137 - 
production. The data quality and speed acquisition 
allow to propose restorers and historians, data of high 
accuracy with a production quality-price ratio five to 
six times lower than traditional photogrammetric 
techniques. 
3.3 Comparison and analysis of Gold quivers: 
The results were far more better than expected. 
Locally, as showed in figure 3, 67 percent of the points 
are within a millimetre, and the other are not differing 
form more than 2 mm. The observed differences are 
due to the global longitudinal curvature of the piece as 
it was explained before. The results found again in the 
other parts of the quiver tends to show great similarities 
between the two models. 
3.3.1 Qualitative comparison 
Difficult is the task to compare two objects old of 
several century’s. Indeed, the question of findings good 
criterions was hard work. 
Figure 3 : difference map in millimetre of a matrice in 
the two quiver models 
This illustration shows that the main parts of the 
matrice is similar (within a gap of one millimeter in 
between ). That is to say that the observed errors are of 
the same order than error measurements and data 
acquisition quality. 
3.3.2 Quantitative comparison 
Statistic indicators are useful for estimation and 
comparison of the two models. Connected with error 
map computed by substracting of one model to another 
to estimate deformations, these statistics illustrate to 
what extents we can compare them with tools as 
minimum and maximum "heights", mean and standard 
deviation. However, further statistical studies will be 
lead with available software 
Table 1: comparisons statistics 
Number of 
points 
Percentage of 
compared points 
Error range 
in 
millimetre 
270620 
22.24 % 
0-0.3 
124885 
10.27% 
0.3 - 0.6 
102033 
8.39 % 
0.6-0.9 
88198 
7.25 % 
0.9- 1.2 
59754 
4.91 % 
1.2- 1.5 
73211 
6.02 % 
1.5 - 1.8 
27384 
2.25 % 
1.8 - 2.1 
3034 
0.25 % 
2.1 -2.3 
733 
0.06 % 
2.3 -2.5 
In this matrice, a total of 750 000 points were 
compared, an average error of 0,85 millimetre was 
observed (0.66 mm standard deviation) with a local 
maximum error of 2.5 mm and a minimum error of 0 
mm. To conclude with these figures, one can say that 
the models are quite the same since 67 percent of the 
points are in a range of one millimetre, and 87 percent 
below 1.5 millimetre; and despite of the global 
curvature error, locally we noticed only four percent of 
points with an error around 2.5 millimetre. 
3.3.3 Profiles 
Computed profiles give relevant information 
concerning models quality. Therefore, we can estimate 
the sensibility to global or local variation of errors if 
they occurred. Profile graphics shows the curved 
effects noticed in the first step of comparison. (They 
are not showed in the present paper) 
4. OUTCOMES 
From the beginning, the main operational purpose of 
the proposed method was to participate actively to the 
conservator’s effort dedicated to historical 
interpretation. 
As the project as being on going for several months, 
the outcomes are still to be raised. However several 
leads are coming out from this study. The first one is to 
notice how well conservated these two object are 
compared to each other. The second is to realize how 
precise there similarities can be. This tends to conclude 
to a same made process or same type mould may have 
been employed. Nevertheless, it is up to the 
conservator and archaeologist to finally decide the 
adequacy of the applied method. Hopefully, these 
results will give them significant material to develop 
explanations and draw final conclusions.