CN106274979A - Rail wear automatic detection device - Google Patents

Abstract

Rail wear automatic detection device, belong to detection field, in order to obtain a kind of detection device that can perform Rail Abrasion Detection System method, it is characterized in that: include that a wordline light beam place plane that a word line laser device, ccd image sensor and microprocessor, a word line laser device send and tested Rail Surface are 60 ° of angles, ccd image sensor is positioned at the surface of a wordline light beam place plane, the input of described microprocessor receives the rail image information that ccd image sensor gathers, and performs Rail Abrasion Detection System method.

Description

Rail wear automatic detection device

Technical field

The invention belongs to detection field, relate to a kind of rail wear automatic detection device, swash particularly to based on a wordline Light image process and microprocessor can effectively to detect rail head of rail surface abrasion depth and width one rail wear automatic Detection device.

Background technology

Railway is the large artery trunks of transportation, compares other means of transportation, and heavy haul railway transport is big with freight volume, low cost Feature develops rapidly all over the world.In track equipment, rail is most important building block, directly bears train and carries Lotus also guides wheel to run.The state of the art of rail is the most intact, and can directly affect train safe by the speed of regulation, flat Steady and continual operation.Railway locomotive be by wheel track between frictional force transmit driving force and brake force, and between wheel track Friction then can cause the generation of rail wear.Along with the high speed of locomotive, heavy duty, high density are run, the abrasion of rail will be quick The increase of degree, particularly sharp radius curve outer rail medial surface abrasion are particularly acute.

The detection technique of rail wear have passed through that monocular of conforming to the principle of simplicity measures ruler class tool detection, digitized instrument detection waited Journey.At present, China has the main method such as contact fixture measurement, EDDY CURRENT, optical triangulation in terms of Rail Abrasion Detection System, The experience that testing result is often depending on detecting the attitude of workman and instrument uses, these methods also exist that detection efficiency is low, inspection Survey the most high problems of precision, can not meet the development need of current high speed.Although occurring in that now that a kind of utilization is swashed The detection equipment of light detection rail surface abrasion, but step-by-step movement detection accurately can not be realized, other detection methods are the most simply stopped Stay theoretical research aspect.

Summary of the invention

In order to obtain a kind of detection device that can perform Rail Abrasion Detection System method, the present invention proposes a kind of rail mill Consumption detection device, to perform rail wear method, is characterized in that: include a word line laser device, ccd image sensor and micro- A wordline light beam place plane and tested Rail Surface that processor, a word line laser device send are 60 ° of angles, and ccd image senses Device is positioned at the surface of a wordline light beam place plane, and the input of described microprocessor receives what ccd image sensor gathered Rail image information, and perform Rail Abrasion Detection System method.

Beneficial effect: The present invention gives a kind of Rail Abrasion Detection System device, and laser is angled with guide rail surface Irradiate, for ensureing that two sections of laser images that rail is formed on surface when having abrasion are being perpendicular to the direction of guide rail not at straight line On；60 ° is in order to two sections of images are wearing away one special triangle of formation on section, the right angled triangle of i.e. 60 °, so Convenient calculating wears away width and the degree of depth.

Accompanying drawing explanation

Fig. 1 is the structured flowchart of rail wear automatic detection device described in embodiment 2；

Fig. 2 is without abrasion laser image schematic diagram；

Fig. 3 is for there being abrasion laser image schematic diagram；

Fig. 4 is brightness curve and disk diameter schematic diagram；

Fig. 5 is the labelling schematic diagram of data point and characteristic quantity.

Detailed description of the invention

Embodiment 1: a kind of rail wear automatic testing method, gathers the laser image of rail, and with complete rail laser Light belt image carries out image comparison, to judge to detect whether rail has abrasion, it is judged that rail has abrasion, to the laser figure gathered As carrying out laser image process, described laser image processes and includes Image semantic classification and Edge extraction, and laser image processes After, select and extract the laser image characteristic quantity relevant to rail wear amount, to be calculated rail wear depth and width.Its In: the laser image characteristic quantity that described extraction is relevant to rail wear amount is more than one in following characteristics amount:

1) length l of two sections of straight line portioies of laser image_AAnd l_B；

2) the stand out e of two sections of linear laser images；

3) lengthwise position difference z of two sections of linear laser images；

4) length l of changeover portion between two sections of linear laser images_C；

5) inclination angle theta of changeover portion between two sections of linear laser images；

Abrasion width and the abrasion degree of depth are referred to as the characteristic quantity of rail wear, select above-mentioned in one or more laser images Characteristic quantity, for calculating the depth and width of rail wear；

And when calculating the depth and width of rail wear, and non-selection whole above-mentioned laser image characteristic quantity is counted Calculate, in order to optimize calculating process, select the combination base as the depth and width calculating rail wear of laser image characteristic quantity Plinth calculates data, selects method during this combination to be: first determine the laser image characteristic quantity relevant to this wear characteristics, and from Laser image characteristic quantity selects preferred features amount, calculates the degree of association system of remaining each laser image characteristic quantity and preferred features amount Number, and seeks its meansigma methods, and this meansigma methods is the threshold value beta that characteristic quantity selects, if wherein between certain two laser image feature Absolute value | rTij | >=β of correlation coefficient, these two laser image features are relevant redundancy features, only select one of them conduct The laser image characteristic quantity that rail wear judges.I.e. when selecting for the combination of the laser image characteristic quantity judged, it is assumed that M is The rail have abrasion pinpoints the feature samples set collected, swashing of the reaction abrasion that this set comprises N number of fixing point Light image characteristic quantity, select correlation coefficient as metric parameter, this parameter be characterized by between similarity, two laser figures As feature is relevant redundancy feature, only select one of them laser image characteristic quantity judged as rail wear, be used for finding Can effectively judge the combination of the minimum laser image characteristic quantity of rail wear width and the degree of depth.

As a kind of embodiment, above-mentioned middle correlation coefficient is as metric parameter, to be characterized by the concrete of a dependency Method is: use sets two groups of different laser image features and is respectively as follows: T_i={ t_ik, k=1,2 ..., n} and T_j={ t_jk, k=1, 2 ..., n}, wherein k represents kth test point, and total n test point, then the correlation coefficient of two groups of laser image features defines such as Under:

${\mathrm{rT}}_{ij}=\frac{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}(t_{ik}-\stackrel{\‾}{t_i})(t_{jk}-\stackrel{\‾}{t_j})}{\sqrt{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{(t_{ik}-\stackrel{\‾}{t_i})}^2}\sqrt{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{(t_{jk}-\stackrel{\‾}{t_j})}^2}}$

In formula,WithIt is respectively two stack features T_iAnd T_jMeansigma methods:With

Correlation coefficient r T_ijReflect two stack features T_iAnd T_jDegree of correlation, rT_ijValue for, time negative, representing two features Negative correlation；rT_ijValue be timing, represent two feature positive correlations；Work as rT_ijWhen=0, it is incoherent between two wear characteristics, Work as rT_ijAbsolute value when being closer to 1, the degree of correlation of two laser image features is the highest, and the redundancy of generation is the biggest, in reaction In the laser image characteristic set of rail wear, utilize the correlation coefficient between each laser image characteristic quantity, threshold value beta is set, if The wherein absolute value of the correlation coefficient between certain two laser image feature | rT_ij| >=β, these two laser image features are relevant Redundancy feature, only selects one of them laser image characteristic quantity judged as rail wear.

In another embodiment, the determination method for threshold value beta described above is: special for certain single laser image The amount of levying, chooses it as preferred features, it is judged that remaining laser image characteristic quantity is that the method for the probability of redundancy feature is: determine After preferred features, by calculate obtain in rail wear width and degree of depth associated laser characteristics of image set with preferred features amount it Between correlation coefficient, the average of these group correlation coefficient data is set to threshold value beta, the determination method of threshold value beta is:

$\mathrm{\β}=\frac{1}{c-1}\underset{j=2}{\overset{c-1}{\mathrm{\Σ}}}\left|{\mathrm{rT}}_{1j}\right|$

Wherein: the quantity of the c amount of being characterized in formula, l is the sequence number of first-selected characteristic quantity, and j is the sequence number of alternative features amount.

Thus, above-described embodiment, try to achieve the correlation coefficient between feature to obtain the probability size of redundancy between feature, should The average of group correlation coefficient data is set to threshold value beta, using this threshold value as the foundation of judging characteristic whether redundancy, thus is sentencing During disconnected two feature redundancy, only select a laser image feature as the depth and width calculating abrasion between redundancy feature, To optimize calculating process, obtain minimal features combination with this.

As a kind of embodiment, the concrete open computational methods wearing away depth and width: two sections of line parts of laser image Length l divided_AAs the preferred features of abrasion width detection, lengthwise position difference z of two sections of linear laser images is deep as abrasion The preferred features of degree detection；Two straight line portioies of laser image are obtained by the Calculation of correlation factor of two groups of laser image features Length l_AAnd l_BAs the characteristic quantity of abrasion width detection, lengthwise position difference z of two sections of linear laser images is as abrasion degree of depth inspection The characteristic quantity surveyed, abrasion width calculation formula is:

$W=l_B+\frac{2}{3}|l-(l_A+l_B)|$

Wherein, l is the width not wearing away rail；

Abrasion depth calculation formula is:

V=z tan 60 °

As a kind of embodiment, described Image semantic classification comprises the steps:

First by image gray processing, draw the rectangular histogram of gray level image, find out gray scale and concentrate scope；

Then use following formula, gray level image is carried out grey level enhancement, make image become apparent from；

$y=255\×\frac{(x-a)}{(b-a)}$

Wherein: a, b are respectively the left and right boundary point of gray value integrated distribution in gray level image rectangular histogram, and x, y represent respectively Gray value before and after grey level enhancement.

As a kind of embodiment, the method for described Edge extraction, comprise the steps:

Appoint the medium filtering brightness curve taking a distribution image vegetarian refreshments in the horizontal direction, divide in these curve peak-peak both sides Qu Chu not change maximum continuity point by brightness step, take the spacing of midpoint p and q, p and q of these two groups of continuity points as detection Template diameter；

If the brightness of image is that (i, j), (c, r) as detection template, wherein c is circle to f to take a round s in picture field The heart, its coordinate is (i_c,j_c), r is radius；

Definition s (c, r) in the set of pixel, and remember the brightness of pixel in round s and be:

$X=\underset{(i,j)\∈X}{\mathrm{\Σ}}f(i,j)$

Moving in making the detection template center of circle little scope in the horizontal direction, in calculating each position detection template, each pixel is bright Degree and, should in the range of brightness and maximum template home position, be a Pixel-level roof edge point of this bright wisp, utilize Little square law fitting a straight line, this straight line is a wordline laser image centrage, and described little scope is a some left side centered by the center of circle The image of right each 2 times of radiuses is interval, to obtain length l of two sections of straight line portioies of laser image_AAnd l_B, two sections of linear laser figures Length l of changeover portion between Xiang_C。

The rail laser image of the present embodiment collection is compared with complete image, to judge whether to gather the steel in image There is abrasion in rail, and when being judged as abrasion, selects characteristic quantity to be calculated abrasion further, first qualitatively judge abrasion, then The thinking of quantitative Analysis abrasion depth and width, and during calculating, characteristic quantity is selected, to optimize the abrasion degree of depth Calculating process with width.

Embodiment 2: supplementing as embodiment 1 technical scheme, or as a kind of individually embodiment: abrasion mainly go out The now head of rail, abrasion include end face abrasion and side wear, the two numerical value must be detected during detection simultaneously, combine Close the wear intensity judging rail.The present embodiment utilizes high intensity narrow beam one word laser beam, laser instrument and a wordline light beam institute Being 60 ° of angles in plane and tested Rail Surface, high-resolution Array CCD sensor is positioned at the surface shooting of laser image Laser image.Bending is occurred in that, the position occurred by bending point and degree of crook at the Rail Surface light beam image having abrasion Determine width and the degree of depth of rail wear.

Rail wear automatic detection device includes: a word line laser device, ccd image sensor, microprocessor, execution list Unit, display and acousto-optic warning unit and interface unit.Ccd image sensor gathers laser image, the image information obtained It is transferred to microprocessor be analyzed processing, extracts image border and center fitting a straight line, form complete rail laser Light belt image outline, converts image information into rail profile parameter, the characteristic quantity of storage rail profile, and joins with complete rail Number is compared, it is judged that whether rail exists abrasion.Abrasion are not had to proceed subsequent point detection；There are abrasion, further determine that Abrasion loss, including the depth and width of abrasion.Performance element accepts the control signal of microprocessor, controls the traveling of detection device Direction and speed, regulation ccd image sensor orientation, the outfan of microprocessor respectively with LCD display and sound and light alarm System connects, and LCD display is for showing current location and the wear intensity of rail, and acoustooptic alarm system is used for pointing out rail to work as There are abrasion in front position, needs to repair.Interface unit is for exchanging information with host computer, and host computer can be further to abrasion position The further fine processing of image put, determines accurately abrasion loss.

The processing stage that Image semantic classification being the early stage of laser image edge extracting, first by image gray processing, draw gray scale The rectangular histogram of image, finds out gray scale and concentrates scope, utilizes formula (1) (gray value during wherein a, b are respectively gray level image rectangular histogram The left and right boundary point of integrated distribution, x, y represent the gray value before and after grey level enhancement respectively) gray level image is carried out grey level enhancement, Image is made to become apparent from.

$y=255\×\frac{(x-a)}{(b-a)}---\left(1\right)$

The rim detection of one wordline laser image uses " ridge-shaped " edge detection method.Based on single pixel brightness Edge detection method noise resisting ability is poor, in order to reduce the interference of picture noise, pixel brightness each in a certain region and As " ridge-shaped " edge distinguishing rule.Owing to circle has each to same tropism, do not affected by ridge-shaped edge direction, therefore, The present invention uses plate way " ridge-shaped " edge detection method.By disk detection template suitable for size at a wordline laser image Move in the certain limit of both sides, when the graded of the brightness sum of pixel each in template meets certain requirements, template Central point is ridge-shaped marginal point.

Appoint the medium filtering brightness curve taking a distribution image vegetarian refreshments in the horizontal direction, divide in these curve peak-peak both sides Qu Chu not change maximum continuity point by brightness step, take the spacing of midpoint p and q, p and q of these two groups of continuity points as detection Template diameter, as shown in Figure 3.

If the brightness of image is that (i, j), (c, r) as detection template, wherein c is circle to f to take a round s in picture field The heart, its coordinate is (i_c,j_c), r is radius.Definition s (c, r) in the set of pixel:

And remember the brightness of pixel in round s and be:

$X=\underset{(i,j)\∈X}{\mathrm{\Σ}}f(i,j)---\left(2\right)$

Moving in making the detection template center of circle little scope in the horizontal direction, in calculating each position detection template, each pixel is bright Degree and, should in the range of brightness and maximum template home position, be a Pixel-level roof edge point of this bright wisp.Utilize Little square law fitting a straight line, this straight line is a wordline laser image centrage.The marginal point detected and the straight line of matching are such as Shown in Fig. 4.

Extract further and rail wear amount associated laser image feature amount, including the system of selection of characteristic quantity and threshold value Determine.

The width of the rail wear that the invention mainly relates to and the degree of depth, may determine that steel by the degree of crook of laser image The width of rail abrasion and the degree of depth, have following characteristic quantity to can be used for selecting:

1) length l of two sections of straight line portioies of laser image_AAnd l_B；

2) the stand out e of two sections of linear laser images；

3) lengthwise position difference z of two sections of linear laser images；

4) length l of changeover portion between two sections of linear laser images_C；

5) inclination angle theta of changeover portion between two sections of linear laser images.

One or more characteristic quantity can be selected for judging the depth and width of rail wear, selecting for judgement During the combination of characteristic quantity, it is desirable to inhomogeneity feature has marked difference, it is to avoid redundancy feature interference judges.Assume that M is to have Pinpointing, on the rail of abrasion, the feature samples set collected, this set comprises the wear characteristics of n fixing point.Select degree of association Coefficient as metric parameter, this parameter can be characterized by between similarity, can effectively judge rail wear width for searching The combination of the minimal features amount of degree and the degree of depth.If two groups of different wear characteristics are respectively as follows: T_i={ t_ik, k=1,2 ..., n} and T_j={ t_jk, k=1,2 ..., n}, wherein k represents kth test point, total n test point, then the correlation coefficient of two stack features It is defined as follows:

${\mathrm{rT}}_{ij}=\frac{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}(t_{ik}-\stackrel{\‾}{t_i})(t_{jk}-\stackrel{\‾}{t_j})}{\sqrt{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{(t_{ik}-\stackrel{\‾}{t_i})}^2}\sqrt{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{(t_{jk}-\stackrel{\‾}{t_j})}^2}}---\left(3\right)$

In formula,WithIt is respectively two stack features T_iAnd T_jMeansigma methods:With

Correlation coefficient r T_ijReflect two stack features T_iAnd T_jDegree of correlation, rT_ijValue for, time negative, representing two features Negative correlation；Value is timing, represents two feature positive correlations.Work as rT_ijWhen=0, it is incoherent between two features.Then rT is worked as_ij Absolute value when being closer to 1, represent that the degree of correlation of two features is the highest, the most issuable redundancy is the biggest.

In the characteristic set of rail wear, utilize the correlation coefficient between each characteristic quantity, threshold value beta be set, if wherein certain The absolute value of the correlation coefficient between two features | rT_ij| >=β, illustrate that the two feature is relevant redundancy feature, can only select One of them characteristic quantity judged as rail wear.

For certain single features, it is the biggest with the direct relation of rail wear depth and width, determination methods is the simplest, uses The highest in the feasibility judging abrasion loss, selected probability is the biggest, and selected feature is as preferred features.Judge a certain Being characterized as the probability of redundancy feature, according to itself and the dependency of preferred features, dependency is the highest, then become relevant redundancy feature Probability the biggest.After determining preferred features, by calculate obtain in rail wear width and degree of depth correlated characteristic set with Correlation coefficient between preferred features amount, is set to threshold value beta by the average of these group data, as shown in formula (4):

$\mathrm{\β}=\frac{1}{c-1}\underset{j=2}{\overset{c-1}{\mathrm{\Σ}}}\left|{\mathrm{rT}}_{1j}\right|---\left(4\right)$

After determining characteristic quantity, calculate the rail wear amount of detection position, store and show, have when transfiniting abrasion and start Acoustic-optic alarm.

Interface unit is for exchanging information with host computer in off-line case, and host computer can be further to abrasion position The further fine processing of image, determines accurately abrasion loss.Owing to using technique scheme, a kind of steel that the present embodiment provides Rail abrasion automatic detection device has such beneficial effect, due to the method using image procossing, in the control of microprocessor Under, the control of PC, device can run under the setting of operator automatically.Equipment has certain integrity and actual effect Property, it is simple to the use of testing staff, the most simple to operate, testing result is accurate, and manufacturing cost is low.

Embodiment 3: a kind of Rail Abrasion Detection System device, including a word line laser device, ccd image sensor and micro-process A wordline light beam place plane and tested Rail Surface that device, a word line laser device send are 60 ° of angles, ccd image sensor position In the surface of a wordline light beam place plane, the input of described microprocessor receives the rail that ccd image sensor gathers Image information, and perform Rail Abrasion Detection System method, this detection method can be the detection method in embodiment 1.Laser with lead The angled irradiation of track surface, for ensureing that two sections of laser images that rail is formed on surface when having abrasion are being perpendicular to guide rail Direction is the most point-blank；60 ° is in order to two sections of images are wearing away one special triangle of formation on section, i.e. one 60 ° Right angled triangle, so convenient calculate abrasion width and the degree of depth.

As a kind of embodiment, detection device also includes LCD display, acoustooptic alarm system and interface unit, described micro- The outfan of processor is connected with LCD display, acoustooptic alarm system, interface unit respectively, and described interface unit is again with upper Machine connects with the information between exchange microprocessor and host computer.

As a kind of embodiment, described detection device also includes performance element, and performance element receives the control of microprocessor Signal, performance element includes the first servomotor and the second servomotor, and microprocessor connects the first servomotor to control steel Rail Abrasion detecting system direct of travel and speed, microprocessor connects the second servomotor, and the second servomotor connects ccd image Sensor, the second servomotor is to control the orientation of ccd image sensor.I.e. sent control signal by processor, control to perform The servomotor of unit, by the fixing velocity form set time, determines next test point, and finely tunes ccd image by motor The angle of sensor, it is ensured that laser beam becomes 60 ° of angles with Rail Surface.

As a kind of embodiment: described microprocessor includes:

Abrasion judge module, gathers the laser image of rail, and carries out image comparison with complete rail laser light belt image, To judge to detect whether rail has abrasion；

Abrasion computing module, it is judged that rail has abrasion, selects and extracts the laser image spy relevant to rail wear amount The amount of levying, to be calculated the rail wear degree of depth and/or width.

As a kind of embodiment, described detection device, also include:

Characteristic quantity selects module, and selecting the described laser image characteristic quantity relevant to rail wear amount is in following characteristics amount More than one:

1) length l of two sections of straight line portioies of laser image_AAnd l_B；

2) the stand out e of two sections of linear laser images；

3) lengthwise position difference z of two sections of linear laser images；

4) length l of changeover portion between two sections of linear laser images_C；

5) inclination angle theta of changeover portion between two sections of linear laser images；

The process of described characteristic quantity selection module selection characteristic quantity is: the characteristic quantity of rail wear includes wearing away width and mill The consumption degree of depth, selects one or more laser image characteristic quantity, for calculating the degree of depth and/or the width of rail wear, is selecting use When the combination of the laser image characteristic quantity judged, it is assumed that M is to pinpoint the feature samples collected on the rail have abrasion Set, this set comprises the laser image characteristic quantity of the reaction abrasion of N number of fixing point, selects correlation coefficient as tolerance ginseng Number, two laser image features are relevant redundancy features, only select one of them laser image judged as rail wear Characteristic quantity, effectively to be judged the combination of the minimum laser image characteristic quantity of rail wear width and the degree of depth, and with this feature Amount combination calculation obtains a certain characteristic quantity of rail wear.

Two laser image features of described judgement are that the method for relevant redundancy feature is: select from laser image characteristic quantity Preferred features amount, calculates the correlation coefficient of remaining each laser image characteristic quantity and preferred features amount, and seeks its meansigma methods, and this is put down Average is the threshold value beta that laser image characteristic quantity selects, if the correlation coefficient wherein between certain two laser image feature is absolute Value | rTij | >=β, then these two laser image features are relevant redundancy features, only select one of them to judge as rail wear Laser image characteristic quantity.

The method calculating described correlation coefficient is:

If two groups of different laser image features are respectively as follows: T_i={ t_ik, k=1,2 ..., n} and T_j={ t_jk, k=1, 2 ..., n}, wherein k represents kth test point, and total n test point, then the correlation coefficient of two groups of laser image features defines such as Under:

${\mathrm{rT}}_{ij}=\frac{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}(t_{ik}-\stackrel{\‾}{t_i})(t_{jk}-\stackrel{\‾}{t_j})}{\sqrt{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{(t_{ik}-\stackrel{\‾}{t_i})}^2}\sqrt{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{(t_{jk}-\stackrel{\‾}{t_j})}^2}}$

In formula,WithIt is respectively two stack features T_iAnd T_jMeansigma methods:With

Correlation coefficient r T_ijReflect two stack features T_iAnd T_jDegree of correlation, rT_ijValue for, time negative, representing two features Negative correlation；rT_ijValue be timing, represent two feature positive correlations；Work as rT_ijWhen=0, it is incoherent between two wear characteristics, Work as rT_ijAbsolute value when being closer to 1, the degree of correlation of two laser image features is the highest, and the redundancy of generation is the biggest；

The method calculating threshold value beta is:

$\mathrm{\β}=\frac{1}{c-1}\underset{j=2}{\overset{c-1}{\mathrm{\Σ}}}\left|{\mathrm{rT}}_{1j}\right|$

Wherein: the quantity of the c amount of being characterized in formula, l is the sequence number of first-selected characteristic quantity, and j is the sequence number of alternative features amount.

Length l of two sections of straight line portioies of laser image_AAs the preferred features of abrasion width detection, two sections of linear lasers Lengthwise position difference z of image is as the preferred features of abrasion depth detection；Calculation of correlation factor by two groups of laser image features Obtain length l of two straight line portioies of laser image_AAnd l_BAs the characteristic quantity of abrasion width detection, two sections of linear laser figures Lengthwise position difference z of picture as the characteristic quantity of abrasion depth detection, abrasion width calculation formula is:

$W=l_B+\frac{2}{3}|l-(l_A+l_B)|$

Wherein, l is the width not wearing away rail；

Abrasion depth calculation formula is:

V=z tan 60 °.

The laser image of the rail gathered, before extracting the laser image characteristic quantity relevant to rail wear amount, has sharp The step that light image processes, described laser image processes and includes Image semantic classification and Edge extraction；

Described Image semantic classification comprises the steps:

First by image gray processing, draw the rectangular histogram of gray level image, find out gray scale and concentrate scope；

Then use following formula, gray level image is carried out grey level enhancement, make image become apparent from；

$y=255\×\frac{(x-a)}{(b-a)}$

Wherein: a, b are respectively the left and right boundary point of gray value integrated distribution in gray level image rectangular histogram, and x, y represent respectively Gray value before and after grey level enhancement；

The method using described Edge extraction, comprises the steps:

Appoint the medium filtering brightness curve taking a distribution image vegetarian refreshments in the horizontal direction, divide in these curve peak-peak both sides Qu Chu not change maximum continuity point by brightness step, take the spacing of midpoint p and q, p and q of these two groups of continuity points as detection Template diameter；

If the brightness of image is that (i, j), (c, r) as detection template, wherein c is circle to f to take a round s in picture field The heart, its coordinate is (i_c,j_c), r is radius；

Definition s (c, r) in the set of pixel, and remember the brightness of pixel in round s and be:

$X=\underset{(i,j)\∈X}{\mathrm{\Σ}}f(i,j)$

Moving in making the detection template center of circle little scope in the horizontal direction, in calculating each position detection template, each pixel is bright Degree and, should in the range of brightness and maximum template home position, be a Pixel-level roof edge point of this bright wisp, utilize Little square law fitting a straight line, this straight line is a wordline laser image centrage, and described little scope is a some left side centered by the center of circle The image of right each 2 times of radiuses is interval, to obtain length l of two sections of straight line portioies of laser image_AAnd l_B, two sections of linear laser figures Length l of changeover portion between Xiang_C。

The above, only the invention preferably detailed description of the invention, but the protection domain of the invention is not Being confined to this, any those familiar with the art is in the technical scope that the invention discloses, according to the present invention The technical scheme created and inventive concept thereof in addition equivalent or change, all should contain the invention protection domain it In.

Claims (10)

1. a Rail Abrasion Detection System device, it is characterised in that include a word line laser device, ccd image sensor and micro-process A wordline light beam place plane and tested Rail Surface that device, a word line laser device send are 60 ° of angles, ccd image sensor position In the surface of a wordline light beam place plane, the input of described microprocessor receives the rail that ccd image sensor gathers Image information, and perform Rail Abrasion Detection System method.

2. Rail Abrasion Detection System device as claimed in claim 1, it is characterised in that also include LCD display, sound and light alarm system System and interface unit, the outfan of described microprocessor is connected with LCD display, acoustooptic alarm system, interface unit respectively, institute State interface unit to be connected with host computer again with the information between exchange microprocessor and host computer.

3. Rail Abrasion Detection System system as claimed in claim 1 or 2, it is characterised in that described detecting system also includes performing Unit, performance element receives the control signal of microprocessor, and performance element includes the first servomotor and the second servomotor, micro- Processor connects the first servomotor and watches with control Rail Abrasion Detection System system direct of travel and speed, microprocessor connection second Taking motor, the second servomotor connects ccd image sensor, and the second servomotor is to control the orientation of ccd image sensor.

4. Rail Abrasion Detection System device as claimed in claim 1, it is characterised in that described microprocessor includes: abrasion judge Module, gathers the laser image of rail, and carries out image comparison with complete rail laser light belt image, to judge that detection rail is No have abrasion；

Abrasion computing module, it is judged that rail has abrasion, selects and extracts the laser image characteristic quantity relevant to rail wear amount, To be calculated the rail wear degree of depth and/or width.

5. Rail Abrasion Detection System device as claimed in claim 1, it is characterised in that also include:

Characteristic quantity selects module, and selecting the described laser image characteristic quantity relevant to rail wear amount is in following characteristics amount More than Zhong:

1) length l of two sections of straight line portioies of laser image_AAnd l_B；

2) the stand out e of two sections of linear laser images；

3) lengthwise position difference z of two sections of linear laser images；

4) length l of changeover portion between two sections of linear laser images_C；

5) inclination angle theta of changeover portion between two sections of linear laser images.

6. Rail Abrasion Detection System device as claimed in claim 5, it is characterised in that described characteristic quantity selects module to select feature The process of amount is: the characteristic quantity of rail wear includes wearing away width and the abrasion degree of depth, selects one or more laser image feature Amount, for calculating the degree of depth and/or the width of rail wear, when selecting for the combination of the laser image characteristic quantity judged, false If M is to pinpoint the feature samples set collected on the rail have abrasion, this set comprises the reaction abrasion of N number of fixing point Laser image characteristic quantity, select correlation coefficient as metric parameter, two laser image features are relevant redundancy features, Only select one of them laser image characteristic quantity judged as rail wear, effectively to be judged that rail wear width is with deep The combination of the minimum laser image characteristic quantity of degree, and a certain characteristic quantity of rail wear is obtained with this feature amount combination calculation.

7. Rail Abrasion Detection System device as claimed in claim 6, it is characterised in that two laser image features of described judgement are The method of relevant redundancy feature is: selects preferred features amount from laser image characteristic quantity, calculates remaining each laser image feature Amount and the correlation coefficient of preferred features amount, and seek its meansigma methods, this meansigma methods is the threshold value beta that laser image characteristic quantity selects, If wherein absolute value | rTij | >=β of the correlation coefficient between certain two laser image feature, then these two laser image features It is relevant redundancy feature, only selects one of them laser image characteristic quantity judged as rail wear.

Rail Abrasion Detection System device the most as claimed in claims 6 or 7, it is characterised in that calculate the side of described correlation coefficient Method is:

If two groups of different laser image features are respectively as follows: T_i={ t_ik, k=1,2 ..., n} and T_j={ t_jk, k=1,2 ..., N}, wherein k represents kth test point, total n test point, then the correlation coefficient of two groups of laser image features is defined as follows:

${\mathrm{rT}}_{ij}=\frac{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}\left(t_{ik}-\stackrel{\‾}{t_i}\right)\left(t_{jk}-\stackrel{\‾}{t_j}\right)}{\sqrt{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left(t_{ik}-\stackrel{\‾}{t_i}\right)}^2}\sqrt{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{\left(t_{jk}-\stackrel{\‾}{t_j}\right)}^2}}$

In formula,WithIt is respectively two stack features T_iAnd T_jMeansigma methods:With

Correlation coefficient r T_ijReflect two stack features T_iAnd T_jDegree of correlation, rT_ijValue for, time negative, representing two feature negatives Close；rT_ijValue be timing, represent two feature positive correlations；Work as rT_ijWhen=0, it is incoherent between two wear characteristics, when rT_ijAbsolute value when being closer to 1, the degree of correlation of two laser image features is the highest, and the redundancy of generation is the biggest；

The method calculating threshold value beta is:

$\mathrm{\β}=\frac{1}{c-1}\underset{j=2}{\overset{c-1}{\Σ}}\left|{\mathrm{rT}}_{1j}\right|$

Wherein: the quantity of the c amount of being characterized in formula, l is the sequence number of first-selected characteristic quantity, and j is the sequence number of alternative features amount.

9. Rail Abrasion Detection System device as claimed in claim 8, it is characterised in that the length of two sections of straight line portioies of laser image Degree l_AAs the preferred features of abrasion width detection, lengthwise position difference z of two sections of linear laser images is as abrasion depth detection Preferred features；Length l of two straight line portioies of laser image is obtained by the Calculation of correlation factor of two groups of laser image features_A And l_BAs the characteristic quantity of abrasion width detection, lengthwise position difference z of two sections of linear laser images is as abrasion depth detection Characteristic quantity, abrasion width calculation formula is:

$W=l_B+\frac{2}{3}|l-\left(l_A+l_B\right)|$

Wherein, l is the width not wearing away rail；

Abrasion depth calculation formula is:

V=z tan 60 °.

10. Rail Abrasion Detection System device as claimed in claim 4, it is characterised in that the laser image of the rail of collection, is carrying Before taking the laser image characteristic quantity relevant to rail wear amount, having the step that laser image processes, described laser image processes Including Image semantic classification and Edge extraction；

Described Image semantic classification comprises the steps:

First by image gray processing, draw the rectangular histogram of gray level image, find out gray scale and concentrate scope；

Then use following formula, gray level image is carried out grey level enhancement, make image become apparent from；

$y=255\×\frac{\left(x-a\right)}{\left(b-a\right)}$

Wherein: a, b are respectively the left and right boundary point of gray value integrated distribution in gray level image rectangular histogram, x, y represent gray scale respectively Gray value before and after enhancing；

The method using described Edge extraction, comprises the steps:

Appoint the medium filtering brightness curve taking a distribution image vegetarian refreshments in the horizontal direction, take respectively in these curve peak-peak both sides Go out the continuity point that brightness step change is maximum, take the spacing of midpoint p and q, p and q of these two groups of continuity points as detection template Diameter；

If the brightness of image be f (i, j), take in picture field a round s (c, r) as detection template, wherein c is the center of circle, its Coordinate is (i_c,j_c), r is radius；

Definition s (c, r) in the set of pixel, and remember the brightness of pixel in round s and be:

$X=\underset{(i,j)\∈X}{\Σ}f(i,j)$

Move in making the detection template center of circle little scope in the horizontal direction, calculate each pixel intensity in the detection template of each position With, in the range of being somebody's turn to do, brightness and maximum template home position, be a Pixel-level roof edge point of this bright wisp, utilizes minimum Square law fitting a straight line, this straight line is a wordline laser image centrage, and described little scope is about putting centered by the center of circle The image of each 2 times of radiuses is interval, to obtain length l of two sections of straight line portioies of laser image_AAnd l_B, two sections of linear laser images Between length l of changeover portion_C。