You are on page 1of 1

Three-dimensional computer vision by phase demodulation and unwrapping procedure

Rigoberto Juarez-Salazar , Carlos Robledo-Sanchez, and Fermin Guerrero Sanchez Benemrita Universidad Autnoma de Puebla - Facultad de Ciencias Fsico-Matemticas

Labor ator io

de In vest iga ci n

s le b la ro t Con s o c i inm D s a Sistem

rjuarezsalazar@gmail.com

Abstract
A fast and robust fringe-pattern processing system to phase guring is proposed. This system reach important applications in numerous areas such as biomedicine, geoscience, optical metrology, machine vision, aeronautical and automotive industries. Experimental results from fringe projection to threedimensional computer vision are presented.

Phase demodulation
We consider a set of K fringe-patterns modelled as Ik (p) = ak (p) + bk (p) cos[(p) + k (p)], (1)

Phase unwrapping
The unwrapped phase data can be modelled as = + 2k, k Z (5)

for k = 0, K 1. The background ak and modulation bk lights are recovered by ak = Aa A a Ik ,


Ab

where k is the jumps function. The gradient of the jumps function can be computed by k = round 1 . 2 (6)

and

b2 k

2Ab Ab (Ik

ak )2 ,

(2)

Introduction
A phase guring is an image processing procedure that consist on two main stages. The rst one is dedicated to extract the wrapped phase (a discontinuous surface) from one or more fringe-patterns (images) [3, 1]. The second one removes the phase jumps from the wrapped phase data to obtain a phase surface and reach a physical meaning information.
!"#$%&'$()*+,-$).#, 6,*27) !"#$%&'() *"+,-.+,') 827*27) /-,0%,1'$%1+) .+'$23%,4) 5+1%&+) 93%,4+0) *':+3,$)

where and are the Moore-Penrose pseudoinverse of a polynomial basis function Aa and Ab , respectively. The data normalization is carried out as k = sat I Ik ak bk = cos( + k ) bk (p) = 0. (3)

A a

Next, the gradient eld integration is formulated as the optimization problem min
k

T kx kL x

2 F

+ ky Ly k

,
F

(7)

Now, there are two main methods to extract the phase k : Fourier fringe analysis [1], function from the data I and Phase-shifting techniques [3]. However, the phase function is obtained up its principal values;i.e., = W , W: [, ), (4)

where Lx and Ly are the difference operator. The solution of (7) leads to the Lyapunovs equation: + kB = C, Ak (8)

?-5+()*(',7)

/0*-$,12.%&(2, =3'**+5) *"'$+) !"#$%&'() %,>-3.';-,)

where W is the wrapping operator. Thus, the unwrapping procedure is necessary.


!"'$+) 2,<3'**%,4)

T T with A = LT L , B = L L , and C = L y y x x y ky + kx Lx . There are efcient numerical methods to solve the linear equation (8) [2]. Finally, the unwrapping procedure is performed by Eq. (5).

!"'$+) 5+.-52(';-,)

Simulation and optical experiments

Conclusions
A phase guring system based on the parameter estimation approach and the rounding-least-squares method was presented. Its feasibility is tested by numeric and optical experiments. Since the proposal is non-iterative, robust, computationally efcient, and no requires user intervention, it may be appropriate to industrial automatic real-time applications.

Do you wish to learn more?


Get the full document, supplementary materials and related papers.

Fig. 1. Three-dimensional computer vision by fringe-projection four-step phase-shifting and phase guring.

References
[1] J. F. Casco-Vasquez, R. Juarez-Salazar, C. Robledo-Sanchez, G. Rodriguez-Zurita, F. G. Sanchez, L. M. Arvalo Aguilar, and C. Meneses-Fabian. Fourier normalized-fringe analysis by zeroorder spectrum suppression using a parameter estimation approach. Optical Engineering, 52(7):074109074109, 2013. [2] G. Golub and C. Van Loan. Matrix Computations. Johns Hopkins Studies in the Mathematical Sciences. Johns Hopkins University Press, third edition, 1996. [3] R. Juarez-Salazar, C. Robledo-Sanchez, C. Meneses-Fabian, F. Guerrero-Sanchez, and L. A. Aguilar. Generalized phaseshifting interferometry by parameter estimation with the least squares method. Optics and Lasers in Engineering, 51(5):626 632, 2013.

Fig. 2. Unwrapping phase for: (1st column) Synthetic data, (2nd column) Optical phase-shifting interferometry, (3rd column) Optical Fourier fringe-normalized analysis, (4th column) Fringe prolometry. (1st-3rd rows) Wrapped phase, estimated phase jumps and unwrapped phase, respectively.

You might also like