CN108917591A - Rail profile autoegistration method and device under a kind of dynamic environment - Google Patents
Rail profile autoegistration method and device under a kind of dynamic environment Download PDFInfo
- Publication number
- CN108917591A CN108917591A CN201810392766.7A CN201810392766A CN108917591A CN 108917591 A CN108917591 A CN 108917591A CN 201810392766 A CN201810392766 A CN 201810392766A CN 108917591 A CN108917591 A CN 108917591A
- Authority
- CN
- China
- Prior art keywords
- roundlet
- circle
- center
- profile
- fitting
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B11/00—Measuring arrangements characterised by the use of optical techniques
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01B—MEASURING LENGTH, THICKNESS OR SIMILAR LINEAR DIMENSIONS; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS
- G01B11/00—Measuring arrangements characterised by the use of optical techniques
- G01B11/24—Measuring arrangements characterised by the use of optical techniques for measuring contours or curvatures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
Abstract
The present invention provides the rail profile autoegistration method and device under a kind of dynamic environment, the method includes：Profile of steel rail optical strip image is acquired, the light stripe centric line in profile optical strip image is extracted；Light stripe centric line is coordinately transformed, multiple space profiles points are obtained；Optimal fitting order value is calculated according to multiple space profiles points；Multiple space profiles points are fitted using optimal fitting order value, obtain contour fitting curve；Contour fitting curve is divided, multiple sections are obtained；The curvature entropy for calculating different sections on the contour fitting curve, obtains corresponding target interval when curvature entropy maximum；The starting point of target interval is as roundlet starting point, and the interval width of target interval is as segmentation width；Roundlet region is determined on contour fitting curve according to roundlet starting point and the segmentation width, and determines the roundlet center of circle according to roundlet region；The roundlet center of circle is aligned with the standard center of circle, so that contour fitting curve is registrated with standard profile.
Description
Technical field
The present invention relates to rail profile autoregistration field, the rail profile under espespecially a kind of dynamic environment is matched automatically
Quasi- method and device.
Background technique
Rail is the main carriers of railway transportation, it is directly supported and the wheel of pilot engine vehicle moves ahead.With train
High speed, heavy duty and high density operation, the abrasion problems of rail become increasingly conspicuous[1].For a long time, for the detection of rail wear,
Railway department mostly uses the mode of Special caliper to carry out artificial regular sampling Detection, this kind of mode low efficiency, precision due to
It manually puts and deviation of reading is difficult to ensure.In recent years, it with the fast development of Laser video camera technology, is surveyed based on structure light vision
The Rail Abrasion Detection System technology of amount is widely used, and domestic and international associated mechanisms develop a plurality of types of vehicle-mounted dynamic steel
Rail measurement of wear system.The technology obtains the high-resolution laser image of rail profile section using vehicle-mounted laser camera assembly,
And by the methods of image procossing, structural light measurement, shape registration, overstep the limit shape and datum mark are identified, then wide with respective standard rail
Datum mark carry out alignment comparison, obtain the wearing valve of measured rail profile.
During measurement of rail wear, choose rail profile that accurately and effectively system survey by benchmark match point and
The alignment of standard rail profile, this process are referred to as the registration of profile, define according to rail wear calculate related wearing valve later.
Curvature threshold split plot design is using curvature as the shape feature of each line segment of contour curve, using fixed curvature threshold come segmentation contour
Each section of circular arc on curve, and radius of curvature is selected to be fitted the benchmark that the center of circle is aligned as profile for the circle segment of 20mm
With point.This method data processing amount is small, can be completed quickly and effectively the real-time measurement to abrasion.Closest approach iteration (ICP,
Iterative Closest Point) algorithm determine web of the rail profile point from measurement coordinate system to Track desigh coordinate system geometry
Correction parameter realizes reference measurement and measurement of wear only with single-sensor.This method measurement reproducibility is high, and is implemented as
This is low.In order to improve efficiency of algorithm, iterative search, which is limited in lesser regional area, to be carried out.
And under train travel environment, due to all by random noise, calibrated error, vehicle disturbance and acquisition equipment itself etc.
There are a large amount of uncertain noise jammings in the outline data of actual measurement in multifactor influence.Therefore, by fixed curvature value point
It is unworkable for cutting roundlet.And ICP algorithm is influenced vulnerable to the noise data on the web of the rail, especially when there are dirty soil on the web of the rail
Etc. convexs spot interference when, measurement accuracy has biggish loss.If furthermore fierce shaking, too small region of search occur for car body
It is easy that iterative solution is made to fall into local optimum, obtains the correction parameter of mistake.
Summary of the invention
In order to solve at present, rail registration technique is asked vulnerable to noise jamming, registration inaccuracy etc. under dynamic driving states
Topic, the embodiment of the present invention provide the rail profile autoegistration method and device under a kind of dynamic environment, the method packet
It includes：
Profile of steel rail optical strip image is acquired, the light stripe centric line in the profile optical strip image is extracted；
The light stripe centric line is coordinately transformed, multiple space profiles points are obtained；
Optimal fitting order value is calculated according to the multiple space profiles point；
The multiple space profiles point is fitted using the optimal fitting order value, obtains contour fitting curve；
The contour fitting curve is divided, multiple sections are obtained；
The curvature entropy for calculating different sections on the contour fitting curve, obtains corresponding target when the curvature entropy maximum
Section；The starting point of the target interval is as roundlet starting point, and the interval width of the target interval is as segmentation width；
According to the roundlet starting point and the segmentation width on the contour fitting curve determination roundlet region, and according to
The roundlet center of circle is determined according to the roundlet region；
The roundlet center of circle is aligned with the standard center of circle, so that the contour fitting curve is registrated with standard profile.
The embodiment of the present invention also provides the rail profile autoregistration device under a kind of dynamic environment, described device packet
It includes：
Acquisition module extracts the light stripe centric line in the profile optical strip image for acquiring profile of steel rail optical strip image；
Coordinate transformation module obtains multiple space profiles points for being coordinately transformed to the light stripe centric line；
Optimal order computing module, for optimal fitting order value to be calculated according to the multiple space profiles point；
Curve fitting module, for being fitted using the optimal fitting order value to the multiple space profiles point,
Obtain contour fitting curve；
Interval division module obtains multiple sections for dividing to the contour fitting curve；
Divide module, for calculating the curvature entropy in different sections on the contour fitting curve, obtains the curvature entropy most
Corresponding target interval when big；The starting point of the target interval is as roundlet starting point, the interval width of the target interval
As segmentation width；
Center of circle fitting module is used for according to the roundlet starting point and the segmentation width in the contour fitting curve
It determines roundlet region, and determines the roundlet center of circle according to the roundlet region；
Registration module, for the roundlet center of circle to be aligned with the standard center of circle, so that the contour fitting curve and standard
Profile registration.
The present invention utilizes the calculating of Optimal order, is accurately fitted the contour curve of rail, and by contour fitting song
Line carries out division and to each interval computation curvature entropy marked off, accurately to determine the roundlet area on contour fitting curve
Domain, and the roundlet center of circle in roundlet region is aligned with the standard center of circle, it is achieved in the autoregistration of rail profile.In the present invention
Method for registering can effectively overcoming noise interference, registration accuracy is high and speed is fast, can quickly and accurately complete rail profile and match
It is quasi-.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, embodiment will be described below
Needed in attached drawing be briefly described, it should be apparent that, the accompanying drawings in the following description is only of the invention some
Embodiment for those of ordinary skill in the art without any creative labor, can also be according to these
Attached drawing obtains other attached drawings.
Fig. 1 is rail profile autoregistration schematic diagram；
Fig. 2A, Fig. 2 B and Fig. 2 C are rail profile structure and each section curvature distribution schematic diagram；
Fig. 3 is the flow chart of the rail profile autoegistration method under a kind of dynamic environment of the embodiment of the present invention；
Fig. 4 A and Fig. 4 B are that profile of steel rail of embodiment of the present invention optical strip image and its corresponding light stripe centric line and space are taken turns
Exterior feature point schematic diagram；
Fig. 5 A and Fig. 5 B are the fitting result schematic diagram of the space profiles point of difference of embodiment of the present invention order；
Fig. 6 A, Fig. 6 B and Fig. 6 C are that the truncation residual error statistic histogram of 4 rank of the embodiment of the present invention and the fitting of 6 ranks and 6 ranks are intended
The truncation residual error statistic histogram and normal distribution of conjunction；
Fig. 7 is the curvature distribution schematic diagram of difference of embodiment of the present invention curve；
Fig. 8 A and Fig. 8 B are that schematic diagram is searched in curvature of embodiment of the present invention Entropy Changes distribution map and maximum curvature entropy section；
Fig. 9 A and Fig. 9 B are roundlet of embodiment of the present invention region segmentation and roundlet center of circle fitting result schematic diagram；
Figure 10 A and Figure 10 B are roundlet of the embodiment of the present invention center of circle and standard center of circle schematic diagram；
Figure 11 is registrated schematic diagram with nominal contour for contour fitting of embodiment of the present invention curve；
Figure 12 A and Figure 12 B are the vertical mill and side grinding numerical statistic result figure of static state of embodiment of the present invention duplicate measurements；
Figure 13 is the radius change statistical results chart that roundlet is fitted in dynamic measurement of the embodiment of the present invention；
Figure 14 is the rail profile autoregistration apparatus structure schematic diagram under a kind of dynamic environment of the embodiment of the present invention.
Specific embodiment
The embodiment of the present invention provides the rail profile autoegistration method and device under a kind of dynamic environment.
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
It is as shown in Figure 1 rail profile autoregistration schematic diagram, rail profile automatically configures first using line-structured light
Vision sensing equipment scans the profile optical strip image that rail profile generates rail；Then Refinement operation is carried out to image, further
Extract the center line of striation；Later, two half sections of profile are synthesized by a tunneling boring profile by coordinate transform, it is specific to grasp
It is to be sat the two dimensional image of the light stripe centric line of extraction according to the parameter matrix for the Laser video camera component demarcated in advance as method
Mark is converted to three dimensional physical coordinate to get to the true three-dimension profile of rail section；Finally, being extracted on the rail exterior feature of the left and right sides respectively
The circular arc sector that radius of curvature is 20mm is fitted two centers of circle as datum mark and carries out spatial registration, by the three-D profile of actual measurement and
Standard rail profile is unified to being compared under the same space coordinate system, calculate measurement point and the match point on corresponding standard gauge it
Between Euclidean distance, obtain the wearing valve of each point on rail profile.Since rail head has abrasion in use, so uncomfortable
Cooperation is with reference of reference, and the web of the rail and rail bottom state are more stable, so choosing the web of the rail and rail bottom junction radius is 20mm's
Two centers of circle of circular fitting carry out rail exterior feature alignment as datum mark.
If Fig. 2A, Fig. 2 B and Fig. 2 C show rail profile structure and each section curvature distribution schematic diagram, as shown in Figure 2 A
For the contour structure figure of standard 60kg/m rail, contour curve is made of multistage straight line and circular arc.Wherein, the web of the rail is by one section of radius
The circular arc DE composition that circular arc C D and radius for 400mm are 20mm.The key of contour registration is from web of the rail data accurately
Extract roundlet region.Curvature can be good at reacting the shape feature of each section of contour curve, and the curvature distribution of each section is as schemed
Shown in 2B.Ideally, exist from orthodrome CD to small arc-shaped DE and from small arc-shaped DE to the curvature of line segment EF significant
Spline smoothing, be easy to extract roundlet region DE using curvature.And under train travel environment, due to being made an uproar at random
The influence of sound, calibrated error, vehicle disturbance and the acquisition factors such as equipment itself, exist in the outline data of actual measurement it is a large amount of not
Determining noise jamming, as shown in Figure 2 C.It therefore, is unworkable by fixed curvature value segmentation roundlet.And ICP algorithm is easy
It is influenced by the noise data on the web of the rail, especially when interfering on the web of the rail there are the convexs spot such as dirty soil, measurement accuracy has
Biggish loss, if furthermore fierce shaking occurs for car body, too small region of search is easy that iterative solution is made to fall into local optimum,
Obtain the correction parameter of mistake.
It is illustrated in figure 3 the process of the rail profile autoegistration method under a kind of dynamic environment of the embodiment of the present invention
Figure, method as shown in the figure include：
Step S1 acquires profile of steel rail optical strip image, extracts the light stripe centric line in the profile optical strip image；
Step S2 is coordinately transformed the light stripe centric line, obtains multiple space profiles points；
Optimal fitting order value is calculated according to the multiple space profiles point in step S3；
Step S4 is fitted the multiple space profiles point using the optimal fitting order value, and it is quasi- to obtain profile
Close curve；
Step S5 divides the contour fitting curve, obtains multiple sections；
Step S6 calculates on the contour fitting curve curvature entropy in different sections, when obtaining the curvature entropy maximum pair
The target interval answered；As roundlet starting point, the interval width of the target interval is used as divides the starting point of the target interval
Cut width；
Step S7 determines roundlet area according to the roundlet starting point and the segmentation width on the contour fitting curve
Domain, and the roundlet center of circle is determined according to the roundlet region；
The roundlet center of circle is aligned, so that the contour fitting curve is matched with standard profile by step S8 with the standard center of circle
It is quasi-.
In the present embodiment, profile of steel rail optical strip image can be acquired by photo structure measuring technique, such as Fig. 4 A and Fig. 4 B
For profile of steel rail of embodiment of the present invention optical strip image and its corresponding light stripe centric line and space profiles point schematic diagram.Wherein, scheme
4A is a part of rail profile optical strip image, to the profile striation image zooming-out light stripe centric line in Fig. 4 A, to optical losses
After line coordinates transformation, available space profiles point.It is as shown in Figure 4 B the corresponding light stripe centric line of Fig. 4 A and space profiles point,
The corresponding space profiles point of light stripe centric line in dotted line frame is shown as along arrow meaning.
It is all presented not from the striation brightness and width that can be seen that in Fig. 4 A in collected profile optical strip image from the web of the rail
Uniform variation, causes the light stripe centric line extracted not smooth enough, further results in as shown in Figure 4 B (in corresponding diagram 4A dotted line frame
Striation), the mixed and disorderly fluctuations unrelated with curvature occur for the space profiles point that obtains after coordinate transform, so that based on curvature
The roundlet extracted region of characteristic becomes extremely difficult.Therefore, it is necessary to pre-process to rail exterior feature data, noise jamming is removed.
By taking the light stripe centric line in Fig. 4 B as an example, carry out curve fitting to it.It is usually adopted in existing curve-fitting method
With the polynomial fitting method based on least square method, when being fitted using least square method, the optimization problem of solution be seek with
The smallest curve of error of sum square of space profiles point distance.It is as shown in table 1 to be taken turns by least square method to space in Fig. 4 B
Exterior feature point carries out the error statistics table after being repeatedly fitted.From table 1 it follows that under the calculating of least square method, when fitting rank
It is secondary that error of fitting is minimum when taking 4, but in fact, due to noise spot interference, residual sum of squares (RSS) minimum is not necessarily mean that optimal
Fitting result.If Fig. 5 A and Fig. 5 B show the contour fitting curve of 4 times and 6 times fittings, although 4 fittings have smaller number
It is worth error, but the curve of 6 fittings has passed through more actual profile points, and curve shape and web of the rail shape are more coincide.Cause
This, the polynomial fitting method based on least square method is obviously inaccurate.
Table 1
If Fig. 6 A and Fig. 6 B show the truncation residual error statistic histogram of Fig. 4 B, it can see 4 fittings from Fig. 6 A
Residual distribution is then similar to be uniformly distributed, and profile point is distributed in matched curve two sides relatively uniformly；And 6 fittings in Fig. 6 B
Residual distribution be then similar to normal distribution, most of profile point is located at or is close to matched curve.Measurement under natural conditions misses
Difference meets normal distribution, illustrates that the result of 6 fittings more tends to be true.
As an embodiment of the present invention, the optimal fitting curve that residual error meets normal distribution is truncated in order to obtain, adopts
With the polynomial fitting method based on truncation residual error histogram, optimal fitting order is found, passes through matched curve as far as possible
More space profiles points, reduces influence of the noise to fitting result.It is calculated according to multiple space profiles points optimal quasi-
Order value is closed, including：
Calculate truncation residual error；
Truncation residual error statistic histogram is determined according to the truncation residual error, is calculated based on the truncation residual error statistic histogram
Optimal fitting order value.
It specifically can be according to following equation calculation optimal fitting order value：
Truncation residual error is calculated according to the following formula：
Wherein, Re is truncation residual error；(xn,yn) it is space profiles point coordinate, n=1,2 ..., N, N is ginseng
With total points of fitting, piFor multinomial coefficient, k >=0, i >=0；
Truncation residual error statistic histogram is determined according to the truncation residual error, is based on the truncation residual error statistic histogram, benefit
Optimal fitting order value is calculated with following formula：
Wherein, koptFor optimal fitting order value, H (x) is qualified function, and [- ε, ε] is that section is truncated in residual error；ε is truncation
Parameter, belongs to experience factor, is arranged excessive vulnerable to noise jamming, and too small will cause the loss of signal.It should preferably to cut
For residual error histogram distribution in disconnected section close to normal distribution, ε value as shown in Figure 6 C is 1.Fig. 6 C is the truncation residual error of 6 ranks fitting
Optimization problem is changed by formula (2) and is sought across the quasi- of most space profiles points by statistic histogram and normal distribution
Close curve.It is possible thereby to which it is 6 that the optimal order sub-value for the space profiles point in Fig. 4 B, which is calculated, and according to 6 ranks to space
Profile point is fitted.
As an embodiment of the present invention, described that the contour fitting curve is divided, multiple sections are obtained, are wrapped
It includes：
Determine N-1 windows of different size；Wherein, N is the number that discrete point is fitted on contour fitting curve；
Each window is subjected to sliding traversal using each fitting discrete point as starting point, obtains multiple sections.
Wherein, contour fitting curve is starting equipped with N number of fitting discrete point, such as with the fitting discrete point of serial number 1
Point, then having N-1 corresponding to the section that the point can divide, i.e. the fitting discrete point of serial number 1 to serial number 2 is first section,
Serial number 1 to serial number 3 be second section ... serial number 1 arrive serial number N be the N-1 section.It follows that the width means pair of window
In the section that should be divided, the number for the fitting discrete point for being included in addition to starting point.Therefore, if the number of fitting discrete point is N,
The section n (n-1)/2 can be so divided altogether.
In the present embodiment, the curvature entropy for calculating different sections on the contour fitting curve includes：
Calculate the curvature of contour fitting curvilinear equation；
Calculate the probability distribution of the curvature；
The curvature entropy in each section is calculated according to the probability distribution.
It is illustrated in figure 7 the curvature distribution schematic diagram of difference curve of the embodiment of the present invention, space profiles point is shown in figure
Curvature distribution, the curvature distribution of the contour fitting curve obtained after fitting and the curvature distribution of standard gauge contour curve.It can
To see, the curvature distribution of the contour fitting curve after fitting is more smooth, effectively reduces noise spot interference.But due at
There are inevitable errors during picture, striation extraction and curve matching etc., and the curvature distribution after leading to fitting is not inconsistent simultaneously
Ideal square wave is closed, great circle (the CD section of Fig. 2A) and roundlet (the DE section of Fig. 2A) region all have certain fluctuation.
It is integrally relatively more steady in view of great circle region curvature, and roundlet regional change is then more violent.Comentropy is by information theory
Father's Shannon in 1948 propose, for describing the uncertainty of signal.The entropy of certain signal is bigger, then the unordered journey of the signal
It spends higher.Regard the rail exterior feature curvature after fitting as one-dimensional signal, is denoted as Sn, n=1,2 ..., N.It is observed from fig. 1 that great circle area
The signal smooth sequential in domain, is similar to be uniformly distributed, and the variation of roundlet regional signal is acutely similar to normal distribution.Therefore, lead to
It crosses to entire signal area { SnSliding window traversal is carried out, the maximum region of curvature entropy is found, that is, can determine roundlet region.Specific mistake
Journey is described as follows,
Calculate contour fitting curvilinear equation x=f (y) curvature be：
Wherein, SnFor curvature, (xn,yn) it is fitting discrete point coordinate, n=1,2 ..., N, N is the total points for participating in fitting；
Curvature probability distribution of the curvature on section (n, n+w) be：
Wherein, PiFor curvature probability distribution, i=n, n+1 ..., n+w；On the section (n, n+w) curvature entropy E (n,
W) it is：
After the definition for giving curvature entropy, the change profile with cut-point position and window size is as shown in Figure 8 A, small
Circle segmentation problem is converted to search for maximum position problems of curvature entropy on curved surface.As can be observed from Figure, larger
The point of value all concentrates on a certain region, therefore suitable starting point is selected to be conducive to improve search efficiency.Accordingly, it is determined that roundlet area
The problem of domain, which is represented by, solves following optimization problem：
Schematic diagram is searched for for maximum curvature entropy section as shown in Figure 8 B, wherein maximum curvature entropy section is target interval,
Its search process can be by algorithmic notation in table 2.
Table 2
By the n being calculatedoptFor the starting point in roundlet region, i.e. roundlet starting point, woptTo divide width.According to
noptWith woptIt can determine roundlet region.
As an embodiment of the present invention, according to the roundlet starting point, the segmentation width and the roundlet region
Interior actual discrete points correct the roundlet starting point.
Wherein, due to the influence of error, the curvature distribution in roundlet region is becoming approximation just from approaches uniformity distribution after fitting
State distribution, and distributed area broadens, and leads to the big roundlet cut-point acquired by algorithm above and actually has slight deviations.If roundlet
The true points in region are D, then utilize formulaTo roundlet starting point noptIt is modified.
As an embodiment of the present invention, described to determine that the roundlet center of circle includes according to the roundlet region：
Objective function is established according to the fitting discrete point in the roundlet region：
The minimum value of the objective function is solved according to least square method, corresponding solve of the minimum value is described small round
The heart.
By any two datum mark can will actual measurement rail profile it is unified under the coordinate system where standard rail profile into
Row, which is overlapped, to be compared, because rail head is gradually worn away in use, is not suitable for choosing datum mark, the web of the rail and rail bottom are not deposited
In abrasion metaboly, so choosing the center of circle of small arc-shaped on the position two sides web of the rail as datum mark is that comparison is reasonable.
The small arc-shaped (for 60kg/m rail) that the web of the rail and rail bottom engaging portion are radius 20mm, it is theoretically all to belong to roundlet
The distance in point to the center of circle all should be 20mm, so establish optimization object function with the distance in the web of the rail fitting discrete point to the center of circle,
This distance and the absolute difference of 20mm radius should obtain minimum.
Therefore, determine that the roundlet center of circle includes according to the roundlet region：
It is (x according to the fitting discrete point in roundlet regioni,yi), establish objective function：
Minf is solved according to least square methodi(x), the solution obtained is the roundlet center of circle.
As an embodiment of the present invention, the roundlet region is divided into left roundlet region and right roundlet region, described small
The round heart is corresponding to be divided into the left roundlet center of circle and the right roundlet center of circle；The standard center of circle is divided into the left standard center of circle and right standard round
The heart；The roundlet center of circle is aligned with the standard center of circle, including：
The left roundlet center of circle and the right roundlet center of circle are divided into the standard center of circle respectively according to spin matrix and translation vector
The left standard center of circle and the alignment of the right standard center of circle.
Roundlet of the embodiment of the present invention region segmentation as shown in Fig. 9 A and Fig. 9 B and roundlet center of circle fitting result schematic diagram, Fig. 9 A
In OLWith ORThe respectively left roundlet center of circle and the right roundlet center of circle, ALFor the starting point in left roundlet region, BLFor left roundlet region
Terminating point, ALWith BLConstitute the segmentation width in left roundlet region.ARFor the starting point in right roundlet region, BRFor right roundlet region
Terminating point, ARWith BRConstitute the segmentation width in right roundlet region.
It is quasi- in Figure 10 A if Figure 10 A and Figure 10 B show roundlet of the embodiment of the present invention center of circle and standard center of circle schematic diagram
The left and right center of circle for closing roundlet can be aligned according to the following formula with the left and right center of circle of the standard roundlet in Figure 10 B：
Wherein, R is spin matrix, and T is translation vector.
After the center of circle of contour fitting curve is aligned with the standard center of circle, available registration schematic diagram as shown in figure 11.It presses
Take a picture should wear away position definition, calculate acquisition rail vertically wear away, at side wear position coordinate and standard gauge coordinate difference
Rail wear value can be obtained in value.
It is the mill point that hangs down that rail head, which is defined, at trackside 25mm, and trackside is side grinding point at rail head 16mm.By to standard
Rail 775 repeated acquisitions of progress, the variation for counting hang down mill value and side grinding value carry out the error and static state repeatability precision of confirmatory measurement.
As Figure 12 A and Figure 12 B show the vertical mill and side grinding numerical statistic result figure of 775 static duplicate measurements.It can be with
See that measurement worst error hangs down mill for 0.024mm, side grinding 0.037mm.Measurement error mean value is obtained through statistics, vertical mill is
0.0042mm, side grinding 0.0084mm；Measurement standard difference hangs down mill for 0.0033mm, side grinding 0.0059mm.
It is as shown in figure 13 the radius change statistical results chart that roundlet is fitted in dynamic measurement of the embodiment of the present invention, institute in figure
It is shown as the roundlet region segmentation and roundlet center of circle fitting result for driving scene by the rail exterior feature image of noise jamming.It can see
Out, method proposed by the present invention has strong anti-interference ability, and effectively overcomes the influence of noise.Here, small using being fitted
Round radius evaluates the stability of segmentation.100m gapless line is chosen at scene, and 4 rail exterior feature sections of every meter of acquisition amount to 400 surveys
Measure section.Small radius of circle statistical result is as shown in figure 13, and radius mean value is 20.05mm, standard deviation 0.0923mm.
Rail profile method for registering in through the invention is accurately fitted the wheel of rail using the calculating of Optimal order
Wide curve, and by carrying out division to contour fitting curve and to each interval computation curvature entropy marked off, accurately to determine
Roundlet region on contour fitting curve out, and the roundlet center of circle in roundlet region is aligned with the standard center of circle, it is achieved in steel
The autoregistration of rail profile.Method for registering in the present invention can effectively overcoming noise interference, registration accuracy is high and speed is fast, can be with
Quickly and accurately complete rail profile registration.
It is as shown in figure 14 the rail profile autoregistration apparatus structure under a kind of dynamic environment of the embodiment of the present invention
Schematic diagram, device as shown in the figure include：
Acquisition module 10 extracts the optical losses in the profile optical strip image for acquiring profile of steel rail optical strip image
Line；
Coordinate transformation module 20 obtains multiple space profiles points for being coordinately transformed to the light stripe centric line；
Optimal order computing module 30, for optimal fitting order value to be calculated according to the multiple space profiles point；
Curve fitting module 40, for being intended using the optimal fitting order value the multiple space profiles point
It closes, obtains contour fitting curve；
Interval division module 50 obtains multiple sections for dividing to the contour fitting curve；
Divide module 60, for calculating the curvature entropy in different sections on the contour fitting curve, obtains the curvature entropy
Corresponding target interval when maximum；For the starting point of the target interval as roundlet starting point, the section of the target interval is wide
Degree is as segmentation width；
Center of circle fitting module 70 is used for according to the roundlet starting point and the segmentation width in the contour fitting curve
Upper determining roundlet region, and the roundlet center of circle is determined according to the roundlet region；
Registration module 80, for the roundlet center of circle to be aligned with the standard center of circle, so that the contour fitting curve and mark
Quasi- profile registration.
Wherein, described optimal fitting order value is calculated according to the multiple space profiles point to include：
Calculate truncation residual error；
Truncation residual error statistic histogram is determined according to the truncation residual error, is calculated based on the truncation residual error statistic histogram
Optimal fitting order value.
Wherein, described that the contour fitting curve is divided, multiple sections are obtained, including：
Determine N-1 windows of different size；Wherein, N is the number that discrete point is fitted on contour fitting curve；
Each window is subjected to sliding traversal using each fitting discrete point as starting point, obtains multiple sections.
The principle and dynamic driving ring solved the problems, such as due to the rail profile autoregistration device under the dynamic environment
Rail profile autoegistration method under border is similar, therefore the reality of the rail profile autoregistration device under the dynamic environment
The implementation that may refer to the autoegistration method of the rail profile under dynamic environment is applied, overlaps will not be repeated.
Rail profile registration apparatus in through the invention is accurately fitted the wheel of rail using the calculating of Optimal order
Wide curve, and by carrying out division to contour fitting curve and to each interval computation curvature entropy marked off, accurately to determine
Roundlet region on contour fitting curve out, and the roundlet center of circle in roundlet region is aligned with the standard center of circle, it is achieved in steel
The autoregistration of rail profile.In through the invention with will definitely effectively overcoming noise interference, registration accuracy is high and speed is fast, can be with
Quickly and accurately complete rail profile registration.
Those of ordinary skill in the art will appreciate that implementing the method for the above embodiments can lead to
Program is crossed to instruct relevant hardware and complete, which can be stored in a computer readable storage medium, such as
ROM/RAM, magnetic disk, CD etc..
Particular embodiments described above has carried out further in detail the purpose of the present invention, technical scheme and beneficial effects
Describe in detail it is bright, it should be understood that the above is only a specific embodiment of the present invention, the guarantor being not intended to limit the present invention
Range is protected, all within the spirits and principles of the present invention, any modification, equivalent substitution, improvement and etc. done should be included in this
Within the protection scope of invention.
Claims (10)
1. the rail profile autoegistration method under a kind of dynamic environment, which is characterized in that the method includes：
Profile of steel rail optical strip image is acquired, the light stripe centric line in the profile optical strip image is extracted；
The light stripe centric line is coordinately transformed, multiple space profiles points are obtained；
Optimal fitting order value is calculated according to the multiple space profiles point；
The multiple space profiles point is fitted using the optimal fitting order value, obtains contour fitting curve；
The contour fitting curve is divided, multiple sections are obtained；
The curvature entropy for calculating different sections on the contour fitting curve, obtains corresponding target area when the curvature entropy maximum
Between；The starting point of the target interval is as roundlet starting point, and the interval width of the target interval is as segmentation width；
Roundlet region is determined on the contour fitting curve according to the roundlet starting point and the segmentation width, and according to institute
It states roundlet region and determines the roundlet center of circle；
The roundlet center of circle is aligned with the standard center of circle, so that the contour fitting curve is registrated with standard profile.
2. the method according to claim 1, wherein described be calculated most according to the multiple space profiles point
Excellent fitting order value includes：
Calculate truncation residual error；
Truncation residual error statistic histogram is determined according to the truncation residual error, is calculated based on the truncation residual error statistic histogram optimal
It is fitted order value.
3. being obtained the method according to claim 1, wherein described divide the contour fitting curve
Multiple sections, including：
Determine N-1 windows of different size；Wherein, N is the number that discrete point is fitted on contour fitting curve；
Each window is subjected to sliding traversal using each fitting discrete point as starting point, obtains multiple sections.
4. the method according to claim 1, wherein described calculate different sections on the contour fitting curve
Curvature entropy includes：
Calculate the curvature of contour fitting curvilinear equation；
Calculate the probability distribution of the curvature；
The curvature entropy in each section is calculated according to the probability distribution.
5. the method according to claim 1, wherein according to the roundlet starting point, the segmentation width and institute
The actual discrete points stated in roundlet region correct the roundlet starting point.
6. the method according to claim 1, wherein described determine that the roundlet center of circle is wrapped according to the roundlet region
It includes：
Objective function is established according to the fitting discrete point in the roundlet region：
The minimum value of the objective function is solved according to least square method, corresponding solve of the minimum value is the roundlet center of circle.
7. the method according to claim 1, wherein the roundlet region is divided into left roundlet region and right roundlet area
Domain, the roundlet center of circle is corresponding to be divided into the left roundlet center of circle and the right roundlet center of circle；The standard center of circle be divided into the left standard center of circle and
The right standard center of circle；The roundlet center of circle is aligned with the standard center of circle, including：
The left roundlet center of circle and the right roundlet center of circle are divided into left mark with the standard center of circle respectively according to spin matrix and translation vector
The director circle heart and the alignment of the right standard center of circle.
8. the rail profile autoregistration device under a kind of dynamic environment, which is characterized in that described device includes：
Acquisition module extracts the light stripe centric line in the profile optical strip image for acquiring profile of steel rail optical strip image；
Coordinate transformation module obtains multiple space profiles points for being coordinately transformed to the light stripe centric line；
Optimal order computing module, for optimal fitting order value to be calculated according to the multiple space profiles point；
Curve fitting module is obtained for being fitted using the optimal fitting order value to the multiple space profiles point
Contour fitting curve；
Interval division module obtains multiple sections for dividing to the contour fitting curve；
Divide module, for calculating the curvature entropy in different sections on the contour fitting curve, when obtaining the curvature entropy maximum
Corresponding target interval；The starting point of the target interval is as roundlet starting point, the interval width conduct of the target interval
Divide width；
Center of circle fitting module, for being determined on the contour fitting curve according to the roundlet starting point and the segmentation width
Roundlet region, and the roundlet center of circle is determined according to the roundlet region；
Registration module, for the roundlet center of circle to be aligned with the standard center of circle, so that the contour fitting curve and standard profile
Registration.
9. device according to claim 8, which is characterized in that described to be calculated most according to the multiple space profiles point
Excellent fitting order value includes：
Calculate truncation residual error；
Truncation residual error statistic histogram is determined according to the truncation residual error, is calculated based on the truncation residual error statistic histogram optimal
It is fitted order value.
10. device according to claim 8, which is characterized in that it is described that the contour fitting curve is divided, it obtains
Multiple sections, including：
Determine N-1 windows of different size；Wherein, N is the number that discrete point is fitted on contour fitting curve；
Each window is subjected to sliding traversal using each fitting discrete point as starting point, obtains multiple sections.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810392766.7A CN108917591A (en) | 2018-04-27 | 2018-04-27 | Rail profile autoegistration method and device under a kind of dynamic environment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810392766.7A CN108917591A (en) | 2018-04-27 | 2018-04-27 | Rail profile autoegistration method and device under a kind of dynamic environment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108917591A true CN108917591A (en) | 2018-11-30 |
Family
ID=64403604
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810392766.7A Pending CN108917591A (en) | 2018-04-27 | 2018-04-27 | Rail profile autoegistration method and device under a kind of dynamic environment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108917591A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110030960A (en) * | 2019-05-07 | 2019-07-19 | 中国铁道科学研究院集团有限公司 | Steel-aluminum conductor rail wear detecting system and method |
| CN111563894A (en) * | 2020-05-12 | 2020-08-21 | 西北工业大学 | Method and system for measuring bending stiffness of continuous fiber reinforced material |
| CN111764208A (en) * | 2020-06-30 | 2020-10-13 | 西京学院 | Steel rail grinding profile rapid prediction method based on interval segmentation |
| CN111859245A (en) * | 2020-07-06 | 2020-10-30 | 东南数字经济发展研究院 | Social e-commerce user group hierarchy dividing method |
| CN111882599A (en) * | 2020-07-31 | 2020-11-03 | 北斗国铁(武汉)智能科技有限公司 | Data processing method and device for steel rail abrasion and storage medium |
| CN111879254A (en) * | 2020-07-31 | 2020-11-03 | 湖北国铁轨道交通研究院有限公司 | Method, device, equipment and medium for processing track measurement data |
| CN113624929A (en) * | 2021-10-12 | 2021-11-09 | 成都千嘉科技有限公司 | Full-range automatic calibration method for gas sensor |
| CN113779672A (en) * | 2021-08-31 | 2021-12-10 | 北京铁科英迈技术有限公司 | Steel rail profile wear calculation method, device, equipment and storage medium |
| CN115305753A (en) * | 2022-10-12 | 2022-11-08 | 中国铁建高新装备股份有限公司 | Method and system for rapidly predicting steel rail profile |
| CN117053718A (en) * | 2023-10-11 | 2023-11-14 | 贵州黔程弘景工程咨询有限责任公司 | Beam bottom linear model generation method based on beam bottom linear measurement |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1776364A (en) * | 2005-11-22 | 2006-05-24 | 北京航空航天大学 | Steel rail near laser visual dynamic measuring device and method |
| CN101532827A (en) * | 2009-04-15 | 2009-09-16 | 北京航空航天大学 | Deviation correction method for measurement of rail wear based on laser vision |
| CN101576375A (en) * | 2009-05-21 | 2009-11-11 | 北京航空航天大学 | Fast processing method of laser vision image of steel rail wear |
| JP4915800B2 (en) * | 2007-03-30 | 2012-04-11 | 中西金属工業株式会社 | Conveyor rail wear inspection device |
| CN104835150A (en) * | 2015-04-23 | 2015-08-12 | 深圳大学 | Learning-based eyeground blood vessel geometric key point image processing method and apparatus |
| CN105956542A (en) * | 2016-04-28 | 2016-09-21 | 武汉大学 | Structure wiring harness counting and matching high-resolution remote-sensing image road extraction method |
-
2018
- 2018-04-27 CN CN201810392766.7A patent/CN108917591A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1776364A (en) * | 2005-11-22 | 2006-05-24 | 北京航空航天大学 | Steel rail near laser visual dynamic measuring device and method |
| JP4915800B2 (en) * | 2007-03-30 | 2012-04-11 | 中西金属工業株式会社 | Conveyor rail wear inspection device |
| CN101532827A (en) * | 2009-04-15 | 2009-09-16 | 北京航空航天大学 | Deviation correction method for measurement of rail wear based on laser vision |
| CN101576375A (en) * | 2009-05-21 | 2009-11-11 | 北京航空航天大学 | Fast processing method of laser vision image of steel rail wear |
| CN104835150A (en) * | 2015-04-23 | 2015-08-12 | 深圳大学 | Learning-based eyeground blood vessel geometric key point image processing method and apparatus |
| CN105956542A (en) * | 2016-04-28 | 2016-09-21 | 武汉大学 | Structure wiring harness counting and matching high-resolution remote-sensing image road extraction method |
Non-Patent Citations (1)
| Title |
|---|
| 王昊等: "行车环境下钢轨轮廓自动配准方法研究", 《北京航空航天大学学报》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110030960A (en) * | 2019-05-07 | 2019-07-19 | 中国铁道科学研究院集团有限公司 | Steel-aluminum conductor rail wear detecting system and method |
| CN111563894A (en) * | 2020-05-12 | 2020-08-21 | 西北工业大学 | Method and system for measuring bending stiffness of continuous fiber reinforced material |
| CN111764208A (en) * | 2020-06-30 | 2020-10-13 | 西京学院 | Steel rail grinding profile rapid prediction method based on interval segmentation |
| CN111859245A (en) * | 2020-07-06 | 2020-10-30 | 东南数字经济发展研究院 | Social e-commerce user group hierarchy dividing method |
| CN111882599A (en) * | 2020-07-31 | 2020-11-03 | 北斗国铁(武汉)智能科技有限公司 | Data processing method and device for steel rail abrasion and storage medium |
| CN111879254A (en) * | 2020-07-31 | 2020-11-03 | 湖北国铁轨道交通研究院有限公司 | Method, device, equipment and medium for processing track measurement data |
| CN111879254B (en) * | 2020-07-31 | 2022-04-22 | 湖北国铁轨道交通研究院有限公司 | Method, device, equipment and medium for processing track measurement data |
| CN113779672B (en) * | 2021-08-31 | 2024-04-12 | 北京铁科英迈技术有限公司 | Rail profile abrasion calculation method, device, equipment and storage medium |
| CN113779672A (en) * | 2021-08-31 | 2021-12-10 | 北京铁科英迈技术有限公司 | Steel rail profile wear calculation method, device, equipment and storage medium |
| CN113624929A (en) * | 2021-10-12 | 2021-11-09 | 成都千嘉科技有限公司 | Full-range automatic calibration method for gas sensor |
| CN115305753B (en) * | 2022-10-12 | 2023-02-07 | 中国铁建高新装备股份有限公司 | Method and system for rapidly predicting steel rail profile |
| CN115305753A (en) * | 2022-10-12 | 2022-11-08 | 中国铁建高新装备股份有限公司 | Method and system for rapidly predicting steel rail profile |
| CN117053718A (en) * | 2023-10-11 | 2023-11-14 | 贵州黔程弘景工程咨询有限责任公司 | Beam bottom linear model generation method based on beam bottom linear measurement |
| CN117053718B (en) * | 2023-10-11 | 2023-12-12 | 贵州黔程弘景工程咨询有限责任公司 | Beam bottom linear model generation method based on beam bottom linear measurement |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108917591A (en) | Rail profile autoegistration method and device under a kind of dynamic environment | |
| CN106338521B (en) | Increasing material manufacturing surface and internal flaw and pattern composite detection method and device | |
| CN113870123B (en) | Automatic detection method for contact net leading height and pulling value based on vehicle-mounted mobile laser point cloud | |
| US8078025B2 (en) | Vehicle dynamic measurement device and method for comprehensive parameters of rail wear | |
| CN104005325B (en) | Based on pavement crack checkout gear and the method for the degree of depth and gray level image | |
| US20190339209A1 (en) | A system for detecting crack growth of asphalt pavement based on binocular image analysis | |
| CN104021676B (en) | Vehicle location based on vehicle dynamic video features and vehicle speed measurement method | |
| CN103487034B (en) | Method for measuring distance and height by vehicle-mounted monocular camera based on vertical type target | |
| CN104236498B (en) | Method for measuring end size of switch rail | |
| CN106846340B (en) | A kind of striation boundary extraction method based on on-fixed characteristic point | |
| CN107301648B (en) | Redundant point cloud removing method based on overlapping area boundary angle | |
| CN103557792B (en) | A kind of vision of drogue target is followed the tracks of and location measurement method | |
| US10223802B2 (en) | Method for determining depth maps from stereo images with improved depth resolution in a range | |
| CN102930540A (en) | Method and system for detecting contour of urban building | |
| CN107792115A (en) | One kind automatically extracts both wired rail crest level methods using three-dimensional laser point cloud | |
| CN102494663B (en) | Measuring system of swing angle of swing nozzle and measuring method of swing angle | |
| CN102884397A (en) | Structured-light measuring method and system | |
| CN104973092A (en) | Rail roadbed settlement measurement method based on mileage and image measurement | |
| CN109341668A (en) | Polyphaser measurement method based on refraction projection model and beam ray tracing method | |
| CN105608703A (en) | Current transformer oil level detection method of intelligent substation inspection robot | |
| CN103632376A (en) | Method for suppressing partial occlusion of vehicles by aid of double-level frames | |
| CN107576325A (en) | A kind of indoor positioning terminal for merging visual odometry and Magnetic Sensor | |
| CN106157338A (en) | A kind of railway tunnel gauge image processing system and method | |
| CN106327536A (en) | Neck line measuring method based on sectional point clouds | |
| CN105890542A (en) | Outside dimension measuring system and method of motor vehicle |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181130 |
|
| WD01 | Invention patent application deemed withdrawn after publication |