Highly efficient three-dimensional image acquisition method based on multi-mode composite encoding and epipolar constraint

Abstract

A highly efficient three-dimensional image acquisition method based on multi-mode composite encoding and epipolar constraint, respectively using a fast imaging mode or a high-precision imaging mode, wherein in the fast imaging mode, two phase maps having different frequencies are obtained by four stripe gratings, and a high-frequency absolute phase is obtained by means of the epipolar constraint and a left-right consistency check, and the three-dimensional image is obtained by means of a mapping relationship between the phase and three-dimensional coordinates; and in the high precision imaging mode, two phases having different frequencies are obtained by means of N+2 stripe gratings, a low-frequency absolute phase is obtained by the epipolar constraint, and the unwrapping of a high-frequency phase is assisted by means of the low-frequency absolute phase, so as to obtain the high-frequency absolute phase, and finally, the three-dimensional image is obtained by the mapping relationship between the phase and the three-dimensional coordinates. In this way, the imaging efficiency is ensured, and the imaging precision is improved.

Description

FIELD OF THE INVENTION

The invention belongs to the field of three-dimensional imaging technology, in particular to a highly efficient three-dimensional image acquisition method based on multi-mode composite encoding and the epipolar constraint

BACKGROUND OF THE INVENTION

In the field of three-dimensional imaging, the fast acquisition of high-precision three-dimensional data of target objects has always been an important technical difficulty. In early days, people used the mechanical three-coordinate measuring machine to detect target objects point by point so as to obtain three-dimensional images, but on the one hand this point-by-point contact image acquisition technique is extremely inefficient, and on the other hand the technique damages the measured object. The disadvantages of this technique make it difficult to be applied in areas such as human detection and cultural relics protection. Compared with the traditional mechanical three-dimensional image acquisition technique, the optical three-dimensional image acquisition technique has been widely used in scientific research, industrial inspection and other fields due to its advantages of non-contact and high efficiency. In recent years, with the development of digital projection equipment, fringe projection techniques in optical three-dimensional imaging methods can realize full-field imaging and has become a research hot spot (S. S. Gorthi and P. Rastogi, “Fringe projection techniques: whither we are?” Opt. Eng. 48, 133-140 (2010).). At present, two mainstream techniques in the field of fringe projection are Fourier profilometry (M. Takeda and K. Mutoh, “Fourier transform profilometry for the automatic measurement of 3-d object shapes,” Applied optics 22, 3977-3982 (1983).) and phase shift profilometry (V. Srinivasan, H.-C. Liu, and M. Halioua, “Automated phase-measuring profilometry of 3-d diffuse objects,” Applied optics 23, 3105-3108 (1984).).

Compared with Fourier profilometry, phase shift profilometry is more suitable for automated high-precision three-dimensional image acquisition due to its insensitivity to ambient light and noise and computational simplicity. Phase shift profilometry obtains the phase of a measured object by projecting a plurality of (at least three) phase shift grating fringes onto the measured object and a synchronously acquiring the grating fringes modulated by the measured object by a camera, and finally the three-dimensional image of the measured object is obtained by means of a phase-height mapping, relationship. How to ensure higher imaging precision at higher imaging efficiency has always been a focus and difficulty in the field of phase shift profilometry. In general, in phase shift profilometry, the more grating fringes that are projected, the more favorable precise phase acquisition is, thereby the more precise three-dimensional image of the measured object can be obtained. However, too many grating fringes will greatly affect the efficiency of phase acquisition and thus affect the speed (efficiency) of three-dimensional image acquisition (Chen Qian; Feng Shijie; Gu Guohua; Zuo Chao; Sun Jiayu; Yu Shiling; Shen Guochen; Li Rubin. A time phase unwrapping method based on dual-frequency three-gray scale sinusoidal grating fringe projection: China, 201410027275.4[1].2013-04-30.). For three-dimensional imaging for dynamic (quasi-static) objects, high imaging efficiency must be required in order to ensure correct imaging. While imaging efficiency has no effect on the final imaging accuracy when measuring static objects, high imaging efficiency still has absolute advantages in the three-dimensional data collection of pipe-lined bulk objects (Long Jiale; Zhang Jianmin; Fan Zhihui. A fast three-dimensional measurement system based on three-wavelength fringe projection: China, 201620177719.7[1].2016-09-07.).

Compared with other methods, three-dimensional imaging technology based on phase shift profilometry has great advantages in measurement mode, imaging efficiency and imaging precision, and it has been widely used in fields such as cultural relic protection and human body detection, but the imaging efficiency and imaging precision must be further improved to overcome the contradiction between the two and thus it can be more widely applied.

SUMMARY OF THE INVENTION

The object of the present invention is to provide a highly efficient three-dimensional image acquisition method based on multi-mode composite encoding and the epipolar constraint, which improves the precision and efficiency of three-dimensional imaging by composite coding and the epipolar constraint.

The technical solution for achieving the object of the present invention is: a highly efficient three-dimensional image acquisition method based on multi-mode composite encoding and the epipolar constraint, with same respectively using a fast imaging mode or a high-precision imaging mode, wherein in the fast imaging mode, two phase maps having different frequencies are obtained by four stripe gratings, and a high-frequency absolute phase is obtained by means of the epipolar constraint and a left-right consistency test, and the three-dimensional image is obtained by means of a mapping relationship between a phase and three-dimensional coordinates; and in the high precision imaging mode, two phases having different frequencies are obtained by means of N+2 stripe gratings, a low-frequency absolute phase is obtained by the epipolar constraint, and the unwrapping of a high-frequency phase is assisted by means of the low-frequency absolute phase, so as to obtain the high-frequency absolute phase, and finally, the three-dimensional image is obtained by the mapping relationship between the phase and the three-dimensional coordinates.

Compared with prior arts, the present invention has significant advantages: (1) The fast imaging mode of the present invention utilizes four grating fringes to obtain a three-dimensional image of the measured object, and compared with the prior art, the combination of the four composite coding grating fringes ensures the high efficiency of three-dimensional image measurement; at the same time, the introduction of the epipolar constraint in the binocular vision enables the technique to obtain high-precision phases with up to 64 fringe periods by using four composite grating stripes, and the high precision of the three-dimensional image is ensured and finally, since the high-frequency absolute phase is directly solved by the epipolar constraint, it does not rely on the low-frequency absolute phase, which avoids the inaccuracy of the three-dimensional image caused by the difference in modulation degree between the two frequency stripe gratings. As shown in FIG. 2, the imaging result of the fast imaging mode of the present invention is finer and has higher imaging precision under the premise of using four fringes. Therefore, the fast imaging mode of the present invention improves imaging accuracy while ensuring imaging efficiency. (2) The high-precision imaging mode of the invention acquires the three-dimensional image of the measured object through the N+2 grating fringes, compared with the prior art, the introduction of the epipolar constraint enables the 8-period low-frequency absolute phase to be directly obtained through the two low-frequency grating fringes, which greatly avoids the redundancy in the prior art that utilizes multi-frame grating fringes (usually much larger than 2) to obtain low-frequency absolute phase, thus improving measurement efficiency. On the other hand, the coding pattern of N 128 fringe periods ensures that the accuracy of the finally acquired three-dimensional image is not lower than the accuracy of the prior art. As shown in FIG. 4, the invention can realize the three-dimensional imaging with an accuracy of 50 um with only five fringes. Therefore, the high-precision imaging mode of the invention greatly improves the imaging efficiency while ensuring imaging accuracy.

The invention is further described in detail below with reference to the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is the flow diagram of the steps of a fast imaging mode.

FIG. 2 is the test result of the fast imaging mode, and FIG. 2(a) and FIG. 2(b) are respectively the three-dimensional image of a dynamic object acquired by the method and the prior art, and FIG. 2(c) and FIG. 2(f) are partial details of FIG. 2(a), FIG. 2(d) and FIG. 2(g) are partial details of FIG. 2(b), and FIG. 2(e) and FIG. 2(h) are the comparison of partial details.

FIG. 3 is the flow diagram of the steps of a high precision imaging mode.

FIG. 4 is the test result of a high-precision imaging mode. FIG. 4(a) shows the result of the three-dimensional imaging of a standard ceramic plate, and FIG. 4(b) is a detail display of a line in FIG. 4(a).

DESCRIPTION OF THE PREFERRED EMBODIMENT

Combining FIG. 1 with FIG. 3, the present invention is based on a highly efficient three-dimensional image acquisition method based on multi-mode composite encoding and the epipolar constraint, respectively using a fast imaging mode or a high-precision imaging mode; the two imaging modes share the same imaging system and the same imaging system calibration results and use the epipolar constraint combing with the auxiliary phase information as the core technology of phase unwrapping. Compared with a traditional multi-frequency time phase unwrapping algorithm, the two imaging modes are greatly improved in efficiency; and compared with a traditional spatial phase unwrapping algorithm, the two imaging modes are greatly improved in efficiency and accuracy. The steps of the fast imaging mode are as follows:

step one, imaging system calibration;

step two, generating, projecting and collecting four dual-frequency grating fringes;

step three, analyzing the grating fringes collected by left camera and right camera respectively so as to obtain a set of high frequency phases and a set of low frequency phases;

step four, searching for each point on the high frequency phase maps of the left camera by using the epipolar constraint, that is, the corresponding point of the original point in space, and removing some error points by the depth constraint;

step five, projecting the remaining spatial corresponding points onto the high-frequency and low-frequency phase maps of the right camera, and determining the final corresponding point by the phase difference between the original points and the corresponding points, so as to obtain a high-frequency absolute phase;step six, acquiring a three-dimensional image according to the absolute phase, so as to realize efficient and precise acquisition of three-dimensional images of dynamic scenesthe high-precision imaging mode comprises the following steps:step one, imaging system calibration;step two, generating, projecting, and acquiring N+2 double-frequency grating stripe patterns;step three, analyzing the grating fringes collected by left camera and right camera respectively so as to obtain a set of high frequency phases and a set of low frequency phases;step four, using the epipolar constraint to search for each point on the high frequency phase maps of the left camera, that is, the corresponding point of the original point in space, and removing some error points by the depth constraint;step five, the remaining space corresponding points are projected to the low-frequency phase map of the right camera, projecting the remaining spatial corresponding points onto the low-frequency phase maps of the right camera, and determining the final corresponding point by the phase difference between the original points and the corresponding points, so as to obtain a high-frequency absolute phase through a low-frequency absolute phase;step six, acquiring a three-dimensional image according to the absolute phase, so as to realize efficient and precise acquisition of three-dimensional images of static scenes.

The process of the two imaging modes is described in detail below.

The flow diagram of the steps of the fast imaging mode of the present invention is shown in FIG. 1, and the specific implementation steps are as follows:

step one, imaging system is calibrated.

The imaging system comprises a computer, a left camera, a right camera, and a projector, wherein the left camera, the right camera and the projector are respectively connected to the computer through data lines, and the projector is connected with the left camera and the right camera through trigger lines; After the imaging system is built, the calibration method mentioned in “A flexible new technique for camera calibration.” by Z. Zhang (Z Zhang, “a flexible new technique for camera calibration.” IEEE Transactions on pattern analysis and machine intelligence. 22 (11), 1330-1334 (2000).) is used for the imaging system calibration so as to obtain calibration parameters of the left camera, the right camera and the projector in a world coordinate system, wherein the calibration parameters comprise a scaling parameter, a translation parameter, a rotation parameter and a distortion parameter between the pixel coordinate system and the world coordinate system.

step two, four dual-frequency grating fringes are generated, projected and collected. the four dual-frequency grating fringes generated by the computer through MatLab are two sinusoidal stripe gratings and two triangular wave stripe gratings, and the stripe gratings are as follows:I₁(x,y)=A(x,y)+B(x,y)sin [πF_H(2x/W−1)]I₂(x,y)=A(x,y)+B(x,y)cos [πF_H(2x/W−1)]I₃(x,y)=A(x,y)+B(x,y)tri[(2F_Lx/W−1)]I₄(x,y)=A(x,y)−B(x,y)tri[(2F_Lx/W−1)]where I_i(x,y) represents the intensity of the grating fringe at the pixel coordinates (x,y) of the generated image, i=1, 2, 3, 4 representing the i-th grating fringe image, A is the image DC component, B is the amplitude, and tri is the triangular wave function with threshold interval [−1, 1], F_H, F_L is the number of fringe periods included in I₁, I₂ and I₃, I₄. W is the pixel width of the entire grating fringe image, A=B=127.5, the values for F, are respectively 64 and 9, and the range of values for x is 0 to W−1; the gratings are synchronously collected by the left camera and the right camera after being projected by the projector. For the sake of simplicity, only the left camera is analyzed here. The analysis process of the right camera is the same as that of the left camera. The grating fringes collected by the left camera are as follows:I₁^c(x^c,y^c)=α(x^c,y^c)[A(x^c,y^c)+B(x^c,y^c)sin Φ_H(x^c,y^c)]+α(x^c,y^c)β₁(x^c,y^c)+β₂(x^c,y^c)I₂^c(x^c,y^c)=α(x^c,y^c)[A(x^c,y^c)+B(x^c,y^c)cos Φ_H(x^c,y^c)]+α(x,y)β₁(x^c,y^c)+β₂(x^c,y^c)I₃x(x^c,y^c)=α(x^c,y^c)[A(x^c,y^c)+B(x^c,y^c)Φ_L(x^c,y^c)]+α(x^c,y^c)β₁(x^c,y^c)+β₂(x^c,y^c)I₄(x^cy^c)=α(x^c,y^c)[A(x^c,y^c)−B(x^c,y^c)Φ_L(x^c,y^c)]+α(x^c,y^c)β₁(x^c,y^c)+β₂(x^c,y^c)where I₁^c(x^c,y^c) is the grating fringe image actually captured by the left camera, i=1, 2, 3, 4. (x^c,y^c) is the pixel coordinates of the image captured by the camera, and α is the surface reflectance of the measured object, β₁ is the reflected ambient light, β₂ is the ambient light directly into the camera, Φ_H(x^c,y^c) is the phase included in the grating fringe diagram I₁^c and the grating fringe diagram I₂^c, Φ_L (x^c,y^c) is the phase included in the grating fringe diagram I₃^c and the grating fringe diagram I₄^c; assumingA^c=α(x^c,y^c)A(x^c,y^c)+α(x^c,y^c)β₁(x^c,y^c)+β₂(x^c,y^c),B^c=α(x^c,y^c)B(x^c,y^c),and leaving out (x^c,y^c) the above four equations can be reduced to:I₁^c=A^c+B^c sin Φ_H I₂^c=A^c+B^c cos Φ_H I₃^c=A^c+B^cΦ_L I₄^c=A^c−B^cΦ_L

The acquisition process of the right camera is the same as the acquisition process of the left camera.

Step three, the grating fringes collected are analyzed by the left camera and the right camera to obtain respectively a set of high-frequency phases and a set of low-frequency phases.

According to the image acquired by the left camera in step two, two sets of phases are obtained as follows:

${\varphi }_H={\tan }^{-1}\frac{2I_1^c-I_3^c-I_4^c}{2I_2^c-I_3^c-I_4^c}$ ${\varphi }_L=\frac{I_3^c-I_4^c}{2B^c},B^c=\frac{\sqrt{{\left(2I_1^c-I_3^c-I_4^c\right)}^2+{\left(2I_2^c-I_3^c-I_4^c\right)}^2}}{2}$ where ϕ_H is a high-frequency wrapped phase and ϕ_L is a low-frequency wrapped phase, and the same steps as those of the left camera can be used to determine that the phases corresponding to the stripe gratings collected by the right camera are ϕ_H′,ϕ_L′.Step four, the epipolar constraint is used to search for each point on the high frequency phase maps of the left camera, that is, the corresponding point of the original point in space, and removing some error points by the depth constraint.

For any point p on ϕ_H, that is, the original point, its cycle order k has F_H possibilities, which means that its absolute phaseΦ_H=ϕ_H+2kπ,k∈[0,F_H−1]has F_H different values and from the following formula:

$x_p=\frac{{\Phi }_{H}W}{2\pi F_H}$ $Z_p=M_Z+\frac{N_Z}{C_Z{x}_p+1}$

It can be seen that p corresponds to F_H points in the three-dimensional space, that is, F_H corresponding points, wherein M_Z, N_Z and C_Z are derived from the calibration parameters obtained in step one (K. Liu, Y. Wang, et al “Dual-Frequency pattern scheme for high-speed 3-D shape measurement.” Optics express. 18(5), 5229-5244 (2010).). There is at most one correct corresponding point in the F_H corresponding points. The key to 3D imaging is to identify the only correct corresponding point among the F_H corresponding points. Considering the limited measurement space range in the actual 3D imaging system, the effective range of the left camera, the right camera and the projector is preset as (Z_min, Z_max), such as Z_min=−200 mm, Z_max=200 mm, all k which make Z exceeds this preset range and its corresponding spatial points will be confirmed as error points and be eliminated: after this step is processed, the correct cycle order k of p and the range of its spatial corresponding points will be reduced to F_H′, where F_H′<<F_H.

Step five, the residuary spatial corresponding points are projected onto the high frequency and low frequency phase maps of the right camera, and the final corresponding point is determined by the phase difference between the original point and the corresponding point, thus obtaining the high frequency absolute phase.

First of all, the residuary spatial points in (Z_min, Z_max) of step four are projected onto the imaging surface of the right camera, and F_H′ two-dimensional corresponding points on the imaging surface of the right camera are obtained. In fact, the wrapped phase ϕ_H′,ϕ_L′ of the correct corresponding point p′ and the wrapped phase ϕ_H, ϕ_L of the original point should be very close, so that, the two-dimensional corresponding points whose ϕ_diff exceed the threshold of 0.5 rad are further excluded by the formulaϕ_diff=ϕ_H(p)−ϕ_H′(p′)(the rad is the radian unit and the threshold size is a prior determination) where p′ is the corresponding point of p on the right camera, ϕ_diff is the difference between the original point and the corresponding point in the wrapped phases ϕ_H and ϕ_H′, the range of the correct corresponding points is reduced from F_H′ to F_H″; finally, selecting the corresponding point that makes ϕ_L(p)−ϕ_L′(p′) the smallest among the residuary F_H″ corresponding points, and considering the point to be the correct corresponding point; then the cycle order k corresponding to the point is the correct cycle order and thus the unique Φ_H of the original point is confirmed.Step six, acquiring a three-dimensional image according to the absolute phase.

The three-dimensional image coordinates are obtained by combining the absolute phase Φ_H obtained in step five with the following formula:

$x_p=\frac{{\Phi }_{H}W}{2\pi N_L}$ $Z_p=M_Z+\frac{N_Z}{C_Z{x}_p+1}$ $X_p=E_X{Z}_p+F_X$ $Y_p=E_Y{Z}_p+F_Y$ Where E_X, F_X, E_Y, F_Y are obtained from the calibration parameters in step one (K. Liu, Y. Wang, et al “Dual-frequency pattern scheme for high-speed 3-D shape measurement. “Optics express. 18 (5), 5229-5244 (2010).), X_P, Y_P, Z_p are the three-dimensional coordinates of the measured object, and thus the three-dimensional image of the measured object can be obtained. It can be seen through the above steps that the fast imaging mode of the present invention utilizes four grating fringes to obtain a three-dimensional image of the measured object, and compared with the prior art, the combination of the four composite coding grating fringes ensures the high efficiency of three-dimensional image measurement; at the same time, the introduction of the epipolar constraint in the binocular vision enables the technique to obtain high-precision phases with up to 64 fringe periods by using four composite grating stripes, and the high precision of the three-dimensional image is ensured and finally, since the high-frequency absolute phase is directly solved by the epipolar constraint, it does not rely on the low-frequency absolute phase, which avoids the inaccuracy of the three-dimensional image caused by the difference in modulation degree between the two frequency stripe gratings.

In order to test the effect of a highly efficient three-dimensional image acquisition method based on multi-mode composite encoding and the epipolar constraint, the present invention gives two sets of test results. FIG. 2(a) and FIG. 2(b) are respectively the three-dimensional image of a dynamic object acquired by the method and the prior art, and FIG. 2(c) and FIG. 2(f) are partial details of FIG. 2(a), FIG. 2(d) and FIG. 2(g) are partial details of FIG. 2(b), and FIG. 2(e) and FIG. 2(h) are the comparison of partial details.

It can be seen from these experimental results that the fast imaging mode of the present invention retains more detailed measurement results than the prior art, and the imaging precision is greatly improved while ensuring high-efficiency three-dimensional imaging.

The flow diagram of the steps of the high-precision imaging mode of the present invention is shown in FIG. 3, and the specific implementation steps are as follows:

Step one, imaging system is calibrated

The imaging system comprises a computer, a left camera, a right camera, and a projector, wherein the left camera, the right camera and the projector are respectively connected to the computer through data lines, and the projector is connected with the left camera and the right camera through trigger lines; After the imaging system is built, the calibration method mentioned in “A flexible new technique for camera calibration.” by Z. Zhang (Z. Zhang, ‘a flexible new technique for camera calibration.” IEEE Transactions on pattern analysis and machine intelligence. 22 (11), 1330-1334 (2000).) is used for the imaging system calibration so as to obtain calibration parameters of the left camera, the right camera and the projector in a world coordinate system.

Step two, N+2 double-frequency grating stripe patterns are generated, projected, and acquired.

N+2 stripe gratings are two low frequency sinusoidal stripe gratings plus N high frequency sinusoidal stripe gratings, where N≤3, for the sake of simplicity, taking N=3 as an example, the stripe gratings are as follows:J₁(x,y)=A(x,y)+B(x,y)cos [πN_H(2x/W−1)]J₂(x,y)=A(x,y)+B(x,y)cos [πN_H(2x/W−1)+2π/3]J₃(x,y)=A(x,y)+B(x,y)cos [πN_H(2x/W−1)+/4π3]J₄(x,y)=A(x,y)+B(x,y)sin [πN_L(2x/W−1)]J₅(x,y)=A(x,y)+B(x,y)cos [πN_L(2x/W−1)]where J_i(x,y) represents the intensity of the grating fringe at the pixel coordinates (x,y) of the generated image, i=1, 2, 3, 4, 5, representing the i-th grating fringe image, and A is the DC component of the image, B is the amplitude, N_H, N_L are the number of fringe periods included in J₁˜J₃ and J₄, J₅ respectively, W is the pixel width of the entire grating fringe image, A=B=127.5, and the values for N_H, N_L are respectively 128 and 8, and the range of values for x is 0 to W−1; the gratings are synchronously collected by the left camera and the right camera after being projected by the projector. For the sake of simplicity, only the left camera is analyzed here. The analysis process of the right camera is the same as that of the left camera. The grating fringes collected by the left camera are as follows:J₁^c(x^c,y^c)=α(x^c,y^c)[A(x^c,y^c)+B(x^c,y^c)cos Ψ_H(x^c,y^c)]+α(x^c,y^c)β₁(x^c,y^c)+β₂(x^c,y^c)J₂^c(x^c,y^c)=α(x^c,y^c)[A(x^c,y^c)+B(x^c,y^c)cos(Ψ_H(x^c,y^c)+2π/3)]+α(x^c,y^c)β₁(x^c,y^c)+β₂(x^c,y^c)J₃^c(x^c,y^c)=α(x^c,y^c)[A(x^c,y^c)+B(x^c,y^c)cos(Ψ_H(x^c,y^c)+4π/3)]+α(x^c,y^c)β₁(x^c,y^c)+β₂(x^c,y^c)J₄^c(x^c,y^c)=α(x^c,y^c)[A(x^c,y^c)+B(x^c,y^c)sin Ψ_I(x^c,y^c)]+α(x^c,y^c)β₁(x^c,y^c)+β₂(x^c,y^c)J₅^c(x^c,y^c)=α(x^c,y^c)[A(x^c,y^c)+B(x^c,y^c)cos Ψ_I(x^c,y^c)]+α(x^c,y^c)β₁(x^c,y^c)+β₂(x^c,y^c)where J₁^c(x^c,y^c) is a grating fringe image actually captured by the left camera, i=1, 2, 3, 4, 5, (x^c,y^c) is the pixel coordinates of the image captured by the camera, and α is the surface reflectance of the measured object, β₁ the reflected ambient light, β₂ is the ambient light directly into the camera, Ψ_H (x^c,y^c) is the phase included in the grating fringe diagram I₁^c and the grating fringe diagram I_x³, Ψ_L (x^c,y^c) is the phase included in the grating fringe diagram I₄^c and the grating fringe diagram I₅^c; assumingA′=α(x^c,y^c)A(x^c,y^c)+α(x^c,y^c)β₁(x^c,y^c)+β₂(x^c,y^c),B^c=α(x^c,y^c)B(x^c,y^c),and leaving out (x^c,y^c), the above four equations can be reduced to:J₁^c=A^c+B^c cos Ψ_H J₂^c=A^c+B^c cos(Ψ_H+2π/3)J₃^c=A^c+B^c cos(Ψ_H+4π/3)J₄^c=A^c+B³ sin Ψ_I J₅^c=A^c+B^c cos Ψ_I

The acquisition process of the right camera is the same as that of the above left camera.

Step three, the grating fringes collected by left camera and right camera respectively are analyzed so as to obtain a set of high frequency phases and a set of low frequency phases, where the phase of the left camera is as follows:

${\psi }_H=\arctan \frac{\sqrt{3}\left(J_1^c-J_3^c\right)}{\left(2J_2^c-J_1^c-J_3^c\right)}$ ${\psi }_L=\arctan \frac{J_4^c-A^c}{J_5^c-A^c},A^c=\frac{J_1^c+J_2^c+J_3^c}{3}$ Where ψ_H is a high-frequency wrapped phase and ψ_L is a low-frequency wrapped phase, and the same steps as those of the left camera can be used to determine that the phases corresponding to the stripe gratings collected by the right camera are ψ_H′,ψ_L′.Step four, using the epipolar constraint to search for the corresponding spatial point of each point (original point) on the high frequency phase map of the left camera, and eliminating some error points by the depth constraint.

For any point q on ϕ_L, that is, the original point, its cycle order k has N_L possibilities, which means that its absolute phaseΨ_L=ψ_L+2π,l∈[0,N_L−1]has N_L different values and from the following formula:

$x_q=\frac{{\Psi }_{L}W}{2\pi N_L}$ $Z_q=M_Z+\frac{N_Z}{C_Z{x}_q+1}$

It can be seen that q corresponds to N_L points in the three-dimensional space, that is, corresponding points, wherein M_Z, N_Z and C_Z are derived from the calibration parameters obtained in step one (K. Liu. Y. Wang, et al “Dual-frequency pattern Scheme for high-speed 3-D shape measurement.” Optics express. 18(5). 229-5244 (2010).); the key to 3D imaging is to identify the only correct corresponding point among the N_L corresponding points. Considering the limited measurement space range in the actual 3D imaging system, the effective range of the left camera, the right camera and the projector is preset as (Z_min, Z_max), such as Z_min=−200 mm, Z_min=200 mm, all l which make Z_q exceeds this preset range and its corresponding spatial points will be confirmed as error points and be eliminated; after this step, the correct cycle level of q and the range of its spatial corresponding points will be reduced to N_L′, where N_L′<<N_L.

Step 5, the residuary spatial corresponding points are projected onto the low frequency phase map of the right camera, and the final corresponding point is determined by the phase difference between the original point and the corresponding point, and the high frequency absolute phase is obtained by means of the low frequency absolute phase.

First of all, the residuary spatial points in (Z_min, Z_max) of step four are projected onto the imaging surface of the right camera, and F_H′ two-dimensional corresponding points on the imaging surface of the right camera are obtained.

In fact, the wrapped phase ψ_L′ the correct corresponding point q′ and the wrapped phase ψ_L of the original point should be very close, since N_L′<<N_L, the corresponding point making ψ_L (p)−ψ_L′(p′) the smallest can be selected from N_L′ corresponding points, and considering that the point is a correct corresponding point, then the corresponding, cycle level l of the correct point is the correct cycle level, so that the unique Ψ_L of the original point is confirmed.

Finally, the final high frequency absolute phase Ψ_H is obtained by the following equation

$l=\mathrm{round}\left(\frac{N_H{\Psi }_L/N_L-{\psi }_H}{2\pi }\right)$ ${\psi }_H={\psi }_H+2l\pi$ Step six, a three-dimensional image is acquired according to the absolute phase.

The coordinates of the three-dimensional image are obtained by combining the absolute phase Ψ_H obtained in step five with the following formula:

$x_q=\frac{{\Psi }_{H}W}{2\pi N_L}$ $Z_q=M_Z+\frac{N_Z}{C_Z{x}_q+1}$ $X_q=E_X{Z}_q+F_X$ $Y_q=E_Y{Z}_q+F_Y$ where E_X, F_X, E_Y, F_Y are obtained from the calibration parameters in step one (K. Liu, Y. Wang, et al “Dual-frequency pattern scheme for high-speed 3-D shape measurement.” Optics express. 18 (5), 5229-5244 (2010).), X_q, Y_q, Z_q are the three-dimensional coordinates of the measured object, and thus obtaining the three-dimensional image of the measured object.

From the above four steps, it can be seen that the high-precision imaging mode of the invention acquires the three-dimensional image of the measured object through the N+2 grating fringes, compared with the prior art, the introduction of the epipolar constraint enables the 8-cycle low-frequency absolute phase to be directly obtained through the two low-frequency grating fringes, which greatly avoids the redundancy in the prior art that utilizes multi-frame grating fringes (usually much larger than 2) to obtain low-frequency absolute phase, thus improving measurement efficiency. On the other hand, the coding pattern of N 128 fringe periods ensures that the precision of the finally acquired three-dimensional image is not lower than the precision of the prior art.

In order to test the effect of a highly efficient three-dimensional image acquisition method based on multi-mode composite encoding and the epipolar constraint, one set of experimental results are presented.

FIG. 4(a) shows the result of the three-dimensional imaging of a static object in a high-precision imaging mode, FIG. 4(b) shows the details of a line in FIG. 4(a), and From FIG. 4, it can be seen the high-precision imaging mode of the invention can achieve imaging precision of the static object up to 50 um.

Claims

1. A highly efficient three-dimensional image acquisition method based on multi-mode composite encoding and epipolar constraint, comprising the step of: using a fast imaging mode, comprising: obtaining two phase maps having different frequencies by four stripe gratings;obtaining a high-frequency absolute phase by the epipolar constraint and a left-right consistency test; andobtaining a three-dimensional image by a mapping relationship between the high frequency absolute phase and three-dimensional coordinates; orusing a high-precision imaging mode, comprising: obtaining two phases having different frequencies by N+2 stripe gratings, wherein the N+2 stripe gratings are two low frequency sinusoidal stripe gratings plus N high frequency sinusoidal stripe gratings, and N is greater than or equal to 3;obtaining a low-frequency absolute phase by the epipolar constraint, and assisting an unwrapping of a high-frequency phase by the low-frequency absolute phase;obtaining the high-frequency absolute phase; andobtaining the three-dimensional image by the mapping relationship between the high-frequency absolute phase and the three-dimensional coordinates.

2. The method according to claim 1, wherein the step of using the fast imaging mode comprises the steps of: 1) performing a calibration of an imaging system;2) generating, projecting and collecting four dual-frequency grating fringes;3) analyzing the grating fringes collected by a left camera and a right camera of the imaging system respectively so as to obtain a set of high frequency phases and a set of low frequency phases;4) searching for each point on a high frequency phase map of the left camera by using the epipolar constraint, said each point to be searched for being corresponding points of an original point in space, and removing some error points with depth constraint;5) a projecting remaining spatial corresponding points onto high-frequency and low-frequency phase diagrams of the right camera, and determining a final corresponding point by a phase difference between original points and the corresponding points, so as to obtain the high-frequency absolute phase; and6) acquiring the three-dimensional image according to the high-frequency absolute phase, so as to realize efficient and precise acquisition of three-dimensional images of dynamic scenes.

3. The method according to claim 2, wherein the imaging system comprises a computer, the left camera the right camera, and a projector, the left camera, the right camera and the projector are respectively connected to the computer through data lines, and the projector is connected with the left camera and the right camera through trigger lines, and wherein the imaging system is calibrated so as to obtain calibration parameters of the left camera, the right camera and the projector in world coordinate system, and the calibration parameters comprise a scaling parameter, a translation parameter, a rotation parameter and a distortion parameter between a pixel coordinate system and the world coordinate system.

4. The method according to claim 2, wherein in the step 2), the four dual-frequency grating fringes generated by the computer through MatLab are two sinusoidal stripe gratings and two triangular wave stripe gratings, and the stripe gratings are as follows: I1(x,y)=A(x,y)+B(x,y)sin [πFH(2x/W−1)]I2(x,y)=A(x,y)+B(x,y)cos [πFH(2x/W−1)]I3(x,y)=A(x,y)+B(x,y)tri[(2FLx/W−1)]I4(x,y)=A(x,y)−B(x,y)tri[(2FLx/W−1)]where Ii(x,y) represents an intensity of the grating fringe at pixel coordinates (x,y) of the generated image, i=1, 2, 3, 4, representing the i-th grating fringe image, A is an image DC component, B is an amplitude, and tri is a triangular wave function with threshold interval [−1, 1], FH, FL is respectively number of fringe periods included in I1, I2 and I3, I4, W is a pixel width of a entire grating fringe image, A=B=127.5, values for FH, FL are respectively 64 and 9 and a range of values for x is 0 to W−1,wherein the gratings are synchronously captured by the left camera and the right camera after being projected by the projector,wherein the grating fringes collected by the left camera are as follows: I1c(xc,yc)=α(xc,yc)[A(xc,yc)+B(xc,yc)sin ΦH(xc,yc)]+α(xc,yc)β1(xc,yc)+β2(xc,yc)I2c(xc,yc)=α(xc,yc)[A(xc,yc)+B(xc,yc)cos ΦH(xc,yc)]+α(x,y)β1(xc,yc)+β2(xc,yc)whereI3c(xc,yc)=α(xc,yc)[A(xc,yc)+B(xc,yc)ΦL(xc,yc)]+α(xc,yc)β1(xc,yc)+β2(xc,yc)I4c(xcyc)=α(xc,yc)[A(xc,yc)−B(xc,yc)ΦL(xc,yc)]+α(xc,yc)β1(xc,yc)+β2(xc,yc)Iic (xc,yc) is a grating fringe image actually captured by the left camera, i=1, 2, 3, 4,(xc,yc) is pixel coordinates of the image captured by the camera, and α is a surface reflectance of a measured object, β1 is a reflected ambient light, β2 is an ambient light directly into the camera, ΦH (xc,yc) is a phase included in the grating fringe diagram I1c and the grating fringe diagram I2c, ΦL(xc,yc) is the phase included in the grating fringe diagram I3c and the grating fringe diagram I4c,wherein assuming Ac=α(xc,yc)A(xc,yc)+α(xc,yc)β1(xc,yc)+β2(xc,yc),Bc=α(xc,yc),B(xc,yc),and omitting (xc,yc), the above four equations can be rewritten as: I1c=Ac+Bc sin ΦH I2c=Ac+Bc cos ΦH,andI3c=Ac+BcΦL I4c=Ac−BcΦL wherein an acquisition process of the right camera is the same as an acquisition process of the left camera.

5. The method according to claim 2, wherein the grating fringes collected by the left camera and the right camera respectively are analyzed so as to obtain a set of high frequency phases and a set of low frequency phases, where the phases of the left camera is as follows:

6. The method according to claim 2, wherein in the step 4), for any point p on ϕH, being the original point, its cycle order k has FH possibilities, and its absolute phase ΦH=ϕH+2kπ,k∈[0,FH−1]has FH different values and from the following formula:

7. The method according to claim 2, wherein the step 5) further comprises the steps of: projecting residuary spatial points in (Zmin, Zmax) in the step 4) onto an imaging surface of the right camera, thus obtaining FH′ two-dimensional corresponding points on the imaging surface of right camera;excluding two-dimensional corresponding points whose ϕdiff exceed a threshold 0.5 rad by ϕdiff ϕH (p)−ϕH′(p′), where p′ is the corresponding point of p on the right camera, ϕdiff is a difference between the original point and the corresponding points in wrapping phases ϕH and ϕH′, a range of correct corresponding points is reduced from FH′ to FH″; andselecting a corresponding point that makes ϕL(p)−ϕL′(p′) the smallest among the residuary FH″ corresponding points, and considering the point to be a correct corresponding point; then a cycle order k corresponding to the point is a correct cycle order, whereby the unique ΦH of the original point is confirmed.

8. The method according to claim 2, wherein the step 6) further comprises the step of: obtaining the coordinates of the three-dimensional image by combining the absolute phase ΦH obtained in the step 5) with the following formula:

9. The method according to claim 1, wherein the step of using the high-precision imaging mode comprises the following steps: a) performing a calibration of an imaging system;b) generating, projecting, and acquiring N+2 double-frequency grating stripe patterns, where N is greater than or equal to 3;c) analyzing the grating fringes collected by a left camera and a right camera of the imaging system so as to obtain respectively a set of high frequency phases and a set of low frequency phases;d) searching for each point on high frequency phase maps of the left camera by using the epipolar constraint, said each point to be searched for being corresponding points of an original point in space, and removing some error points with depth restraint;e) projecting remaining spatial corresponding points onto low-frequency phase diagrams of the right camera, and determining a final corresponding point by a phase difference between the original points and the corresponding points, so as to obtain a high-frequency absolute phase through a low-frequency absolute phase; andf) acquiring a three-dimensional image according to the absolute phase, so as to realize efficient and precise acquisition of three-dimensional images of static scenes.

10. The method according to claim 9, wherein in the step b), N+2 the N+2 stripe gratings are two low-frequency sinusoidal stripe gratings plus N high-frequency frequency sinusoidal stripe gratings, where N≥3, and when N=3, the stripe gratings are as follows: J1(x,y)=A(x,y)+B(x,y)cos [πNH(2x/W−1)]J2(x,y)=A(x,y)+B(x,y)cos [πNH(2x/W−1)+2π/3]J3(x,y)=A(x,y)+B(x,y)cos [πNH(2x/W−1)+4π/3]J4(x,y)=A(x,y)+B(x,y)sin [πNL(2x/W−1)]J5(x,y)=A(x,y)+B(x,y)cos [πNL(2x/W−1)]where Ji(x,y) represents an intensity of the grating fringe at pixel coordinates (x,y) of the generated image, i=1, 2, 3, 4, 5,representing the i-th grating fringe image, and A is a DC component of the image, B is an amplitude, NH, NL are a number of fringe periods included in J1˜J3 and J4, J5 respectively, W is a pixel width of the entire grating fringe image, A=B=127.5, and values for NH, NL are respectively 128 and 8, and a range of values for x is 0 to W−1,wherein the gratings are synchronously collected by the e camera and the right camera after being projected by the projector,wherein the grating fringes collected by the left camera are as follows: J1c(xc,yc)=α(xc,yc)[A(xc,yc)+B(xc,yc)cos ΨH(xc,yc)]+α(xc,yc)β1(xc,yc)+β2(xc,yc)J2c(xc,yc)=α(xc,yc)[A(xc,yc)+B(xc,yc)cos(ΨH(xc,yc)+2π/3)]+α(xc,yc)β1(xc,yc)+β2(xc,yc)J3c(xc,yc)=α(xc,yc)[A(xc,yc)+B(xc,yc)cos(ΨH(xc,yc)+4π/3)]+α(xc,yc)β1(xc,yc)+β2(xc,yc)J4c(xc,yc)=α(xc,yc)[A(xc,yc)+B(xc,yc)sin ΨL(xc,yc)]+α(xc,yc)β1(xc,yc)+β2(xc,yc)J5c(xc,yc)=α(xc,yc)[A(xc,yc)+B(xc,yc)cos ΨL(xc,yc)]+α(xc,yc)β1(xc,yc)+β2(xc,yc)where Jic(xc,yc) is a grating fringe image actually captured by the left camera, i=1, 2, 3, 4, 5 (xc,yc), is pixel coordinates of the image captured by the camera, and α is a surface reflectance of a measured object, β1 is a reflected ambient light, β2 is an ambient light directly into the camera, ΨH (xc,yc) is a phase included in the grating fringe map I1c and the grating fringe map I3c, ΨL(xc,yc) is a phase included in the grating fringe diagram I4c the grating fringe diagram I5c,wherein assuming Ac=α(xc,yc)A(xc,yc)+α(xc,yc)β1(xc,yc)+β2(xc,yc),Bc=α(xc,yc)B(xc,yc),and omitting (xc,yc), the above equations can be rewritten as: J1c=Ac+Bc cos ΨH J2c=Ac+Bc cos(ΨH+2π/3)J3c=Ac+Bc cos(ΨH+4π/3),andJ4c=Ac+Bc sin ΨL J5c=Ac+Bc cos ΨL wherein an acquisition process of the right camera is the same as an acquisition process of the left camera.

11. The method according to claim 9, wherein in the step c), the grating fringes collected by left camera and right camera respectively are analyzed so as to obtain a set of high frequency phases and a set of low frequency phases, where the phase of the left camera is as follows:

12. The method according to claim 9, wherein in the step d), for any point q on ϕL, being the original point, its cycle order k has NL possibilities, and its absolute phase ΨL=ψL+2lπ,l∈[0,NL−1]has NL different values and from the following formula:

13. The method according to claim 9, wherein the step e) further comprises the steps of: projecting residuary spatial points in (Zmin, Zmax) in step d) onto an imaging surface of the right camera, and obtaining NL′ two-dimensional corresponding points on the imaging surface of the right camera;confirming an unique ΨL of the original point, wherein since NL′<<NL, a corresponding point making ψL(p)−ψL′(p′) the smallest can be selected from NL′ corresponding points, and considering that the point is a correct corresponding point, then the corresponding cycle order l of the correct point is a correct cycle order ΨL; andobtaining the final high frequency absolute phase ΨH by the following equation:

14. The method according to claim 9, wherein the step f) further comprises reconstructing the three-dimensional image according to the absolute phase by: obtaining the coordinates of the three-dimensional image by combining the absolute phase ΨH obtained in the step e) with the following formula: