Statistical Comparison between Low-Cost Methods for 3D Characterization of Cut-Marks on Bones

In recent years, new techniques for the morphological study of cut marks have become essential for the interpretation of prehistoric butchering practices. Different criteria have been suggested for the description and classification of cut marks. The methods commonly used for the study of cut marks rely on high-cost microscopy techniques with low portability (i.e., inability to work in situ), such as the 3D digital microscope (3D DM) or laser scanning confocal microscopy (LSCM). Recently, new algorithmic developments in the field of computer vision and photogrammetry, have achieved very high precision and resolution, offering a portable and low-cost alternative to microscopic techniques. However, the photogrammetric techniques present some disadvantages, such as longer data collection and processing time, and the requirement of some photogrammetric expertise for the calibration of the cameras and the acquisition of precise image orientation. In this paper, we compare two low-cost techniques and their application to cut mark studies: the micro-photogrammetry (M-PG) technique presented, developed, and validated previously, and a methodology based on the use of a structured light scanner (SLS). A total of 47 experimental cut marks, produced using a stainless steel knife, were analyzed. The data registered through virtual reconstruction was analyzed by means of three-dimensional geometric morphometrics (GMM).


Introduction
Recently, archaeological research has substantially changed, due to the application of new technologies and multidisciplinary approaches. The application and implementation of the latest technological advances have enhanced research, scientific dissemination, and archaeological heritage management, promoting the creation of new research methodologies. Consequently, a more precise archaeological investigation is possible nowadays, including the performance of very accurate analyses with better graphic and visual designs that also promote the dissemination of the findings to a wider audience.
Geotechnologies, especially photogrammetry and laser scanning, have long played an important role in archaeological research. Photogrammetry, as a geotechnology of taking measurements and Low-cost systems arose for the videogame industry. The most commonly used low-cost system is the Kinect system (depth sensor, real-time Mobile 3d), which can be used alone, or can be incorporated to other systems, implying a large cost increase (e.g., Google Tango ( Figure 1f) (Table 1), DPI-8, eyesmap, etc.). The resolution reached by these systems is around 8 mm at best [48]. In archeology, the use of these tools is proliferating for the close capture of sceneries as a complement to the terrestrial laser scanning [49]. It has also been used for documentation of stone carvings [50]. The stability and metric quality of low-cost systems (resolutions of 8 mm) cannot compete with high-or medium-range systems (resolutions of 0.02 mm and 0.05 mm, respectively).
In order to alleviate the disadvantages of the M-PG techniques, as well as those related to the active handheld laser scanner systems, here, we present a methodology based on a low-cost and improved structured light scanner (SLS-2) technique as an alternative to the study of cut marks on bone surfaces. We also compare, statistically, this methodology (in terms of precision and resolution) with a previously validated M-PG technique [19,51]. The comparison of these two techniques permits elucidating their simultaneous applicability in archaeological contexts. In addition, using structured light laser scanner (SLS-2) techniques, it is possible to obtain high resolution images that allow the study of taphonomic marks in detail and in real time, as SLS-2 requires short data collection and processing times. Thus, SLS-2 technique presents two fundamental features that enable the study of microscopic morphological details of larger samples. In this paper, we also test the capacity of SLS-2 systems to assess differences among cut marks on bones with different morphological characteristics.
Low-cost systems arose for the videogame industry. The most commonly used low-cost system is the Kinect system (depth sensor, real-time Mobile 3d), which can be used alone, or can be incorporated to other systems, implying a large cost increase (e.g., Google Tango ( Figure 1f) (Table  1), DPI-8, eyesmap, etc.). The resolution reached by these systems is around 8 mm at best [48]. In archeology, the use of these tools is proliferating for the close capture of sceneries as a complement to the terrestrial laser scanning [49]. It has also been used for documentation of stone carvings [50]. The stability and metric quality of low-cost systems (resolutions of 8 mm) cannot compete with highor medium-range systems (resolutions of 0.02 mm and 0.05 mm, respectively).
In order to alleviate the disadvantages of the M-PG techniques, as well as those related to the active handheld laser scanner systems, here, we present a methodology based on a low-cost and improved structured light scanner (SLS-2) technique as an alternative to the study of cut marks on bone surfaces. We also compare, statistically, this methodology (in terms of precision and resolution) with a previously validated M-PG technique [19,51]. The comparison of these two techniques permits elucidating their simultaneous applicability in archaeological contexts. In addition, using structured light laser scanner (SLS-2) techniques, it is possible to obtain high resolution images that allow the study of taphonomic marks in detail and in real time, as SLS-2 requires short data collection and processing times. Thus, SLS-2 technique presents two fundamental features that enable the study of microscopic morphological details of larger samples. In this paper, we also test the capacity of SLS-2 systems to assess differences among cut marks on bones with different morphological characteristics.

Sample
For the purpose of this study, experimental cut marks were produced using a stainless steel knife, model Molybdenum Vanadio C 0.5 CR 14 MO 0.5 VA 0.25, which allows the control of certain variables (see Maté-González et al. [19] for further details) ( Figure 2). A total of 22 cut marks on an Ovis aries radius and humerus, were registered using SLS and M-PG techniques, creating 2 homologous sets of cut marks to assess differences in virtual reconstruction techniques. A second set of cut marks, reproduced using SLS, was analyzed to evaluate possible differences among cut marks on long bone epiphyses and flat bones. Radius and humerus are long bones with a more or less oval shaft, which determines cut mark section, shape, and form, when compared to cut marks performed of flat axial bones, where the curvature of bones is less, or not pronounced. In this case, a total of 21 cut marks on long bone shafts, and 26 cut marks on a scapula, were virtually reconstructed and statistically analyzed. The bones used in this second part of the study also belong to a young Ovis aries individual.

Sample
For the purpose of this study, experimental cut marks were produced using a stainless steel knife, model Molybdenum Vanadio C 0.5 CR 14 MO 0.5 VA 0.25, which allows the control of certain variables (see Maté-González et al. [19] for further details) ( Figure 2). A total of 22 cut marks on an Ovis aries radius and humerus, were registered using SLS and M-PG techniques, creating 2 homologous sets of cut marks to assess differences in virtual reconstruction techniques. A second set of cut marks, reproduced using SLS, was analyzed to evaluate possible differences among cut marks on long bone epiphyses and flat bones. Radius and humerus are long bones with a more or less oval shaft, which determines cut mark section, shape, and form, when compared to cut marks performed of flat axial bones, where the curvature of bones is less, or not pronounced. In this case, a total of 21 cut marks on long bone shafts, and 26 cut marks on a scapula, were virtually reconstructed and statistically analyzed. The bones used in this second part of the study also belong to a young Ovis aries individual. Only cut marks where the limits of the mark were clear, e.g., the start and end of the mark could be easily identified, were selected for the study. A very abnormal cut mark, according to both registration techniques (SLS and M-PG), was excluded from the analysis to avoid statistical noise.

Reflex Camera and Macro-Lens
Cut marks were first registered using micro-photogrammetry (M-PG) and computer vision techniques, so as to create high-resolution 3D models of each cut mark. Precise metrical models were generated using images taken with vertical and oblique photography, using a CANON EOS 700D with a 60 mm macro lens (Table 2), and following the specified protocol explained in Maté-González et al. [18]. The camera was self-calibrated to simultaneously compute the interior and exterior camera parameters [52]. A total of 9-13 photos were taken for each mark, depending on the geometry of the bone and the shape of the mark (Figure 3a). The 3D reconstruction of a single mark took 25-30 min depending on the number of photos acquired. Photographs were processed with the open-source photogrammetric reconstruction software GRAPHOS (inteGRAted PHOtogrammetric Suite) [53,54] to generate a 3D model for each mark (Figure 3b). The scaled 3D models were exported as PLY files, and subsequently registered and studied by means of Geometric Morphometrics. Only cut marks where the limits of the mark were clear, e.g., the start and end of the mark could be easily identified, were selected for the study. A very abnormal cut mark, according to both registration techniques (SLS and M-PG), was excluded from the analysis to avoid statistical noise.

Reflex Camera and Macro-Lens
Cut marks were first registered using micro-photogrammetry (M-PG) and computer vision techniques, so as to create high-resolution 3D models of each cut mark. Precise metrical models were generated using images taken with vertical and oblique photography, using a CANON EOS 700D with a 60 mm macro lens (Table 2), and following the specified protocol explained in Maté-González et al. [18]. The camera was self-calibrated to simultaneously compute the interior and exterior camera parameters [52]. A total of 9-13 photos were taken for each mark, depending on the geometry of the bone and the shape of the mark (Figure 3a). The 3D reconstruction of a single mark took 25-30 min depending on the number of photos acquired. Photographs were processed with the open-source photogrammetric reconstruction software GRAPHOS (inteGRAted PHOtogrammetric Suite) [53,54] to generate a 3D model for each mark (Figure 3b). The scaled 3D models were exported as PLY files, and subsequently registered and studied by means of Geometric Morphometrics.

Structured Light Scanner
Cut marks were also digitalized with a DAVID structured-light scanner SLS-2 (Table 3), a newer version of the former DAVID SLS-1 which was improved with a macro lens getting better resolutions. This equipment, located at the C.A.I. of Archaeometry at the Complutense University of Madrid, consists in a camera, a projector, and a calibration marker board that in the first phase, needs to be calibrated ( Figure 4). To carry out this process, a DAVID USB CMOS Monochrome camera is positioned and fit with a macro lens alongside an ACER K132 projector, both facing towards the calibration marker board at an angle between 15 • and 25 • (Figure 5a). The projection produced by the projector has to cover the entire calibration marker board, in our case, the size and calibration pattern corresponds to a 15 mm scale. Within the DAVID software, the scale is introduced as displayed on the calibration marker board, the camera's exposure is adjusted accordingly while the focus of all the single instruments is adjusted. The equipment is then calibrated. During this process, the camera, as well as the projector, must remain fixed and stable.
The second phase consists in substituting the calibration marker board for the bone we intend to scan. The DAVID structured-light scanner SLS-2 can produce a density of up to 1.2 million points. The use of this scanning process provides a real reproduction of the bone external topography (Figure 5b). In this case, the matt polished surface of the bones avoids problems related to light intensity, or the contrast of lights and shadows during data collection. The active sensor reduces data capture time to less than 1 min. The DAVID structured-light scanner SLS-2 used in this experiment produced a higher quality resolution than the scanner used in Maté-González et al. [18].

Structured Light Scanner
Cut marks were also digitalized with a DAVID structured-light scanner SLS-2 (Table 3), a newer version of the former DAVID SLS-1 which was improved with a macro lens getting better resolutions. This equipment, located at the C.A.I. of Archaeometry at the Complutense University of Madrid, consists in a camera, a projector, and a calibration marker board that in the first phase, needs to be calibrated ( Figure 4). To carry out this process, a DAVID USB CMOS Monochrome camera is positioned and fit with a macro lens alongside an ACER K132 projector, both facing towards the calibration marker board at an angle between 15° and 25° (Figure 5a). The projection produced by the projector has to cover the entire calibration marker board, in our case, the size and calibration pattern corresponds to a 15 mm scale. Within the DAVID software, the scale is introduced as displayed on the calibration marker board, the camera's exposure is adjusted accordingly while the focus of all the single instruments is adjusted. The equipment is then calibrated. During this process, the camera, as well as the projector, must remain fixed and stable.
The second phase consists in substituting the calibration marker board for the bone we intend to scan. The DAVID structured-light scanner SLS-2 can produce a density of up to 1.2 million points. The use of this scanning process provides a real reproduction of the bone external topography (Figure 5b). In this case, the matt polished surface of the bones avoids problems related to light intensity, or the contrast of lights and shadows during data collection. The active sensor reduces data capture time to less than 1 min. The DAVID structured-light scanner SLS-2 used in this experiment produced a higher quality resolution than the scanner used in Maté-González et al. [18].

Statistical Comparison
For the purpose of this study, 13 landmarks located on the exterior and interior of the mark groove were used to describe the cut marks, following the model proposed by Courtenay et al. [55] ( Figure 6). This model has been proved to better capture the morphological information of cut marks when using 3D techniques, including the opening angle of the mark, and the location of its central deepest point. The landmarking step was performed in Avizo (Visualisation Sciences Group, USA).
Landmarks are homologous points that contain shape and size information in the form of Cartesian coordinates, allowing the comparison among different elements [56][57][58][59]. Landmark configurations can be analyzed by means of geometric morphometric procedures based on a Procrustes superimposition, also called generalized procrustes analysis (GPA). This technique takes the landmark data, and normalizes the form information by the application of superimposition procedures (translation, rotation and scaling). After GPA, the remaining differences between the structures, defined by landmarks, expose patterns of variation and covariation that can be studied using common multivariate statistics [60][61][62].

Statistical Comparison
For the purpose of this study, 13 landmarks located on the exterior and interior of the mark groove were used to describe the cut marks, following the model proposed by Courtenay et al. [55] (Figure 6). This model has been proved to better capture the morphological information of cut marks when using 3D techniques, including the opening angle of the mark, and the location of its central deepest point. The landmarking step was performed in Avizo (Visualisation Sciences Group, USA).
Landmarks are homologous points that contain shape and size information in the form of Cartesian coordinates, allowing the comparison among different elements [56][57][58][59]. Landmark configurations can be analyzed by means of geometric morphometric procedures based on a Procrustes superimposition, also called generalized procrustes analysis (GPA). This technique takes the landmark data, and normalizes the form information by the application of superimposition procedures (translation, rotation and scaling). After GPA, the remaining differences between the structures, defined by landmarks, expose patterns of variation and covariation that can be studied using common multivariate statistics [60][61][62]. Several principal component analyses (PCA) performed in MorphoJ [63] and Morphologika 2.5 [56] were used to assess patterns of variation among the data in shape and form space, to study shape and size differences among cut marks. Form spaces, containing size and shape information, were obtained by re-scaling data using the natural logarithm of centroid size. Changes in shape and form space were visualized with the aid of transformation grids and warpings [64] computed using thin-plate splines.
Several tests were performed to assess differences and similarities among (1) registration techniques, and (2) cut marks on cylindrical and flat bones. The presence of defined groups was statistically tested using a multiple variance analysis (MANOVA) on the principal components (PC) scores, and analyses of variance (ANOVA) when the condition of variance homogeneity was fulfilled according to the Levene's test. When samples showed unequal variances, the Welch's t-test was performed to test the null hypothesis of equal means among groups. Variance analyses were performed in the free software R [65].
Linear discriminant analyses (LDA) were performed using Morpho J [63] to determine the shape features that best distinguish between the established groups [66]. Permutation tests were computed to assess differences between group means.

SLS vs. M-PG
The whole sample, consisting of 22 homologous 3D models (i.e., pairs of 3D models obtained with SLS and M-PG), was subjected to a PCA, where two outliers representing the same cut mark (N 5) can be observed (Figure 7a). Since differences between this mark (N 5) and the rest of the sample are due to the nature of the cut mark itself, as it appears clearly separated when registered using both techniques, tests were run leaving the outlier out of the study, to avoid noise. However, numerical results obtained with and without the outlier are similar, and do no show significant differences between methodologies.
The cut marks registered by means of M-PG and SLS-2, overlap considerably in the second PCA plot, where less than 50% of the total variance is explained (Figure 7b). However, a slight difference can be noticed, as cut mark models generated using M-PG show a greater dispersion that expresses larger variance. Nevertheless, this larger variance among cut marks registered using M-PG methods seems to be mainly produced by only two specific marks that are clearly separated from the main scatter range.  [63] and Morphologika 2.5 [56] were used to assess patterns of variation among the data in shape and form space, to study shape and size differences among cut marks. Form spaces, containing size and shape information, were obtained by re-scaling data using the natural logarithm of centroid size. Changes in shape and form space were visualized with the aid of transformation grids and warpings [64] computed using thin-plate splines.

Several principal component analyses (PCA) performed in MorphoJ
Several tests were performed to assess differences and similarities among (1) registration techniques, and (2) cut marks on cylindrical and flat bones. The presence of defined groups was statistically tested using a multiple variance analysis (MANOVA) on the principal components (PC) scores, and analyses of variance (ANOVA) when the condition of variance homogeneity was fulfilled according to the Levene's test. When samples showed unequal variances, the Welch's t-test was performed to test the null hypothesis of equal means among groups. Variance analyses were performed in the free software R [65].
Linear discriminant analyses (LDA) were performed using Morpho J [63] to determine the shape features that best distinguish between the established groups [66]. Permutation tests were computed to assess differences between group means.

SLS vs. M-PG
The whole sample, consisting of 22 homologous 3D models (i.e., pairs of 3D models obtained with SLS and M-PG), was subjected to a PCA, where two outliers representing the same cut mark (N 5) can be observed (Figure 7a). Since differences between this mark (N 5) and the rest of the sample are due to the nature of the cut mark itself, as it appears clearly separated when registered using both techniques, tests were run leaving the outlier out of the study, to avoid noise. However, numerical results obtained with and without the outlier are similar, and do no show significant differences between methodologies.
The cut marks registered by means of M-PG and SLS-2, overlap considerably in the second PCA plot, where less than 50% of the total variance is explained (Figure 7b). However, a slight difference can be noticed, as cut mark models generated using M-PG show a greater dispersion that expresses larger variance. Nevertheless, this larger variance among cut marks registered using M-PG methods seems to be mainly produced by only two specific marks that are clearly separated from the main scatter range. Variance analyses (MANOVA, ANOVA) conducted on the principal component (PC) scores and assessing a greater variance percentage than the one observed in the PCA plot, support the results depicted in the graph. Both the MANOVA and ANOVA results (Table 4) highlight that there is no significant difference (p-value > 0.05) between the models generated by means of M-PG and the SLS-2. In fact, the Levene's test performed on the sample confirmed the homogeneity of the variance in both groups (p-value > 0.05). Equally, differences between registration techniques were assessed in form space, where changes in size are also considered alongside shape, resulting in no significant differences (Table 4).  Variance analyses (MANOVA, ANOVA) conducted on the principal component (PC) scores and assessing a greater variance percentage than the one observed in the PCA plot, support the results depicted in the graph. Both the MANOVA and ANOVA results (Table 4) highlight that there is no significant difference (p-value > 0.05) between the models generated by means of M-PG and the SLS-2. In fact, the Levene's test performed on the sample confirmed the homogeneity of the variance in both groups (p-value > 0.05). Equally, differences between registration techniques were assessed in form space, where changes in size are also considered alongside shape, resulting in no significant differences (Table 4). A jackknife cross-validated LDA was performed to assess minimal variance within cut mark groups, and maximal variance between groups, extracting a confusion matrix that indicates high confusion rates between the cut marks registered with M-PG techniques and the SLS-2 (Table 5). Though misclassified marks do never surpass the 50%, the Procrustes and Mahalanobis distances calculated between the groups, and the permutation tests run to assess their significance, stress that the two groups cannot be distinguished (Procrustes distance p-value = 0.44, Mahalonobis distance p-value = 0.47).
According to our statistical results, cut marks registered using M-PG and SLS-2 show no significant difference, and can thus be equally applied for the study of cut marks. In this way, the use of the SLS method would be validated in taphonomic contexts, becoming a supplementary technique that can be added to previously investigated microscopic and photogrammetric techniques.

Cut Marks on Long Bone Diaphyses vs. Cut Marks on Flat Bones
The PCA of the cut marks in shape space shows a non-polarized morphospace, defined by a high number of PCs. The scatter-plot in Figure 8 expresses only 53.1% of the total variance. In shape space, cut marks on the scapula (flat bone) appear more widely distributed than the cut marks produced on the humerus and the radius (long bones) that are mainly concentrated in the center of the graph. Though both groups clearly overlap, the shape of the cut marks on flat bones tends to be more variable, especially in relation to changes in shape expressed by PC1, that is described in its positive axis by flat and narrow cut mark shapes, and in its negative limit by wider and more curved shapes. Most cut marks registered on the scapula are closer to the narrow and flat shape, but variance in this sample is still quite large. The cut marks generated on long bone diaphyses tend to be closer to the mean shape, and mostly change along PC2, which is explained by changes in the direction of the curvature of the mark. However, in most cases, cut marks on long bones are distributed along the negative x-axis, indicating that these marks tend to be wider, deeper, and more curved. This description is consistent with the nature of long bone shafts.
The PCA in form space is quite different, with both groups of cut marks similarly distributed along the graph, especially along PC1 (Figure 9). The first principal component is closely related to changes in centroid size that are remarkable as changes in width. Despite the scattering of the sample in the plot, two tendencies can be noticed in relation to the nature of the bones, where cut marks haven been generated: cut marks on long bones tend to gather mostly in the central or negative area of the x-axis, which corresponds to wider, deeper, and more curved forms, while 69.2% of cut marks on the scapula appear on the positive area of the x-axis, equivalent to narrower and flatter marks. Such changes in form expressed in PC1 make up 97.6% of the total variance, while PC2 only explains 0.68% of the variance and is determined by the same variables that characterize PC1 in shape space. Cut marks on long bones are similarly scattered along the positive and negative ranges of the second principal component, whereas cut marks on flat bones mostly fall in the negative range of the y-axis, determined by flatter and narrower mark forms.  The visual overlapping of cut mark groups in the PCA plots needs to be further investigated, since those graphs are two-dimensional representations of the first two PCs. Thus, the scatter-plots, first, fail to explain the entirety of the variance expressed by the sample, and, second, cannot display the distance among groups in the z-axis.
The MANOVAs performed on the PC scores to assess the differences among group means show no significant differences in shape (Wilks' Lambda = 0.8529, F = 1.414, p = 0.2395) and form space (Wilks' Lambda = 0.9137, F = 0.7747, p = 0.536). However, the Levene's test indicates that variance among groups is unequal in shape and form space, and Welch's tests were needed to confirm the similarity of the group means (Table 4). On top of that, the confusion matrix calculated shows that confusion rates between cut marks on flat and long bones are not that high and that most of the marks can be correctly classified (Table 5).
Despite graphical overlapping and statistical non-significant results in some cases, it cannot be disregarded that there is an interesting difference in the variance range of each sample, with cut marks on long bones being more concentrated in a specific area of the graph, in response to a   The visual overlapping of cut mark groups in the PCA plots needs to be further investigated, since those graphs are two-dimensional representations of the first two PCs. Thus, the scatter-plots, first, fail to explain the entirety of the variance expressed by the sample, and, second, cannot display the distance among groups in the z-axis.
The MANOVAs performed on the PC scores to assess the differences among group means show no significant differences in shape (Wilks' Lambda = 0.8529, F = 1.414, p = 0.2395) and form space (Wilks' Lambda = 0.9137, F = 0.7747, p = 0.536). However, the Levene's test indicates that variance among groups is unequal in shape and form space, and Welch's tests were needed to confirm the similarity of the group means (Table 4). On top of that, the confusion matrix calculated shows that confusion rates between cut marks on flat and long bones are not that high and that most of the marks can be correctly classified (Table 5).
Despite graphical overlapping and statistical non-significant results in some cases, it cannot be disregarded that there is an interesting difference in the variance range of each sample, with cut marks on long bones being more concentrated in a specific area of the graph, in response to a The visual overlapping of cut mark groups in the PCA plots needs to be further investigated, since those graphs are two-dimensional representations of the first two PCs. Thus, the scatter-plots, first, fail to explain the entirety of the variance expressed by the sample, and, second, cannot display the distance among groups in the z-axis.
The MANOVAs performed on the PC scores to assess the differences among group means show no significant differences in shape (Wilks' Lambda = 0.8529, F = 1.414, p = 0.2395) and form space (Wilks' Lambda = 0.9137, F = 0.7747, p = 0.536). However, the Levene's test indicates that variance among groups is unequal in shape and form space, and Welch's tests were needed to confirm the similarity of the group means (Table 4). On top of that, the confusion matrix calculated shows that confusion rates between cut marks on flat and long bones are not that high and that most of the marks can be correctly classified (Table 5).
Despite graphical overlapping and statistical non-significant results in some cases, it cannot be disregarded that there is an interesting difference in the variance range of each sample, with cut marks on long bones being more concentrated in a specific area of the graph, in response to a stronger influence of bone morphology. Therefore, further experimentations are necessary to assess differences in variance distribution depending on the characteristics of the bone. The future enlargement of the experimental sample would help characterize cut marks on different bones and identify possible patterns related to bone morphology.

Conclusions
In this paper, we present a low-cost, portable and accurate technology for the analysis of marks on bone surfaces. The use of structured light scanners in this research area has been statistically validated by comparison with the results obtained using photogrammetric methods. Structured light scanners present remarkable advantages over photogrammetric methods, especially the time needed for data collection and processing. Particularly, the time required for the reconstruction of cut marks using SLS-2 is lower (only a few seconds) than the time required with M-PG (around 25 min). Regarding the quality of results (precision and resolution) obtained, the incorporation of a specific system of lenses to the low-cost David SLS-2 system allows us to reach a resolution of 0.016 mm and a precision equivalent to the M-PG (±0.017 mm [51]).
As shown in this paper, results obtained with both techniques are statistically identical, allowing the combination of SLS and M-PG for the study of certain taphonomic traces. Beyond that, structured light scanners do not only permit the application of the technique to the study of marks, but they also provide sufficient resolution to assess differences among very similar marks. Three-dimensional reconstructions generated with an improved DAVID structured light scanner have sufficient resolution so as to compare cut marks produced with the same raw material (stainless steel knife) on cylindrical (long bones) and flat bones (axial element). The resolution obtained with SLS-2 technique opens a new range of possibilities for taphonomic studies, making it possible to observe and analyze morphological differences of microscopic marks in much greater detail ( Figure 10). The implementation of SLS to archaeological research would, thus, enhance statistical studies and enable the proposal of a broader range of taphonomic questions, including hypotheses on subtle morphological features.
Here, we assessed the performance of SLS-2 in morphological studies on bone surface modifications. However, the method could be extended to other types of surfaces to study the shape and size features of different elements, such as engravings on portable or parietal art.
Although the results obtained in this study are very satisfactory and demonstrate the validity of this low-cost method for the study of conspicuous and well-defined marks (e.g., cut marks), the SLS-2 technique still shows the same difficulties when carrying out studies of more loosely defined marks (e.g., trampling marks), as it was the case with M-PG techniques. It is true that the SLS-2 technique achieves a better resolution than M-PG ( Figure 10), but still lacks sufficient resolution to register the subtleties of fine microscopic details.
Researchers now have a low-cost method at their disposal that overcomes the limitations encountered when using microscopes (e.g., restricted access due to high costs), providing fast and accurate results. Accordingly, analytical costs and time might reduce, facilitating an increase in processed sample sizes.