07.11.2014 Views

Application and Optimisation of the Spatial Phase Shifting ...

Application and Optimisation of the Spatial Phase Shifting ...

Application and Optimisation of the Spatial Phase Shifting ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

166 Improvements on SPS<br />

This procedure requires a storage interval ∆t for <strong>the</strong> phase maps ϕ O (x, y, t) which is matched to <strong>the</strong><br />

possibly varying velocity <strong>of</strong> object deformation <strong>and</strong> displacement. The implicit fringe counting capability<br />

<strong>of</strong> TPU lends itself for driving such a matched course <strong>of</strong> ∆t automatically. As far as I know, this issue has<br />

only once been dealt with before on <strong>the</strong> basis <strong>of</strong> speckle decorrelation analysis [Gül93]; here however, <strong>the</strong><br />

quantity <strong>of</strong> interest that we want to limit is <strong>the</strong> number <strong>of</strong> fringes in ∆ϕ(x, y) instead <strong>of</strong> <strong>the</strong> speckle<br />

decorrelation [Bur00b].<br />

In practice, ESPI <strong>of</strong>ten deals with objects consisting <strong>of</strong> several independent parts that may undergo<br />

different displacements <strong>and</strong> deformations. However, sawtooth fringes do not allow to determine rigid<br />

body movements or <strong>the</strong> sign <strong>of</strong> <strong>the</strong> deformation itself, unless <strong>the</strong> fringe orders are tracked by additional<br />

devices like, for instance, a phase stabilisation unit [Bro00]. We will see that temporal unwrapping<br />

delivers <strong>the</strong>se data for each object part automatically.<br />

6.7.1 Temporal unwrapping <strong>of</strong> speckle phases<br />

The use <strong>of</strong> temporal unwrapping is not entirely straightforward in speckle interferometry; we will<br />

<strong>the</strong>refore briefly consider <strong>the</strong> cumulative impact <strong>of</strong> speckle noise on displacement data.<br />

Not surprisingly, badly modulated pixels cause problems also in this application. The statistical<br />

fluctuations <strong>of</strong> <strong>the</strong> calculated phase should yield a displacement <strong>of</strong> zero when monitored over a sufficient<br />

number <strong>of</strong> frames. It was however observed that even for longer observation sequences with hundreds <strong>of</strong><br />

frames, some <strong>of</strong> <strong>the</strong>se pixels seemed to change <strong>the</strong>ir phase constantly in one direction; both signs <strong>of</strong><br />

displacement were present. In a fringe counting procedure, <strong>the</strong>se pixels would trigger data storage even<br />

when no actual displacement has occurred. Therefore such outliers have to be suppressed; <strong>and</strong> as usual in<br />

speckle interferometry, a low-pass filter can serve to do so. This may be objectionable in TPS, because it<br />

impairs <strong>the</strong> spatial resolution; but for larger speckle sizes, as used in SPS, <strong>the</strong> resolution will not suffer<br />

greatly.<br />

There are sophisticated <strong>and</strong> well-founded filtering schemes [Hun97] that give excellent rejection <strong>of</strong> noise<br />

in <strong>the</strong> displacement map over long, albeit not infinite, times <strong>of</strong> observation [Cog99]. For reasons <strong>of</strong><br />

processing speed, a simpler filtering scheme is used here. The accumulated phase Φ(x, y, t) <strong>of</strong> a pixel at<br />

time t is considered faulty when it differs by more than π from <strong>the</strong> accumulated phase <strong>of</strong> at least one out<br />

<strong>of</strong> its nearest neighbours. In that case,<br />

( )<br />

( x y t) = ( x y − t) + ( x − y t) + ( x + y t) + ( x y + t)<br />

Φ , , : Φ , 1, Φ 1, , Φ 1, , Φ , 1, / 4 (6.26)<br />

<strong>and</strong> <strong>the</strong> outlier is eliminated.<br />

By <strong>the</strong> selection criterion, filtering takes place only when necessary, <strong>and</strong> processing time is saved. This<br />

helps to obtain a high frame rate, which is very important since also temporal unwrapping relies on <strong>the</strong><br />

sampling <strong>the</strong>orem, as detailed above; <strong>and</strong> due to <strong>the</strong> cumulative nature <strong>of</strong> <strong>the</strong> process, errors due to missed<br />

fringes (violation <strong>of</strong> <strong>the</strong> sampling condition) will last in <strong>the</strong> map <strong>of</strong> Φ(x, y, t) until it is cleared. The phase

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!