Jamet - 4 
The restricted number of test sites doesn't allow to derive an absolute average 
productivity: this first go only shows some tendencies. The influence of the operator 
will be utterly discarded: moreover, we haven't any evaluation of the difficulty for 
learning to use this kind of tool. However, on an experiment run by the designer of the 
interface himself, we can assume that we are very near from an optimal ergonomy (for 
the given operator of course!). If the capture time on the semi-automatic prototype 
cannot be confronted with the production line capture time, a self comparison can be 
done with a fully manual mode implemented on the prototype. 
Site 
(plott. mode) 
Image 
scale 
Road netw. 
length 
Overall 
time 
Total 
capt. time 
Automatic 
capt. time 
Manual 
capt. time 
Montpellier 
(manual) 
1:30 000 
49.0 
49:03 
31:56 
• 
31:56 
Montpellier 
(semi-auto) 
1:20 000 
49.0 
47:44 
18:24 
12:32 
5:52 
Valreas 1 
(manual) 
1:20 000 
17.2 
18:22 
12:38 
- 
12:38 
Valreas 1 
(semi-auto) 
1:20 000 
17.2 
17:11 
6:09 
4:41 
1:28 
Valr6as 2 
(manual) 
1:30 000 
16.1 
12:27 
9:31 
- 
9:31 
ValrSas 2 
(semi-auto) 
1:30 000 
16.1 
9:29 
4:07 
2:43 
1:24 
St.Cast 
(manual) 
1:30 000 
19.7 
26:46 
19:15 
- 
19:15 
St.Cast (semi 
auto) 
1:30 000 
19.7 
16:22 
6:31 
5:11 
1:20 
Table 1: Capture time for the different test sites 
(length in km, times in mn:ss) 
The table 1 gathers synthetic results on the capture times in manual and semi-automatic 
modes. The total time represents the overall duration of the acquisition for the full test 
site, for which we indicate the length of the road network in kilometres. The capture 
times correspond to the phases when the operator is actually digitising roads, either with 
the automatic tool, or in manual mode. The remaining difference consists of the time for 
error correction and topology manipulations, and the time between captures modes 
(reaction delays of the operator, image scrolling without capture, etc...). 
Theses figures lead to two main observations. First, the use of the semi-automatic road 
tracker does decrease significantly the capture time. Secondly, according to the 
landscape, this does or doesn't reflect on the overall time. More precisely, the use of the