CN110248179A - Camera pupil aberration correcting method based on light field coding - Google Patents

Abstract

The invention belongs to Technology of Precision Measurement field, specially a kind of camera pupil aberration correcting method based on light field coding.Step of the present invention are as follows: in order to overcome the pupil aberration of pinhole camera, determine every imaging ray, using four-dimensional function (x, y, u, v) description imaging light field, wherein (x, y) indicates picpointed coordinate, (u, v) it indicates coordinate of the light on pupil, namely indicates radiation direction；Binaryzation pupil coding is carried out using digital micromirror array；Based on compressed sensing thought, light field matrix is solved using the optimization of full variation；Calculate the corresponding pupil coordinate (u, v) of each picture point (x, y), in conjunction with camera calibration as a result, can effectively correcting camera pupil aberration, so that building one meets the camera model of actual imaging process.The present invention can effectively eliminate pupil aberration caused by pin-hole imaging is assumed, overcome the direction ambiguousness of monocular vision, the accurate direction for calculating imaging ray；It is significant for the measurement accuracy for improving photogrammetric technology.

Description

Camera pupil aberration correcting method based on light field coding

Technical field

The invention belongs to Technology of Precision Measurement field, specially a kind of camera pupil aberration correction side based on light field coding Method.

Background technique

In modern precision measurement, photometrology is common measuring three-dimensional morphology technology.Photometrology is mainly divided For two classes: for irreflexive surface three dimension measuring technique and for the surface three dimension measuring technique of mirror-reflection.The former is with three Angle is Typical Representative, and especially fringe projection technology is in industrial Precision measurement using very extensive.Its three-dimensional reconstruction Method mainly includes stereoscopic vision and phase height mapping.And it is directed to the surface of mirror-reflection, the most commonly used is phase measurement deviations Art.Since its measuring system is simple, dynamic range is big, strong antijamming capability, can be used for complex-curved measurement, obtains in recent years Extensive concern.Its principle is generation rule striped over the display, and striped deforms after measured surface reflects, using CCD Camera shooting deformation pattern, the surface graded distribution that can calculate tested surface shape is derived by geometrical relationship, then by integrating To face shape height.

In photometrology and camera calibration, camera is generally reduced to ideal pin-hole imaging model, it is believed that camera The light of each pixel of CCD passes through same point (camera pupil principal point).But since the pupil of camera lens has a scale Very little, this hypothesis can generate pupil aberration, and ray position calculated and direction can contain apparent deviation, then pin-hole model The object-image relation that camera cannot perfectly be expressed, causes to seriously affect to three-dimensional measurement precision.Therefore need it is a kind of it is easier, have The high-precision pupil aberration correcting method of effect.

Summary of the invention

The purpose of the present invention is to provide a kind of method that can effectively carry out camera pupil aberration correction, with realize camera at As the reconstruction of light field.

The present invention proposes a kind of camera pupil aberration correcting method, is based on light field coding techniques, the specific steps are as follows:

(1) it in order to overcome the pupil aberration of pinhole camera, determines every imaging ray, is retouched using four-dimensional function (X, Y, u, v) State imaging light field L:

(1)

Wherein, (X, Y) indicates coordinate of the light at image planes Z, (u, v) and indicate that coordinate of the light on pupil is (right The z coordinate answered be 0), namely indicate radiation direction；

(2) building light field encodes optical path；It include primary mirror, 2 relay lenses, digital micromirror arrays in the light field coding optical path (DMD), CCD, as shown in Figure 1；Wherein, use LCD screen as object, object space light is focused on through primary mirror to be invented in image planes, then Switch to parallel light projection on DMD for light is converged through the first relay lens, which carries out coded sample by DMD, and realizes The turnover of optical path；Directional light after coding is focused in CCD image planes by the second relay lens, realizes Image Acquisition；It is carried out on DMD Hadamard binaryzation coding, realizes effective pupil encoded acquisitions；In Fig. 1, L1 is object distance, L2=f1+k1+f2, In, f1 be primary mirror focal length, f2 be relay lens focal length, k1 be diffraction distance, L3 be relaying mirror spacing, L4 be relay lens focal length away from From；

(3) by pupil pixel (u, v) row of the vectorization as light field matrix L, by pixel coordinate (X, Y) vectorization as light field square The column of battle array L, and each element numerical value in light field matrix L is that formula (1) determines that all light travel to (X, Y) pixel from (u, v) Overall strength；Based on compressed sensing principle, observing matrix Φ is determined according to the corresponding binary-coding of DMD, by corresponding even numbers small echo Base forms sparse matrix Ψ；If light field matrix L=Ψ α, α is coefficient of the light field matrix L under wavelet basis, then following target is defined Function:

(2)

Wherein, image shot by camera when β is calibration for cameras under difference pupil coding,Indicate vector or matrixkNorm,Indicate full variation operation, λ is weight coefficient, is determined by the noise level of image；

(4) alternating direction multipliers method solving optimization problem (2) is used, obtains light field matrix L, and utilize ginseng mark outside participating in camera It is fixed, accurate ray position and direction are obtained, the corresponding pupil coordinate (u, v) of each picture point (X, Y) is calculated, in conjunction with camera Calibration result, can effectively correcting camera pupil aberration, so that building one more meets the camera mould of actual imaging process Type.

It is described to use alternating direction multipliers method (ADMM) solving optimization problem (2), it is to be converted into augmentation Lagrange Equation:

(3)

Wherein, D is the corresponding differential operator of full variation,tFor Lagrange multiplier, ρ is weight coefficient；To formula (3), with unknown quantityα,s,tAlternative optimization is grouped to solve.

The present invention is based on light field coding and compressed sensing principles, and sparse matrix and observing matrix therein are all freely to set It sets, suitable sparse matrix and observing matrix can be designed according to the distribution character of practical light field matrix.Meanwhile optimizing mesh Two of scalar functions, the respectively similitude of Metric reconstruction light field and acquisition data, and the clarity of reconstruct target, can basis Actual needs adjustment weight coefficient obtains accurate light field matrix.

The present invention can effectively eliminate pupil aberration caused by pin-hole imaging is assumed, overcome the direction ambiguousness of monocular vision, The accurate direction for calculating imaging ray.It is significant for the measurement accuracy for improving photogrammetric technology.

Detailed description of the invention

Fig. 1 is that the light field that the present invention constructs encodes optical path diagram.

Fig. 2 is the Hadamard matrix encoded for DMD.

Fig. 3 is Imaging Simulation figure.

Fig. 4 is the three-dimensional measurement situation before and after pupil aberration correction.Wherein, (a) is the three-dimensional after pupil aberration correction Measurement result is (b) the three-dimensional measurement error after pupil aberration correction, is (c) three-dimensional measurement before pupil aberration correction Error.

Specific embodiment

The present invention is further illustrated below by embodiment combination attached drawing.

Embodiment 1: in measurement process, suitable imaging optical path is built, as shown in Figure 1.Object space light is through primary mirror in optical path It focuses on and invents in image planes, then switch to parallel light projection on DMD for light is converged through the first relay lens, DMD turns optical path Folding converges on CCD via the second relay lens to be imaged, and is the mirror image of the second relay lens and CCD in dashed box.L1 is object distance, L2 =f1+k1+f2, wherein f1 is primary mirror focal length, and f2 is relay lens focal length, and k1 is diffraction distance, and L3 is relaying mirror spacing, during L4 is After mirror focal distance.Specific optical path parameter is as shown in table 1.(i.e. observing matrix is encoded using the Hadamard that DMD carries out 27 × 27 Φ), the image after acquisition coding, coding pattern are as shown in Figure 2.CCD pixel resolution ratio used is 3840 × 3840 in experiment, institute With the dimension of light field matrix L for 729 × 14745600 ((27 × 27) × (3840 × 3840)).Since actual imaging process is not Stringent pinhole imaging system, the light of single object point only can be by part pupil imagings on a certain CCD pixel, and as light field is divided Cloth, as shown in Figure 3.By the sparse matrix Ψ of the suitable light field matrix of complete dictionary method training excessively, managed by compressed sensing By with formula (4), available specific light field matrix L=Ψ α.Parameter in calculating process is chosen as shown in table 2, uses ADMM method is iterated optimization to objective function.It realizes the sampling to pupil, constructs the mapping relations of pupil and CCD, auxiliary The pupil aberration of Zhang Zhengyou peg model effectively correcting camera, building one more meet the camera model of actual imaging process. In order to prove the validity of pupil aberration correction, a three-dimensional measurement experiment is devised, is surveyed using the phase in photometrology Measuring deviation art method and measuring a radius of curvature is 1000mm, and bore is the spherical reflector of 60mm.Fig. 4 (a) is pupil aberration Three-dimensional measuring result after correction.Shown in three-dimensional measurement error such as Fig. 4 (b) after pupil aberration correction, whole RMS 67.91 nm.Shown in three-dimensional measurement error such as Fig. 4 (c) before pupil aberration correction, whole 230.51 nm of RMS.From reality Testing result and can be seen that the correction of pupil aberration can be effectively improved caused by the camera model excessively constrained in photometrology Measurement error.

(3)

Wherein, D is the corresponding differential operator of full variation,tFor Lagrange multiplier, ρ is weight coefficient；To formula (4), with unknown quantityα,s,tAlternative optimization is grouped to solve.

The main optical path parameter of table 1.

。

2. calculating process major parameter of table

。

Claims (2)

1. a kind of camera pupil aberration correcting method based on light field coding, which is characterized in that specific step is as follows:

(1) using four-dimensional function (X, Y, u, v) description imaging light field L, i.e. coordinate representation of the light at the z of section are as follows:

(1)

Wherein, (X, Y) indicates coordinate of the light at image planes Z, (u, v) and indicate coordinate of the light on pupil, it is right The z coordinate answered is 0, namely indicates radiation direction；

(2) building light field encode optical path: light field encode optical path in include primary mirror, 2 relay lenses, digital micromirror array (DMD), CCD；Wherein, use LCD screen as object, object space light is focused on through primary mirror to be invented in image planes, then will be converged through the first relay lens Caustic switchs to parallel light projection on DMD, which carries out coded sample by DMD, and realizes the turnover of optical path；In second The directional light after coding is focused in CCD image planes after mirror, realizes Image Acquisition；Hadamard binaryzation volume is carried out on DMD Code, realizes effective pupil encoded acquisitions；

(3) by pupil pixel (u, v) row of the vectorization as light field matrix L, by pixel coordinate (X, Y) vectorization as light field square The column of battle array L, and each element numerical value in light field matrix L is that formula (1) determines that all light travel to (X, Y) pixel from (u, v) Overall strength；Based on compressed sensing principle, observing matrix Φ is determined according to the corresponding binary-coding of DMD, by corresponding even numbers small echo Base forms sparse matrix Ψ；If light field matrix L=Ψ α, α is coefficient of the light field matrix L under wavelet basis, following target letter is defined Number:

(2)

Wherein, image shot by camera when β is calibration for cameras under difference pupil coding,Indicate vector or matrixkNorm,Indicate full variation operation, λ is weight coefficient, is determined by the noise level of image；

(4) alternating direction multipliers method solving optimization problem (2) is used, obtains light field matrix L, and utilize ginseng mark outside participating in camera It is fixed, accurate ray position and direction are obtained, the corresponding pupil coordinate (u, v) of each picture point (X, Y) is calculated, in conjunction with camera Calibration result, can effectively correcting camera pupil aberration, so that building one more meets the camera mould of actual imaging process Type.

2. the camera pupil aberration correcting method according to claim 1 based on light field coding, which is characterized in that described It is to be converted into augmentation Lagrange's equation using alternating direction multipliers method solving optimization problem (2):

(3)

Wherein, D is the corresponding differential operator of full variation,tFor Lagrange multiplier, ρ is weight coefficient；To formula (3), with unknown quantityα,s,tAlternative optimization is grouped to solve.