CN1295651C - Composite calibration method of mold surface optical measurement system - Google Patents

Abstract

The present invention relates to a synthetic calibration method of a mold surface optical measurement system. A grating is calibrated by a calibration plate with unique design and a projector of corresponding rhombic lattice arrays, and the black rhombic lattices are uniformly distributed on the white ground color of the calibration plate. The position of the calibration plate is positioned and corrected by laser to make the rhombic lattices projected by the projector fall in the white gap of the calibration plate. The images of the calibration plate before and after the projector is closed are recorded by a camera to be used as calibration images; essential data calibrated by the camera is obtained by processing the calibration images. The calibration of the camera is realized by a two-step calibration method; the spatial coordinate of the calibration point of the projector is figured out according to the calibration result of the camera; the spatial coordinate is matched with the coordinate of the corresponding projector image plane calibration point to be used as essential data for realizing the calibration of the projector. The calibration of the geometric parameters of the system is realized according to a conversion relation between a projector coordinate system and a camera coordinate system. The present invention solves the precise and automatic extraction problem of calibration points, realizes the synthetic calibration of the camera, the projector and the measurement system, and has significant engineering practical value.

Description

Profile optical measuring system composite calibration method

Technical field:

The present invention relates to a kind of profile optical measuring system composite calibration method, particularly a kind of method that is used for camera calibration, projector calibrating and system's geometrical parameter calibration of structured light projection three dimensional type planar survey system belongs to mechanical engineering and field of optical measuring technologies.

Background technology:

Structured light projection three dimensional type planar survey system is based on principle of triangulation, form by video camera and projector, the optical model of video camera and projector all adopts pin-hole model, its optical axis intersects mutually, the line of photocentre is parallel with reference planes, photocentre is all influential to measurement result to the distance L of reference planes and projector or the geometric parameters such as visual field width FW of video camera on reference planes apart from d, photocentre, and therefore video camera, projector inside and outside parameter and the system's geometric parameter based on pin-hole model all needs to demarcate.How to realize the accurate demarcation of internal and external parameter and system's geometric parameter of video camera and projector, to satisfy the basic demand of triangulation, precision and the reliability that guarantees the profile three-dimensional measurement had great important, also is the problem that numerous researchers pay special attention to.

In the tradition scaling method, two step standardizations are widely used in Camera calibration, its method is under certain camera model, based on specific experiment condition such as shape, the known calibrated reference of size, through its image is handled, the basic data that obtains to demarcate is the pixel coordinate and the space physics coordinate of calibration point, utilize a series of mathematic(al) manipulations and computing method then, ask for the inner parameter and the external parameter of camera model, but finish the demarcation of whole measuring system, also must take certain additional measure.In the prior art, Jean-YvesBouguet is at Http: //propose to adopt the semi-automatic method of man-machine interactively to obtain the demarcation basic data in www.vision.caltech.edu/bouguetj/index.html one literary composition, operate simple and easy, but artificial factor has obviously influenced precision and efficient that basic data is obtained, and is unfavorable for the demarcation of projector and systematic parameter.Adopt red pattern scaling board calibrating camera in Andrew Certain and Li Zhang Hierarchical Stripe Scanning one literary composition in http://www.cs.washington.edu/homes/lizhang/projects/hstripe, the blue pattern of projection then is to scaling board, by the colour TV camera record, keep blue method of patterning by the filtering red pattern and come the labeling projection instrument, this obviously and be not suitable for the B commonly used in the three-dimension measuring system and the demarcation of physical grating projector.

Summary of the invention:

For overcoming the deficiencies in the prior art and defective, the present invention proposes a kind of profile optical measuring system composite calibration method, guaranteeing on the basis that the demarcation basic data is accurately obtained automatically, realize the composite calibration of video camera, projector and systematic parameter, this scaling method is applicable to the structured light projection three dimensional type planar survey system that meets principle of triangulation.

The present invention adopts unique plane reference plate design, and scaling board is a background color with white, the black diamonds grid array that is evenly distributed on it, and the grid summit is as calibration point; The projector calibrating grating adopts identical rhombus grid array pattern, and number of arrays is lacked 1 row, 1 row than scaling board rhombus grid array.Design an anchor point at the scaling board center, a pair of laser tube is installed on measuring system simultaneously, timing signal makes a pair of laser spots that laser tube sends overlap substantially with anchor point on the scaling board, fine setting scaling board position makes the rhombus grid array pattern of projector projection just in time drop in the white gap of scaling board, close forward and backward scaling board image respectively as projector and Camera calibration image by the camera record projector, so just guaranteed the unification of the world coordinate system of projector and camera calibration.The camera calibration image is carried out angle point extract processing automatically, obtain the basic data of camera calibration, adopt two step calibration algorithms to realize Camera calibration.Again the projector calibrating image is carried out angle point and extract processing automatically, because the rhombus grid of projection and the rhombus grid of scaling board are not overlapping, can remove the camera calibration point that has extracted the front, accurately obtain the pixel coordinate of projector calibrating unique point, calculate the volume coordinate of calibration point according to the camera calibration result, realize the demarcation of projector as basic data with corresponding projector image plane reference point coordinate coupling.In the demarcation that realizes system's geometric parameter on video camera and projector calibrating result's the basis according to the transformational relation between projector coordinates system and the camera coordinate system.Concrete demarcating steps is as follows:

1) obtains basic data automatically: scaling board is placed measured zone, adjust its position by a pair of laser spots and location round dot, make the rhombus grid array of projection just in time be positioned at the white gap of scaling board, close forward and backward scaling board image respectively as projector and Camera calibration image by the camera record projector.With value district (SUSAN) method the camera calibration image is carried out angle point based on canny best edge detection criteria and minimal absorption nuclear and extract automatically, acquisition Camera calibration point pixel coordinate; Again the projector calibrating image is carried out same processing, because the rhombus grid of projection and the rhombus grid of scaling board are not overlapping, can remove the angle point that has extracted the front exactly, obtain only to have the corner pixels coordinate of projector calibrating pattern, be the calibration point pixel coordinate of projector.The pixel coordinate and the volume coordinate of camera calibration point are mated the basic data that obtains camera calibration automatically.

2) camera calibration: as input, realize the demarcation of the inside and outside parameter of video cameras by two step calibration algorithms with the basic data of camera calibration.At first, adopt the linear least-squares estimation technique to calibrate camera parameters as much as possible according to the transformational relation between camera coordinate system and the world coordinate system; Adopt nonlinear optimization method to calibrate whole camera parameters then, make observed feature point for calibration image and the unique point image that goes out by Model Calculation reach optimum matching.

3) projector calibrating: obtain the corresponding volume coordinate of projector calibrating point according to the camera interior and exterior parameter of having demarcated, mate with the calibration point coordinate on known projector grating plane, generate the basic data of projector calibrating, adopt 2) in calibration algorithm realize the demarcation of projector inside and outside parameter.

4) system's geometrical parameter calibration: obtain transformational relation between projector coordinates system and the camera coordinate system according to the transformational relation between projector coordinates system and camera coordinate system and the world coordinate system, thus the calibration result of the system of acquisition geometric parameter.

The present invention adopts design unique scaling board and projector calibrating grating, and takes the uncalibrated image acquisition methods based on laser positioning, has solved the accurately automatic extraction problem of calibration point well, has remedied the deficiency of color filtering method; And by under same scaling board position, obtaining video camera and projector calibrating image, guarantee the unification of world coordinate system in camera calibration and the projector calibrating, realized the composite calibration of video camera and projector and measuring system, fully guarantee the precision and the reliability of profile three-dimensional measurement, have the important engineering practical value.

Description of drawings:

The scaling board synoptic diagram that Fig. 1 designs for the present invention.

Fig. 2 is a projector calibrating image synoptic diagram.

Fig. 3 is the measuring system structural representation.

1 is scaling board rhombus grid array among the figure, and 2 is anchor point, and 3 is the rhombus grid array of projector projection, and 4 is projector, and 5 is video camera, and 6 is laser tube.

Embodiment:

Below in conjunction with accompanying drawing concrete enforcement of the present invention is further described.

As Fig. 1, Fig. 2, shown in Figure 3, profile optical measuring system composite calibration method is as follows to the concrete implementation step that a structured light projection three dimensional type planar survey system carries out composite calibration according to the present invention:

1. make scaling board and projector calibrating grating: according to the design of Fig. 1, according to the characteristic of measuring system itself, make the plane reference plate of corresponding size, be background color with white, the black diamonds grid array 1 that is evenly distributed on it, the scaling board center indicates a black anchor point 2.The pattern of projector calibrating grating adopts identical rhombus grid array, its spacing, big or small corresponding with scaling board rhombus grid array 1, and number of arrays is lacked delegation's one row than scaling board rhombus grid array 1;

2. obtain the projector calibrating image: the measured zone that scaling board is placed the profile optical measuring system, adjust the scaling board position by laser spots and scaling board anchor point 2 that a pair of laser tube 6 sends, make the rhombus grid array 3 of projector projection just in time be positioned at the white gap of scaling board, by video camera 5 records scaling board image at this moment, be the projector calibrating image;

3. obtain the camera calibration image: close projector 4, conceal the rhombus grid array 3 of projection on the scaling board,, be the camera calibration image by video camera 5 records scaling board image at this moment;

4. obtain the camera calibration point pixel coordinate: the camera calibration image is handled, obtain the border of each rhombus grid in the scaling board rhombus grid array 1 according to canny edge extracting criterion, and, obtain rhombus grid angular coordinate according to the shape facility filtering noise of rhombus grid; With the angular coordinate is that search window is set up at the center, obtains according to SUSAN method principle to absorb the respective pixel point coordinate of nuclear with value district (USAN) area minimum in the window, and this coordinate is the camera calibration point pixel coordinate;

5. obtain the camera calibration basic data: the distribution characteristics according to the scaling board rhombus grid array of taking 1 sorts to the calibration point pixel coordinate, obtain calibration point pixel coordinate collection A, sorted by same rule in the summit of the rhombus grid array 1 on the scaling board in the world coordinate system simultaneously, obtain corresponding calibration point volume coordinate collection B, generate the basic data of camera calibration by the coupling of A and B automatically;

6. linear calibration's camera parameters: as known conditions, adopt the linear least-squares estimation technique to calibrate camera parameters rotation matrix R, translation matrix T and equivalent focal length f according to following fundamental relation formula with the basic data of camera calibration _c:

$\left(\begin{array}{c}{x}_c\\ y_c\\ z_c\end{array}\right)=R\left(\begin{array}{c}{x}_w\\ y_w\\ z_w\end{array}\right)+T$

In the formula, (x _c, y _c, z _c) be the coordinate of calibration point in camera coordinate system, (x _w, y _w, z _w) be the volume coordinate of calibration point in world coordinate system, R is a rotation matrix, T is a translation matrix.

7. optimize and demarcate whole camera parameters: on the basis of step 6, adopt Marquardt (Levenberg-Marquardt) nonlinear optimization method to be optimized, make the value minimum of following expression, obtain the calibration result of camera interior and exterior parameter:

$\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{(x_{\mathrm{Ii}}-x_{\mathrm{Pi}})}^2+\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{(y_{\mathrm{Ii}}-y_{\mathrm{Pi}})}^2$

In the formula, x _Ii, y _IiBe the pixel coordinate of calibration point, x _Pi, y _PiFor utilizing the calculating pixel coordinate of the anti-corresponding calibration point of obtaining of calibrating parameters.

8. obtain projector calibrating point pixel coordinate: the projector calibrating image is carried out the processing same with step 4, remove the camera calibration point that has extracted in the step 4 simultaneously, obtain only to have the angular coordinate of projector rhombus grid array 3 patterns, be projector calibrating point pixel coordinate;

9. obtain the projector calibrating basic data: because projector calibrating point is positioned at same plane with camera calibration point, can calculate the world coordinates of projector calibrating point according to the video camera internal and external parameter of having demarcated, obtain projector calibrating space of points coordinate set B with reference to step 5 _p, with the calibration point coordinate set A on known projector grating plane _pBe complementary and generate the projector calibrating basic data automatically;

10. projector calibrating: the inside and outside parameter with reference to step 6,7 pairs of projector is demarcated;

11. profile optical measuring system geometrical parameter calibration: because the world coordinate system identical (the scaling board position is identical) of camera calibration and projector calibrating, the transformational relation that therefore can get projector coordinates and be between Xp and camera coordinate system Xc is:

$X_P=R_P{R}_C^{-1}{X}_C+T_P-R_P{R}_C^{-1}{T}_C$

In the formula, R _P, T _PBe rotation matrix and the translation matrix of changing between the projector coordinates of having demarcated system and the world coordinate system, R _C, T _CBe rotation matrix and the translation matrix of changing between the camera coordinate system demarcated and the world coordinate system, order $R^{\′}=R_P{R}_C^{-1}=\left[\begin{array}{ccc}{r}_{11}^{\′}& r_{12}^{\′}& r_{13}^{\′}\\ r_{21}^{\′}& r_{22}^{\′}& r_{23}^{\′}\\ r_{31}^{\′}& r_{32}^{\′}& r_{33}^{\′}\end{array}\right],$ $T^{\′}=T_P-R_P{R}_C^{-1}{T}_C=\left[\begin{array}{c}{t}_x^{\′}\\ t_y^{\′}\\ t_z^{\′}\end{array}\right],$ Then get the calibration result of system's geometric parameter:

The photocentre distance: $d=O_C{O}_P^{\′}=\sqrt{{(r_{13}^{\′}{Z}_T+t_x^{\′})}^2+{(r_{23}^{\′}{Z}_T+t_y^{\′})}^2+{(r_{33}^{\′}{Z}_T+t_z^{\′})}^2}$

Photocentre arrives the distance of reference planes: L=dctg α

The visual field width of video camera on reference planes: $\mathrm{FW}=\frac{d-f_C}{f_C}s$

In the formula, $Z_T=-\frac{r_{13}^{\′}{t}_x^{\′}+r_{23}^{\′}{t}_y^{\′}+r_{33}^{\′}{t}_z^{\′}}{r_{13}^{\′2}+r_{23}^{\′2}+r_{33}^{\′2}},$ α=arcsin (r ₁₃'), s is the width value on camera review plane, f _cBe the video camera equivalent focal length.

Claims (1)

1. profile optical measuring system composite calibration method is characterized in that composite calibration method of the present invention may further comprise the steps:

1) makes scaling board and projector calibrating grating: according to the characteristic of measuring system itself, make the plane reference plate of corresponding size, with white is background color, the black diamonds grid array (1) that is evenly distributed on it, the scaling board center indicates a black anchor point (2), the pattern of projector calibrating grating adopts identical rhombus grid array, its spacing, big or small corresponding with scaling board rhombus grid array (1), and number of arrays is lacked delegation's one row than scaling board rhombus grid array (1);

2) obtain the projector calibrating image: the measured zone that scaling board is placed the profile optical measuring system, adjust the scaling board position by laser spots and scaling board anchor point (2) that laser tube (6) sends, make the rhombus grid array (3) of projector projection just in time be positioned at the white gap of scaling board, by video camera (5) the record scaling board image of this moment, be the projector calibrating image;

3) obtain the camera calibration image: close projector (4),, be the camera calibration image by video camera (5) the record scaling board image of this moment;

4) obtain the camera calibration point pixel coordinate: according to canny criterion and SUSAN method the camera calibration image is carried out rhombus grid angle point and extract processing automatically, obtain the border of each rhombus grid in the scaling board rhombus grid array (1) earlier by canny edge extracting criterion, and according to the shape facility filtering noise of rhombus grid, obtain rhombus grid angular coordinate, with the angular coordinate is that search window is set up at the center, obtain the respective pixel point coordinate that absorbs the same value district (USAN) of nuclear area minimum in the window according to SUSAN method principle, be the camera calibration point pixel coordinate thereby obtain the corner pixels coordinate;

5) obtain the camera calibration basic data: according to the distribution characteristics of scaling board rhombus grid array (1), the basic data that the pixel coordinate and the volume coordinate of camera calibration point are mated automatic generation camera calibration;

6) linear calibration's camera parameters: with the camera calibration basic data as known conditions, according to the transformational relation between camera coordinate system and the world coordinate system:

$\left(\begin{array}{c}{x}_c\\ y_c\\ z_c\end{array}\right)=R\left(\begin{array}{c}{x}_w\\ y_w\\ z_w\end{array}\right)+T$

In the formula, (x _c, y _c, z _c) be the coordinate of calibration point in camera coordinate system, (x _w, y _w, z _w) be the volume coordinate of calibration point in world coordinate system, R is a rotation matrix, T is a translation matrix,

Adopt the linear least-squares estimation technique to calibrate camera parameters: rotation matrix R, translation matrix T and equivalent focal length f _c

7) optimize calibrating camera parameters: on the basis of step 6), adopt the Levenberg-Marquardt nonlinear optimization method to be optimized, make the value minimum of following expression, obtain the calibration result of camera interior and exterior parameter:

$\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{(x_{\mathrm{Ii}}-x_{\mathrm{Pi}})}^2+\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{(y_{\mathrm{Ii}}-y_{\mathrm{Pi}})}^2$

In the formula, x _Ii, y _IiBe the pixel coordinate of calibration point, x _Pi, y _PiFor utilizing the calculating pixel coordinate of the anti-corresponding calibration point of obtaining of calibrating parameters;

8) obtain projector calibrating point pixel coordinate: the projector calibrating image is carried out the processing same with step 4), remove the camera calibration point that has extracted in the step 4) simultaneously, obtain the corner pixels coordinate of projector rhombus grid array (3), be projector calibrating point pixel coordinate;

9) obtain the basic data of projector calibrating: calculate the volume coordinate of projector calibrating point according to the camera parameters of having demarcated, with the calibration point coordinate on the known projection grating plane basic data of generation projector calibrating automatically that is complementary;

10) projector calibrating: with reference to step 6), 7) inside and outside parameter of projector (4) is demarcated;

11) profile optical measuring system geometrical parameter calibration: according to projector coordinates is that to obtain projector coordinates be that transformational relation between Xp and the camera coordinate system Xc is for transformational relation between Xp and camera coordinate system Xc and the world coordinate system:

$X_P=R_P{R}_C^{-1}{X}_C+T_P-R_P{R}_C^{-1}{T}_C$

In the formula, R _P, T _PFor the projector coordinates of having demarcated is rotation matrix and the translation matrix of changing between Xp and the world coordinate system, R _C, T _CBe rotation matrix and the translation matrix of changing between the camera coordinate system Xc that demarcated and the world coordinate system, order $R^{\′}=R_P{R}_C^{-1}=\left[\begin{array}{ccc}{r}_{11}^{\′}& r_{12}^{\′}& r_{13}^{\′}\\ r_{21}^{\′}& r_{22}^{\′}& r_{23}^{\′}\\ r_{31}^{\′}& r_{32}^{\′}& r_{33}^{\′}\end{array}\right],$ $T^{\′}=T_P-R_P{R}_C^{-1}{T}_C=\left[\begin{array}{c}{t}_x^{\′}\\ t_y^{\′}\\ t_z^{\′}\end{array}\right],$ Then get the calibration result of system's geometric parameter:

The photocentre distance: $d=O_C{O}_P^{\′}=\sqrt{{(r_{13}^{\′}{Z}_T+t_x^{\′})}^2+{(t_{23}^{\′}{Z}_T+t_y^{\′})}^2+{(r_{33}^{\′}{Z}_T+t_z^{\′})}^2},$

Photocentre arrives the distance of reference planes: L=dctg α,

The visual field width of video camera on reference planes: $\mathrm{FW}=\frac{d-f_C}{f_C}s$

In the formula, $Z_T=-\frac{r_{13}^{\′}{t}_x^{\′}+r_{23}^{\′}{t}_y^{\′}+r_{33}^{\′}{t}_z^{\′}}{r_{13}^{\′2}+r_{23}^{\′2}+r_{33}^{\′2}},$ α=arcsin (r ₁₃'), s is the width value on camera review plane, f _cBe the video camera equivalent focal length.

Images