CN103234482A - Structured light measuring system calibration method based on sinusoidal grating - Google Patents

Abstract

The invention discloses a structured light measuring system calibration method based on sinusoidal grating. The method includes producing coding grating, setting up measuring system to capture calibrating board flat images, projecting coding grating, collecting, storing, confirming phase-unfolding initial points, solving wrapped phase and unfolding, restoring the calibrating board flat images of projector imaging planes, calibrating projectors according to the multi-group projector calibrating board flat images at different positions, calibrating cameras according to multiple camera calibrating board flat images at different positions, and acquiring rotation matrixes from projector coordinate systems to camera coordinate system according to calibrating results of the projectors and the cameras. The phase-unfolding initial point can be effectively confirmed by utilizing coding disc images, and image quality of image region is guaranteed to be high; and the designed phase-unfolding method has a noise-proof function, 'line-pulling' conditions in conventional phase-unfolding methods are overcome, and the designed structured light measuring system calibration method has the advantages of simple operation, strong robustness, high measurement accuracy and the like.

Description

A kind of structured light measurement system scaling method based on sinusoidal grating

Technical field

A kind of scaling method of the structured light measurement system based on sinusoidal grating belongs to the computer measurement technical field.

Background technology

The structure light vision measurement is a kind of measuring method of new object dimensional surface configuration, with noncontact, convenient quick, higher characteristics such as precision, has obtained in fields such as industrial detection, virtual reality, historical relic's protection and engineering in medicine in recent years using widely.As a kind of active vision measuring method, structure light vision is measured the pattern (pattern) that will design in advance by projector and is projected the testee surface, increase vision matching characteristic information, thereby solve a stereoscopic vision corresponding point matching difficult problem, realize obtaining of object surface depth information by triangulation then.

The structured light system calibrating is the key factor that influences The measuring precision, and it mainly comprises the demarcation of the inside and outside parameter of video camera and projector, and the determining of transformation matrix between video camera and the projector coordinates system.

For traditional system calibrating method, all be to begin phase unwrapping from the image section start mostly, yet often quality is lower in the image border, noise spot is more, with the section start of such zone as phase unwrapping, introducing error that can be too early causes the failure of phase unwrapping; In the expansion process of phase place, traditional system calibrating method all is to use the ranks pointwise to get parcel, and the disadvantage of this method is exactly: in the phase unwrapping path, the later each point that more preceding mistake can continue always causes " backguy " phenomenon.

Summary of the invention

The objective of the invention is to disclose a kind of structured light measurement system scaling method based on sinusoidal grating, at first in scaling board plane projection one secondary circular bright area, and this zone carried out binaryzation, this zone generally is positioned at the center of entire image, this regional picture quality is often than higher, and with the starting point of its center as phase unwrapping; Compare traditional phase developing method again, use a kind of method of deploying with anti-noise function, namely mark the positive and negative limit in the phase place main value image earlier, according to pole location cut-off rule is set then, use the flood completion method that wrapped phase is carried out phase unwrapping then.Beneficial effect of the present invention is: can determine the phase unwrapping starting point easy, effectively, and strong robustness, the precision height.

Concrete implementation step of the present invention is:

1) makes encode grating; Traditional encode grating only uses sinusoidal grating or Gray code grating, when phase unwrapping, use the reference position of image as the starting point of phase unwrapping usually, yet, original position picture quality at image is often lower, has increased the risk of phase unwrapping failure.This paper adopts a secondary coding pie chart picture, determines the phase unwrapping starting point well to have solved this problem.

2) building structure light measurement system uses video camera to catch the scaling board image;

The employed structured light measurement system of this paper mainly comprises: a video camera, a projector, a computing machine and a scaling board.Projector projects deformed grating on the scaling board, and video camera is deposited to computing machine after it is gathered, and computing machine is to its processing of decoding.

3) use projector successively with the encode grating image projection to scaling board, and use video camera to catch preservation respectively;

4) determine the phase unwrapping starting point; In view of traditional method can not well be determined the starting point of phase unwrapping, this paper obtains its center by the coding pie chart is looked like to handle, and its center as launching starting point.

5) to the encode grating decoding, obtain the wrapped phase image, and wrapped phase is launched;

If $d_i=\mathrm{int}\left[\frac{\mathrm{\Δ\φ}\left(i\right)}{2\mathrm{\π}}\right](i=\mathrm{1,2,3,4})---\left(1\right)$

(1) in the formula, Δ φ (i) is wrapped phase poor between the neighbor pixel, int[] computing represents to get the nearest integer of numerical value in the bracket, regulation int| ± 0.5|=0; Then have

$d_i=\left\{\begin{array}{cc}-1& \mathrm{\&Delta;\&phi;}\left(i\right)<-\mathrm{\&pi;}\\ 0& |\mathrm{\&Delta;\&phi;}\left(i\right)\&le;\mathrm{\&pi;}|\\ +1& \mathrm{\&Delta;\&phi;}\left(i\right)>\mathrm{\&pi;}\end{array}\right.---\left(2\right)$

If the closed loop path to four pixels of a 2*2 is asked all d _iAfterwards, the gained result has three kinds of possibilities ,-1,0 and 1.When being-1, represent to have a negative pole point in these four points, when being 1, represent to have a positive limit, represent not exist limit when being zero; When

The time, must there be a limit in representing at 4; When

The time, calculate the d of three points again respectively _iIf sum is the d of certain three point _iSum is zero, and then the another one point is exactly limit, specifically is positive limit or negative pole point, depends on

If its value-1, then that point is exactly the negative pole point, otherwise then is positive limit;

By above-mentioned rule, determine the positive and negative limit among the view picture wrapped phase figure, write code and between two nearest positive and negative limits, dispose a cut-off rule; Use the flood filling algorithm to begin to carry out phase unwrapping from starting point; Finally obtained view picture phase unwrapping image;

6) according to the uniqueness of absolute phase, recover to obtain the image of scaling board plane on the projector imaging plane.

7) repeating step 3), 4), 5), 6), obtain projector and the camera calibration plate images of many group diverse locations, uses traditional camera marking method, according to the projector calibrating board plane image of diverse location, acquisition projector inside and outside parameter;

8) the camera calibration board plane image of the how secondary diverse location that obtains according to step 7) is finished the inside and outside parameter of video camera is demarcated;

9) calibration result of video camera and projector comprehensive step 7), 8) can determine finally that camera coordinates is tied to the transformation matrix of projector coordinates system.

The present invention uses coding pie chart picture to determine the starting point of phase unwrapping effectively, and, can guarantee that the residing image-region quality of this starting point is higher; And the phase developing method with anti-noise function of the present invention design has solved " backguy " phenomenon of conventional phase method of deploying effectively; And simple to operate, strong robustness has very high measuring accuracy, for the accurate demarcation of structured light measurement system is laid a good foundation.

Description of drawings

The scaling method process flow diagram of a kind of structured light measurement system based on sinusoidal grating of Fig. 1

The row sinusoidal grating that Fig. 2 generates

The row sinusoidal grating that Fig. 3 generates

The coding pie chart picture that Fig. 4 generates

The scaling board plane picture that Fig. 5 video camera is caught

Fig. 6 video camera is caught projects the capable sinusoidal grating image of three width of cloth on the scaling board plane

Fig. 7 video camera is caught projects three width of cloth row sinusoidal grating images on the scaling board plane

The coding pie chart picture that projects the scaling board plane that Fig. 8 video camera is caught

The pie chart that the step 9) of Fig. 9 obtains looks like to carry out binaryzation

Coding pie chart after the binaryzation of Figure 10 looks like to carry out centering

The phase difference value of adjacent four pixels of Figure 11

Figure 12 limit searching route synoptic diagram

The wrapped phase image of Figure 13 line direction

The expansion phase image of Figure 14 line direction

The projector calibrating board plane image that Figure 15 recovery obtains

The outer parameter calibration result of Figure 16 projector and video camera

The relative position relation figure of the projection plate plane picture of Figure 17 5 secondary diverse locations

Embodiment

Provide preferred implementation of the present invention below, and be illustrated by reference to the accompanying drawings.

As shown in Figure 1, a kind of structured light measurement system scaling method based on sinusoidal grating of the present invention's proposition mainly comprises following nine steps:

1) makes encode grating; Comprising the capable sinusoidal grating of three width of cloth phase shifts 120 degree, row sinusoidal grating and a secondary coding pie chart picture that is used for definite phase unwrapping starting point of three width of cloth phase shifts, 120 degree:

When the light intensity of the image that goes out when projector projects satisfied Sine distribution, its light intensity then can be represented with (1) formula so:

I ₀(x,y)=r ₀(x,y){A(x,y)+B(x,y)cosφ(x,y)} （1）

Wherein, r ₀(x, y) (x y) is the background light intensity, and (x y) is fringe amplitude to B, and (x y) is initial phase to φ for reflectivity, the A of expression body surface.

Adopt three step phase-shift methods, so just need the bar graph of three width of cloth phase intervals, 120 degree.If T is the sampling period, (i j), gives the some assignment to each the some I in the scan image

127.5+127.5cos(i×2π/T+0.5×k×π) （2）

Wherein k gets 0,1,2, represents 3 phase shifts respectively.The row sinusoidal grating image that obtains, as shown in Figure 2; In like manner can get the capable sinusoidal grating image of three width of cloth, as shown in Figure 3.

Following surface analysis is the sampled point i in the formula (2) and the selection rule of sampling period T once:

1. sampling interval pixel preferably, the demonstration point of each projector just can be corresponding one to one with sampling pixel points like this.If sampling interval will cause losing of precision more than a pixel, because a plurality of demonstration points in the image that projector is launched will corresponding same sampled point.If be less than a pixel at interval, obviously will increase the rise time of sinusoidal grating image, yet extra sampled point also be the point of having no idea to be mapped on the projector.

2. the selection in sampling period need be according to the surperficial complexity of testee decision.If it is milder that curved surface changes, the sampling period with regard to eligible suitably more greatly, and if the testee concave-convex surface is very violent, then the sampling period just should be selected smallerly, the sampling period used herein is 32s.

Coding pie chart picture mainly is to determine starting point to be about to its centralized positioning starting point when wrapped phase is carried out phase unwrapping.Its image as shown in Figure 4.

2) building structure light measurement system uses video camera to catch the scaling board image; The scaling board image of catching as shown in Figure 5.

3) use projector successively the capable sinusoidal grating of three width of cloth, three width of cloth row sinusoidal gratings and coding pie chart to be looked like to project on the scaling board, and use video camera to catch preservation respectively;

The capable sinusoidal grating image of three width of cloth of catching as shown in Figure 6.Three width of cloth row sinusoidal gratings of catching as shown in Figure 7, the pie chart picture of catching is as shown in Figure 8.

4) determine the phase unwrapping starting point; The projection pie chart picture that step 3) is caught, as shown in Figure 8; Use the otsu method that it is at first carried out binaryzation, use the region growing method to determine disc area then, as shown in Figure 9; And then definite disc centre, as shown in figure 10; And with its starting point as phase unwrapping.

5) derive and utilize three step phase-shift methods that one group of image that step 3) obtains is found the solution, obtain the wrapped phase on the row, column direction, and the wrapped phase that utilizes the branch cutting method to fill the row, column direction in conjunction with flood carries out phase unwrapping;

Suppose to throw altogether the secondary sinusoidal grating image of n, then the phase difference value of two adjacent amplitude grating images is 2 π/N, if I _nRepresenting the light intensity put on the n width of cloth image then has (3) formula:

$I_n(x,y)=a(x,y)+b(x,y)\cos [\mathrm{\&phi;}(x,y)+\frac{2\mathrm{\&pi;}(n-1)}{N}]---\left(3\right)$

In following formula, a (x y) is the background light intensity, and b (x y) then is the reflectivity on testee surface, φ (x, y) phase place of the deformed grating of catching for ccd video camera.Under the known situation of light intensity, still have three unknown parameters, so if try to achieve φ (x, y), then N will equal 3 at least, the system of equations that (3) formula of finding the solution is formed can obtain:

$\mathrm{\&phi;}(x,y)=\arctan \frac{\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{I}_n(x,y)\sin [2\mathrm{\&pi;}(n-1)/N]}{\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{I}_n(x,y)\cos [2\mathrm{\&pi;}(n-1)/N]}---\left(4\right)$

In order to improve measuring speed, use three step phase-shift methods to measure, namely N is taken as 3, brings (4) Shi Deshi (5) into:

$\mathrm{\&phi;}(x,y)=\arctan \left(\frac{\sqrt{3}(I_2-I_3)}{2I_1-I_2-I_3}\right)---\left(5\right)$

By (5) formula as can be known, the calculating of phase place need be carried out arctangent computation, so (x, value y) always between [π ,+π], is referred to as wrapped phase to φ.Parcel thinks that phase place does not have uniqueness, so must launch to obtain to launch phase place to absolute phase, absolute phase just is because it is only the phase place that needs, have absolute uniqueness.

Owing to causing in the raster image of shooting, a variety of causes such as dust, experimental facilities in the air have various noise spots, these noise spots also can be brought among the wrapped phase figure, the existence of noise spot can cause difficulty to the expansion of phase place undoubtedly, so, the identification of the noise spot particular importance that seems.Phase value this as a physical quantity, therefore, the absolute phase of arbitrfary point all should be unique in the phase diagram, this and phase unwrapping path independence.Namely along deployment paths arbitrarily wrapped phase is launched, resulting result should be identical.If carry out phase unwrapping along a closed path, then saltus step and should be zero, otherwise must have noise spot in this closed path.

Fig. 3 .7 is 2 * 2 pixel regions, defines deployment paths by counterclockwise direction, as shown in figure 11.

Suppose $d_i=\mathrm{int}\left[\frac{\mathrm{\&Delta;\&phi;}\left(i\right)}{2\mathrm{\&pi;}}\right](i=\mathrm{1,2,3,4})---\left(6\right)$

(6) in the formula, Δ φ (i) is wrapped phase poor between the neighbor pixel, int[] computing represents to get the nearest integer of numerical value in the bracket, regulation int| ± 0.5|=0.Then have

$d_i=\left\{\begin{array}{cc}-1& \mathrm{\&Delta;\&phi;}\left(i\right)<-\mathrm{\&pi;}\\ 0& |\mathrm{\&Delta;\&phi;}\left(i\right)\&le;\mathrm{\&pi;}|\\ +1& \mathrm{\&Delta;\&phi;}\left(i\right)>\mathrm{\&pi;}\end{array}\right.---\left(7\right)$

If the closed loop path to four pixels of a 2*2 is asked all d _iAfterwards, the gained result has three kinds of possibilities ,-1,0 and 1.When being-1, represent to have a negative pole point in these four points, when being 1, represent to have a positive limit, represent not exist limit when being zero.When

The time, must there be a limit in representing at 4, how to identify this point once with regard to simple analysis below.

When

The time, calculate the d of three points again respectively according to the path among Figure 12 _iIf sum is the d of certain three point _iSum is zero, and then the another one point is exactly limit, specifically is positive limit or negative pole point, depends on

If its value-1, then that point is exactly the negative pole point, otherwise then is positive limit.

Determine positive and negative limit among the view picture wrapped phase figure by said method, write code below and between two nearest positive and negative limits, dispose a cut-off rule, use the flood filling algorithm to begin to carry out phase unwrapping from starting point then, in the process of launching, deployment paths can not be passed through cut-off rule, must get around.The wrapped phase image of Figure 13 for using three step phase-shift methods to obtain, and Figure 14 finds the solution the phase unwrapping figure that obtains for using said method.

6) absolute phase that obtains according to step 5) and the uniqueness principle of absolute phase recover to obtain the projector image of the scaling board image correspondence that step 3) obtains.

Can obtain capable absolute phase and the row absolute phase of camera review by step 5), by the ranks absolute phase values of any this point in obviously can unique definite two-dimentional absolute phase image, be easy to obtain the ranks absolute phase of projector image, uniqueness by absolute phase, can realize the coupling of the point of projector image and camera review, by this matching relationship, just be easy to obtain step 2) the corresponding projector image of scaling board image of catching, as shown in figure 15.

7) repeating step 2), 3), 4), 5), 6) obtain projector and the camera calibration plate images of many group diverse locations, use traditional camera marking method, according to the projector calibrating board plane image of diverse location, it is as follows to obtain the projector inside and outside parameter:

The intrinsic parameter matrix is: ${\mathrm{KK}}_c=\left[\begin{array}{ccc}1799.10145& 0& 466.52306\\ 0& 1794.41509& 663.93531\\ 0& 0& 1\end{array}\right]$

" camera lens " distortion parameter: kc=[-0.02649 0.04410-0.00476-0.00365 0.00000]

8) the camera calibration board plane image of the how secondary diverse location that obtains according to step 7) is finished the inside and outside parameter of video camera is demarcated, and the result is as follows:

The intrinsic parameter matrix is: $\left[\begin{array}{ccc}3832.40327& 0& 789.69111\\ 0& 3839.23190& 619.39854\\ 0& 0& 1\end{array}\right]$

Lens distortion parameter: kc=[-0.13806 2.52811 0.00087 0.00037 0.00000]

9) calibration result of video camera and projector comprehensive step 7), 8) determines that finally camera coordinates is tied to the transformation matrix of projector coordinates system.

Translation vector: T=[-464.48770-266.91702 18.83379]

Rotation matrix: $R=\left[\begin{array}{ccc}0.9615& -0.0180& 0.2742\\ 0.0215& 0.9997& -0.0096\\ -0.2740& 0.0151& 0.9616\end{array}\right]$

The video camera that Figure 16 obtains for the scaling board that uses 5 diverse locations and the outer parameter calibration result of projector, the therefrom relative space position of video camera and projector relation as can be seen.Figure 17 is the enlarged drawing on scaling board plane.

As can be seen, the present invention uses special fgs encoder pie chart picture effectively to determine the starting point of phase unwrapping, and, can guarantee that the residing image-region quality of this starting point is higher; And the phase developing method with anti-noise function of the present invention design has solved " backguy " phenomenon of conventional phase method of deploying effectively; And simple to operate, strong robustness has very high measuring accuracy, for the accurate demarcation of structured light measurement system is laid a good foundation.

Claims (3)

1. structured light measurement system scaling method based on sinusoidal grating is characterized in that may further comprise the steps:

1) makes encode grating, comprising the capable sinusoidal grating of three width of cloth phase shifts, 120 degree, row sinusoidal grating and a secondary coding pie chart picture that is used for determining the phase unwrapping starting point of three width of cloth phase shifts, 120 degree;

2) building structure light measurement system uses video camera to catch the scaling board image;

3) use projector successively the capable sinusoidal grating of three width of cloth, three width of cloth row sinusoidal gratings and pie chart to be looked like to project on the scaling board, and use video camera to catch preservation respectively;

4) determine the phase unwrapping starting point, to the coding pie chart picture that step 3) is caught, use the otsu method that it is at first carried out binaryzation, use the region growing method to determine coding disk image-region then, then determine its center, with its starting point as phase unwrapping;

5) derive and utilize three step phase-shift methods that one group of image that step 3) obtains is found the solution, obtain the wrapped phase on the row, column direction, and utilize the branch cutting method in conjunction with the flood completion method wrapped phase of row, column direction to be carried out phase unwrapping;

6) absolute phase that obtains according to step 5) and the uniqueness principle of absolute phase recover to obtain step 2) projector image of the scaling board image correspondence that obtains;

7) repeating step 2), 3), 4), 5), 6) obtain projector and the camera calibration plate images of many group diverse locations, uses traditional camera marking method, according to the projector calibrating board plane image of diverse location, acquisition projector inside and outside parameter;

8) the camera calibration board plane image of the how secondary diverse location that obtains according to step 7) is finished the inside and outside parameter of video camera is demarcated;

9) calibration result of video camera and projector comprehensive step 7), 8) determines that finally camera coordinates is tied to the transformation matrix of projector coordinates system.

2. a kind of structured light measurement system scaling method based on sinusoidal grating according to claim 1 is characterized in that: use coding pie chart picture in step 1).

3. a kind of structured light measurement system scaling method based on sinusoidal grating according to claim 1 is characterized in that: carry out phase unwrapping in step 5);

Suppose to throw altogether the secondary sinusoidal grating image of n, then the phase difference value of two adjacent amplitude grating images is 2 π/N, if I _nRepresenting then has the light intensity put on the n width of cloth image

$I_n(x,y)=a(x,y)+b(x,y)\cos [\mathrm{\&phi;}(x,y)+\frac{2\mathrm{\&pi;}(n-1)}{N}]---\left(3\right)$

In the formula, (x y) is the background light intensity to a, and (x y) is the reflectivity on testee surface to b, φ (x, y) phase place of the deformed grating of catching for ccd video camera; Make N will equal 3 at least, find the solution the system of equations that (3) formula is formed, obtain

$\mathrm{\&phi;}(x,y)=\arctan \frac{\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{I}_n(x,y)\sin [2\mathrm{\&pi;}(n-1)/N]}{\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{I}_n(x,y)\cos [2\mathrm{\&pi;}(n-1)/N]}---\left(4\right)$

Measure with three step phase-shift methods, namely N is taken as 3, and substitution (4) formula obtains

$\mathrm{\&phi;}(x,y)=\arctan \left(\frac{\sqrt{3}(I_2-I_3)}{2I_1-I_2-I_3}\right)---\left(5\right)$

If $d_i=\mathrm{int}\left[\frac{\mathrm{\&Delta;\&phi;}\left(i\right)}{2\mathrm{\&pi;}}\right](i=\mathrm{1,2,3,4})---\left(6\right)$

(6) in the formula, Δ φ (i) is wrapped phase poor between the neighbor pixel, int[] computing represents to get the nearest integer of numerical value in the bracket, regulation int| ± 0.5|=0; Then have

$d_i=\left\{\begin{array}{cc}-1& \mathrm{\&Delta;\&phi;}\left(i\right)<-\mathrm{\&pi;}\\ 0& |\mathrm{\&Delta;\&phi;}\left(i\right)\&le;\mathrm{\&pi;}|\\ +1& \mathrm{\&Delta;\&phi;}\left(i\right)>\mathrm{\&pi;}\end{array}\right.---\left(7\right)$

If the closed loop path to four pixels of a 2*2 is asked all d _iAfterwards, the gained result has three kinds of possibilities ,-1,0 and 1; When being-1, represent to have a negative pole point in these four points, when being 1, represent to have a positive limit, represent not exist limit when being zero; When

The time, must there be a limit in representing at 4; When

The time, calculate the d of three points again respectively _iIf sum is the d of certain three point _iSum is zero, and then the another one point is exactly limit, specifically is positive limit or negative pole point, depends on

If its value-1, then that point is exactly the negative pole point, otherwise then is positive limit;

Determine the positive and negative limit among the view picture wrapped phase figure, write code and between two nearest positive and negative limits, dispose a cut-off rule; Use the flood filling algorithm to begin to carry out phase unwrapping from starting point; Finally obtained view picture phase unwrapping image.

Images