Fluorescence lidar techniques offer considerable potential for remote, non-invasive diagnostics of stone cultural heritage in the outdoor environment. Here we present the results of a joint Italian-Swedish experiment, deploying two hyperspectral fluorescence lidar imaging systems, for the documentation of past conservation interventions on the Colosseum, Rome. Several portions of the monument were scanned and we show that it was possible to discriminate among masonry materials, reinforcement structures and protective coatings inserted during past conservation interventions, on the basis of their fluorescence signatures, providing useful information for a first quick, large-scale in situ screening of the monument.
© 2008 Optical Society of America
The documentation and scientific investigation of materials play a key role for the conservation and management of the cultural heritage worldwide . The knowledge of the history of a monument and its status, together with the acquisition of information about previous conservation interventions, are all important factors for a thorough understanding of its present decay. Studying the behavior of a monument treated in the past gives us valuable information for its future conservation and also a good opportunity for a sound selection of suitable conservation solutions.
In this context, modern documentation methods based on new, non-invasive technologies are fundamental for the conservation sector [2,3]. Nowadays, the first assessment before intervention is often an image-based evaluation. Besides traditional photographic methods, spectral imaging has also rapidly developed as a non-invasive technique within conservation science over the last decades [4,5]. However, when large surfaces in the outdoor environment must be tackled, the application of the spectral imaging methods is not as straightforward as for movable heritage and its application must often be limited to the analysis of samples in the laboratory [4,6].
The assessment of the status of a major monument, such as the Colosseum in Rome, regarding materials inserted during previous conservation interventions, protective coatings and reinforcement structures, but also regarding new or re-worked travertine blocks, can be troublesome without extensive testing in a laboratory environment. From this point of view, the use of fluorescence lidar imaging can play a crucial role to gain knowledge about the materials on the surface of the monument and thus provide the conservation sector with valuable, accessible information.
Fluorescence lidar imaging is a non-invasive, remote sensing technique that extends the application of the LIF (Laser Induced Fluorescence) technique to the outdoor environment and combines it with the advantages offered by imaging. Although initially developed for marine and vegetation monitoring applications [7,8], fluorescence lidar remote sensing has already been applied to the monitoring of the cultural heritage for the last decade: first experiments date back to the mid 1990s, when characterization of biodeteriogens and lithotypes of stone monuments were performed in point measurements on the Cathedral and Baptistery of Parma, Italy [9,10]. These initial experiments have then been extended to fluorescence lidar imaging campaigns with acquisition of hyperspectral fluorescence maps on several targets [11–13]. Main fields of investigation have been the detection and characterization of different lithotypes [10,12,14], of biodeteriogens [15,16] and of protective coatings [12–17], showing good potential of the technique for an extensive monitoring and investigation of the stone cultural heritage. So far these results have also demonstrated the possibility to distinguish materials coming from different extraction areas as well as to detect the presence of protective coatings and other extraneous materials on the monument’s surface. These outcomes can thus be exploited for the analysis of past conservation interventions.
This paper illustrates main results of a joint Italian-Swedish measurement campaign carried out on the Colosseum, Rome, and focuses on the application of the fluorescence lidar imaging to the documentation of past conservation interventions and their consequences. Specifically, this paper addresses the following aspects:
- Detection of masonry materials used in conservation interventions (e.g., mortar, cement), but also characterization of stone blocks with the specific purpose to help identification of re-worked material.
- Mapping of reinforcement structures on the surface, such as old and new clamps, and their characterization especially as far as corrosion-inhibiting coatings are concerned.
The measurement campaign has also concerned the investigation of some degradation processes of the monument, such as biological contamination and soiling of surfaces: these are discussed separately in another work .
The measurements were conducted in January 2005 within the framework of a joint Italian-Swedish lidar campaign held at the Colosseum, Rome. The measurements presented here were carried out on selected areas of the monument, chosen in agreement with the Soprintendenza Archeologica di Roma to analyze interventions carried out in past and recent times for the external reinforcement and conservation of the monument. Measurements were performed at night to avoid interference with the intense diurnal activity at the site.
Two different fluorescence lidar systems were deployed for these measurements: the Italian CNR-IFAC fluorescence lidar system and the Swedish LTH lidar system. Both systems are housed in vans used as mobile laboratories on site. The two lidar systems were placed at a distance of about 18 m and 65 m from the monument, respectively.
Hyperspectral fluorescence data were acquired on the selected areas and then analyzed with multivariate statistical techniques to investigate the spots on the monument’s surface interesting with regard to past restoration or reinforcement interventions. Traditional documentation methods, such as digital photography, surveying and in situ studies, as well as the study of conservation reports and archive materials, have been exploited to complement hyperspectral fluorescence data.
The following subsections give a description of the investigated materials, the instrumentation, the experimental conditions and the data processing methods used in this paper.
2.1 The monument: examined areas and materials
The Flavian Amphitheatre, world-wide known as the Colosseum, is one of the best known historical monuments of its kind. The Flavian Emperor Vespasian (69–79 AD) planned and began the construction while his son Titus (79–81 AD) completed and inaugurated the arena in 80 AD . The amphitheatre served as an arena for spectacular scenery with its famous gladiator games. In the 5th century the public interest in these games and the politics changed with the Christian emperors, leading to the suspension of the games in AD 438. The building that was then in the center of the city of Rome rapidly fell into abandonment and the centuries that followed included reuse of the material in the building.
The amphitheatre suffered from demolitions, earthquakes, fires, and water drainage problems during the centuries. In Europe, a renewed interest in the Antiquity and the period of classical architecture led to a more systematic excavation and restoration of the arena, which was performed during the nineteenth and twentieth centuries . Due to its structural collapse, many reinforcements have been carried out on the façades, also including new materials and change of materials. Although many interventions and conservation projects have been undertaken with documentation in text and photographs, the knowledge of previous projects is sometimes hard to encompass.
The section of the monument selected for these fluorescence measurements is on its north side, facing Via dei Fori Imperiali (Fig. 1(a)): it ranges from arch L to arch LIV (Fig. 1(d)) and comprehends both heavily soiled areas (on the left of the picture) and areas cleaned recently with fine water mist, without any additive, during a restoration project carried out in 1999 . In the latter areas black crusts and soiled parts have thus been removed from the travertine and the ageing of the stone has now left it with a browner staining on some parts.
This section of the external ring of Colosseum has undergone several reconstruction and conservation interventions in the last centuries, including the previous restoration project and erection of the second buttress by the architect Giuseppe Valadier during 1823–1826. The main stone material is travertine, cut in large blocks held together by metal clamps. At first visual inspection, the travertine can be regarded as dating back to the first building phase, although there are indications suggesting that some parts of the lower half on the left side of the studied area contain blocks dating from a modern restoration period. In general, different techniques of craftsmanship forming the ashlars, tools used, boundaries, faces of the stone with different surface finish bearing the marks of the masons, etc., are used to estimate the period to which the examined material dates back. However, the reuse of older blocks, widespread especially in the past, is hard to distinguish on the basis of these methods, and its detection usually require collection of samples.
Besides areas undergone past restoration interventions, this section of the monument features also several mortar joints, holes filled up with cement due to previous conservation and several metal clamps used to prevent structural decay and to strengthen semi-columns and arches. These structural reinforcement interventions date from the last two centuries and are located both in the soiled and in the cleaned areas.
The areas of this section that were mapped with the lidar systems are indicated in Fig. 1(d) with a summary of the main relevant features of each area. Besides hyperspectral fluorescence images of these areas, point fluorescence measurements were also acquired in some spots selected for their peculiar features, and their locations are specified in the Results and Discussion section as needed.
The CNR-IFAC lidar system, the FLIDAR, is an excimer-based fluorescence lidar system that has been repeatedly used for fluorescence remote sensing of monuments since 1994 (see, e.g., [10,12]). It uses an excimer laser (XeCl @ 308 nm; energy per pulse: 30 mJ) as an excitation source. This is typically operated with a 2-Hz pulse repetition rate and has a pulse duration of 10 ns. The signal is collected with a 25-cm diameter 1-m focal length Newtonian telescope. A fiber bundle conveys the signal from the telescope focal plane to the input slit of a 275-mm focal length spectrometer (Acton Research, 1235) coupled to an intensified gated 512-photodiode array detector (EG&G, 1420R) which can detect a 300-nm wide spectrum in the 300–800 nm spectral range with an effective 2.4-nm spectral resolution. Long-pass optical filters are used to reject the laser line and spectrometer higher orders. Control electronics, data acquisition and storing are controlled via a personal computer. The system is housed inside a small van (FIAT, Ducato Maxi) whose dimensions (w×l×h) are 2.1 m×5.5 m×2.5 m. A detailed description of the system can be found in [10,22,23].
The FLIDAR has been recently upgraded with a automatic target scanning system to perform hyperspectral fluorescence imaging. In addition, the system is equipped with a target pointing system in the visible, coaxial with the lidar system, to reference the acquired fluorescence images on the target: the position of the pointing laser spot is recorded on a photo acquired at the same time. The final FLIDAR data is thus a referenced hyperspectral fluorescence image of the scanned surface.
The LTH lidar system was originally intended for atmospheric differential absorption lidar (DIAL) measurements, but is also well adapted for fluorescence measurements and has been employed in several fluorescence lidar campaigns [11–13,24–26]. The system in its most recent implementation is thoroughly described in . The system can perform fluorescence imaging by scanning a folding mirror in a roof-top dome to direct the laser radiation where desired. The laser beam is sent out coaxially with the receiving telescope, so they are always aligned. The laser used is a frequency tripled Nd:YAG laser (Spectra Physics, GCR-290) at 355 nm running with 20 Hz repetition frequency. By Q-switched operation 4–5 ns pulses with high energy, in our case limited to about 30 mJ, are created. The beam is expanded to 4 cm diameter before it is transmitted to the target. Alternatively, radiation from a modified optical parametric oscillator (OPO) system (Spectra Physics, MOPO-730) , can be tuned to emit laser radiation in the wavelength range 220 nm-1.7 µm. The receiver is a 40-cm-diameter Newtonian telescope which collects the fluorescence radiation. The collected signal is guided through a long-pass filter focused onto the tip of a 600 µm optical fiber and connected to an optical multichannel analyzer (OMA) system . The OMA is equipped with a spectrometer (Oriel, MS125) with a gated and intensified CCD detector (Andor Technology, DH501-25U-01), which can detect the spectrum in the wavelength range 280–810 nm with a 2.2 nm resolution. The system is housed in a Volvo F610 truck of dimensions (w×l×h) 2.5 m×8.0 m×3.1 m and is powered by a 40 kVA motor generator pulled by the truck.
2.3 Experimental set-up
Figure 1 shows the experimental conditions at the site: measurements were taken at night to avoid interference with the usual intense activity at the site during daytime. Fig. 1(a) shows the planimetry of the Colosseum and also indicates the section of the monument chosen for the experiment. The locations of the two mobile lidar laboratories with respect to the monument are also marked in the map. In Fig. 1(b) and Fig. 1(c), the two mobile laboratories hosting the lidar systems are shown while deployed during the experiment.
The CNR-IFAC mobile lidar laboratory, shown in Fig. 1(b), was placed at about 18 m from the target. At such distance the diameter of the spot effectively measured on the target was about 2 cm. The horizontal and vertical spatial resolution of the acquired images (i.e. the distance between the center of one measured spot and the following one) varied between 6 cm and 12 cm, depending on the purpose of the measurement, the features of the target and the time required to perform it. Fluorescence data were acquired with 308-nm excitation. A fluorescence spectrum (840 channels distributed over the 300–800 nm spectral range) was acquired for each pixel of the fluorescence image. The fluorescence spectrum was obtained by merging two separate 300-nm wide fluorescence measurements, one from 300 nm to 600 nm and the other from 500 nm to 800 nm. Each fluorescence measurement was accumulated over 30 laser shots. The acquisition of the fluorescence spectrum, its storage and then aiming the next point with the FLIDAR system takes about 1 min in these conditions. Before the acquisition of each fluorescence map, the FLIDAR laser pointing system was used to define the effective area where the fluorescence map would be acquired next; this is achieved by acquiring digital photos while the pointing system is aiming in turn at each vertex of the area to be scanned next and by storing them in the computer together with the relevant fluorescence data.
The LTH mobile lidar laboratory, shown in Fig. 1(c), was parked about 65 m from the target. The spot size on the target was about 4 cm and the resolution of the scans varied between 9 and 20 cm, depending on the purpose of the measurement, the features of the target and also the time required. The overall dimensions of the scanned area varied as well according to the issues to investigate. Fluorescence data were acquired with the 355-nm excitation. The LTH system acquired a complete fluorescence spectrum from 400 nm to 800 nm at each laser pulse. Each spectrum was averaged over 100 or 200 shots, depending on the area considered and the accuracy needed. Measuring one spot, storing the spectrum and moving the dome to the next spot takes about 10 seconds.
2.4 Data analysis
All fluorescence spectra from both lidar systems were corrected for the wavelength response of the relevant optical system.
Most evaluations were carried out using multivariate statistical techniques, specifically either Principal Component Analysis (PCA) or Cluster Analysis (CA). These techniques offer several advantages to extract information out of a complex set of data such as those obtained with hyperspectral fluorescence lidar imaging.
A thorough description of multivariate techniques, such as the PCA and CA, can be found in, e.g., .
The thematic maps shown in the next section, i.e. false-color maps displaying the spatial pattern concerning a specific theme or specific target attributes (e.g., the spatial distribution of protective coatings over a surface), were obtained using one of the following methods:
- calculating a ratio between two selected spectral bands of the fluorescence spectra and plotting the ratio values as a function of the corresponding (x,y) positions in a false-color coded map;
- applying the PCA technique to the fluorescence data set of a given area and plotting either a PC score or the ratio between two PC scores as a function of the corresponding (x,y) position in a false-color coded map;
- applying the PCA technique to the fluorescence data set of a given area, selecting three PC scores (usually the first three) and associating to each of them one of the three channels of a Red-Green-Blue (RGB) coded image; in this way the final RGB map will contain information referring to all the three selected PCs;
- applying a CA method to a fluorescence data set of a given area, associating to each identified cluster a specific color and plotting them as a function of the corresponding (x,y) position in a false-color coded map. In particular, the maps shown in this paper were obtained applying a hierarchical agglomerative clustering to the data.
3. Results and discussion
This section is organized in two parts: the first part focuses on the characterization of masonry materials inserted during restorations carried out in the past, such as travertine blocks and joints of mortar or cement. The second part deals with the identification of strengthening elements, such as metal clamps, on the surface and their characterization in terms of protective coatings.
3.1 Characterization of masonry materials
The analysis of the surface concerning the historical background of the building and the time aspect of the initial construction and later reconstructions may be a troublesome process when it comes to the interpretation of the stones being used. When observing the surfaces, interpretation of techniques used by the masons, geological identification, mortars and other information from archives are all useful indicators but often need additional documentation and scientific investigations on site and in the laboratory. The fluorescence lidar technique as a non-invasive technique can provide additional data without requiring samples or even the use of scaffolds or lift, as required by a thorough visual inspection or any sampling procedure.
To demonstrate the potential of the fluorescence technique to support experts in the identification of non-original blocks, an area (D in Fig. 1(d)) containing both original Roman blocks, as identified by the experts according to the block’s surface finishing, and blocks placed during past restorations, was selected and measured with 308-nm excitation. Due to the practice, widespread in the past, of re-working original blocks, a visual inspection of these blocks by experts could not univocally determine if some of the blocks were inserted during previous restorations or were in fact in situ original blocks processed again by craftsmen.
Figure 2(a) shows the scanned area together with the grid automatically generated by the FLIDAR software to reference the fluorescence data on the target. This area comprehends four travertine blocks: Blocks I was placed in the present position during a restoration intervention carried out in the 1950s; Blocks III and IV were original blocks dating back to the Roman period; Block II on one hand showed signs of processing carried out in recent times - as inferred from the finish of the stone surface clearly done with a modern tool - on the other hand it featured a distribution of the holes in the stone typical of the old Roman blocks. Thus, Block II could also be an original Roman block that was re-worked by restorers to harmonize the other newly inserted blocks.
Figure 2(b) shows a map with the spectral integrals of the intensity of the non-normalized fluorescence spectra: the map highlights a different spectral behavior of Block I with respect to the other three blocks (Blocks II, III, IV), as pointed out by a prominence of lighter colors for Block I, indicating a definitely higher fluorescence intensity with respect to the other three blocks. On the other hand, Block II shows a spectral behavior very similar to Blocks III and IV. This leads to believe that Block II, although showing several indications of interventions carried out in recent times, is in fact an original Roman block. This is further supported by the comparison of the fluorescence spectra shown in Fig. 2(c): spectra belonging to Block I (labeled as I and J) have similar profiles and are definitely more intense than the spectra of Block II and IV (labeled as K and L). The spatial locations of these fluorescence spectra are indicated in the map of Fig. 2(b) and labeled accordingly.
An extended portion (Area F in Fig. 1(d)), containing the same area as discussed in Fig. 2(b) (Area D in Fig. 1(d)), was analyzed with the 355-nm excitation, from a greater distance (65 m). The map of Fig. 3(a), superimposed to the photo of the examined area, was obtained from the mean values of the fluorescence spectral intensity in the 410–750 nm spectral range. Also in this case, the spectral behavior of Block II is very similar to that of the blocks belonging to the central column, identified by the experts as being original Roman blocks according to their surface finishing. On the contrary, all the other blocks on the left part of this area generally show a more intense fluorescence intensity, as indicated by the lighter colors in the map. Figure 3(b) shows some typical fluorescence spectra referring to different spots of the scanned area: the relevant pixels are squared with the corresponding colors in the map of Fig. 3(a). Note that also with 355-nm excitation Block I (upper left in area D) showed a high fluorescence intensity with respect to all the other blocks, as also found with 308-nm excitation (see Fig. 2). The fluorescence spectral integrals, in fact, feature a certain variance also within a single block. This is clear by comparing the map in Fig. 3(a) with the photo in Fig. 3(c) where the different blocks are marked in different colors: Blocks A, I and II were blocks placed or re-worked later, during the interventions carried out in the 1950s; Blocks B, III and IV are original Roman blocks. Figure 3(d) reports a box and whisker plot for the data set of fluorescence integrals grouped by different blocks. The notches in the boxes represent a robust estimate of the uncertainty about the medians for box-to-box comparison . The graph in Fig. 3(d) indicates that the true medians of Block A and I differ, with 95% confidence, from those referring to Blocks B, II and IV, since their notches do not overlap. Block III has a greater spread and its notches slightly overlap with those of block A. To further investigate the feasibility to distinguish the blocks on the basis of their fluorescence spectral intensity integrals, the data set has been analyzed with a multiple comparison test using the Tukey-Kramer method . The mean values and corresponding uncertainties were evaluated for each block and reported in Fig. 3(e): Blocks A and I (blue markers in the graph) can be considered different from the other blocks (red markers in the graph), with a 95% confidence, since their bars do not overlap.
Area D, a detail of which is shown in Fig. 4(a), also features an apparent joint of grain mortar containing also limestone powder: the joint is located between Block I and Block II of Area D (Fig. 2(a)), in correspondence of line 7 (first four pixels) of the map of Fig. 2(b). The same joint has also been scanned in Area F with the 355-nm excitation as shown in Fig. 3(a), where it is marked with an orange oval. Figure 4(b) shows the fluorescence spectra of the first four pixels of line 7 of the map in Fig. 2(b): the spectra are compared with some reference point fluorescence spectra acquired separately in a selected spot of the joint and with the fluorescence spectra corresponding to the lines above and below the joint in the map. In general, the mortar shows a higher fluorescence contribution at lower wavelengths than the spectra of stone: the mortar spectra are both brighter and flatter between 450–520 nm than the spectra from the adjoining block. This is also confirmed by the 355-nm excited fluorescence spectra shown in Fig. 4(c): these are the fluorescence spectra corresponding to the first six pixels of line 17 of the map on Area F shown in Fig. 3(a): according to their shape, the first two spectra (red lines) refer to the stone rather than to the mortar. This can be inferred also from the photo in Fig. 4(a) where it is clear that the joint is almost not present in the first pixels of the line. The 355-nm excited fluorescence spectra are noisier than the 308-nm excited spectra also because of the greater distance from the target (65 m instead of 18 m).
The feasibility of detecting historic joints with the fluorescence lidar technique was also investigated in heavily soiled areas, as that shown in Fig. 5(a). The area to scan was chosen on the border of a recently cleaned section of the monument. This region, corresponding to Area B in Fig. 1(d), features an extensive, old smeared cement joint and comprehends both a heavily soiled portion of the façade and a portion of travertine recently cleaned. The soiled portion has a very rough surface due to weathering and natural ageing of the travertine, but also due to previous fire incidents, that caused stress to the stone. Figure 5(b) shows a false-color map obtained by a ratio of the of the spectral intensity integral in the wavelength range 485–505 nm divided by spectral intensity integral in the wavelength range 525–735 nm. Comparing the map in Fig. 5(b) with the photo in Fig. 5(a) it is clear that the brightest pixels correspond to an old smeared cement joint. This is further confirmed by the fact that the fluorescence spectra relative to the joint have a remarkably different spectral shape with respect to the fluorescence spectra of both the non-cleaned, soiled travertine and the cleaned one, as shown in Fig. 5(c). The green curves correspond to spectra from the joint, while the cyan curves are reference spectra from the underlying stone as marked by the squares of the same colors in Figs 5(a) and 5(b). Note that this old cement joint has a spectral shape remarkably different from that of the mortar joint containing limestone powder shown in Fig. 4(c).
3.2 Detection and characterization of metal reinforcement structures
The section of the monument selected for the fluorescence campaign included the presence of several past and recent interventions for the reinforcement of the original structure with metal clamps. During the measurements, it was noted that some of the metal clamps gave characteristic spectral signals, including a sharp peak around 380 nm, interpreted as the corrosion-inhibiting coating applied to the clamps.
An example is shown in Fig. 6. Figure 6(a) shows a photo of the investigated area and indicates two measured points (A and B in the photo), on different metal clamps. Both clamps belong to the same period, dating back to the consolidation interventions carried out by Valadier. In Fig. 6(b), their corresponding fluorescence spectra are shown. The sharp peak in the ultraviolet region is clearly seen in point A, while it does not occur in point B, indicating that the clamp in B has not been recently treated with the corrosion-inhibiting coating.
A scan over a cleaned area containing metal clamps was performed with 308-nm excitation and results are reported in Fig. 7. This area corresponds to Area E in Fig. 1(d). In Fig. 7(a), the area with the measured points is shown. In Fig. 7(b) a CA has been performed, and it is clearly seen that points on the metal clamps are different from the surrounding travertine stone. The same feature can be seen by plotting the spectral intensity integrals in the wavelength range 360–400 nm, as can be seen in Fig 7(c). The metal clamps have been treated, so they are clearly identified by the peak at around 380 nm, and the spectra corresponding to the colored points are seen in Fig. 7(e). The curves have the same color as the corresponding pixels in Fig. 7(c). In Fig. 7(d), a PCA-RGB analysis has been applied to the spectra from cluster 1 in Fig. 7(b), i.e., the ones corresponding to stone points. Spectra were normalized to the standard deviation as for spectral intensity and to the maximum as for spectral band variance prior to processing. One can notice two areas with purple pixels, marked by red circles, which seem to differ from the surrounding stone. By comparison with Fig. 7(a), it is possible to assign these points to damaged pits in the wall, where there are patches containing cement and gravel. Yellowish green areas are interpreted as travertine with a high level of alveolization. Examples of spectra from the area are given in Fig. 7(f), where the colors of the spectra are the same as the colors of the pixels in Fig. 7(d).
The section of the monument considered in the measurement campaign comprises several other portions where metal clamps were applied to strengthen its structure. Figure 8 shows an area (Area C in Fig. 1(d)) where titanium clamps were applied during a restoration carried out recently, in 1999. This area has also been cleaned with fine water mist during the recent restoration project mentioned previously. The area was measured using 308 nm excitation and is shown in Fig. 8(a) together with the measured points. This area comprehends three titanium clamps placed in different positions, even under the arch. Figure 8(b) shows a CA-based map where three pixels (light blue and yellow pixels) clearly stand out against the vast majority of the brown-marked pixels referring to the travertine. The positions of the former on the scanned area are shown in Fig. 8(c) where it is clear that they correspond to the position of the clamps. The relevant fluorescence spectra are shown in Fig. 8(d) together with typical fluorescence spectra from the travertine (brown): the distinctive feature of the spectra referring to the clamps (spectra marked in light blue and yellow) is the 380-nm peak due to the corrosion-inhibiting coating, which is present both in the blue and the yellow spectra. In addition, while the blue fluorescence spectrum clearly refers to a spot largely centered on the metal clamp, the other two spectra marked in yellow show an intermediate behavior between the fluorescence features of the corrosion-inhibiting coating and those of the travertine because in this case the spot was not centered on the metal clamp. Another consideration can be done about the spatial resolution adopted for the scan of the area: in this case, a resolution of 10 cm was enough to detect the position of the clamps and to assess the presence of the corrosion-inhibiting coating on them. This shows as a medium-to-low spatial resolution can be used for a first screening to detect roughly the position of metal bars on the surface: the same technique can be then exploited to perform scan in the areas of major interest with a higher spatial resolution.
The method has been applied also in heavily soiled areas to check the feasibility of detecting metal clamps on them. Figure 9 reports the results of fluorescence lidar mapping on an area characterized by heavily soiled and weathered travertine (area A in Fig. 1(d)). Iron clamps were attached during a past external reinforcement intervention, carried out by Valadier, to strengthen stressed and affected stones. The area has an exceptionally irregular surface indicated by cavities and large material losses, where soot and different deposits on the surface have created a very dark crust and patina which covers not only the travertine but also the iron clamps. The scanned area is shown in Fig. 9(a), with the grid indicating the spots where fluorescence measurements were taken. Data were acquired with 308-nm excitation. An analysis of the fluorescence data with the CA technique to extract the position of the iron clamps in the scanned area provided the map shown in Fig. 9(b). For this CA analysis the ‘Euclidean’ distance and the ‘single’ linkage method were used. Light blue pixels identify the vast majority of spots corresponding to soiled and weathered travertine, while the presence of the soiled iron clamps is highlighted by the blue pixels in the map. A set of fluorescence spectra corresponding to the identified classes are shown in Fig. 9(c), where the colors of the spectra correspond to the color of the relevant class as shown in Fig. 9(b). Here fluorescence spectra are not normalized. From this graph it is clear that the spectra corresponding to the iron clamps (blue spectra) are characterized by a very low intensity with respect to all the other fluorescence spectra. In addition, the iron clamp spectra, which are shown separately and rescaled in Fig. 9(d), show a spectral shape that is considerably different from those referring to the soiled stones (see Fig. 9(c)), being characterized by a sharp fluorescence contribution around 450 nm. Such spectral shape is very different from the fluorescence spectra obtained from the travertine. Also in a heavily soiled area, such as this one, where there is a higher variability of the spectral shape of the fluorescence due to the presence of many contaminants on the surface, it is not a shape expected for travertine stone. Since the scanned area was relatively easy to be observed from the ground (about 6 m above the ground level), the position of the iron clamps could be also verified by visual inspection (yellow circles in Fig. 9(a); the green circle, on the contrary, refers to a deep crack and spalling of the stone). However, the fluorescence imaging can give the same information from any portion of the façade, where visual inspection is limited unless scaffolding or lifts are used. Another piece of information that can be retrieved from the spectra shown in Fig. 9(d) is that the clamps are not recently treated with the corrosion-inhibiting coating detected on those previously analyzed because of the lack of the typical fluorescence peak centered at about 380 nm.
A further aspect to be considered for the detection and characterization of metal clamps on a monument’s façade is a proper choice of the scanning spatial resolution. In this case, the 8-cm resolution was quite coarse for a thorough detection of the clamps, as points easily miss the narrow features. Only a few points on the clamps were found, as can be inferred from the comparison of Fig. 9(a) and Fig. 9(b). On the other hand, a trade-off between spatial resolution and time required for the scanning must often be achieved, especially for a first screening of an extensive surface. Further detailed analysis can be then carried out on areas selected on the basis of a first screening and adopting a higher spatial resolution such as that of Fig. 7.
A further area containing old iron clamps was scanned and the results are shown in Fig. 10. Figure 10(a) indicates the measured area with a red frame; an analysis to single out the treated metal clamps has been superimposed. The values were obtained by using a ratio between the mean value of the spectral intensities in the 385–405 nm and the mean value of the spectral intensities in the 480–625 nm wavelength band. The color scale indicates the value of this ratio (T is the applied threshold used to suppress low-value spectra). Some example spectra are given in Fig. 10(b) and these points are marked with squares in Fig. 10(a). The red squares correspond to the points on metal clamps, and these spectra are drawn as red curves. Similarly, the green squares correspond to travertine, and these are drawn as green curves.
This fluorescence lidar study clearly shows under several aspects the potential of the technique to locate historical conservation interventions. Firstly, the technique was demonstrated to provide useful information for supporting the experts in the identification of original stone blocks that were re-worked during undocumented conservation interventions carried out in the past. Secondly, fluorescence-based false color maps were also exploited to point out joints and strengthening clamps over the monument’s surface. A further aspect that was successfully investigated was the assessment of the distribution of the corrosioninhibiting coating applied on metal clamps placed to reinforce the structure.
Although limited in providing analytical information, the technique can be particularly suitable for a first quick screening of a monument’s extended surfaces: in fact, a specific advantage of the fluorescence lidar technique is that it operates on the remote-sensing principle and thus makes areas of high elevation accessible without using scaffolds, as traditionally needed to collect samples for laboratory analysis. Fluorescence lidar can provide a non-invasive investigation of large areas without the use of any infrastructure and this could open the way for a routine, timely in situ monitoring of the vast outdoor stone cultural heritage.
The concurrent deployment of the two systems was also used to compare the performances of the two different lasers used in this experiment. In general, the excimer laser with emission at 308 nm provided additional spectral information at lower wavelengths (320–370 nm) where a fluorescence contribution is often found, e.g. in case of heavily soiled stones. On the other hand, the Nd:YAG laser offered the advantage of a higher pulse repetition rate (20 Hz with respect to 2 Hz), which minimized the time needed for image acquisition, and a higher reliability and ruggedness (e.g., no dead times required for laser cavity refilling and higher energy stability).
The authors wish to acknowledge the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS), the Swedish Research Council (VR) as well as the Swedish Institute in Rome for supporting this study and the El.En. S.p.A. (Italy) for funding the CNR lidar upgrading. They are finally indebt to Prof. A. La Regina and Arch. G. Martines for the access to the Colosseum site to carry out this study.
References and links
1. A. Moropoulou, N.P. Avdelidis, and E.T. Delegou, “NDT and planning on historical buildings and complexes for the protection of cultural heritage,” in Cultural Heritage Conservation and Environmental Impact Assessment by Non-Destructive Testing and Micro-Analysis, R. Grieken Van and K. Janssens, eds. (Taylor & Francis Group, London, UK, 2005), pp. 67–76.
2. R. Dallas, Guide for Practitioners 4: Measured Survey and Building Recording for Historic Buildings and Structures, (Historic Scotland, Edinburgh, 2004). [PubMed]
3. A. Aldrovandi, E. Buzzegoli, A. Keller, and D. Kunzelman, “Investigation of painted surfaces with a reflected UV false color technique,” in Proceedings of Art’05 - 8th International Conference on Non Destructive Investigations and Microanalysis for the Diagnostics and Conservation of the Cultural and Environmental Heritage, C. Parisi ed. (ICR, Brescia, Italy, 2005), pp. 3–18.
4. C. Fischer and I. Kakoulli, “Multispectral and hyperspectral imaging technologies in conservation: current research and potential applications,” Reviews in Conservation 7, 3–16 (2006).
5. E. Ciliberto and G. Spoto, Modern Analytical Methods in Art and Archaeology - Vol. 155 in Chemical Analyses, (John Wiley & Sons, New York, 2000).
6. M. Laurenzi Tabasso and S. Simon, “Testing methods and criteria for the selection/evaluation of products for the conservation of porous building materials,” Reviews in Conservation 7, 67–82 (2006).
7. S. Svanberg, “Fluorescence spectroscopy and imaging of lidar targets,” in Laser Remote Sensing, T. Fujii and T. Fukuchi, eds. (CRC Press, Boca Raton, 2005), pp. 433–467.
8. F.E. Hoge, “Oceanic and terrestrial lidar measurement,” in Laser Remote Chemical Analysis, R.M. Measures, ed. (John Wiley&Sons, New York, 1988), pp. 409–503.
9. V. Raimondi, L. Masotti, G. Cecchi, and L. Pantani, “Remote sensing of cultural heritage: a new field for lidar fluorosensors,” in Proceedings of the 1st International Congress on Science and Technology for the Safeguard of Cultural Heritage in the Mediterranean Basin (Tipolitografia Luxograph s.r.l., Palermo, Italy, 1998) vol. II, pp. 935–938.
10. V. Raimondi, G. Cecchi, L. Pantani, and R. Chiari, “Fluorescence lidar monitoring of historic buildings,” Appl. Opt. 37, 1089–1098 (1998). [CrossRef]
11. P. Weibring, T. Johansson, H. Edner, S. Svanberg, B. Sundnér, V. Raimondi, G. Cecchi, and L. Pantani, “Fluorescence lidar imaging of historical monuments,” Appl. Opt. 40, 6111–6120 (2001). [CrossRef]
12. D. Lognoli, G. Cecchi, I. Mochi, L. Pantani, V. Raimondi, R. Chiari, Th. Johansson, P. Weibring, H. Edner, and S. Svanberg, “Fluorescence lidar imaging of the cathedral and baptistery of Parma,” Appl. Phys. B 76, 457–465 (2003). [CrossRef]
13. L. Pantani, G. Cecchi, D. Lognoli, I. Mochi, V. Raimondi, D. Tirelli, M. Trambusti, G. Valmori, P. Weibring, H. Edner, T. Johansson, and S. Svanberg, “Lithotypes characterization with a fluorescence lidar imaging system using a multi-wavelength excitation source,” Proc. SPIE 4886, 151–159 (2003). [CrossRef]
14. G. Cecchi, L. Pantani, V. Raimondi, L. Tomaselli, G. Lamenti, P. Tiano, and R. Chiari, “Fluorescence lidar technique for the remote sensing of stone monuments,” J. Cult. Heritage 1, 29–36 (2000). [CrossRef]
15. G. Cecchi, L. Pantani, V. Raimondi, D. Tirelli, L. Tomaselli, G. Lamenti, M. Bosco, and P. Tiano, “Fluorescence lidar technique for the monitoring of biodeteriogens on the cultural heritage,” Proc. SPIE 2960, 137–147 (1996). [CrossRef]
16. D. Lognoli, G. Lamenti, L. Pantani, D. Tirelli, P. Tiano, and L. Tomaselli, “Detection and characterisation of biodeteriogens on stone cultural heritage by fluorescence lidar,” Appl. Opt. 41, 1780–1787 (2002). [CrossRef] [PubMed]
17. G. Ballerini, S. Bracci, L. Pantani, and P. Tiano, “Lidar remote sensing of stone cultural heritage: Detection of protective treatments,” Opt. Eng. 40, 1579–1583 (2001). [CrossRef]
18. J. Hällström, Architectural Conservation and Restoration, Lund University, P.O. Box 118, SE- 221 00 Lund, Sweden, and K. Barup, R. Grönlund, A. Johansson, S. Svanberg, L. Palombi, D. Lognoli, V. Raimondi, G. Cecchi, and C. Conti, are preparing a manuscript to be called “Documentation of façades previously cleaned: A case study on the Colosseum, Rome, using hyperspectral imaging fluorescence lidars”.
19. A. Gabucci, The Colosseum, (Electa, Milan, 2000).
20. M. Jonsson, La Cura dei Monumenti alle Origini. Restauro e Scavo di Monumenti Antichi a Roma 1800–1830, Acta Instituti Romani Regni Sueciae, Series altera in 8°, XIV (Stockholm, 1986).
21. C. Conti, “Anfiteatro Flavio: Il restauro delle superfici in travertino,” Arkos: Scienza e Restauro 2, 22–27 (2001).
22. G. Cecchi, P. Mazzinghi, L. Pantani, R. Valentini, D. Tirelli, and P. De Angelis, “Remote sensing of chlorophyll a fluorescence of vegetation canopies: 1. Near and far field measurement techniques,” Remote Sens. Environ. 47, 18–28 (1994). [CrossRef]
23. G. Cecchi, L. Pantani, B. Breschi, D. Tirelli, and G. Valmori, “FLIDAR: A multipurpose fluorosensorspectrometer,” EARSeL Advances in Remote Sensing 1, 72–78 (1992).
24. H. Edner, J. Johansson, S. Svanberg, E Wallinder, M. Bazzani, B. Breschi, G. Cecchi, L. Pantani, B. Radicati, V. Raimondi, D. Tirelli, G. Valmori, and P. Mazzinghi, “Laser-induced fluorescence monitoring of vegetation in Tuscany,” EARSeL Advances in Remote Sensing 1, 119–130 (1992).
26. R. Grönlund, J. Hällström, S. Svanberg, and K. Barup, “Fluorescence lidar imaging of historical monuments - Övedskloster, a Swedish case study,” in Lasers in the Conservation of Artworks: LACONA VI Proceedings, Vienna/Austria, Sept. 21–25, 2005, J. Nimmrichter, W. Kautek, and M. Schreiner, eds. (Springer, Berlin, Germany, 2007) pp. 583–592.
28. P. Weibring, J.N. Smith, H. Edner, and S. Svanberg, “Development and testing of a frequency-agile optical parametric oscillator system for differential absorption lidar,” Rev. Sci. Instrum. 74, 4478–4484 (2003). [CrossRef]
29. C. af Klinteberg, M. Andreasson, O. Sandström, S. Andersson-Engels, and S. Svanberg, “Compact medical fluorosensor for minimally invasive tissue characterization,” Rev. Sci. Instrum. 76, 034303 (2005). [CrossRef]
30. A.C. Rencher, Methods of Multivariate Analysis (Wiley Interscience, New York, 2002). [CrossRef]
31. P.F. Velleman and D.C. Hoaglin, Applications, Basics, and Computing of Exploratory Data Analysis, (Duxberry Press, Boston, 1981).
32. Y. Hochberg and A.C. Tamhane, Multiple Comparison Procedures, (Wiley Interscience, New York, 1987). [CrossRef]