CN103247044A - Defective condition detection method based on curved and dotted singularity characteristics of insulator of overhead contact system of high-speed rail - Google Patents

Abstract

The invention discloses a defective condition detection method based on curved and dotted singularity characteristics of an insulator of an overhead contact system of a high-speed rail. The method comprises the steps as follows: the overhead contact system is shot at night, so that an image of a to-be-detected insulator is obtained; the image of the to-be-detected insulator is subjected to contrast ratio enhancement and Gaussian filter and noise reduction processing; an approximate direction of the insulator in the image is determined through Radon transformation; then the insulator is determined and an angle is finely adjusted by using the curved singularity characteristics of the second generation curvelet transform; and finally a defective condition of the insulator is judged by the curved singularity characteristics of the second generation curvelet transform and the dotted singularity characteristics of the wavelet transform. The method can distinguish the insulator rapidly and correctly, and the state whether the insulator is damaged, a defective foreign matter blended state, and a type of the defective state are judged effectively. The actual input quantity is the image to-be-detected, the actual output quantity is a picture of the insulator which is positioned and result data of the defective state detection, the distinguishing effect is good, and the judgment result is accurate and reliable.

Description

Defective mode detection method based on high ferro contact net insulator curve-like and point-like singularity feature

Technical field

The present invention relates to electrified high-speed railway contact net Intelligent Measurement field, specifically refer to a kind of detection method of high ferro contact net insulator defective mode.

Background technology

In electric railway, the transmission line of electricity to the special shape of electric locomotive power supply that set up in the downline sky is called as contact net.Wherein, insulator is the important component part of supporting in the contact net with suspender.The main effect of insulator is to be responsible for the insulation of contact net etc.This shows whether insulator state has well directly determined the normal power supply of electric locomotive.Along with the growing growth of quantity and the scale of China Express Railway, the state-detection task of high ferro insulator also is tending towards heavy and difficulty.

The manual observation method is the main method that traditional insulator defective mode detects.There are the artificial drawbacks such as dangerous higher, artificial detection uncertainty, circuit circumstance complication of living in of observing in the artificial observational technique of seeing.Along with development of science and technology, also produced the detection method of many unartificial observations, for example infrared imaging method, ultraviolet imagery method, power plant's mensuration, ultrasonic Detection Method etc.But these methods but exist, and the surveying work amount is big, detection accuracy is low, big, the problems such as expense is expensive, anti-external interference ability of operation spinoff.Be a kind of expense new-type insulator detection method low, simple to operate and detect based on the insulator of image, can carry out real-time detection and localization, in time find the situation of defective mode.Existing two generations bent ripple location insulator and detect damage state based, Wavelet Detection insulator and be mingled with the foreign matter technology, but do not see that utilizing bent wave conversion to obtain matrix of coefficients generates the technology that judgment matrix and matched curve detect, do not see utilizing curve-like singularity and the small echo point-like Singularity Detection of Qu Bo to detect insulator breakage and the integrated application that is mingled with foreign matter respectively, do not see and utilize the adjustment of bent wave system matrix number to determine the technology of insulator angle.

Document " Li Bo. the research of insulator online test method both at home and abroad. electrical technology; 2011,9:1-5 " introduced the application in the insulator context of detection of direct observational method, ultraviolet imagery method, infrared imaging method, ultrasonic Detection Method, equivalent salt density method, steep wave test method(s), potential measurement method, electric-resistivity method, electric field method, leakage current method and pulse current method.Narrate ultimate principle, implementation method and the calibration method of corresponding poor insulator and the merits and demerits of every kind of method of every kind of method.

Document " Chen Tao. based on the research of contactless deterioration insulator detection method. University Of Chongqing doctorate paper .2006 " introduced by measure near the insulator in the certain distance scope some electric field intensity, according to location insulators such as the size of tested insulator, shapes.Utilize the relation of relative air humidity RH, realize the detection of contaminated insulator.

Document " Wang Huiyan. based on the rim detection of wavelet transformation and in the porcelain insulator Application in Fault Diagnosis. Automation of Electric Systems .2004; 28 (15): 76-79 " introduced the multi-scale edge feature of selecting for use second order batten biorthogonal wavelet to extract the porcelain vase image, a kind of scheme of passing through the insulator rim detection of dynamic process insulator image realization is proposed.

Patent " based on two generations bent wave system count the insulator breakage failure detection method of morphology band energy method. the patent No.: 201210174805, inventor: Liu Zhigang, Han Zhiwei; Applicant: Southwest Jiaotong University " utilize bent wave conversion of two generations, obtain the energy bar of anisotropic filter factor matrix for analyzing, obtain insulator information, judge the situation of insulator breakage.

Summary of the invention

The objective of the invention is to overcome the shortcoming and deficiency that above-mentioned prior art exists, a kind of applied to high-speed railway touching net insulator failure detection method is provided.This method is to utilize major light and high-definition camera on the railway combined inspection vehicle to be engaged in the image of night contact net being taken, carry out correct one or more insulator part of identification fast, and accurately judge the integrated operation of defective mode whether insulator breakage takes place or be mingled with foreign matter.Its core is to utilize the curve-like singularity on insulator border and bent wave conversion to have the characteristics of susceptibility for the curve-like singularity respectively, has realized judgement and the damaged detection of insulator position with bent wave conversion; And utilize one dimension gray-scale pixels point between the insulator sheet to change obviously and rule and small echo have the characteristics of susceptibility to the point-like singularity of one dimension pixel, with small echo the horizontal gray-scale value of image between the insulator sheet is done the wavelet singular value and detect the judgement insulator and be mingled with the foreign matter situation.The insulator recognition effect is good, and judged result can reach one or more contact net insulator of identification and can judge that insulator breakage and the defective mode that is mingled with foreign matter detect integrated purpose accurately and reliably.

The objective of the invention is to be achieved through the following technical solutions:

Defective mode detection method based on high ferro contact net insulator curve-like and point-like singularity feature the steps include:

(1), image acquisition and input: utilize major light and high-definition camera on the railway combined inspection vehicle to be engaged in night to the shooting of contact net, obtain and import insulator image to be detected;

(2), the image pre-service: (1) gained insulator image to be detected degree of comparing strengthened and the gaussian filtering denoising obtain gray level image f (x, y);

(3), the insulator direction is determined: the bracket place direction at insulator place is 0 °, 45 °, 135 ° near zones substantially in gray level image; (angle of the maximum value correspondence of integrated value is high gray-scale value straight line place angle for x, near y) each angle integration three angles to utilize the Radon conversion to ask image f

To judge that this angle may be the roughly direction of insulator place bracket in the image, i.e. the roughly direction of insulator;

(4), insulator location: (x y) rotates to be in (3) roughly angle of three of gained respectively with image f (x y), and carries out following all operations to this three width of cloth image respectively: with image f ' (x to obtain three width of cloth image f ', y) carry out the bent wave conversion of two generations of 64 directions, the matrix of coefficients that obtains is carried out statistical treatment, obtain judgment matrix, judge the insulator position according to the judgment matrix feature;

(5), fine setting insulator angle: relatively the matrix of coefficients of the bent wave conversion of 6 directions is closed in the insulator position, chooses the matrix of coefficients of insulator coefficient of region matrix variance minimum, and with the angle of this matrix of coefficients correspondence original image is rotated to this angle;

(6), judge insulator defective mode kind: for the insulator position that has judged in (4), utilize the curve-like singularity of bent wave conversion detection of two generations defective mode, the insulator system matrix number that has obtained in (4) is added up judgement, judge the situation of insulator porcelain patticoat breakage, the point-like singularity that the recycling wavelet transformation has, the gray scale sudden change situation between insulator sheet crack obtained judges that insulator is mingled with the situation of foreign matter, exports oriented picture and testing result data to the subsequent treatment unit.

Compared with prior art, beneficial effect of the present invention is as follows:

1, insulator breakdown detection method of the present invention is based on image processing techniques, is to be detected object with the insulator image, by the picture of field real-time acquisition is handled to extract the insulator feature, has realized the detection of insulator defective mode.

2, the present invention utilizes the Radon conversion to ask each angle integration of specific 0 °, 45 °, the 135 ° near zone of image.The angle of the maximum value correspondence of integrated value is high gray-scale value straight line place angle, judges that this angle may be the roughly direction of insulator place bracket in the image, the i.e. roughly direction of insulator.Realized obtaining of insulator place angle.

3, the curve-like singularity refers to two dimension or the non-homogeneous zone of higher-dimension image pixel, and there is the phenomenon of uneven distribution in the curve-like pixel.The insulator border is the slick and sly high gray-scale value curve of two dimension, and the background very low with gray-scale value has singularity, shows the curve-like singularity.Bent wave conversion has susceptibility for the curve-like singularity, pixel curve-like to the insulator border changes very sensitive, the testing image that the present invention separates by Qu Bofen with bent wave conversion, generate matrix of coefficients, the matrix of coefficients statistical treatment is generated judgment matrix, realized the judgement of insulator position.

4, the present invention separates testing image by Qu Bofen, generates matrix of coefficients.To the insulator position of having oriented, relatively 3 angles are closed in this position increases the matrix of coefficients that directions and 3 angles reduce the bent wave conversion of direction, choose the matrix of coefficients of this position parameter matrix variance minimum, and with the angle of this matrix of coefficients correspondence with the original image rotation to this angle, realized the fine setting of insulator direction.

5, the point-like singularity refers to that there is the phenomenon of non-uniform Distribution such as sudden change in the distribution of one dimension pixel.Along the insulator direction, the obvious and rule of grey scale change belongs to the point-like singularity that the one dimension pixel suddenlys change between the insulator sheet.Small echo has susceptibility to the point-like singularity of one dimension pixel.Because be mingled with the even regularity of distribution that foreign matter has changed gray scale between sheet, so the present invention does the wavelet singular value with small echo to the horizontal gray-scale value of image and detects, with the processing of making zero of the horizontal gray-value image minimal value of gained, calculate adjacent zeros distance and band mean distance, judge that according to distance relation insulator is mingled with the foreign matter situation.Realized that insulator breakage and the defective mode that is mingled with foreign matter detect.

As mentioned above, the method that the present invention adopts at insulator have the characteristics of curve-like singularity along the porcelain patticoat direction, utilize bent wave conversion to catch the linear feature of insulator image accurately.At the curve-like singularity features of damaged insulator, effectively judge the damaged situation of insulator with bent wave conversion.Be mingled with the crosswise spots singularity of foreign matter at ceramics, use wavelet transformation can effectively judge the situation that is mingled with foreign matter of insulator.Realize that insulator breakage and the defective mode that is mingled with foreign matter detect.

Description of drawings

Fig. 1 operational flowchart.

Fig. 2 testing image.

Fig. 3 angle rotate effect figure.

Fig. 4 judgment matrix matched curve figure.

Fig. 5 finishes the insulator signal of location.

The local figure of the damaged insulator of Fig. 6.

The damaged insulator judgment matrix of Fig. 7 matched curve figure.

Fig. 8 is mingled with the local figure of foreign matter insulator.

Fig. 9 is mingled with the horizontal gray-scale statistical figure of foreign matter insulator.

Embodiment

Below in conjunction with accompanying drawing embodiments of the present invention are described in further detail.

Fig. 1 is operational flowchart of the present invention.

Fig. 2 is testing image.

A kind of embodiment step of the present invention is as follows:

(a) image acquisition

Utilize major light and high-definition camera on the railway combined inspection vehicle to be engaged in night to the shooting of contact net, obtain the less insulator image to be detected of undesired signal and be input in the follow-up unit.

(b) image pre-service

Insulator image degree of comparing to be detected strengthened and the gaussian filtering denoising obtain gray level image f (x, y).

The picture contrast Enhancement Method is prior art, as adopting down surface function to partly degree of the comparing enhancing operation of image insulator.Contrast enhancement processing has suppressed between the gray area of insulator junction relatively, and has given prominence to insulator panel region in the image.The definition contrast strengthens computing:

M(m,n)=(I(m,n)-L _in)((H _out-L _out)/(H _in-L _in))+L _out （1）

L _in≤I(m,n)≤H _in （2）

Wherein, I is original image, and M is the image after the contrast adjustment.L _InAnd H _InBe the pixel coverage that strengthens in the original image, L _OutAnd H _OutBe the pixel coverage after strengthening.

Because there is partial noise in the gained image, in order to reduce interference of noise, the present invention has used existing gaussian filtering, and noise is handled.

The gaussian filtering method is prior art, and the Gaussian function with two variablees has following citation form:

$h(x,y)=e^{-\frac{x^2+y^2}{{2\mathrm{\σ}}^2}}---\left(3\right)$

In the following formula, σ is standard deviation.Utilize following formula can produce a m * n filter template, can carry out denoising to image.

(c) the insulator direction is determined

There is tangible high gray-scale value in the bracket at insulator place in the gray level image, and bracket place direction is 0 °, 45 °, 135 ° near zones substantially.(angle of the maximum value correspondence of integrated value is high gray-scale value straight line place angle for x, near y) each angle integration three angles to utilize the Radon conversion to try to achieve gray level image f

Can judge that this angle may be the roughly direction of insulator place bracket in the image, the i.e. roughly direction of insulator.

The Radon transform method is prior art, with the Radon conversion be designated as Rf (θ, t), known:

$\mathrm{Rf}(\mathrm{\θ},t)=\∫f(x,y)\mathrm{ds}={\∫}_{-\∞}^{\∞}f(t\cos \mathrm{\θ}+s\sin \mathrm{\θ},t\sin \mathrm{\θ}-s\cos \mathrm{\θ})\mathrm{ds}---\left(4\right)$

In the formula

: xcos θ+ysin θ=t

Be that the Radon conversion is that (x y) is mapped to Rf (θ, t) plane with processing back gray level image f.When existing and straight line L _{T, θ}Vertical high gray scale straight-line segment, then (θ t) produces a bigger peak value to the Rf of Dui Ying point.To selected specific three angular interval in the picture.Be defined as follows:

Only top three angular interval are detected.Get the maximum value of Rf (θ, x '), note θ at this moment is insulator angle place angle

(d) insulator location

(x y) rotates to be in (c) 3 roughly angles respectively with image f

Obtain three width of cloth image f ' (x, y).This moment, the insulator angle was roughly adjusted near 0 °.And respectively three width of cloth images are carried out following all operations.Fig. 3 is the angle rotate effect figure in wherein 45 ° of direction intervals.

Bent wave conversion is prior art, and (x y) is mapped as coefficient sequence α with image f ' _μThe function of (μ ∈ M).Under the rough situation

Definition:

${\mathrm{\α}}_{\mathrm{\μ}}=\⟨{\mathrm{\φ}}_{k_1,k_2},P_{0}f\⟩,\mathrm{\μ}=(k_1,k_2)\∈M^{\′}\backslash M---\left(6\right)$

Accurately situation is given a definition:

α _μ=<△ _sf,ψ _μ>,μ∈M _s （7）

By (6) (7) formula as can be known, bent wave conversion has multiple dimensioned, many decomposition, anisotropic characteristics, and on the insulator detected image, insulator porcelain vase slabbing is placed successively, has good direction consistance, owing to have the feature of the curve-like singularity sensitivity of Qu Bo, thus the Qu Bofen solution to insulator porcelain vase position than other position sensings.(x y) carries out the bent wave conversion of two generations of 64 directions, obtains the two-dimensional coefficient matrix of 64 directions, finds out the vertical dimension coefficients matrix with image f '

Choose the rectangle frame bigger slightly than insulator, to whole matrix of coefficients

The circulation of element is chosen one by one, and the matrix of at every turn choosing is carried out obtaining an one dimension judgment matrix β along the adding up of rectangular array direction _ξ, ask β _ξThe mean value of all ordinates is H, match judgment matrix β _ξObtain judging curve.If judge that crest value that curve contains continuous 5 rules is less greater than the variance of horizontal ordinate distance between the peak value of 1.5H and crest, can think that then this matrix frame region may be the insulator zone.Be the judgement curve synoptic diagram of insulator among Fig. 3 as Fig. 4, picture meets the demands among visible Fig. 4, and this zone is the insulator region.The insulator synoptic diagram of Fig. 5 for having been fixed, wherein the white edge position is for locating the insulator position of coming out.

(e) fine setting insulator angle

Relatively the matrix of coefficients of the bent wave conversion of 6 directions is closed in the insulator position, chooses the matrix of coefficients of insulator coefficient of region matrix variance minimum, and with the angle of this matrix of coefficients correspondence original image is rotated to this angle.

(f) judge insulator defective mode kind

For the insulator zone of locating out in (d), judgment matrix β _ξThe mean value that calculates all ordinates is H, supposes that numerical value is n greater than the quantity of the peak value of 1.5H, and the matrix result who adds up is put into cartesian coordinate system, from left to right crest is carried out numbering from 1 integer, makes the position of its horizontal ordinate be respectively T ₁... T _nThe distance of calculating between all adjacent peak horizontal ordinates is respectively S ₁... S _N-1, calculate S ₁... S _N-1Mean value be S _aIf S _i＞1.5S _aWherein (i=1.2...n) then judges T _iWith T _I+1There is the breakage of insulator between the insulator magnetic skirt of representative.Be the partial schematic diagram of damaged situation insulator as Fig. 6, Fig. 7 is this regional judgment matrix matched curve figure, wherein S ₅＞1.5S _a, there is disrepair phenomenon in the insulator magnetic skirt that reflects after the 5th.

Otherwise, the insulator image of orienting is carried out local contrast again regulate, the horizontal gray-scale value of image after the calculating contrast adjustment is based on the some singularity of wavelet transformation, the horizontal gray-scale value of image is done the wavelet singular value detect, with the processing of making zero of the horizontal gray-value image minimal value of gained.Suppose total m zero point, calculate the adjacent zeros distance L _j(wherein j=1...m) and band mean distance

The definite threshold value of singular value detection according to small echo is namely worked as

The time (wherein j=1...m), can judge that j insulator magnetic skirt may exist and be mingled with the foreign matter defective mode, otherwise, can judge that insulator state is good.Be the part figure that is mingled with the foreign matter insulator as Fig. 8.Fig. 9 does wavelet transformation for this zone and carries out rezero operation horizontal parametric statistics figure afterwards again, can find

Namely the 5th magnetic skirt place is mingled with foreign matter, exports oriented picture and testing result data.

Claims (4)

1. based on the defective mode detection method of high ferro contact net insulator curve-like and point-like singularity feature, the steps include:

(1), image acquisition and input: utilize major light and high-definition camera on the railway combined inspection vehicle to be engaged in night to the shooting of contact net, obtain and import insulator image to be detected;

(2), the image pre-service: (1) gained insulator image to be detected degree of comparing strengthened and the gaussian filtering denoising obtain gray level image f (x, y);

(3), the insulator direction is determined: the bracket place direction at insulator place is 0 °, 45 °, 135 ° near zones substantially in gray level image; (angle of the maximum value correspondence of integrated value is high gray-scale value straight line place angle for x, near y) each angle integration three angles to utilize the Radon conversion to ask image f

To judge that this angle may be the roughly direction of insulator place bracket in the image, i.e. the roughly direction of insulator;

(4), insulator location: (x y) rotates to be in (3) roughly angle of three of gained respectively with image f

(x y), and carries out following all operations to this three width of cloth image respectively: with image f ' (x to obtain three width of cloth image f ', y) carry out the bent wave conversion of two generations of 64 directions, the matrix of coefficients that obtains is carried out statistical treatment, obtain judgment matrix, judge the insulator position according to the judgment matrix feature;

(5), fine setting insulator angle: relatively the matrix of coefficients of the bent wave conversion of 6 directions is closed in the insulator position, chooses the matrix of coefficients of insulator coefficient of region matrix variance minimum, and with the angle of this matrix of coefficients correspondence original image is rotated to this angle;

(6), judge insulator defective mode kind: for the insulator position that has judged in (4), utilize the curve-like singularity of bent wave conversion detected image of two generations, the insulator system matrix number that has obtained in (4) is added up judgement, judge the situation of insulator porcelain patticoat breakage, the point-like singularity of recycling wavelet transformation detected image, obtain point-like singularity rule situation of change between insulator sheet crack and judge that insulator is mingled with the situation of foreign matter, exports oriented picture and testing result data to the subsequent treatment unit.

2. high ferro contact net insulator defective mode detection method as claimed in claim 1 is characterized in that, the curve-like singularity refers to two dimension or the non-homogeneous zone of higher-dimension image pixel in described (6) step, and there is the phenomenon of uneven distribution in the curve-like pixel.

3. high ferro contact net insulator defective mode detection method as claimed in claim 1 is characterized in that, judges that the concrete grammar of insulator breakage in described (6) step is: to the insulator judgment matrix β that obtains in (4) _ξ, ask β _ξThe mean value of all ordinates is H, supposes that numerical value is n greater than the quantity of the peak value of 1.5H, from left to right crest is carried out numbering from 1 integer, makes the position of its horizontal ordinate be respectively T ₁... T _nThe distance of calculating between all adjacent peak horizontal ordinates is respectively S ₁... S _N-1, calculate S ₁... S _N-1Mean value be S _a, if S _i＞1.5S _aWherein (i=1.2...n) then judges T _iWith T _I+1May there be the breakage of insulator between the insulator magnetic skirt of representative, otherwise, may there be the disrepair phenomenon of insulator.

4. high ferro contact net insulator defective mode detection method as claimed in claim 1 is characterized in that, the point-like singularity refers to that there is the phenomenon of non-uniform Distribution such as sudden change in the distribution of one dimension pixel in described (6); Described concrete determination methods with wavelet transformation location insulator is: the horizontal gray-value image of the insulator of orienting is done the wavelet singular value detect, with the processing of making zero of the horizontal gray-value image minimal value of gained; Suppose total m zero point, calculate the adjacent zeros distance L _j(wherein j=1...m) and band mean distance

The definite threshold value of singular value detection according to small echo is namely worked as

The time (wherein j=1...m), judge that j insulator magnetic skirt may exist and be mingled with the foreign matter defective mode, otherwise, can judge that insulator may not exist and be mingled with the foreign matter phenomenon.

Images