The principle of image correlation is based on tracking information from a so-called “reference” image in the following images, often called “distorted images”. All of the images form a film, from which we seek to extract a measure.
We can therefore see the correlation of images as following points over the entire observed surface: following the movements of a surface of a part is equivalent to following the movements of all the points of this surface. This requires the part to be “textured”: if there is only one black point, we cannot measure the displacement and the deformation of the white zone which surrounds it. For a point to be recognized, you need a random texture that allows you to recognize the area around the point of interest. This texture is called a speckle pattern.
Most of the time, speckles are sprayed. We just spray spots of black mat paint, of random sizes, on a white background, also mat to avoid reflections that could interfere with the measurement. A speckle of white dots on a black background is also possible.
What is a “good speckle pattern”?
To make a good speckle, it must be:
- With task sizes adapted to the measurement carried out.
To check that the speckle pattern is random and does not contain too much black or too much white, we can use grey level histograms. An optimal histogram is a “flat” one with an equivalent distribution of white and black.
It is also necessary to take care to adapt the size of the spots, this results from a compromise between the phenomenon studied (fine spots) and the resolution of the imager (size large enough to be visible by the sensor).
Speckle pattern using spray paint
Different types of speckles encountered during Eikosim’s tests
It is sometimes possible to do without the speckle pattern step. Indeed, certain materials, such as concrete, even have a “natural” texture which has sufficient heterogeneity to act as a natural speckle pattern, as shown in Fig. 4. With this type of surface, it is possible to directly process the images without painting the structure.
For the cases where one wishes to study displacements of rigid body only, without deformation, it is possible to digitally create a speckle pattern and to print it on adhesive sheets, which will be deposited on the structure to be studied. The advantage of this technique is that the randomness as well as the management of the size of the spots are entirely controlled by the user who can choose himself the min and max diameters of his speckle spots.
Finally, there are still other methods, like the one developed by Alphanov :
written and illustrated by Girolamo MINCUZZI and Simon NOURRY
It is well known that after irradiating the surface of a material like metal, semiconductor or even transparent dielectric with few ultrashort pulses is possible to modify the surface morphology with the generation of subwavelength ripples. These ripples are perpendicular to the polarization direction and are referred as LIPSS laser induced periodic surface structures (see Fig.1 (a) and (b)).
By increasing the number N of pulses LIPSS tends to be transformed in micro-grooves which are parallel to the polarization direction (see Fig.1 (c) and (d)). Finally, by further increasing N spikes with characteristic size of ≈10 µm appear on the surface (see Fig.1 (e) and (f)).
Thanks to these surface structures it is possible to control the surface reflectivity R over a wide range of values. Interestingly, in the case of surfaces covered with spikes light experience multiple reflection enabling a light trapping mechanism reducing R < 10%. This effect is enhanced when spikes surface consists of nanometric pores. In the last case (see Fig.2 left) values of R < 5% can be obtained yielding a “deep black” effect (see Fig.2 right).
Moreover, thanks to the synchronized use of laser beam scanners and translation stages it is possible to efficiently process surfaces larger than a single pulse spot and varying from some tens of microns up to nearly 1m.
Here we carried out a preliminary, systematic investigation of how process parameters like, scan speed, hatch, number of successive scans, pulse energy will impact the morphology and the R value of a mirror-like, stainless-steel surface. For all tests we used a 350-fs pulsed laser emitting at λ = 1030 nm with repetition rate of 500 kHz (Tangerine from Amplitude Systemes) and spot size 2ω0 ≈ 35 µm. Values of R ≈ 20% corresponding to a grey color were found for surfaces having a size comprised between 1cm2 and 0.2mm × 0.2mm utilizing the same set of process parameters.
On the contrary, for R< 5% we observed that the size of the processed surface has a huge bearing on the final result. For instance, Fig. 3 shows the case of two similar surface morphologies and color (R<4%). But, whilst Fig. 3 left is relative to a 1cm2 and has been processed with a pulse energy of E = 50 µJ, Fig.3 right is relative to 100µm × 200µm and has been processed with E = 32 µJ.
Finally, we utilised all the results obtained from this preliminary investigation to process a surface of 10cm × 10cm. We made two different patterns having features with different size and varying from some cm2 to ≈100 µm×100 µm. Results are shown in the Fig. 4.
We were finally able to create a speckle with different levels of grey (by mixing the random creation of white, grey and black), on a mirrored steel surface. This specific speckle pattern was then used in a stereo-correlation study in collaboration with EikoSim.
13 RUE SAINT HONORE