ART-06 - General Information

“Edge detection combined with optical and infrared NDT techniques: an aid for
wooden samples with complex surface and subsurface defects”
D. Ambrosini1, K. Mouhoubi2, D. Paoletti1
[email protected]
of Industrial and Information Engineering and Economics (DIIIE), Las.E.R. Laboratory, University of L’Aquila,
L’Aquila, Italy
2GRESPI/ECATHERM, UFR Sciences Exactes et Naturelles, Reims cedex 02, France
Stefano Sfarra
To explore wood is to realize its complexity, its diversity, and its variability. The union of wood and paint is as old as
the human desire to protect an object, or simply to decorate a surface. The link between paint and wood is therefore
at the heart of any approach to conservation of these objects. To the conservator, the analysis of conditions and
problems involves a familiarity with the physical structure of the wood as a material and with its surface interaction
with the applied paint, as well as with the behaviour of the wood after paint application. Because paints are
mixtures of pigments and binders, degradation is only partially determined by the nature of the binder. Many paints
based on the binders are very fragile; consequently, their susceptibility to physical or mechanical damage is
generally a more important cause of conservation problems than chemical changes in the binders themselves.
Panel paintings are increasingly being investigated using advanced non-destructive infrared and optical
measurement techniques. In the present work, a wooden sample having a complex surface and realized following
the Cennino Cennini rules, containing natural and fabricated defects (Mylar® inserts), was investigated by stimulated
thermography, near-infrared reflectography, double-exposure (DE) and sandwich holographic (SH) interferometry.
The stimulated thermography technique consists in depositing energy, whatever be the means of deposition and the
type of energy (sun, flash lamp, laser, and hot air flow), into the observed system (in the present case a wooden
sample with complex surface and subsurface defects) and in monitoring the temporal and/or local evolution of the
surface temperature field of the system caused by this thermal stimulation. Infrared reflectography is a nondestructive testing imaging technique based on the different optical behaviour of visible and near-infrared (NIR)
radiation through a thin pictorial layer. This effect is a consequence of both lower NIR absorption and reduced NIR
scattering due to the particle size smaller than the wavelength. The acquisition of NIR images using LED lamps
working at different wavelengths, seems a very promising method in this field. However, NIR and DE are not
dynamic techniques, while SH is a dynamic technique. In the latter, a number of holograms can be made, each one
recording a single state of the object, in a temporal sequence.
Since enhancing the edge of a detached area identified by SH, means improving the detection of the defect’s
position, this idea was applied in the present research. Instead, the defect’s depth was retrieved working with phase
analysis, i.e. using the pulsed phase thermography (PPT) technique.
Finally, the results coming from optical and infrared NDT techniques were compared each other in order to explore
the advantages and disadvantages of the methods used.
Stefano Sfarra
1. Introduction;
2. Description of the sample;
3. Principles of Non-Destructive Testing (NDT):
Active InfraRed Thermography (IRT);
Principal Component Thermography (PCT);
Pulsed Phase Thermography (PPT);
Holographic Interferometry (HI: DE – SH);
III. Near-InfraRed Reflectography (NIRR);
IV. Ultraviolet (UV) Imaging;
4. The Canny edge detector;
Stefano Sfarra
Experimental results;
1. Introduction
Stefano Sfarra
2. Description of the sample
Figure: (a) sketch of the
sample, (b) α side. The
position of the defects are
marked, and (c) β side.
The position of the
defects are marked.
The inspected sample is an hollow cylinder [height (l): 180 mm] that is composed of a support in
poplar wood [thickness (r – ri): 16.5 mm], several defects in Mylar® [thickness: 0.2 mm], a layer
of gypsum [thickness: 2 mm, α = 4.7 m2/s x 10-7] and a layer of paint [thickness: 0.5 mm, α =
0.87 m2/s x 10-7]. The layers of paint and gypsum are given by re – r (Fig. a).
In particular, the defects of the α side are five (Fig. b), while the defects of the β side (Fig. 1c)
are seven, ranging from 10 mm of diameter (α4) to 4.5 x 2 mm2 (α2). In this Fig. it is possible to
see the gypsum layer, while in Fig. a the painted sample.
Stefano Sfarra
3. Principles of Non-Destructive Testing
Stefano Sfarra
3.I.a. IRT – Principal Component Thermography (PCT)
In the square pulse configuration, the specimen surface is
submitted to a long square pulse (from a few seconds to
several minutes), and the temperature rise and decay is
registered using an infrared camera and stored as a 3D
matrix composed by N thermograms, where x and y are the
spatial coordinates, and t is the time.
SPT data is generally processed to improve defect visibility
and to performed quantitative characterization of defects.
Arndt proposes an adaptation of the pulsed phase
thermography algorithm to detect and characterize defects
from SPT experiments. In this work, we propose to use
principal components to process SPT data.
Stefano Sfarra
Rajic N. “Principal component thermography for flaw contrast enhancement and flaw depth
characterization in composite structures,” Compos. Struct., 58:521-528, 2002.
3.I.b. IRT – Pulsed Phase Thermography (PPT)
Mathematically, a pulse can be decomposed into
a multitude of individual sinusoidal components.
In that respect, when a specimen is pulse
heated, thermal waves of various amplitudes
and frequencies are launched into the specimen,
in a transient mode. Going back and forth
between temporal and frequency domains is
possible with mathematical tools such as the
well known Fourier transform.
Figure 1 shows an example of correspondence
between these domains. In fact for PT thermal
waves of different frequencies are launched in
the specimen simultaneously in the transient
regime while in LT, one thermal wave of single
frequency is tested in stationary regime. The
possibility to link both techniques was then
found interesting and this was called pulsed
phase thermography (PPT) processing.
Stefano Sfarra
3.II. Holographic Interferometry (HI: DE)
Experimental configuration for Double Exposure (DE) HI
Stefano Sfarra
3.II. Holographic Interferometry (HI: DE)
In Double-Exposure holographic interferometry, two
holograms of the two object waves occurring
sequentially in time are recorded on a single
photographic plate. The interference between these
images produces interference fringes overlaid on the
image of the object. The interference fringes are
indicative of deformation, displacement, rotation and
change in refractive index or thickness of the object. The
fringes appear to be localized in space, not necessarily
on the object. When viewing direction is altered, the
fringes shift and change their form. A hologram of the
object in its initial unstressed state is recorded by
exposing the photographic plate. Without removing the
photographic plate from the setup, the object is stressed
and a second exposure is made. The plate is then
developed and reconstructed by the reconstruction
wave to observe the interference.
Stefano Sfarra
3.II. Holographic Interferometry (HI: SH)
Stefano Sfarra
The basic idea of the Sandwich HI technique
consists in recording of two individual
holograms for two different stages of the
object. Both holograms are reconstructed at
the same time. Unlike the DE method, here it
is possible to change mutual position of the
records of the reconstructed wavefronts
during the reconstruction process, because
each of them regenerates independently. The
relative change in position of one hologram
with respect to the other is equivalent to the
relevant change of the object between the
exposures, and vice versa. In this manner, it is
possible to compensate the specific changes of
the object additionally in the process of
reconstruction. The technical realization of this
method is based on the use of special
kinematical equipment to place the holograms
exactly back to the places where they were
perpendicular to the surface plane of the
investigated object; but it can also be applied
to measure any of its deformations.
3.III. Near-InfraRed Reflectography (NIRR)
Near-infrared reflectography (NIRR) is a non-destructive imaging
technique based on the different optical behaviour of visible and
near infrared (NIR) radiation through a thin pictorial layer. This
effect is a consequence of both lower NIR absorption and reduced
NIR scattering due to the particle size smaller than the wavelength,
whilst near-infrared transmittography (NIRT) provides information
about the internal layers, i.e. the fibre distribution.
Stefano Sfarra
3.IV. Ultraviolet (UV) imaging
Grease stains on
countertop: left –
color, right – near-UV
A repainted front
driver’s side fender
on a Toyota Prius:
left – color, right –
Stefano Sfarra
4. The Canny edge detector
The Canny edge detector is an edge detection operator that uses a multi-stage algorithm to
detect a wide range of edges in images. It was developed by John F. Canny in 1986. Canny's aim
was to discover the optimal edge detection algorithm. In this situation, an "optimal" edge
detector means:
good detection – the algorithm should mark as many real edges in the image as possible.
good localization – edges marked should be as close as possible to the edge in the real image.
minimal response – a given edge in the image should only be marked once, and where possible,
image noise should not create false edges.
To satisfy these requirements Canny used the calculus of variations – a technique which finds the
function which optimizes a given functional. The optimal function in Canny's detector is described
by the sum of four exponential terms, but it can be approximated by the first derivative of a
The Canny algorithm contains a number of adjustable parameters, which can affect the
computation time and effectiveness of the algorithm.
For example, it is possible to remember: the size of the Gaussian filter and the thresholds.
Stefano Sfarra
5. Experimental results 1/6
Seeing the NIR and UV results inherent
the α side (Figs. b-g) and comparing
them with the VIS image (Fig. a), no
information can be retrieved about the
preparatory drawing and the author’s
signature, that is positioned at the
bottom right corner of this side.
Figure: (a) the painted sample; NIR result working with a filter at: (b) 715 nm, (c) 850 nm, (d) 1000 nm; (e) UV result; NIR
result working with a LED lamp at: (f) 850 nm, (g) 940 nm.
Stefano Sfarra
5. Experimental results 2/6
More information about the inner defects linked to the α and β sides are identifiable using
the stimulated infrared thermography approach (Figs. a, b). In fact, all the defects (from α1
to α5) are detected in Fig. a, while two subsurface defects (β1 and β4, i.e. the smaller defects
of the β side) are not detectable in Fig. b. Very interesting to note: 1) the presence of two
unknown defects, surrounded by red dotted ovals in Fig. b, probably located inside the
wooden support, and 2) the red colour of the α5 defect (Fig. a) if compared to the other
Figure. Stimulated infrared thermography results: (a) α side, and (b) β side.
Stefano Sfarra
5. Experimental results 3/6
A comparison among NDT has involved the use of the DE (Figs. a, b) and SH techniques with
the latter pattern of fringes enhanced by the Canny edge detector (Figs. c, d). Several
fabricated defects can be detected using the integrated approach, as well as the use of the
Canny’s algorithm (Fig. c, d) seems very promising to exalt the position and the shape of the
defects, if compared to the classical DE holographic recording (Figs. a, b).
Figure. α side: (a) 1st configuration:
DE-HI texp = 3 s – heating time = 4
min, (b) 2nd configuration: DE-HI texp
= 3 s – heating time = 3 min; β side:
(c) 1st configuration: SH-HI, (d) 2nd
configuration: SH-HI.
Stefano Sfarra
5. Experimental results 4/6
Comparing the results coming from the stimulated thermography and the active
thermography methods, it seems that the first one appears more suitable to retrieve all the
subsurface defects; also in the latter case, an interesting link is due to the α5 defect
characterized to a bright spot, i.e. the most visible defect detected. The same consideration is
adapt to link the β3 and β7 defects of the β side.
Figure. Active thermography results - α side: (e) PCT-EOF4, β side: (f) PCT-EOF3.
Stefano Sfarra
5. Experimental results 5/6
Since the depth of the defects are known and identical, i.e. they are located 2.5 mm beneath the paint
surface, the defect’s depth was also calculated from a relationship of the form:
where fb [Hz] is the blind frequency defined as the limiting frequency at which a defect located at a
particular depth presents enough (phase or amplitude) contrast to be detected on the frequency
Defect contrast is enhanced using the phase allowing deeper probing. Conventional experimental C1
values when using the phase from lock-in thermography experiments range between 1.5 to more
than 2 with a value of C1 = 1.82 typically adopted in experimental studies. PPT results agree with
these numbers. In this way, the inversion problem in PPT is reduced to the estimation of fb from the
phase, while α can be considered as a combined diffusivity (αc), according to:
It is well-known that noise content present in phase data is considerable, especially at high
frequencies. This causes a problem for the determination of the blind frequency. A de-noising step is
therefore often required. The combination of PPT and TSR has proven to be very effective for this
matter, reducing noise and allowing the depth retrieval for a defect. Taking into account the data
obtained working with the phase, as well as the input data, i.e. Δt = 1 s, heating time = 300 s, cooling
time = 900 s, processed thermograms = 600, fmax = 0.5 Hz, fb = 0.07 Hz, the estimated depth is 2.4 mm,
very close to the real depth.
Stefano Sfarra
5. Experimental results 6/6
Figure. α side: (a) thermogram at t = 3 s with cold image subtracted, (b) temperature evolution for two reference points
marked in (a), (c) phase contrast curve: Δϕ = ϕdefected –ϕsound, (d) phasegram at f = 0.0048 Hz linked to the red square
detail in (a), (e) thermographic signal reconstruction (TSR) – 5th order, and (f) phase-frequency curve.
α3 defect, was chosen as the reference defect for the depth retrieval with the phase,
considering the difficulty of detecting its position using the PCT technique.
Stefano Sfarra
6. Conclusions
In the cultural heritage field, there is an interest in inspecting objects having a complex
surface. From the present research, several conclusions can be drawn: a) the large
amounts of energy absorbed by the carbon of the pigments used in the drawing is
testified from the results and taking into account that the underdrawings were realized
using graphite, b) the processing of the data inherent the stimulated infrared
thermography approach, it is very effective in order to retrieved the position of the
subsurface defects, c) the use of the Canny’s algorithm on the SH results can be
considered an interesting application both to enhance the contrast of the fringes
pattern and to underline the shape of the defects, d) the integration between the PCT
technique and the stimulated infrared thermography method has been able to
characterize the presence of the α5 defect as dissimilar to the other defects at least in
the nature, e) the deviation [%error = 4% = (zest – z)/z] between the real depth of the
defects and the depth retrieved with the phase analysis provides a useful information
to the restorer before the starting of the restoration procedure.
Stefano Sfarra
Stefano Sfarra

similar documents