows:
P in formulaiIndicate that gray level is the probability that i occurs, i=0,1,2 ... ... t；PtIndicate that gray level is the general of t appearance Rate.
Step 2.3, if the region that pixel of the gray level higher than t is constituted is background area B, the then table of background area B entropy Up to formula are as follows:
I=t+1, t+2, t+3 ... .L-1；
P in formulaiIndicate that gray level is the probability that i occurs, i=0,1,2 ... ... t；PtIndicate that gray level is the general of t appearance Rate.
Step 2.4, max-thresholds are calculated,
It is defined according to entropy function,
In formulaI=0 therein, 1,2 ... ... t；I=0,1 therein, 2 ... ... L-1；
Then, when entropy function C (t) is maximum,
Calculate t=arg max { C (t) } (2-4)
Mean that the max-thresholds of gray scale t.
Step 3, the texture feature vector that insulator is extracted in pretreated image was carried out from step 2, it is specific to walk It is rapid as follows:
Step 3.1, using the statistical method in texture characteristic extracting method to 3 color spaces of former RGB color image Calculate separately mean value, variance, the degree of bias, kurtosis, energy, entropy parameter of totally 6 color space features as textural characteristics.
Specific formula for calculation is as follows:
(1) mean value
(2) variance
(3) degree of bias
(4) kurtosis
(5) energy
(6) entropy
Wherein g (i) is the gray value of i-gray level, pgIt (i) is the probability of i-th of gray value.
Former RGB color image is just divided into 18 feature vectors by this 6 features in total.The single features different to image into Normalization inside feature is carried out when row feature extraction, is carried out this 18 feature vectors inside feature using Gaussian normalization method [- 1,1] are normalized to, other element value is reduced and the distribution of the element value after normalization is had an impact.Feature vector is returned One changes FiCalculation formula it is as follows:
Wherein μiIndicate the mean value of target image characteristics vector to be identified, σiIndicate target image characteristics vector to be identified Standard deviation, fiFor 18 one dimensional histograms initial characteristic values and fi∈[f1,f2,f3,.......f18]；I=1,2,3 ... ... 18.
Step 3.2, by the texture feature vector extracted in step 3.1 and the line in established sample database in advance Reason feature vector compares matching, calculates its similarity, method particularly includes:
The matched similarity of textural characteristics in target image to be identified and characteristic image library is calculated using chessboard distance formula Distance, the normalization characteristic vector of two images are that 1 18 dimensional vector may be expressed as:
Fxi[Fx1,Fx2,Fx3,......Fx18]；
Fyi[Fy1,Fy2,Fy3,......Fy18]；
Then, it is calculated according to chessboard distance formula:
Wherein S1Indicate the matched similarity distance of textural characteristics in target image to be identified and characteristic image library, FxiAnd Fyi Indicate the normalization characteristic vector of two images, S1Value is bigger, and the similarity of two images is higher.
Step 4, insulator color feature vector to be identified is extracted, to the HSV color of insulation subgraph collected in step 1 Color space carries out unequal interval quantization, is specifically implemented according to the following steps:
Step 4.1, the space tone H is divided into 8 parts, saturation degree S and the space brightness V are divided into 3 parts, the tone H of image Range is [0,360 °], and the range of saturation degree S and brightness V are [0,1].
Step 4.2, color quantization is carried out according to the different range of H, S, V, specific as follows shown:
3 color components after quantization are synthesized one-dimensional characteristic vector:
I=HQsQv+SQv+V (4-4)
I=HQsQvQ in+SQ formulas=3 be the quantization series of component S；Qv=3 be component V quantization series when 3-4 can table It is shown as:
I=9H+3S+V (4-5)
The subspace that HSV space after quantization resolves into obtains 72 one dimensional histograms.
Step 4.3, the feature vector Gaussian normalization of the color histogram after quantifying in step 4.2 is subtracted to [- 1,1] Few other element value has an impact the distribution of the element value after normalization, the normalized F of feature vectorlCalculation formula is such as Under:
Wherein μlIndicate the mean value of target image characteristics vector to be identified, σlIndicate target image characteristics vector to be identified Standard deviation, flFor 72 one dimensional histograms initial characteristic values and fl∈[f1,f2,f3,.......f72]；L=1,2,3 ... ... 72.
Step 4.4, target image to be identified is calculated using chessboard distance formula to match with color characteristic in characteristic image library Similarity distance.The normalization characteristic of two images may be expressed as: to for 2 72 dimensional vectors
Fpl[Fp1,Fp2,Fp3,......Fp72]；Fql[Fq1,Fq2,Fq3,......Fq72]；
Wherein chessboard distance calculation formula:
Wherein S2Indicate the matched similarity distance of textural characteristics in target image to be identified and characteristic image library, FplAnd Fql Indicate the normalization characteristic vector of two images, S2Value is maximum, and the similarity of two images is higher.
Step 5, insulator shape eigenvectors to be identified are extracted, the shape feature of insulator is unrelated with external environment, It is that the external most stable information of object and image are most intuitive, most directly visualizes, using based on Hu in region description method Square algorithm extracts insulator shape feature, is specifically implemented according to the following steps:
Step 5.1, the p+q rank square of acquired image f (x, y) is defined are as follows:
Then, the central moment of p+q rank square are as follows:
In formulaIndicate the center of gravity of image-region.
Step 5.2, the central moment in (5-2) formula can be returned to obtain image property unrelated with scaling itself One changes, and the central moment after normalization indicates are as follows:
In formulaP+q=2,3,4 ... ... normalized centers are away from translation, scaling and the rotation to object Remain unchanged.
Step 5.3, image f (x, y) second order and third central moment are calculated using using formula (5-2), obtained,
Wherein second-order moment around mean μ02And μ20Respectively indicate the inertia around the vertically and horizontally axis by gray scale mass center Square, third central moment μ03And μ30Asymmetric degree of the mensurable institute's analyzed area of amplitude to vertically and horizontally axis.
7 invariant moments are constructed, translation, scaling and invariable rotary can be kept under conditions of consecutive image.Described 7 are not Bending moment is defined respectively as:
Central moment expression formula after second order normalized:
Central moment expression formula after the normalization of three ranks:
φ1=η20+η02 (5-8)
φ2=(η20+η02)2+4η2 11 (5-9)
φ3=(η30-3η12)2+(3η21-η03)2 (5-10)
φ4=(η30+η12)2+(η21+η03)2 (5-11)
φ5=(η30-3η12)(η30+η12)[(η20+η12)2-3(η21+η03)2]+(3η21+η03)(η21+η03)[3(η30+η12)2- (η21+η03)] (5-12)
φ6=(η20-η02)[(η30+η12)2-(η21+η03)2]+4η11(η30+η12)(η21+η03) (5-13)
φ7=(3 η21-η03)(η30+η12)[(η30+η12)2-3(η21+η03)2]+(3η21-η03)(η21+η03)[3(η30+η12)2- (η21+η03)2] (5-14)
φ therein1~φ7For the calculation formula of 7 invariant moments, each η value indicates the center after second order, the normalization of three ranks Square.
Step 5.4,7 shape spies of insulator to be identified are extracted by Hu square algorithm (5-8)~(5-14) calculation formula Vector is levied, then to each feature vector Gaussian normalization to [- 1,1].By shape in target image to be identified and characteristic image library Feature normalization is that 27 dimensional vectors may be expressed as:
Fsi[Fs1,Fs2,Fs3,......Fs7]；Fti[Ft1,Ft2,Ft3,......Ft7]；
Equally, using chessboard distance formula calculate target image to be identified in characteristic image library shape feature it is matched Similarity distance S3:
S3Value is maximum, and the similarity of two images is higher.
Step 5.5, the color, shape and textural characteristics of the insulator extracted are subjected to Fusion Features.If three kinds of features The similarity (distance) of Vector Fusion be D, three kinds of single features vectors (feature of color, shape and texture) similarity (away from From) S1,S2,S3Corresponding weight is respectively w1,w2,w3, then the similarity (distance) of three kinds of Fusion Features vectors be,
D=s1w1+s2w2+s3w3 (5-16)
∑ w in formulai=1；I=1,2,3.
Step 6,3 kinds of features that insulator is extracted in step 3, step 4, step 5 are merged, three kinds of single features When fusion, each characteristic quantity is normalized before similarity measurement, adjust and scheme in target image to be identified and database As the weight between three features.Its weight is determined with analytic hierarchy process (AHP), calculates consistency ratio CR, verifies judgment matrix three Whether the weight of Fusion Features meets the requirements, and is specifically implemented according to the following steps:
Step 6.1, if texture, color and the shape eigenvectors of a target image i to be identified are [Oi1,Oi2,Oi3], N number of image in recognition target image i and database is treated by chessboard distance and carries out similarity calculation, is calculated distance value and is remembered respectively For Di1,Di2,......DiN；DiNIt is also three characteristic distances of any one image in target image to be identified and data picture library:
DiN=[d1,iN,d2,iN,d3,iN] (6-1)
Step 6.2, the mean μ of 3N distance value is calculatedDAnd standard deviation sigmaD, specific formula for calculation are as follows:
F thereina、fb、fcRespectively represent texture, color and shape eigenvectors；
Step 6.3, N width image (N takes any positive integer) similarity distance D in recognition target image and feature database is treatedi1, Di2,......DiNIt is normalized, keeps three kinds of different characteristic vectors roughly the same on similarity calculation, DiNIt is worth normalizing [0,1] section, expression formula are largely fallen in after change are as follows:
Step 6.4, the weight of three Fusion Features is determined using analytic hierarchy process (AHP).With 1-9 scaling law to Fusion Features after Three feature description indexes importance of insulation subgraph are judged, export three using the judgment matrix A that policymaker provides The weight s of featurei, then the consistency met with consistency ratio CR test and judge matrix, three features are exported to verify Weighted value accuracy, specific step is:
Step 6.4.1, according to 1-9 scaling law to three feature weight s1,s2,s3Importance proportion quotiety two-by-two carries out Assignment forms judgment matrix:
A1=(apq)n×nWherein apqProportion quotiety for pth feature with respect to q feature importance,
Wherein, 1-9 scaling law is as shown in the table:
Proportion quotiety Define explanation
1 Two feature weights are strong on an equal basis
3 One feature weight is slightly stronger than another
5 One feature weight is stronger than another
7 One feature weight is important very stronger than another
9 One feature weight is absolute stronger than another
2,4,6,8 It trades off between two feature weight scales
Step 6.4.2 takes width insulation subgraph, in hierarchy Model, only takes first layer to this insulation subgraph It is secondary, and this level can be divided into three classifications, i.e. texture (C), color (S) and shape (T), then corresponding judgment matrix indicates Are as follows:
It is thereinRespectively color and shape feature proportion quotiety, texture and color characteristic ratio Scale, texture and shape feature proportion quotiety.Generally according to the interpretation of result of three features of user's priori knowledge and extraction, shape Other two aspect ratios of aspect ratio are great.
When starting, one initial value of scale value is provided, that is, is setJudgement is consistent Sex rate CR.
Step 6.4.3, to A1In every row element multiplication and open cube and obtain vector Yi=(y1,y2,y3), calculation formula is such as Under:
To YiIt is normalized to obtain normalized weight vector si=(s1,s2,s3)
Step 6.4.4, to A1In every column element sum to obtain Zq=(Z1,Z2,Z3)
Step 6.4.5 calculates judgment matrix A1Maximum value characteristic value
Step 6.5, consistency ratio CR is calculated, and whether is met unanimously using consistency ratio CR test and judge matrix Property, and then determine the weighted value of Fusion Features,
Specific step are as follows:
Step 6.5.1 calculates judgment matrix A1Coincident indicator CI,
Wherein n is judgment matrix A1Order, the Maximum characteristic root λ of the smaller judgment matrix of CImaxMore meet completely the same Property, CI is bigger, and the degree for illustrating judgment matrix deviation crash consistency is bigger.
Judgment matrix A is calculated according to 6-101Other two characteristic value is denoted as respectively: λ2、λ3These three maximum eigenvalue are sought Average valueCalculation formula is as follows:
Step 6.5.2 calculates Aver-age Random Consistency Index RI according to formula (9-10)
Step 6.5.3 calculates consistency ratio CR using calculated CI and RI:
The judgment matrix A as CR < 0.11With satisfactory consistency, i.e. weight vectors siMeet matrix equation: A1si=λmaxsi There is maximum eigenvalue λmax, so judgment matrix A1The weighted value of three Fusion Features meets the requirements, the power used when calculating at this time The weighted value used when weight values are image co-registration in this method exports fused insulator identification image.
As CR >=0.1, judgment matrix A1Without satisfactory consistency, then goes to step 6.4 and re-start assignment decision It is right that person needs againThree scales carry out assignment according to 1-9 scaling law, with assignment when preceding primary calculating It is adjusted after comparison, constructs new judgment matrix A1, then calculate separately weight vector si, judgment matrix maximum feature λmaxAnd one Cause property index CI, finally calculates consistency ratio CR, until CR < 0.1 is set up.
A kind of insulator recognition detection method based on multicharacteristic information integration technology of the invention, by insulator The identification of textural characteristics shape feature color characteristic, can be effectively detected insulator whether there is or not falling string, and inside and outside that whether there are cracks is damaged, be It is no to there are the open defects such as self-destruction, by the identification of color characteristic, detect degree filthy on insulator.The texture of insulator Feature, color characteristic and shape feature effectively combine, and carry out the detection of multicharacteristic information fusion recognition to target image, and pass through Go out final target insulator with three property data base similarity matching identifications of established insulator, and judges it online Operating status whether there is filth, reveal, fall the failures such as string, crackle breakage.With this come solve in insulator identification process due to Acquisition of information is inaccurate, object recognition rate caused by uncertain and incomplete etc. factors is low and fault diagnosis in time-consuming and laborious etc. ask Topic.

Claims (1)

1. a kind of insulator recognition detection method based on multicharacteristic information integration technology, which is characterized in that specifically according to following Step is implemented:
Step 1, it insulate on the transmission line of electricity acquired using the focusing video camera with holder installed on steel tower at the scene or iron The picture signal of son；
Step 2, image is collected to step 1 to pre-process, carry out image segmentation using optimal entropic threshold method, pass through analysis chart As the entropy of grey level histogram, optimal threshold is found；The step 2 is specifically implemented according to the following steps:
Step 2.1, the tonal range of acquired image in step 1 is denoted as { 0,1,2 ... ... L-1 }；
Step 2.2, if the region that pixel of the gray level lower than t is constituted is target area A, the then expression formula of target area A entropy Are as follows:
P in formulaiIndicate that gray level is the probability that i occurs, i=0,1,2 ... ... t；ptIndicate that gray level is the probability that t occurs；
Step 2.3, if the region that pixel of the gray level higher than t is constituted is background area B, the then expression formula of background area B entropy Are as follows:
I=t+1, t+2, t+3 ... .L-1；
P in formulaiIndicate that gray level is the probability that i occurs, i=0,1,2 ... ... t；ptIndicate that gray level is the probability that t occurs；
Step 2.4, max-thresholds are calculated,
It is defined according to entropy function,
In formulaI=0 therein, 1,2 ... ... t；I=0 therein, 1,2 ... ... L-1；
Then, when entropy function C (t) is maximum,
Calculate t=argmax { C (t) } (2-4)
Mean that the max-thresholds of gray scale t；
Step 3, the texture feature vector that insulator is extracted in pretreated image was carried out from step 2；Step 3 tool Body follows the steps below to implement:
Step 3.1, distinguished using 3 color spaces of the statistical method in texture characteristic extracting method to former RGB color image Mean value, variance, the degree of bias, kurtosis, energy, entropy parameter of totally 6 color space features as textural characteristics are calculated,
Specific formula for calculation is as follows:
(1) mean value
(2) variance
(3) degree of bias
(4) kurtosis
(5) energy
(6) entropy
Wherein g (i) is the gray value of i-gray level, pgIt (i) is the probability of i-th of gray value；
Former RGB color image is just divided into 18 feature vectors by this 6 features in total；The single features different to image carry out special Normalization inside feature is carried out when sign is extracted, this 18 feature vectors are carried out by normalizing inside feature using Gaussian normalization method Change to [- 1,1], reduces other element value and the distribution of the element value after normalization is had an impact；The normalization of feature vector FiCalculation formula it is as follows:
Wherein μiIndicate the mean value of target image characteristics vector to be identified, σiIndicate the standard of target image characteristics vector to be identified Difference, fiFor 18 one dimensional histograms initial characteristic values and fi∈[f1,f2,f3,.......f18]；I=1,2,3 ... ... 18；
Step 3.2, the texture feature vector extracted in step 3.1 and the texture in established sample database in advance is special Sign vector compares matching, calculates its similarity, method particularly includes:
The matched similarity distance of textural characteristics in target image to be identified and characteristic image library is calculated using chessboard distance formula, The normalization characteristic vector of two images is that 1 18 dimensional vector may be expressed as:
Fxi[Fx1,Fx2,Fx3,......Fx18]；
Fyi[Fy1,Fy2,Fy3,......Fy18]；
Then, it is calculated according to chessboard distance formula:
Wherein S1Indicate the matched similarity distance of textural characteristics in target image to be identified and characteristic image library, FxiAnd FyiIt indicates The normalization characteristic vector of two images, S1Value is bigger, and the similarity of two images is higher；
Step 4, insulator color feature vector to be identified is extracted, it is empty to the HSV color of insulation subgraph collected in step 1 Between carry out unequal interval quantization；The step 4 is specifically implemented according to the following steps:
Step 4.1, the space tone H is divided into 8 parts, saturation degree S and the space brightness V are divided into 3 parts, the tone H range of image For [0,360 °], the range of saturation degree S and brightness V are [0,1]；
Step 4.2, color quantization is carried out according to the different range of H, S, V, specific as follows shown:
3 color components after quantization are synthesized one-dimensional characteristic vector:
I=HQsQv+SQv+V (4-4)
Q in formulas=3 be the quantization series of component S；Qv=3 be component V quantization series when (4-
4) it may be expressed as:
I=9H+3S+V (4-5)
The subspace that HSV space after quantization resolves into obtains 72 one dimensional histograms；
Step 4.3, it by the feature vector Gaussian normalization of the color histogram after quantifying in step 4.2 to [- 1,1], reduces each Other element value has an impact the distribution of the element value after normalization, the normalized F of feature vectorlCalculation formula is as follows:
Wherein μlIndicate the mean value of target image characteristics vector to be identified, σlIndicate the standard of target image characteristics vector to be identified Difference, flFor 72 one dimensional histograms initial characteristic values and fl∈[f1,f2,f3,.......f72]；L=1,2,3 ... ... 72
Step 4.4, target image to be identified and the matched phase of color characteristic in characteristic image library are calculated using chessboard distance formula Like degree distance；The normalization characteristic of two images may be expressed as: to for 2 72 dimensional vectors
Fpl[Fp1,Fp2,Fp3,......Fp72]；Fql[Fq1,Fq2,Fq3,......Fq72]；
Wherein chessboard distance calculation formula:
Wherein S2Indicate the matched similarity distance of textural characteristics in target image to be identified and characteristic image library, FplAnd FqlIt indicates The normalization characteristic vector of two images, S2Value is maximum, and the similarity of two images is higher；
Step 5, insulator shape eigenvectors to be identified are extracted, the shape feature of insulator is unrelated with external environment, is object In vitro most stable information and image is most intuitive, most direct visualization, calculated using based on Hu square in region description method Method extracts insulator shape feature；The step 5 is specifically implemented according to the following steps:
Step 5.1, the p+q rank square of acquired image f (x, y) is defined are as follows:
Then, the central moment of p+q rank square are as follows:
In formulaIndicate the center of gravity of image-region；
Step 5.2, normalizing can be carried out to the central moment in formula (5-2) to obtain image property unrelated with scaling itself Change, the central moment after normalization indicates are as follows:
In formulaP+q=2,3,4 ... ... normalized centers are kept away from translation, scaling and the rotation to object It is constant；
Step 5.3, image f (x, y) second order and third central moment are calculated using using formula (5-2), obtained
Wherein second-order moment around mean μ02And μ20Respectively indicate the moment of inertia around the vertically and horizontally axis by gray scale mass center, three ranks Central moment μ03And μ30Asymmetric degree of the mensurable institute's analyzed area of amplitude to vertically and horizontally axis；
7 invariant moments are constructed, translation, scaling and invariable rotary can be kept under conditions of consecutive image；
The 7 invariant moments are defined respectively as:
Central moment expression formula after second order normalized:
Central moment expression formula after the normalization of three ranks:
φ1=η20+η02 (5-8)
φ2=(η20+η02)2+4η2 11 (5-9)
φ3=(η30-3η12)2+(3η21-η03)2 (5-10)
φ4=(η30+η12)2+(η21+η03)2 (5-11)
φ5=(η30-3η12)(η30+η12)[(η20+η12)2-3(η21+η03)2]+(3η21+η03)(η21+η03)[3(η30+η12)2-(η21 +η03)] (5-12)
φ6=(η20-η02)[(η30+η12)2-(η21+η03)2]+4η11(η30+η12)(η21+η03) (5-13)
φ7=(3 η21-η03)(η30+η12)[(η30+η12)2-3(η21+η03)2]+(3η21-η03)(η21+η03)[3(η30+η12)2-(η21 +η03)2] (5-14)
φ therein1~φ7For the calculation formula of 7 invariant moments, each η value indicates the central moment after second order, the normalization of three ranks；
Step 5.4, by Hu square algorithm (5-8)~(5-14) calculation formula extract 7 shape features of insulator to be identified to Amount, then to each feature vector Gaussian normalization to [- 1,1]；By shape feature in target image to be identified and characteristic image library Being normalized to 27 dimensional vectors may be expressed as:
Fsi[Fs1,Fs2,Fs3,......Fs7]；Fti[Ft1,Ft2,Ft3,......Ft7]；
Equally, using chessboard distance formula calculate target image to be identified in characteristic image library shape feature it is matched similar Spend distance S3:
S3Value is maximum, and the similarity of two images is higher；
Step 5.5, the color, shape and textural characteristics of the insulator extracted are subjected to Fusion Features；If three feature vectors The similarity (distance) of fusion is D, the similarity (distance) of three kinds of single features vectors (feature of color, shape and texture) S1,S2,S3Corresponding weight is respectively w1,w2,w3, then the similarity (distance) of three kinds of Fusion Features vectors be,
D=s1w1+s2w2+s3w3 (5-16)
∑ w in formulai=1；I=1,2,3.；
Step 6,3 kinds of features that insulator is extracted in step 3, step 4, step 5 are merged, three kinds of single features fusions When, each characteristic quantity is normalized before similarity measurement, adjust image three in target image to be identified and database Weight between a feature；Its weight is determined with analytic hierarchy process (AHP), calculates consistency ratio CR, verifies three features of judgment matrix Whether the weight of fusion meets the requirements；The step 6 is specifically implemented according to the following steps:
Step 6.1, if texture, color and the shape eigenvectors of a target image i to be identified are [Oi1,Oi2,Oi3], pass through Chessboard distance treats N number of image in recognition target image i and database and carries out similarity calculation, calculates distance value and is denoted as respectively Di1,Di2,......DiN；DiNIt is also three characteristic distances of any one image in target image to be identified and data picture library:
DiN=[d1,iN,d2,iN,d3,iN] (6-1)
Step 6.2, the mean μ of 3N distance value is calculatedDAnd standard deviation sigmaD, specific formula for calculation are as follows:
F thereina、fb、fcRespectively represent texture, color and shape eigenvectors；
Step 6.3, N width image similarity distance D in recognition target image and feature database is treatedi1,Di2,......DiNReturned One change processing, N take any positive integer, keep three kinds of different characteristic vectors roughly the same on similarity calculation, DiNAfter value normalization Major part falls in [0,1] section, expression formula are as follows:
Step 6.4, the weight of three Fusion Features is determined using analytic hierarchy process (AHP)；With 1-9 scaling law to insulating after Fusion Features Three feature description indexes importance of subgraph are judged, export three features using the judgment matrix A that policymaker provides Weight si, then the consistency met with consistency ratio CR test and judge matrix, to verify the power for exporting three features Weight values accuracy；The step 6.4 is specifically implemented according to the following steps:
Step 6.4.1, according to 1-9 scaling law to three feature weight s1,s2,s3Importance proportion quotiety carry out assignment, formed Pairwise comparison matrix:
A1=(apq)n×nWherein apqProportion quotiety for pth feature with respect to q feature importance,
Step 6.4.2 takes width insulation subgraph, in hierarchy Model, only takes the first level to this insulation subgraph, and This level can be divided into three classifications, i.e. texture (C), color (S) and shape (T), then corresponding judgment matrix indicates are as follows:
It is thereinRespectively color and shape feature proportion quotiety, texture and color characteristic proportion quotiety, Texture and shape feature proportion quotiety；According to the interpretation of result of three features of user's priori knowledge and extraction, shape feature is than it His two aspect ratios are great,
When starting, one initial value of scale value is provided, that is, is setJudge consistency ratio Rate CR；
Step 6.4.3, to A1In every row element multiplication and open cube and obtain vector Yi=(y1,y2,y3), calculation formula is as follows:
To YiIt is normalized to obtain normalized weight vector si=(s1,s2,s3)
Step 6.4.4, to A1In every column element sum to obtain Zq=(Z1,Z2,Z3),
Step 6.4.5 calculates judgment matrix A1Maximum value characteristic value
Step 6.5, consistency ratio CR is calculated, and whether meets consistency using consistency ratio CR test and judge matrix, into And determine the weighted value of Fusion Features；The step 6.5 is specifically implemented according to the following steps:
Step 6.5.1 calculates judgment matrix A1Coincident indicator CI,
Wherein n is judgment matrix A1Order, the Maximum characteristic root λ of the smaller judgment matrix of CImaxMore meet crash consistency, CI is got over Illustrate that the degree of judgment matrix deviation crash consistency is bigger greatly；
Judgment matrix A is calculated according to 6-101Other two characteristic value is denoted as respectively: λ2、λ3The flat of these three maximum eigenvalue is sought Mean valueCalculation formula is as follows:
Step 6.5.2 calculates Aver-age Random Consistency Index RI according to formula (6-13),
Step 6.5.3 calculates consistency ratio CR using calculated CI and RI:
The judgment matrix A as CR < 0.11With satisfactory consistency, i.e. weight vectors siMeet matrix equation: A1si=λmaxsiHave most Big eigenvalue λmax, so judgment matrix A1The weighted value of three Fusion Features meets the requirements, the weighted value used when calculating at this time The weighted value as used when image co-registration in this method exports fused insulator identification image；
As CR >=0.1, judgment matrix A1Without satisfactory consistency, then goes to step 6.4 and re-start assignment policymaker's needs Again rightThree scales carry out assignment according to 1-9 scaling law, after assignment comparison when preceding primary calculating It is adjusted, constructs new judgment matrix A1, then calculate separately weight vector si, judgment matrix maximum feature λmaxAnd consistency refers to CI is marked, consistency ratio CR is finally calculated, until CR < 0.1 is set up.
CN201610452634.XA 2016-06-21 2016-06-21 A kind of insulator recognition detection method based on multicharacteristic information integration technology Active CN106127756B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610452634.XA CN106127756B (en) 2016-06-21 2016-06-21 A kind of insulator recognition detection method based on multicharacteristic information integration technology

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610452634.XA CN106127756B (en) 2016-06-21 2016-06-21 A kind of insulator recognition detection method based on multicharacteristic information integration technology

Publications (2)

Publication Number Publication Date
CN106127756A CN106127756A (en) 2016-11-16
CN106127756B true CN106127756B (en) 2019-03-26

Family

ID=57470387

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610452634.XA Active CN106127756B (en) 2016-06-21 2016-06-21 A kind of insulator recognition detection method based on multicharacteristic information integration technology

Country Status (1)

Country Link
CN (1) CN106127756B (en)

Families Citing this family (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106960178B (en) * 2017-02-23 2020-02-07 中国科学院自动化研究所 Training method of insulator recognition model and insulator recognition and positioning method
CN106908452A (en) * 2017-04-24 2017-06-30 武汉理工大学 Engine lubricating oil quality monitoring device based on machine vision
CN108108772B (en) * 2018-01-06 2021-08-10 天津大学 Insulator pollution flashover state detection method based on aerial image of distribution line
CN110046630A (en) * 2018-01-16 2019-07-23 上海电缆研究所有限公司 Defect mode identification method/the systems/devices and readable storage medium storing program for executing of object
CN108470141B (en) * 2018-01-27 2021-08-10 天津大学 Statistical feature and machine learning-based insulator identification method in distribution line
CN108680833B (en) * 2018-03-30 2021-06-11 中科吉芯(秦皇岛)信息技术有限公司 Composite insulator defect detection system based on unmanned aerial vehicle
CN108765373B (en) * 2018-04-26 2022-03-22 西安工程大学 Insulator abnormity automatic detection method based on integrated classifier online learning
CN109345503A (en) * 2018-08-06 2019-02-15 浙江大学 A kind of capsule gastroscope image blutpunkte recognition methods based on image multiple features fusion
CN111462057B (en) * 2020-03-23 2023-02-21 华南理工大学 Transmission line glass insulator self-explosion detection method based on deep learning
CN111914720B (en) * 2020-07-27 2021-08-17 长江大学 Method and device for identifying insulator burst of power transmission line
CN113469953B (en) * 2021-06-10 2022-06-14 南昌大学 Transmission line insulator defect detection method based on improved YOLOv4 algorithm
CN114022747B (en) * 2022-01-07 2022-03-15 中国空气动力研究与发展中心低速空气动力研究所 Salient object extraction method based on feature perception
CN114455255A (en) * 2022-01-27 2022-05-10 山东仁功智能科技有限公司 Abnormal cigarette sorting error detection method based on multi-feature recognition
CN114724042B (en) * 2022-06-09 2022-09-02 国网江西省电力有限公司电力科学研究院 Automatic detection method for zero-value insulator in power transmission line
CN115542100B (en) * 2022-11-29 2023-03-28 广东电网有限责任公司东莞供电局 Insulator fault detection method, device, equipment and medium
CN115972468B (en) * 2022-11-30 2024-03-12 东方(天津)新材料科技有限公司 Preparation method of anti-corrosion sleeve
CN117593285A (en) * 2023-12-14 2024-02-23 江苏恒兆电缆有限公司 Quality detection system and method for flexible mineral insulation flexible fireproof cable

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101403676A (en) * 2008-10-28 2009-04-08 华北电力大学 Insulator hydrophobicity rank amalgamation judging method based on D-S evidence theory
CN103745234A (en) * 2014-01-23 2014-04-23 东北大学 Band steel surface defect feature extraction and classification method
CN103810704A (en) * 2014-01-23 2014-05-21 西安电子科技大学 SAR (synthetic aperture radar) image change detection method based on support vector machine and discriminative random field
CN104392209A (en) * 2014-11-07 2015-03-04 长春理工大学 Evaluation model for image complexity of target and background
CN104571468A (en) * 2013-10-11 2015-04-29 中国移动通信集团广东有限公司 Method and device for processing characteristics of digital image

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101403676A (en) * 2008-10-28 2009-04-08 华北电力大学 Insulator hydrophobicity rank amalgamation judging method based on D-S evidence theory
CN104571468A (en) * 2013-10-11 2015-04-29 中国移动通信集团广东有限公司 Method and device for processing characteristics of digital image
CN103745234A (en) * 2014-01-23 2014-04-23 东北大学 Band steel surface defect feature extraction and classification method
CN103810704A (en) * 2014-01-23 2014-05-21 西安电子科技大学 SAR (synthetic aperture radar) image change detection method based on support vector machine and discriminative random field
CN104392209A (en) * 2014-11-07 2015-03-04 长春理工大学 Evaluation model for image complexity of target and background

Also Published As

Publication number Publication date
CN106127756A (en) 2016-11-16

Similar Documents

Publication Publication Date Title
CN106127756B (en) A kind of insulator recognition detection method based on multicharacteristic information integration technology
CN106023185B (en) A kind of transmission facility method for diagnosing faults
CN112734692B (en) Defect identification method and device for power transformation equipment
Wang et al. Automatic fault diagnosis of infrared insulator images based on image instance segmentation and temperature analysis
CN112380952B (en) Power equipment infrared image real-time detection and identification method based on artificial intelligence
Davari et al. Intelligent diagnosis of incipient fault in power distribution lines based on corona detection in UV-visible videos
Wang et al. Recognition and drop-off detection of insulator based on aerial image
CN108765373A (en) A kind of insulator exception automatic testing method based on integrated classifier on-line study
CN108537154A (en) Transmission line of electricity Bird&#39;s Nest recognition methods based on HOG features and machine learning
CN103247039B (en) A kind of high voltage cable live detection method based on composite vision
CN105957088B (en) Transformer composite insulator casing monitoring method and system based on computer vision
CN108072667A (en) Insulator contamination level detection method and system based on EO-1 hyperion
CN109872317A (en) A kind of defect identification method based on power equipments defect identification learning model
CN106326808A (en) Method for detecting bird nests in power transmission line poles based on unmanned plane images
CN108108772A (en) A kind of insulator contamination condition detection method based on distribution line Aerial Images
CN107257351A (en) One kind is based on grey LOF Traffic anomaly detections system and its detection method
CN112115770A (en) Method and system for identifying autonomous inspection defects of unmanned aerial vehicle of overhead line
CN111507277B (en) Construction method and application of workflow normalization recognition model
Zhang et al. Detection and condition assessment of icicle bridging for suspension glass insulator by image analysis
Jadin et al. Thermal condition monitoring of electrical installations based on infrared image analysis
CN116824517B (en) Substation operation and maintenance safety control system based on visualization
CN111259736A (en) Real-time pedestrian detection method based on deep learning in complex environment
CN110889435A (en) Insulator evaluation classification method and device based on infrared image
CN111310899B (en) Power defect identification method based on symbiotic relation and small sample learning
Lu et al. An image recognition algorithm based on thickness of ice cover of transmission line

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
TR01 Transfer of patent right

Effective date of registration: 20210816

Address after: 710065 No. 11301, block B, Huajing Business Plaza, No. 20, Fenghui South Road, Zhangba street, high tech Zone, Xi'an, Shaanxi Province

Patentee after: XI'AN JIN POWER ELECTRICAL Co.,Ltd.

Address before: 710048 No. 19 Jinhua South Road, Shaanxi, Xi'an

Patentee before: XI'AN POLYTECHNIC University

TR01 Transfer of patent right