6A-2-2 
.- ?.- r > r •’■ •-' rr>r- vr/f'- ' v 
The main tool for 3D data acquisition was the mobile mapping 
van On-Sight developed by Transmap Corp. The 3D co-ordinates 
of objects were measured with help of recently developed 
methods for automated analysis of stereo georeferenced images 
(see Section 3). 
2 ACQUISITION OF THE DATA 
On-Sight is MMS that integrates a Global Positioning System 
(GPS) receiver, with an inertial navigation system (INS) and up to 
four digital cameras. 
Figure 1. On-Sight mobile mapping van from Transmap Corp. 
GPS and INS determine the position and attitude of the van at any 
time, while the digital cameras capture high-resolution colour 
images showing road and road surrounding on the left side and in 
front of the van. In the post-process colour images are 
georeferenced. With the help of stereo photogrammetry and 
interior, and exterior orientation parameters one can calculate 3D 
co-ordinates of any object visible in stereo images. 
3 EXTRACTION OF 3D OBJECTS 
The 3D objects were extracted and measured from georeferenced 
images using automated photogrammetrical methods 
(Gajdamowicz, 1998, Gajdamowicz and Ohman, 1998, 
Gajdamowicz, 1999). 
75?< '* 
. - 4 
MMS 
trajectory 
’yr- 
\ "«-V r —* 
Actual position of MMS 
X .. § fxm 
the recent position of the system. By selecting any point on 
trajectory curve the user can access a corresponding image (Fig 
3). 
jHfflTwirm—— 
■i .101-1 
0IB £dfc Doted 
ftSndow fclafp 
1 « 1 » 1 
a 1 u. i «»b. i «— 1’ ^ 1 *- 1 
♦ 
D#»ert tnc-owa 
C 30 Port mmjctner* 
(? 3DV«aoi fwmww 
Figure 3. Georeferenced image corresponding to the position of 
MMS 
The 3D co-ordinates of any feature in the image are typically 
calculated in a stepwise process. In the first step the homologues 
image co-ordinates of the desired feature have to be measured. 
Typically, corresponding features in left and right image are 
measured manually by a human operator (Transmap CorpGeofit, 
1996). This approach introduces automation. The operator 
initialises the measurement in one image. Then the algorithm that 
employs epipolar geometry and cross-correlation (CC) in colour 
space finds a correspondence. The set of transformation enables 
the image co-ordinates to be expressed in global co-ordinate 
space (Fig 4). 
Sie Edn tfew ßetao iÿndow 0e»> 
1 « 1 » I g 1 »>» 
Sf-ldT»l 
*1 V 
■ ■ 
Figure 2. Path of the MMS visualised on a digital map. 
In the initial step the MMS data is visualised on a digital map (Fig 
2). Such visual presentation enables easy access to any stereo 
image par. The black curve in Figure 2 indicates the trajectory of 
the MMS, where the light dot (marked wit circle) corresponds to 
Figure 4. Results of automated measurements. 
The colour information is used not only to find a correspondence, 
but also to automatically segment objects like the road centreline, 
road width, or road signs. The segmentation algorithm is initiated 
by thresholding. Next, morphological filtering and Connected 
Component Analysis are used to find the regions of interest 
(ROI). Further analysis ofROIs consists of edge detection and 
edge linking with Hough transformation. In the last step the image 
co-ordinates of the corresponding object are automatically 
extracted and transformed to the global co-ordinates frame.