CN102749034A - Railway switch gap offset detection method based on image processing - Google Patents

Abstract

The invention discloses a railway switch gap offset detection method based on image processing, which comprises the following steps of: S11) preprocessing original input images to obtain binary images of the original input images; S12) designing a gap template by adopting a statistic method, matching the binary images of the original input images obtained in the step S11 with the gap template through a template matching algorithm, and finding image areas which are least different from the gap template in the binary images; and S13) conducting image segmentation to the image areas obtained in the step S12, analyzing the segmented image areas, removing the image areas without gaps, finding the image areas with gaps, realizing the accurate positioning of the gaps and calculating the offset of the left and right sides of the gaps. The railway switch gap offset detection method based on image processing has the advantages that the gap offset data can be effectively calculated in real time, the cover is not required to be opened for inspection and the goal of high gap monitoring intuitiveness and accuracy is realized.

Description

Railway switch breach offset checking method based on Flame Image Process

Technical field

The present invention relates to a kind of railway switch breach offset checking method, be specially adapted to the measurement of ZD6 sections way trouble goat breach side-play amount based on Flame Image Process.

Background technology

Along with the develop rapidly of railway construction in China, at a high speed, highdensity driving section constantly increases, for guaranteeing traffic safety, to the utilization quality of track switch and state stability require increasingly high.Railway switch trailed switch machine is the emphasis equipment of realizing normal conversion, guaranteeing traffic safety, and the indication notch side-play amount of goat, the conversion resistance force of track switch, the vibration acceleration of goat etc. are the important parameters of reflection goat operate as normal." railway signal maintenance regulation " regulation: maintenance was no less than 2 times in every month to the track switch conversion equipment.And the timely potential utilization hidden danger of quality of discovering device of simple tour.At present; Domestic track switch conversion equipment breach is detected mainly adopts mechanical contact-type or photodetection formula method to detect, and the inspection of need uncapping mostly is at the scene in the complex environment application process; Have not directly perceived, the more high situation of false alarm rate, result of use is not satisfactory.

Summary of the invention

Technical matters to be solved by this invention provides a kind of railway switch breach offset checking method based on Flame Image Process of the inspection that need not to uncap; This method can be carried out the automatic identification of breach and the measurement of side-play amount in real time effectively, realizes the intuitive and the accuracy of breach monitoring.

For solving the problems of the technologies described above, the technical scheme that the present invention adopted is to disclose a kind of railway switch breach offset checking method based on Flame Image Process, it is characterized in that may further comprise the steps:

S11 carries out the image pre-service to original input picture, obtains the bianry image of original input picture;

S12; Adopt the method design breach template of statistics; The bianry image of the original input picture that step S11 is obtained matees said bianry image and breach template with template matching algorithm, in said bianry image, finds and image-region that breach template difference is minimum;

S13, the said image-region that step S12 is obtained carries out image segmentation, and analyzes cutting apart good image-region; Get rid of unnotched image-region; Find image-region jaggy, realize the accurate location of breach, and calculate the side-play amount of breach the right and left.

As preferably, the pre-service of image described in the step S11 may further comprise the steps:

S211 is a grayscale mode with the rgb color mode-conversion of original image;

S212 carries out binaryzation to the grayscale mode of original image;

S213, filtering noise is removed the small size connected domain, obtains the bianry image of the original input picture behind the filtering noise.

As preferably, adopt the method design breach template of statistics may further comprise the steps described in the step S12:

S311, statistics thousand sheets samples pictures obtains the gap width data;

S312, the gap width data that these samples pictures are obtained are averaged and are obtained first template T ₁

S313, to the gap width data less than first template T ₁The gap width data that obtain of all samples pictures of gap width data average and obtain second template T ₂

S314, to gap width greater than first template T ₁The gap width data that obtain of all samples pictures of gap width data average and obtain the 3rd template T ₃

S315 adjusts three template T ₁, T ₂, T ₃Final width and height.

As preferably, may further comprise the steps with template matching algorithm described in the step S12:

S411 travels through in the bianry image of original input picture with the breach template, finds the subgraph with the bianry image of the minimum original input picture of breach template difference;

S412 asks the similarity of subgraph of bianry image of the original input picture of each breach template coupling corresponding with it, according to the subgraph ordering of similarity order from big to small with the bianry image of the original input picture of each coupling, is designated as P successively _M1, P _M2, P _M3, and record P _M1, P _M2, P _M3Position in original input picture.

Beneficial effect: this method adopts the method for pattern-recognition that the breach image is discerned automatically and detected; Finally obtain the side-play amount of gap position; Can calculate the breach offset data in real time effectively, the inspection that need not to uncap realizes intuitive and accuracy that breach is monitored.

Description of drawings

In conjunction with accompanying drawing, other characteristics of the present invention and advantage can become clearer from the explanation of following preferred implementation of coming by way of example principle of the present invention is made an explanation.

Fig. 1 is a kind of schematic flow sheet of embodiment that the present invention is based on the railway switch breach offset checking method of Flame Image Process;

Fig. 2 the present invention is based on original input picture synoptic diagram in a kind of embodiment of railway switch breach offset checking method of Flame Image Process;

Fig. 3 the present invention is based on the pretreated schematic flow sheet of image in a kind of embodiment of railway switch breach offset checking method of Flame Image Process;

Fig. 4 is the synoptic diagram that the present invention is based on template matching algorithm in a kind of embodiment of railway switch breach offset checking method of Flame Image Process;

Fig. 5 is the another synoptic diagram that the present invention is based on template matching algorithm in a kind of embodiment of railway switch breach offset checking method of Flame Image Process;

Fig. 6 the present invention is based on the schematic flow sheet that side-play amount is calculated in a kind of embodiment of railway switch breach offset checking method of Flame Image Process.

Embodiment

To combine accompanying drawing that embodiment of the present invention is described in detail below:

As shown in Figure 1, based on the concrete realization of the railway switch mouth offset checking method of Flame Image Process can be divided into original input picture, image pre-service, detect gap regions, side-play amount is calculated these several big steps.

S11, the image pre-service

This step is that original input picture is carried out the image pre-service, obtains the bianry image of original input picture.Original input picture refers to railway switch conversion monitoring system through being installed in the miniature digital image sensor in switch indication rod notch zone, and is to the image that the regional image real-time acquisition of switch indication rod notch obtains, as shown in Figure 2.

The image pre-service may further comprise the steps:

S211 is the grayscale mode of original image with the rgb color mode-conversion of original image, and conversion formula is:

Gray=R*0.299+G*0.587+B*0.114 （1）

Wherein, R represents red component, and G represents green component, and B represents blue component.

S212 carries out binaryzation to the grayscale mode of original image.Can adopt maximum variance between clusters to try to achieve the adaptive threshold of the binaryzation of original image gray areas.The binaryzation formula is:

$\mathrm{bw}(x,y)=\left\{\begin{array}{ccc}1& \mathrm{if}& \mathrm{Gray}(x,y)\&GreaterEqual;\mathrm{Th}\\ 0& \mathrm{if}& \mathrm{Gray}(x,y)<\mathrm{Th}\end{array}\right.---\left(2\right)$

Wherein (x y) represents gray-scale map to Gray, and (x y) represents the gray-scale value of the original image after the binaryzation to bw, and Th is a threshold value.

S213, filtering noise is removed the small size connected domain, obtains the bianry image of the original image behind the filtering noise.Remove the small size connected domain; Can adopt 8 neighborhood connected component labeling methods, add up the area of each connected domain, will all compose less than the pixel value of the connected domain of threshold value L is 0; Finally obtain the bianry image of the original image behind the filtering noise, particular flow sheet is as shown in Figure 3.

S12 detects gap regions

This step is the method design breach template that adopts statistics; The bianry image of the original input picture that step S11 is obtained; With template matching algorithm the bianry image and the breach template of original input picture are mated, in the bianry image of original input picture, find and the minimum image-region of breach template difference.

The design of breach template is to adopt statistical method, may further comprise the steps:

S311, statistics thousand sheets samples pictures obtains the gap width data;

S312, the gap width data that these samples pictures the are obtained first mould T by the time that averages ₁

S313, to the gap width data less than the first template T ₁The gap width data that obtain of all samples pictures of gap width data average and obtain the second template T ₂

S314, to gap width greater than the first template T ₁The gap width data that obtain of all samples pictures of gap width data average and obtain the 3rd template T ₂

S315 adjusts three template T ₁, T ₂, T ₃Final width and height; Three templates are referred to as the breach template, and the breach template also is a bianry image, and the breach template is made up of three rectangular block of pixels; The rectangular block of pixels pixel value on the left side is 1 entirely; Middle rectangular block of pixels is a breach, and pixel value is 0 entirely, and the rectangular block of pixels on the right and the rectangular block of pixels on the left side are just the same.

Template matching algorithm may further comprise the steps:

S411 travels through in the bianry image of original input picture with the breach template, finds the subgraph with the bianry image of the minimum original input picture of breach template difference.Concrete steps are: travel through in the bianry image of original input picture with the breach template, find the subgraph with the bianry image of the minimum original input picture of breach template difference.With breach template T ₁, T ₂, T ₃In the bianry image of original image, travel through, when just beginning, the first template T ₁The upper left angle point of bianry image of upper left angle point and original image overlap, take the first template T ₁A subgraph in the bianry image of the original image under covering with first template is compared, and calculates the number of different pixel between them, then the first template T ₁Move to next pixel, carry out same operation, all compared, find out that minimum piece of different pixel up to all positions, be exactly we to seek with the first template T ₁Subgraph in the bianry image of the original image that matees most.Can be imagined as, when just beginning, the first template T ₁Be to be stacked on the bianry image of original image, the alignment of their upper left corner, that piece zone, just subgraph and first template T of at this moment being covered by template on the bianry image of original image ₁Size is identical, calculates the first template T ₁With this piece zone, the just number of both different pixels of subgraph; Then, the monoblock first template T ₁On the bianry image of original image, move, move the position of a point, the subgraph of at this moment getting back calculates the number of different pixel again, moves again, calculates, at last up to the first template T again ₁The position, the bianry image lower right corner of position, the lower right corner and original image overlaps.Then to the second template T ₂With the 3rd template T ₃Carry out identical operations, find the subgraph that matees most separately, each breach template all finds a subgraph that matees most on the bianry image of original image.Three breach templates find three subgraphs that mate most altogether.Idiographic flow is as shown in Figure 4.

Calculate the number of different pixel with following formula:

$D(x,y)=\underset{m=x}{\overset{M+x-1}{\mathrm{\&Sigma;}}}\underset{n=y}{\overset{N+y-1}{\mathrm{\&Sigma;}}}{(P_{\mathrm{xy}}(m,n)-T(m-x,n-y))}^2---\left(3\right)$

Wherein, (x, y) position of the upper left angle point of indication notch template in the bianry image of original image; (coordinate in the bianry image upper left corner of original image is (0 for m, the n) coordinate of represent pixel point on the bianry image of original image; 0), T represents the breach template, and M represents the height of breach template; N represents the width of breach template, P _XyRepresent the subgraph of the bianry image of the original image under the covering of breach template.Because; This moment, the bianry image of breach template and original image all was the bianry image of representing with 0,1; The result of calculation of formula (3) is exactly the number of different pixel of the subgraph of breach template and the breach template bianry image that covers original image down, and when D was 0, the subgraph of the bianry image of the original image under indication notch template and the covering of breach template was identical; D is more little, and expression is coupling more.

S412 asks the similarity of subgraph of bianry image of the original input picture of each breach template coupling corresponding with it, according to the subgraph ordering of similarity order from big to small with the bianry image of the original input picture of each coupling, is designated as P successively _M1, P _M2, P _M3, and record P _M1, P _M2, P _M3Position in the bianry image of image.The formula of similarity degree of subgraph that calculates the bianry image of the original image of each breach template under covering with its breach template of mating most is:

$R_i=\frac{\underset{m=x}{\overset{M_i+x-1}{\mathrm{\&Sigma;}}}\underset{n=y}{\overset{N_i+y-1}{\mathrm{\&Sigma;}}}{T}_i(m-x,n-y)*P_i(m,n)}{\sqrt{\underset{m=x}{\overset{M_i+x-1}{\mathrm{\&Sigma;}}}\underset{n=y}{\overset{N_i+y-1}{\mathrm{\&Sigma;}}}{T}_i(m-x,n-y)*T_i(m-x,n-y)}*\sqrt{\underset{m=x}{\overset{M_i+x-1}{\mathrm{\&Sigma;}}}\underset{n=y}{\overset{N_i+y-1}{\mathrm{\&Sigma;}}}{P}_i(m,n)*P_i(m,n)}}---\left(4\right)$

Wherein, (i=1,2,3), T ₁Represent first template, T ₂Represent second template, T ₃Represent the 3rd template, P _iBe breach template T _iThe subgraph of the bianry image of the original image of behind the bianry image of the whole original image of traversal, finding out that matees most.R _iBig more, represent similarly more, work as R _i=1 o'clock, breach template the and subgraph of the bianry image of the original image of coupling is identical.According to R _iSize the subgraph of the bianry image of the original image of coupling is sorted, making number one of coupling is designated as P successively _M1, P _M2, P _M3Concrete process flow diagram is as shown in Figure 5.

S13, side-play amount is calculated

This step mainly is to obtain P _M1, P _M2, P _M3After, accurate locating notch and calculate the side-play amount of breach the right and left.Concrete steps are:

S131 is to P _MiCarry out the projection of vertical direction, obtain projection value P _CiComputing method are:

$P_{\mathrm{Ci}}\left(n\right)=\underset{m=0}{\overset{H_i-1}{\mathrm{\&Sigma;}}}{P}_{\mathrm{Mi}}(m,n)(i=\mathrm{1,2,3})---\left(5\right)$

The H here _iBe P _MiHighly.

S132 chooses appropriate threshold T _S, to projection value P _CiCarry out binaryzation, obtain the projection value PB after the binaryzation _Ci:

${\mathrm{PB}}_{\mathrm{Ci}}\left(n\right)=\left\{\begin{array}{ccc}1& \mathrm{if}& P_{\mathrm{Ci}}\left(n\right)\&GreaterEqual;T_P\\ 0& \mathrm{if}& P_{\mathrm{Ci}}\left(n\right)<T_P\end{array}\right.---\left(6\right)$

The T here _PCan get 0.5*H _i, H _iBe P _MiHighly.

S133 finds the projection value PB after the binaryzation _CiFirst negative edge and last rising edge, respectively as the position, the left and right sides of breach;

S134, the side-play amount of position, the left and right sides in the bianry image of whole original image of calculating breach is designated as LeftPos and RightPos respectively, and calculates the distance of the side-play amount of breach the right and left, is designated as Length, Length=RightPos-LeftPos;

S135 is if Length is more than or equal to threshold value T _S, then LeftPos and RightPos are exactly final result, and algorithm finishes, if Length is less than threshold value T _S, repeating step S131 to S134 then, but be to P at this moment _M2Operate, if pass through P _M2Data are handled the Length that obtains more than or equal to threshold value T _S, then this time obtaining LeftPos and RightPos is exactly final result, and algorithm finishes, if Length is less than threshold value T _S, then repeating step S131 to S134 is to P at this moment _M3Operate, if pass through P _M3Data are handled the Length that obtains more than or equal to threshold value T _S, LeftPos that then this time obtains and RightPos are exactly final result, and algorithm finishes, if Length is less than threshold value T _S, then final result gets P _M1, P _M2, P _M3Those maximum group data of middle Length.Wherein step S131 to S133 is the process that the image-region that step S12 obtains is cut apart as image.Threshold value T _SSetting be in order to get rid of some special circumstances, such as as gap width during, P greatly to certain program _M1Might be 1 image block for pixel value entirely, this image block be minimum template (the second template T of gap width ₂) the subgraph of the bianry image of the original image of coupling behind the bianry image of the whole original image of traversal, found out.Generally to data P _M1, P _M2After handling, algorithm just can finish.Particular flow sheet is as shown in Figure 6.

Though described embodiment of the present invention in conjunction with the accompanying drawings, those of ordinary skills can make various distortion or modification within the scope of the appended claims.

Claims (4)

1. railway switch breach offset checking method based on Flame Image Process is characterized in that may further comprise the steps:

S11 carries out the image pre-service to original input picture, obtains the bianry image of original input picture;

S12; Adopt the method design breach template of statistics; The bianry image of the original input picture that step S11 is obtained matees said bianry image and breach template with template matching algorithm, in said bianry image, finds and image-region that breach template difference is minimum;

S13, the said image-region that step S12 is obtained carries out image segmentation, and analyzes cutting apart good image-region; Get rid of unnotched image-region; Find image-region jaggy, realize the accurate location of breach, and calculate the side-play amount of breach the right and left.

2. the railway switch breach offset checking method based on Flame Image Process according to claim 1 is characterized in that the image pre-service may further comprise the steps described in the step S11:

S211 is a grayscale mode with the rgb color mode-conversion of original image;

S212 carries out binaryzation to the grayscale mode of original image;

S213, filtering noise is removed the small size connected domain, obtains the bianry image of the original input picture behind the filtering noise.

3. the railway switch breach side-play amount detection algorithm based on Flame Image Process according to claim 1 is characterized in that adopting described in the step S12 method design breach template of statistics may further comprise the steps:

S311, statistics thousand sheets samples pictures obtains the gap width data;

S312, the gap width data that these samples pictures are obtained are averaged and are obtained first template T ₁

S313, to the gap width data less than first template T ₁The gap width data that obtain of all samples pictures of gap width data average and obtain second template T ₂

S314, to gap width greater than first template T ₁The gap width data that obtain of all samples pictures of gap width data average and obtain the 3rd template T ₃

S315 adjusts three template T ₁, T ₂, T ₃Final width and height.

4. the railway switch breach side-play amount detection algorithm based on Flame Image Process according to claim 1 is characterized in that may further comprise the steps with template matching algorithm described in the step S12:

S411 travels through in the bianry image of original input picture with the breach template, finds the subgraph with the bianry image of the minimum original input picture of breach template difference;

S412 asks the similarity of subgraph of bianry image of the original input picture of each breach template coupling corresponding with it, according to the subgraph ordering of similarity order from big to small with the bianry image of the original input picture of each coupling, is designated as P successively _M1, P _M2, P _M3, and record P _M1, P _M2, P _M3Position in original input picture.

Images