CN110246192A - Binocular crag deforms intelligent identification Method - Google Patents
Binocular crag deforms intelligent identification Method Download PDFInfo
- Publication number
- CN110246192A CN110246192A CN201910537671.4A CN201910537671A CN110246192A CN 110246192 A CN110246192 A CN 110246192A CN 201910537671 A CN201910537671 A CN 201910537671A CN 110246192 A CN110246192 A CN 110246192A
- Authority
- CN
- China
- Prior art keywords
- point
- image
- crag
- video camera
- coordinate
- 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
- G01B11/16—Measuring arrangements characterised by the use of optical techniques for measuring the deformation in a solid, e.g. optical strain gauge
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
- G06T7/85—Stereo camera calibration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
- G06T2207/10012—Stereo images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30232—Surveillance
Abstract
Binocular crag provided by the invention deforms intelligent identification Method, include the following steps, initially set up the left and right camera model with first order radial distortion, one is found by optimum combination that all parameters of left and right video camera are constituted to define stereo visual system using particle swarm algorithm, and left and right video camera is demarcated；It is then based on circular indicia point and carries out images match, clearly measurement target；Finally on the basis of images match, three-dimensional coordinate recovery is carried out to identification point matched in image, compare the three-dimensional coordinate of identification point under the state of front and back, the in-plane displacement deviation, acoplanarity displacement deviation and the strain value of each mark dot center of mark dot center are calculated, finally to determine that whether there is or not deformations on object structures surface.This method has many advantages, such as that simple, lossless, real-time is good, non-contact, precision is high, can effectively overcome the problems, such as that traditional object deformation measurement method precision is insufficient, can play an important role in the practical engineering application of side slope crag detection.
Description
Technical field
The present invention relates to crag technology for deformation monitoring fields, and in particular to binocular crag deforms intelligent identification Method.
Background technique
When the states such as the stability of side slope are monitored, the surface layer deformation of side slope is that characterization side slope state changes
Become it is the most significant a bit.Highway Administration of Guangdong Province is imaged in one kind filed on September 4th, 2012 based on twin camera
The highway High and dangerous slope monitoring method (application number: CN201210324273.2) of technology, disclose following steps: the first step is
Side slope three-dimensional is rebuild, and is installed video camera along length of slope directional spreding to be monitored, is made each scheduled monitoring point while being in two
The coverage of video camera；Three-dimensional reconstruction specifically includes: being imaged using linear transformation method or perspective projection transformation Matrix Solving
Machine parameter, and orthogonal spin matrix and translation matrix of the world coordinate system relative to camera coordinate system are calculated, then from Liang Tai
Characteristic point is extracted in the image of video camera shooting, using based on the transformation model and grab sample one for rotating and scaling invariant
Cause method carries out Corresponding matching to characteristic point, realizes that side slope three-dimensional is rebuild.Second step is slope monitoring, is imaged according to two
The image of machine shooting calculates the space coordinate for the monitoring point laid in side slope, obtains the space coordinate point set of monitoring point, periodically counts
The Euclidean distance between the three-dimensional space of two neighboring monitoring point is calculated, slope surface situation of change is monitored；It is pre- finally to carry out dispute disaster
Slope displacement or slip value that monitoring obtains are compared, at preset rules by report with the default critical value allowed
Reason.
Side slope three-dimensional deformation is observed using circular markers a kind of filed on April 8th, 2015 by Dalian University of Technology
Method (application number CN201410817771.X) discloses following technical scheme: proposing to use binocular vision system, and passes through mark
Note circular feature point deforms to observe side slope three-dimensional.A series of regular circles reached the same goal to label on the slope of side slope
Characteristic point carries out captured in real-time, obtains the two-dimensional coordinate of all mark points of each moment, then passes through program for two-dimensional coordinate two
Two matchings, obtain 3 d space coordinate.
Above-mentioned technical proposal has the following problems: (1) not accounting for linear transformation in side slope three-dimensional reconstruction process
In journey the problem of first order radial distortion, the accuracy of side slope crag micro-strain detection is inadequate；(2) in images match process
In the problem of not accounting for image border precision.
Summary of the invention
For the defects in the prior art, the present invention provides a kind of binocular crag deformation intelligent identification Method, can solve existing
There is the problem that accuracy is inadequate in the detection of side slope crag in technology.
Binocular crag provided by the invention deforms intelligent identification Method, includes the following steps,
S1. establish the left and right camera model for having first order radial distortion, using particle swarm algorithm find one by it is left,
Optimum combination that all parameters of right video camera are constituted defines stereo visual system, demarcates to left and right video camera；
S2. images match is carried out based on circular indicia point, clearly measurement target；
S3. on the basis of images match, three-dimensional coordinate recovery, relatively front and back are carried out to identification point matched in image
The three-dimensional coordinate of identification point under state finally calculates in-plane displacement deviation, acoplanarity displacement deviation and each mark of mark dot center
The strain value of dot center is known, to determine that whether there is or not deformations on object structures surface.
Further, the left and right camera model with first order radial distortion is established method particularly includes:
S11. the three-dimensional system of coordinate of left and right video camera is constructed using the pin-hole model with first order radial distortion；
S12. the image coordinate system on left and right video camera imaging face is constructed；
S13. the computer picture coordinate system of left and right video camera is constructed；
S14. the three-dimensional coordinate transformation of one point P of object under test surface is obtained into computer picture coordinate system by double
The mathematical model for the stereo visual system that CCD camera is constituted；
S15. the mathematical model of the stereo visual system constituted according to double CCD cameras finds out the minimum of point P three-dimensional coordinate
Two multiply solution, find the optimum combination being made of all parameters of left and right video camera using particle swarm algorithm.
Further, by the three-dimensional coordinate transformation of one point P of object under test surface to computer picture coordinate system in S14
In specific steps are as follows:
(1) the Rigid Body In Space evolution from three-dimensional system of coordinate to camera coordinate system is described with homogeneous coordinates；
(2) ideal image coordinate system is transformed to from camera coordinate system；
(3) from ideal image coordinate system transformation to real image coordinate system, distortion model is established；
(4) from real image coordinate system transformation to computer picture coordinate system.
Further, one is found using particle swarm algorithm in S15 to be made of most all parameters of left and right video camera
Excellent combined specific steps are as follows:
(1) optimization of the full algorithm of particle is established according to the mathematical model for the stereo visual system being made of double CCD cameras
Model；
(2) stereopsis constituted using the three-dimensional coordinate of one point P of the body surface of actual measurement and by double CCD cameras
The residual error between three-dimensional coordinate that the mathematical model of feel system is calculated establishes objective function.
Further, images match is carried out based on circular indicia point in S2, clearly measures target method particularly includes:
S21. Canny algorithm coarse positioning image border is utilized；
S22. image border is accurately determined using Zernike square operator；
S23. ellipse fitting method is utilized, circular indicia point feature is extracted；
S24. the characteristic point of extraction is matched, then determines object to be measured.
It further, is to be filtered by two-dimensional Gaussian function first differential with image convolution using Canny in S21
Then wave finds local maximum to filtered image, specifically:
(1) using Gaussian function to image filtering；
(2) gradient magnitude of each pixel is obtained to image convolution using the first differential of Gaussian function | G | and gradient
Direction θ；
(3) gradient direction is divided into four areas, measurement point gradient magnitude is compared with adjacent pixel gradient value on gradient direction
Compared with whether label measurement point is marginal point；
(4) gradient magnitude of each pixel is counted, and calculates gradient mean value D and variances sigma, by gradient mean value and variance
High threshold of the sum as edge detection, using 0.4 times of high threshold as Low threshold；
(5) edge connection is carried out, it is rough to determine measurement target.
Further, image border is accurately determined using Zernike square operator in S22 specifically:
(1) determine whether the point is marginal point by calculating four parameters of each pixel, the four of the pixel
A parameter is respectively as follows: image background gray scale h, the step height k of gray scale, the height l of central point to edge, central point to edge
Vertical line and x-axis included angle；
(2) measurement neighborhood of a point each point is mapped to the inside of unit circle, according to the point f's (x, y) in discrete picture
The A of Zernike orthogonal moment calculating measurement pointnm；
(3) after the square A11 and A20 that obtain marginal point, central point is calculated to the height l and central point at edge to edge
Vertical line and x-axis included angle；
(4) then, the subpixel coordinates of marginal point are obtained.
Further, it is matched in characteristic point of the S24. to extraction specifically:
(1) epipolar-line constraint matches: matching left video camera by EP point analytic equation and normalizes corresponding points P in plane
The corresponding right video camera imaging plane of characteristic point on coordinate；
(2) tracking identification point is carried out using Kalman filtering algorithm: the position of first coarse positioning subsequent time identification point, so
Carry out neighborhood search again afterwards.
As shown from the above technical solution, beneficial effects of the present invention: binocular crag deformation intelligent identification Method, including with
Lower step initially sets up the left and right camera model with first order radial distortion, using particle swarm algorithm find one by it is left,
Optimum combination that all parameters of right video camera are constituted defines stereo visual system, demarcates to left and right video camera；Then
Images match is carried out based on circular indicia point, clearly measurement target；Finally on the basis of images match, to being matched in image
Identification point carry out three-dimensional coordinate recovery, the three-dimensional coordinate of identification point, finally calculates mark dot center relatively under the state of front and back
The strain value of in-plane displacement deviation, acoplanarity displacement deviation and each mark dot center, to determine that whether there is or not changes on object structures surface
Shape.This method has many advantages, such as that simple, lossless, real-time is good, non-contact, precision is high, can effectively overcome traditional object to become
The problem of shape measurement method precision deficiency can play an important role in the practical engineering application of side slope crag detection.
Detailed description of the invention
It, below will be to tool in order to illustrate more clearly of the specific embodiment of the invention or technical solution in the prior art
Body embodiment or attached drawing needed to be used in the description of the prior art are briefly described.In all the appended drawings, similar member
Part or part are generally identified by similar appended drawing reference.In attached drawing, each element or part might not be drawn according to actual ratio
System.
Fig. 1 is the flow diagram that binocular crag of the present invention deforms intelligent identification Method.
Fig. 2 is the left and right camera model that first order radial distortion is had in the embodiment of the present invention.
Fig. 3 is plane sub-pixel edge step illustraton of model in the embodiment of the present invention.
Fig. 4 is epipolar geom etry constraint principles figure in the embodiment of the present invention.
Specific embodiment
It is described in detail below in conjunction with embodiment of the attached drawing to technical solution of the present invention.Following embodiment is only used
In clearly illustrating technical solution of the present invention, therefore it is only used as example, and cannot be used as a limitation and limit protection of the invention
Range.
It should be noted that unless otherwise indicated, technical term or scientific term used in this application should be this hair
The ordinary meaning that bright one of ordinary skill in the art are understood.
If Fig. 1 is the flow diagram that binocular crag provided in this embodiment deforms intelligent identification Method, specific steps
Are as follows:
S1. the left and right camera model for having first order radial distortion is established as shown in Figure 2, is found using particle swarm algorithm
One defines stereo visual system by the optimum combination that all parameters of left and right video camera are constituted, and carries out to left and right video camera
Calibration；
Establish the left and right camera model with first order radial distortion method particularly includes:
S11. the three-dimensional system of coordinate of left and right video camera is constructed using the pin-hole model with first order radial distortion.Wherein,
Optic center point of the left and right video camera in three-dimensional system of coordinate is respectively OclAnd OcR, OclCoordinate in three-dimensional system of coordinate is
(Xcl,Ycl,Zcl),OcCoordinate of the r in three-dimensional system of coordinate is (Xcr,Ycr,Zcr), wherein ZclAxis and ZcrAxis is taken the photograph with left and right respectively
Camera optical axis coincidence.Coordinate of the one point P of object under test surface in three-dimensional system of coordinate is (Xw, Yw, Zw).
S12. the image coordinate system on left and right video camera imaging face is constructed, O is enabledilXlYlAnd OirXrYrIt is respectively left and right to take the photograph
Image coordinate system on camera imaging surface, wherein picture centre OilAnd OirRespectively optical axis ZclAnd ZcrWith left and right cameras image
The intersection point of plane, XclAxis is respectively and XlAxis, XrAxis is parallel, YclAxis respectively with YlAxis, YrAxis is parallel；
S13. the computer picture coordinate system for constructing left and right video camera, enables OlulvlAnd OrurvrRespectively left and right camera shooting
The computer picture coordinate system of machine, origin OlPositioned at the upper left corner in left video camera imaging face, u and v respectively indicate pixel and are located at number
The columns and line number of group.(Xu, Yu) is image coordinate of the P point under ideal national forest park in Xiaokeng.(Xd, Yd) is the reality of P point
Image coordinate deviates from its position ideal image coordinate (Xu, Yu) because of the radial deformation of lens.
S14. by the three-dimensional coordinate (Xw, Yw, Zw) of one point P of object under test surface transform to computer picture coordinate (u,
v)；Specific shift step are as follows:
(1) the Rigid Body In Space evolution from three-dimensional system of coordinate to camera coordinate system is described with homogeneous coordinates, with next
Coordinate (formula 1) description:
In formula (1), the orthogonal rotational transformation matrix that R is 3 × 3, the translation vector that T is 3 × 1.R and T have been determined
Orientation of the left and right video camera relative to three-dimensional system of coordinate, the matrix element of R can there are three Eulerian angles γ, β and α to indicate, then
R, T are as follows:
T=[Tx, Ty, Tz]T (2)
(2) ideal image coordinate system is transformed to from camera coordinate system；
(3) from ideal image coordinate system transformation to real image coordinate system, distortion model is established；
In formula (4)K is first order radial distortion coefficient.
(4) from real image coordinate system transformation to the conversion of computer picture coordinate system:
In formula (5), (cx, cy) be point O pixel coordinate, that is, principal point coordinate, (dx, dy) is respectively on the plane of delineation
Distance on the direction x, y between unit pixel, Sx are to compare between the two, i.e. aspect ratio.
Joint type (1), (3), (4), (5) obtain:
Principle, which must be inspected, according to CCD camera is obtained it is found that being observed by left and right two video cameras space object
Two images are obtained, computer picture coordinate uses (u respectivelyl,vl) and (ur,vr) indicate, the sky of object can be uniquely determined
Between three-dimensional position (Xw, Yw, Zw) therefore, the model of left and right two video cameras of comprehensive analysis is available by double CCD cameras
The mathematical model of the stereo visual system of composition, is indicated with matrix form:
It is denoted as:
The least square solution of three-dimensional coordinate (Xw, Yw, Zw) can be found out by formula (8):
[Xw, Yw, Zw]T=(BTB)-1BTD (9)
The specific method for the optimum combination that one is made of all parameters of left and right video camera is found using particle swarm algorithm
Are as follows:
The optimization mould of the full algorithm of particle is established according to the mathematical model for the stereo visual system being made of double CCD cameras
Type, the parameter in Optimized model include video camera all in stereo visual system mathematical model outwardly and inwardly totally 24 ginsengs
Number (including video camera external parameter: α1,β1,γ1,Tx1, Ty1, Tz1,αr,βr,γ r,Txr, Tyr, TzrWith inner parameter f1,sx1,
dy1,k1, cx1, cy1,fr,sxr,dyr,kr, cxr, cyrTotally 24).Their constraint condition, i.e., the value range of each variable, this 24
A parameter can be obtained by actual measurement or reference product specification.
The stereoscopic vision constituted using the calibration point three-dimensional coordinate (Xw, Yw, Zw) of actual measurement and by double CCD cameras
Three-dimensional coordinate (the X ' that the mathematical model of system is calculatedw, Y 'w, Z 'w) between residual error establish following objective function:
In formula (10), N represents the quantity of calibration point, and x represents parameter all in vision measurement system.
Then:
X=[α1,β1,γ1,Tx1,Ty1, Tz1, f1, sx1, dy1,k1,cx1,
cy1,αr, βr, γr,Txr,Tyr,Tzr,fr,sxr,dyr, kr, cxr, cyr]T (11)
Simply it is denoted as:
X=[x1, x2..., x24]T (12)
S2. images match is carried out based on circular indicia point, clearly measurement target；
Specifically:
S21 utilizes Canny algorithm coarse positioning image border；
Canny algorithm is to be filtered by two-dimensional Gaussian function first differential with image convolution, then to filtered
Image finds local maximum, specifically:
(1) using Gaussian function to image filtering；
(2) gradient magnitude of each pixel is obtained to image convolution using the first differential of Gaussian function | G | and gradient
Direction θ；
Dimensional Gaussian convolution function are as follows:
The gradient magnitude of pixel are as follows:
Wherein,
(3) gradient direction is divided into four areas, measurement point gradient magnitude is compared with adjacent pixel gradient value on gradient direction
Compared with whether label measurement point is marginal point.If specifically: the pixel gradient amplitude is less than consecutive points gradient magnitude, marks
Measurement point is non-edge point, gradient value is set to zero, otherwise as candidate marginal.
(4) gradient magnitude of each pixel is counted, and calculates gradient mean value D and variances sigma, by gradient mean value and variance
High threshold Th, i.e. Th=D+ σ of the sum as edge detection；Using 0.4Th as Low threshold T1.
(5) edge connects；
Each pixel is labeled as strong edge point and weak marginal point first with dual threshold, strong edge point refers to gradient magnitude
Greater than the pixel of high threshold, weak marginal point refers to pixel of the gradient magnitude between high threshold and Low threshold, is otherwise non-edge
Point.
Then, line trace connection is clicked through to strong edge, it, can if there is strong edge point in four neighborhoods of weak marginal point
It is handled using connecting as marginal point, otherwise it is assumed that it is non-edge point.
S22. image border is accurately determined using Zernike square operator；
Zernike square operator is to determine whether the point is marginal point by calculating four parameters of each pixel.Picture
Four parameters of vegetarian refreshments are respectively as follows: image background gray scale h, the step height k of gray scale, the height l of central point to edge, center
Point arrives the vertical line at edge and the included angle of x-axis.
It is illustrated in figure 3 plane sub-pixel edge step illustraton of model, the Zernike of the point f (x, y) in discrete picture
Orthogonal moment are as follows:
According to formula (15) as can be seen that in order to calculate the A of measurement pointnm, need the neighborhood of a point each point being mapped to list
The inside of circle of position, wherein Vnm(ρ, θ) is indicated are as follows:
In Zernike edge detection algorithm specific implementation process, since in four parameters of marginal point, value is used
Central point to edge height l and central point to the vertical line at edge and the included angle of x-axis, so need to only derive 2 moulds
Plate, i.e. A11 and A20.
Only 7 × 7 neighborhood gray values of template and pixel need to be subjected to convolution in calculating, operation is simply easy to
It realizes.After obtaining the square A11 and A20 of marginal point, calculate central point to edge height l and central point to edge vertical line
With the included angle of x-axis.
Then, the subpixel coordinates of marginal point are available:
S23. ellipse fitting method is utilized, circular indicia point feature is extracted；
The general expression of elliptic curve mode are as follows:
Ax2+Bxy+Cy2+ Dx+Ey+F=0 (19)
The edge of circular indicia point image is accurately positioned during (S21) above, can using least square method
To obtain six parameters of A, B, C, D, E, F in formula (19).
Establish objective function:
Wherein,
Utilize G pairs of objective functionDerivation is acquired when objective function G is minimum value
The canonical form of elliptic equation are as follows:
By x=xcos θ+ysin θ；Y=-xsin θ+ycos θ is substituted into (18), is obtained:
x2(Acos2θ-Bcosθ+Csin2θ)+xy(2Acosθsinθ+B(cos2θ…
-sin2θ)-2Ccosθsinθ)+y2(Asin2θ+Bcosθsinθ+Ccos2θ)…
+ x (Dcos θ-Esin θ)+y (Dsin θ+Ecos θ)+F=0 (24)
(21) and the comparison of (22) formula are available:
2Acosθsinθ+B(cos2θ-sin2θ) sin θ=0-2Ccos θ
θ=1/2 × arctan (B/ (C-A)) (25)
Wherein, A '=Acos2θ-Bcosθsinθ+Csin2θ, C '=Asin2θ+Bcosθsinθ+ Ccos2θ, D '=Dcos
θ-Esin θ, E '=Dcos θ+Esin θ.
It, can be further to each edge on the basis of above scheme after obtaining least square fitting elliptic equation
For point to fitted ellipse apart from given threshold, iteration improves the precision at fitted ellipse center, rejects 5% marginal point every time, directly
It is less than threshold position to criterion distance difference, thus can effectively controls elliptical fitting precision.
S24 Feature Points Matching
(1) limit epipolar-line constraint matches,
Fig. 4 show epipolar geom etry constraint principles figure, and object P normalizes empty imaging plane in left and right video camera in Fig. 4
Picture point on П l and П r is respectively pl and pr, and the optical center O1O2 of two cameras of left and right returns with object point P formed plane and left and right camera
One changes plane intersection, obtains left outside polar curve elpl and right EP point erpr, el and er are baseline O1O2 and the normalization of two cameras is flat
The crosspoint in face, while being also subpoint of two camera photocentres in normalization plane, referred to as Multi- extended.It can from Fig. 3
Object point P1, P2, P3 on left camera optical axis correspond to same point Pl in left video camera normalization plane out, but in right camera shooting
Three different points are incident upon in machine normalization plane, and on right polar curve erpr, this is to say if in left figure
In have a characteristic point, then the projection one in its corresponding right video camera imaging plane is scheduled on corresponding right polar curve.
If p=(x, y, 1), p'=(x', y', 1) is respectively that object point normalizes on empty imaging plane in left and right video camera
Picture point.Then EP point analytic equation are as follows:
In formula 26, E is the eigenmatrix of outer polar plane: E=[t]xR.Spin matrix of the R between two cameras, t
For the translation vector between two cameras, symbol [t]xThe antisymmetric matrix defined for vector t:
(2) Kalman tracks
Tracking identification point is carried out using Kalman filtering algorithm, firstly, the position of coarse positioning subsequent time identification point, so
Afterwards, then neighborhood search is carried out, can reduce search radius in this way, saved and calculate the time.
Kalman filtering is a special case of Bayesian filter, is known and line in system mode function and observation function
When property, when process noise, observation noise and posterior probability density function Gaussian distributed, available best posterior probability.
System state equation are as follows: xk=AKxk-1+AKuk-1+wk-1
Observational equation are as follows: sk=Hkxk+vk
Wherein it is assumed that process noise wk-1 and observation noise vk be it is independent identically distributed, shown in following formula: wk~N (0,
Qk), vk~N (0, Rk) wherein Qk, RkThe respectively covariance matrix of process noise and observation noise.
Specific step are as follows:
Prior estimate
Prior estimate error covariance
Kalman gain
Posterior estimator
Posterior estimator error covariance pK=(I-KKH)P- k (31)
Firstly, being initialized to Kalman filtering, if the state vector (X-direction) of target are as follows:
X (t)=[x (t), v (t), a (t)]T (32)
State matrix is initialized as:
Observing matrix is initialized as: HK=[1 0 0]
The estimation initial value that state is established using the observation of target in first three frame image, is considered respectively in the estimation
The displacement of target, speed, acceleration.
Then Posterior estimator error covariance are as follows:
σ in formula (34)xRWith σxQRespectively observation noise and process noise.
S5. on the basis of images match, three-dimensional coordinate recovery, relatively front and back are carried out to expression point matched in image
The three-dimensional coordinate of identification point under state finally calculates in-plane displacement deviation, acoplanarity displacement deviation and each mark for indicating dot center
The strain value of dot center is known, to determine that whether there is or not deformations on object structures surface.
Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations；
Although present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its
It is still possible to modify the technical solutions described in the foregoing embodiments, or special to some or all of technologies
Sign is equivalently replaced；And these are modified or replaceed, various embodiments of the present invention that it does not separate the essence of the corresponding technical solution
The range of technical solution should all cover within the scope of the claims and the description of the invention.
Claims (8)
1. binocular crag deforms intelligent identification Method, it is characterised in that: include the following steps,
S1. the left and right camera model for having first order radial distortion is established, one is found using particle swarm algorithm and is taken the photograph by left and right
Optimum combination that all parameters of camera are constituted defines stereo visual system, demarcates to left and right video camera；
S2. images match is carried out based on circular indicia point, clearly measurement target；
S3. on the basis of images match, three-dimensional coordinate recovery is carried out to identification point matched in image, relatively under the state of front and back
The three-dimensional coordinate of identification point finally calculates in the in-plane displacement deviation, acoplanarity displacement deviation and each identification point for identifying dot center
The strain value of the heart, to determine that whether there is or not deformations on object structures surface.
2. binocular crag according to claim 1 deforms intelligent identification Method, it is characterised in that: establish radial with single order
The left and right camera model of distortion method particularly includes:
S11. the three-dimensional system of coordinate of left and right video camera is constructed using the pin-hole model with first order radial distortion；
S12. the image coordinate system on left and right video camera imaging face is constructed；
S13. the computer picture coordinate system of left and right video camera is constructed；
S14. the three-dimensional coordinate transformation of one point P of object under test surface is obtained being imaged by double CCD into computer picture coordinate system
Mechanism at stereo visual system mathematical model；
S15. the mathematical model of the stereo visual system constituted according to double CCD cameras finds out the least square of point P three-dimensional coordinate
Solution, finds the optimum combination being made of all parameters of left and right video camera using particle swarm algorithm.
3. binocular crag according to claim 2 deforms intelligent identification Method, it is characterised in that: by determinand in S14
The three-dimensional coordinate transformation of one point P of body surface face is to the specific steps in computer picture coordinate system are as follows:
(1) the Rigid Body In Space evolution from three-dimensional system of coordinate to camera coordinate system is described with homogeneous coordinates；
(2) ideal image coordinate system is transformed to from camera coordinate system；
(3) from ideal image coordinate system transformation to real image coordinate system, distortion model is established；
(4) from real image coordinate system transformation to computer picture coordinate system.
4. binocular crag according to claim 2 deforms intelligent identification Method, it is characterised in that: utilize particle in S15
Group's algorithm finds the specific steps for the optimum combination that one is made of all parameters of left and right video camera are as follows:
(1) Optimized model of the full algorithm of particle is established according to the mathematical model for the stereo visual system being made of double CCD cameras；
(2) stereo visual system constituted using the three-dimensional coordinate of one point P of the body surface of actual measurement and by double CCD cameras
The mathematical model three-dimensional coordinate that is calculated between residual error establish objective function.
5. binocular crag according to claim 1 deforms intelligent identification Method, it is characterised in that: based on round mark in S2
Know point and carry out images match, clearly measures target method particularly includes:
S21. Canny algorithm coarse positioning image border is utilized；
S22. image border is accurately determined using Zernike square operator；
S23. ellipse fitting method is utilized, circular indicia point feature is extracted；
S24. the characteristic point of extraction is matched, then determines object to be measured.
6. binocular crag according to claim 5 deforms intelligent identification Method, it is characterised in that: utilize Canny in S21
To be filtered by two-dimensional Gaussian function first differential with image convolution, local maxima then is found to filtered image
Value, specifically:
(1) using Gaussian function to image filtering；
(2) gradient magnitude of each pixel is obtained to image convolution using the first differential of Gaussian function | G | and gradient direction θ；
(3) gradient direction is divided into four areas, measurement point gradient magnitude is compared with adjacent pixel gradient value on gradient direction, mark
Remember whether measurement point is marginal point；
(4) count the gradient magnitude of each pixel, and calculate gradient mean value D and variances sigma, by gradient mean value and variance and make
For the high threshold of edge detection, using 0.4 times of high threshold as Low threshold；
(5) edge connection is carried out, it is rough to determine measurement target.
7. binocular crag according to claim 5 deforms intelligent identification Method, it is characterised in that: utilized in S22
Zernike square operator accurately determines image border specifically:
(1) determine whether the point is marginal point by calculating four parameters of each pixel, four ginsengs of the pixel
Number is respectively as follows: image background gray scale h, the step height k of gray scale, the height l of central point to edge, the vertical line of central point to edge
With the included angle of x-axis；
(2) measurement neighborhood of a point each point is mapped to the inside of unit circle, according to the point f's (x, y) in discrete picture
The A of Zernike orthogonal moment calculating measurement pointnm；
(3) after the square A11 and A20 that obtain marginal point, calculate central point to edge height l and central point to edge vertical line
With the included angle of x-axis；
(4) then, the subpixel coordinates of marginal point are obtained.
8. binocular crag according to claim 5 deforms intelligent identification Method, it is characterised in that: in S24. to the spy of extraction
Sign point is matched specifically:
(1) epipolar-line constraint matches: matching the feature that left video camera normalizes corresponding points P in plane by EP point analytic equation
Coordinate in the corresponding right video camera imaging plane of point；
(2) carry out tracking identification point using Kalman filtering algorithm: the position of first coarse positioning subsequent time identification point, then again into
Row neighborhood search.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910537671.4A CN110246192A (en) | 2019-06-20 | 2019-06-20 | Binocular crag deforms intelligent identification Method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910537671.4A CN110246192A (en) | 2019-06-20 | 2019-06-20 | Binocular crag deforms intelligent identification Method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110246192A true CN110246192A (en) | 2019-09-17 |
Family
ID=67888390
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910537671.4A Pending CN110246192A (en) | 2019-06-20 | 2019-06-20 | Binocular crag deforms intelligent identification Method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110246192A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103323209A (en) * | 2013-07-02 | 2013-09-25 | 清华大学 | Structural modal parameter identification system based on binocular stereo vision |
| CN104501735A (en) * | 2014-12-23 | 2015-04-08 | 大连理工大学 | Method for observing three-dimensional deformation of side slope by utilizing circular marking points |
| CN106504284A (en) * | 2016-10-24 | 2017-03-15 | 成都通甲优博科技有限责任公司 | A kind of depth picture capturing method combined with structure light based on Stereo matching |
-
2019
- 2019-06-20 CN CN201910537671.4A patent/CN110246192A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103323209A (en) * | 2013-07-02 | 2013-09-25 | 清华大学 | Structural modal parameter identification system based on binocular stereo vision |
| CN104501735A (en) * | 2014-12-23 | 2015-04-08 | 大连理工大学 | Method for observing three-dimensional deformation of side slope by utilizing circular marking points |
| CN106504284A (en) * | 2016-10-24 | 2017-03-15 | 成都通甲优博科技有限责任公司 | A kind of depth picture capturing method combined with structure light based on Stereo matching |
Non-Patent Citations (2)
| Title |
|---|
| 廉小磊等: "基于粒子群算法的双目立体视觉系统标定", 《计算机工程与应用》 * |
| 申宇: "基于双目立体视觉的结构变形监测技术研究", 《中国优秀硕士学位论文全文数据库（电子期刊）信息科技辑》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103411553B (en) | The quick calibrating method of multi-linear structured light vision sensors | |
| CN109035200B (en) | Bolt positioning and pose detection method based on single-eye and double-eye vision cooperation | |
| CN111223133B (en) | Registration method of heterogeneous images | |
| CN104318548B (en) | Rapid image registration implementation method based on space sparsity and SIFT feature extraction | |
| CN105716539B (en) | A kind of three-dimentioned shape measurement method of quick high accuracy | |
| CN104063702B (en) | Three-dimensional gait recognition based on shielding recovery and partial similarity matching | |
| CN103278138B (en) | Method for measuring three-dimensional position and posture of thin component with complex structure | |
| CN109579695B (en) | Part measuring method based on heterogeneous stereoscopic vision | |
| CN104121902B (en) | Implementation method of indoor robot visual odometer based on Xtion camera | |
| CN106204544A (en) | A kind of automatically extract index point position and the method and system of profile in image | |
| CN109859272A (en) | A kind of auto-focusing binocular camera scaling method and device | |
| CN106485757A (en) | A kind of Camera Calibration of Stereo Vision System platform based on filled circles scaling board and scaling method | |
| CN110763204B (en) | Planar coding target and pose measurement method thereof | |
| CN110879080A (en) | High-precision intelligent measuring instrument and measuring method for high-temperature forge piece | |
| CN102788572A (en) | Method, device and system for measuring attitude of lifting hook of engineering machinery | |
| CN105335699B (en) | Read-write scene is read and write intelligent identification and the application thereof of element three-dimensional coordinate | |
| CN108007345A (en) | A kind of digger operating device measuring method based on monocular camera | |
| CN107145890A (en) | A kind of pointer dashboard automatic reading method under remote various visual angles environment | |
| CN106097430B (en) | A kind of laser stripe center line extraction method of more gaussian signal fittings | |
| CN111462198A (en) | Multi-mode image registration method with scale, rotation and radiation invariance | |
| CN115861448A (en) | System calibration method and system based on angular point detection and characteristic point extraction | |
| CN110245634A (en) | Multiposition, multi-angle crag deformation judgement and analysis method | |
| CN107679542A (en) | A kind of dual camera stereoscopic vision recognition methods and system | |
| Datondji et al. | Rotation and translation estimation for a wide baseline fisheye-stereo at crossroads based on traffic flow analysis | |
| CN103400377B (en) | A kind of three-dimensional circular target based on binocular stereo vision detects and determination methods |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190917 |
|
| RJ01 | Rejection of invention patent application after publication |