(STEMS 
| Ricerca in 
eration: one 
| cameras as 
and a small 
updating the 
e these tools 
=] Economic 
own that the 
>, a Piaggio 
the means of 
(em for Land 
le to supply 
the aid from 
ion Azimuth 
on Distance 
st-processing 
nsor health, 
mance and 
ince. Sensor 
; a Kalman 
2 
  
of Trieste 
  
The LITTON LN-200 fiber optic gyro IMU is the core of the 
system. It is made up by three accelerometers and three fiber 
optic gyros, and is able to supply 0.01? Pitch and Roll, and 
0.04? true Heading accuracy in real time; 0.005? Pitch and Roll, 
and 0.02° true Heading accuracy after post-processing. 
Primary GPS double frequency data are necessary to fix initial 
coordinates, and to bound IMU temporal drift, tested to be next 
to 5?/hour. The whole system positioning accuracy depends on 
GPS mode: metric level using a good DGPS input, decimetric 
level using RTK. The system uses both Primary and Secondary 
GPS single frequency data for GAMS (GPS Azimuth 
Measurement Subsystem), which permits to improve IMU 
heading performances. 
DMI aids IMU on GPS outages, making the trajectory a nearly 
continuous line, as position data can be output at 200 Hz rate. 
The van-mounted Applanix POS/LV is coupled with two 
imaging subsystems, each composed by a Matrox 4Sight 
computer and a Basler CCD B/W camera. 
Beyond that, we are testing auxiliary instrumentation: lateral 
Laser Distance Meters (LDM) have been tested for 
implementation in improving slope data, and a Laser Scanner 
has recently been acquired. 
2.2 The Standard Working Process 
  
e Data 
Download 
e Data 
Processing 
  
e GIS 
Generation 
  
  
  
  
  
  
  
  
  
Figure 2. The Standard Process 
For road axis determination, the MMS surveyed trajectory is 
used, applying the axis offset detected by photogrammetric 
measurements. 
Positioning data acquisition rate is set before the mission start, 
and it can go from 1 to 200 Hz, which is the IMU maximum 
sampling rate. 
Road slopes are given by the MMS attitude data, measured by 
the IMU components. Attitude data noise, due to the vehicle 
elastic suspensions, is less than measurement maximum error. 
183 
Road width is obtained by measuring the distance between the 
road sides, on the rectified monoscopic images coming from 
one of the front mounted cameras; maximum error in this 
determination is on decimetre level. 
The MMS runs each surveyed road back and forward. The 
reason for this is: 
e Some road axis topology constraints exist: there are 
bifurcation, roundabouts, lane separations and so on, and a 
one-way survey is not enough; 
e The road slope (transversal gradient) may vary for each 
lane; 
e A complete photo coverage is needed; 
e In this way, at least for some road segments, it is possible 
to achieve two independents axis estimates, with enhanced 
accuracy assessment. 
At the end of the survey campaign, the obtained data for each 
surveyed road is: 
e Two complete trajectographic solutions in opposite 
directions, as sequences of records containing, among 
others, time and distance tags, the WGS84 ellipsoidal 
coordinates of the MMS reference point, the Euler angles 
(heading, pitch and roll angles of the body frame in the 
navigation frame); 
* A number of image files, taken from each camera and 
time-tagged. 
At the end of the photogrammetric measurement process, each 
trajectographic solution is provided with its offset from road 
axis with decimetric accuracy in the vehicle body frame. 
The global coordinates and Euler angles, provided by the 
trajectographic subsystem, are then used to perform the 
transformations from body frame to navigation frame and from 
navigation frame to geocentric WGS84 frame. Finally, a 
coordinate transformation from geographic WGS84 to plane 
Gauss-Boaga is performed using IGM (Istituto Geografico 
Militare: the Italian national mapping agency) parameters. 
If possible, all the road axis measurements are taken on the road 
centerline sign. 
The Road Cadastre introduces the GDF (Geographic Data Files) 
European standard in the representation of the road network and 
road related informations. 
In the GDF model, a road is typically a sequence of 
“Junctions”, each two of these are connected by one or more 
"Road Element". 
First of all, the “Junctions” are identified by intersection of road 
axis and, if necessary, the intersection points are added to the 
point sequence. Subsequently, a road axis segmentation is done 
using these points. For each point the curvimetric distance is 
recalculated with respect to the segment startpoint. Each 
segment identifies the geometry needed to describe a “Road 
Element” feature in the GDF dataset. 
The relevant process here is the “Junctions” identification. A 
great accuracy is needed, because often the “Junctions” are 
common to two or more roads, which must not necessarily be 
surveyed at the same time using the same technique. As a final 
task, the segmented attributes set, obtained from the 
photogrammetric measurement process and provided with the 
curvimetric distances, is projected on the road graph. Each 
attribute is provided with a Road Element code, the class, value 
and direction codes from the Road Cadastre lookup tables, the 
curvimetric distance of the start-point and end-point and the 
references to the start-point and end-point in the point sequence. 
All the intermediate processes described here are performed in a 
relational DBMS, whose structure is based upon the table