The system we have designed is contactless, capable of scanning shapes of about 1.40 × 0.90 m2, and able to be configured and set up for any situation, thanks by the Cartesian robots used for the lasers scansion. Among the 3D imaging sensors [26, 27] we decided to use the single point laser triangulation methodology due to: (i) the high accuracy in a given field of view, (ii) the inherent insensitivity to the environmental illumination conditions, and (iii) the high acquisition rate [28].
Taking into account the irreplaceable nature of the pieces and the potentially damage caused by the laser beam and the induced heat, an overriding concern was to avoid harming the masterpieces. Therefore, being required that the system operated at the highest safety level, we choose a red semiconductor laser with a power of 0.5 W and it is moved continuously during scanning.
The acquisition system was composed of: (i) two laser displacement-sensors (PM Instruments, LDS, FSO 30 mm, 0.01%), and (ii) a two-axes XY-robot (Robostar, ROK stroke 300 mm, payload 40 kg) for the sensor positioning. The system was assembled using an industrial mounting system (Montech, Quickset series) characterized of high flexional rigidity and low perpendicularity tolerance. An industrial pc (National Instruments, PXI-8105) equipped with a 16 bit DAQ acquisition device, and with an custom LabVIEW application (NI, US), triggered the motors and acquired signals. In order to prevent any possible damage, we built the control software to preclude that the laser beam could remain still on the tested artifact for hardware or software bad functioning. Thus, to preserve the integrity of the surface, a disaster recovery procedure was implemented: the measurement heads are automatically moved in a safe location outside the painting and switched off when an alarm code is identified by the in-house developed control software.
The described system was assembled in two different configurations, as a function of the analyzed artwork, as reported below. For both the cases, the test were performed in a controlled microclimate environment.
“Annunciazione”
The automatic system was used to monitor the horizontal displacements of the canvas fibers during the tensioning tests. The experiments aimed to provide detailed information for the analysis of the residual elasticity of the fibers, so to optimally set the tensioning of the canvas.
As the Fig. 4c shows, two surfaces of the masterpiece were scanned, one area on the front of the painting and one area on back. In the frontal side of the canvas, the full measurement area, 0.275 × 0. 580 m2 (Fig. 4) was divided into two sub-areas of 0.275 × 0.290 m2 (identified in the figure as areas A and B) due to the robot stroke, which was limited to 0.300 × 0.300 m2. A reference plane for the canvas deflection at the frontal side was obtained by placing a series of vertical whiteboard tapes, on the supporting frame of the artwork, with a distance of 3 mm from the canvas. In the posterior area, a surface of 0.275 × 0.290 m2 (area C) was acquired. Before starting the tests, the parallelism between the painting surface and the laser’s head was controlled by comparing the relative distances at the four corners and at the center of the scanned area. The frame of the mechatronic system was considered to have been correctly positioned when the all the relative distances were less than 0.5 mm. We computed the relative displacement along the orthogonal direction of the canvas, assuming as a reference the plane given by the added whiteboard tapes, for both the frontal and posterior scanned areas.
Figure 4b report the laser trajectory performed by the robot end effector: each of the areas A, B and C was divided into a grid of 14 horizontal lines with vertical spacing of 20 mm. In order to limit the vibrations induced by the movement of the robot, the motor speed was limited to 17.00 ± 0.02 mm/s. The velocity profile of the robot was trapezoidal, i.e. (1) acceleration, (2) constant speed, and (3) deceleration, but only in (4) the laser output was considered valid.
The acquisitions were conducted in parallel with the verification tests performed by the laboratory staff of the ISCR; the purpose of these tests was to estimate the tensioning level the canvas. The tensioning tests consisted of a micrometer screw, positioned at the center of the rear part of the canvas, used to impose horizontal displacements to the center of the canvas. The actual position of the screw and the correspondent reaction force of the canvas were monitored by a linear variable differential transformer (LVDT) and load cell, respectively (Fig. 5). Because the experiments aimed only to estimate the stress condition of the canvas, the displacement level imposed to the center of the canvas was minimal. In addition, any operation have been authorized by the head of the department and the director of the Institute.
Two different tensioning procedures have been implemented. In the first procedure, the micrometer screw advanced with the sequence (0, 1, 2) mm. In the second procedure, the screw advanced with the sequence (0, 3.75, 6.25, 8.75, 10) mm. The difference between the two selected ranges was suggested by the team in charge for the restoration. The position of 0 mm was identified with the simple contact of the screw with the canvas. The laser scan was performed on the areas A and B, at the front of the canvas, during the first tensioning test, and on the area C, at the back of the canvas, during the second one. The time needed for the acquisition was of about 7 and 4 m for the two tests, respectively.
The experimental procedure was articulated in: (i) the imposition of a selected out-of-plane condition via the micrometer screw; (ii) the collection of LVDT data; and (iii) the starting of the surface scanning and the gathering of laser sensor outputs.
“Portiera Oddi-Montesperelli” and “Paliotto di San Domenico”
The “Portiera Oddi-Montesperelli” was analyzed after the conservation treatments to verify the capability of the new support, manufactured by the ISCR, to follow the canvas deformations due to the change of the environmental conditions, i.e. temperature and RH, that could damage the gilding and the stamps [5]. Actually, the fluctuations of temperature and RH are considered one of the main factors that contribute to the deterioration of artifacts.
The “Paliotto di San Domenico” was monitored before the conservation to permit the design of a new specific support frame, and to use the collected data for a comparison after the expected conservation treatments.
In both the cases, the system was equipped with a temperature and RH transducer (HMT-310, Vaisala, FI).
The robot configuration was similar to the one adopted for “Annunciazione”, but with an additional vertical motor M3, having a stroke of 1200 mm (Fig. 6). The horizontal motor M2 was not used in the present case. Moreover, because the system had to monitor the vertical displacements of the vertical edges of the canvas, two laser sensors were mounted on a horizontal bar fixed to the end-effector of the robot. Two series of markers were positioned on both sides of the “Portiera Oddi-Montesperelli”, but only one series on the left side of the “Paliotto di San Domenico” in Fig. 6a (the right side was too much deteriorated). Each 8 consisted of a small plastic clip with square-section 10 mm × 10 mm and a length 50 mm. In the case of “Paliotto”, 8 markers was attached to the canvas vertical edges, and one marker (named F in Fig. 6a, b) was placed on the fixed frame, close to the canvas. The nominal distanced between each marker pair was 100 mm. In the case of “Portiera”, the used markers were 9 + 1 in total.
The test procedure can be schematized as it follows: (i) the acquisition of the environmental temperature and relative humidity; (ii) the laser signal acquisition to verify the parallelism between the acquisition system and the panel (parallelism dimensional tolerance ≤0.5 mm); (iii) the acquisition of the laser signal (scanning speed ~ 17.00 ± 0.02 mm/s) for detecting the marker edges, in order to measure the relative distances between each pair of contiguous markers.
The relative displacements were evaluated by acquiring the elapsed time between two consecutive detections of markers and multiplying it for the motor speed. We computed the distance-variation of each pair of markers, defined as the difference between the maximum and the minimum relative distance measured during the day. The marker fixed on the support frame was used as a reference for measuring the total stretching of the two canvases.
The implemented procedure for the monitoring of “Portiera Oddi-Montesperelli” consisted of eight acquisitions from 10:00 a.m. to 4:00 p.m. for three consecutive days. For the “Paliotto di San Domenico”, instead, we performed three acquisitions each day, for three non-consecutive days in the same week, from 12:00 p.m. to 2:00 p.m. For both the cases, the time interval between each acquisition was 40 min. The tests were scheduled automatically in climate-controlled environment.
System characterization
The uncertainty estimation can be performed (i) by means of repeated measures on a suitable standard reference object (Type-A approach [29]), or, if sufficient and reliable information are available about the relevant error sources, (ii) by a Type-B evaluation approach, or (iii) or by a mixed approach (Type-A plus Type-B).
The use of a reference object is certainly the most reliable method for accuracy estimation of the system. However, the simplicity of the measuring system makes the procedure (ii) sufficiently reliable. The apparatus, in fact, consists of an XY-cartesian robot and a laser displacement sensor, therefore, the only knowledge of the accuracies of both of the devices, certified by the calibrations documents, is basically sufficient to estimate the overall uncertainty value.
The uncertainty has to be estimated differently for both the two configurations previously described, i.e. for the canvas painting and for the leather paintings; however, few preliminary considerations need to be carried out.
The main relevant inaccuracy sources for the overall acquisition system are: (i) the inaccuracy of the robot in reproducing a selected target trajectory; (ii) the uncertainty associated to the laser displacement-sensors; (iii) the precision in assembling the support frame; (iv) the sensitivity of the assembly to environmental effects such as temperature and vibrations. The first two sources can be evaluated with Type-B uncertainty estimation approach via the manufacturer specifications. The effects of the third error source could be reduced and estimated by using additional knowledge on the mounting tolerance. In particular, the apparatus mounting inaccuracy, which affects the displacement measures with both systematic and predictable components, can be corrected by using additional geometrical references, acquired during the artwork laser scanning. The assembly sensitivity to temperature and vibrations, instead, cannot be modeled, and, therefore, this phenomenon has to be reduced as much as possible by strengthening the frame and reducing the robot speed at minimum.
“Annunciazione”
The uncertainty associated to the laser displacement sensor (±0.005 mm) combines with the planarity tolerance of the robot (±0.01 mm). Considering independent the two sources of uncertainties, by adding them in quadrature, the uncertainty in the planarity of the scanned surface can be estimated equal to ±0.02 mm. The displacement data were collected while the robot was moving horizontally at a constant sampling (1 kHz), it follows that the capability of the robot to maintain a constant speed influenced the position estimation of the laser within the scan area. In order to reduce the effect of the speed inaccuracy, the external whiteboard tapes have been taken as references for the scanned areas; more precisely, the whiteboard edged has been used to normalize the signals collected for each horizontal line of the scan. The normalization could not eliminate the inaccuracy due to the velocity variation around the average value, but the effect in term of horizontal trajectory inaccuracy is less than the maximum tracking-error admitted by the controller (max admitted deviation ±0.05 mm). As regards the vertical position of the laser, because it was constant during the scan of each line of the grid, the associated inaccuracy was 0.05 mm, which is negligible in comparison with the horizontal accuracy. In conclusion, the inaccuracy of laser position in the surface scan can be estimated as equal to 0.05 mm.
“Portiera Oddi-Montesperelli” and “Paliotto di San Domenico”
In these cases the sensors were used only for the detection of the edges of the markers, thus, the accuracy of the laser sensors and the planarity of the robot did not affect the results. The main cause of error can be found in the robot velocity inaccuracy along the vertical direction. Actually, because the aim of the experiment was to monitor only the variation of the vertical relative distances between the markers, and not the absolute displacements, the positional inaccuracy of the robot assumed a minor role. As previously mentioned, the inaccuracy can be considered lower than the tracking error, i.e. 0.05 mm. However, in order to take into account the effect of the temperature on the robot support, an estimation of the robot frame stability had to be performed. Therefore, six markers were attached to a fixed rigid reference frame, simulating a constant marker displacement input for the acquisition system. Two markers were positioned at the higher side of the frame, two at the middle side and two in the lower side. The standard deviation of repeated measures during the daily acquisitions was calculated for the distance between each pair of adjacent markers. The value obtained was 3σ = 0.2 mm. In conclusion, the observed deformation of the robot frame during the monitoring can be assumed not greater than 0.2 mm.