The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B3h. Beijing 2008 
9. CONCLUSION 
MMS are efficient and cost effective tools for building and 
updating GIS databases. However, manual measurements of 
GIS features in MMS are still manpower demanding procedure. 
We have initiated a wide scope project for automated GIS 
features extraction, to decrease and possibly eliminate the most 
of the human work in the post-processing. In this paper, we 
present ARVEE, a robust automatic functional road geometry 
extraction system for MMS. There are three innovations in 
ARVEE. First, instead of only extracting the central lane line or 
the nearby lane line pair, our system extracts all the visible lane 
lines in the georeferenced image sequences. Second, the lane 
line attributes are recognized, so the output is a functional 
description of the road geometry. Third, the output is the high 
accurate absolute-georeferenced models which are compatible 
to the GIS database. Test over massive real mobile mapping 
demonstrate that ARVEE are ready for the real world 
applications. 
10. REFERENCE 
El-Sheimy, N. and K.P., S., 1999. Navigating Urban Areas by 
VISAT - A Mobile Mapping System Integrating 
GPS/INS/Digital Cameras for GIS Applications. Navigation, 
Journal of the USA Institute of Navigation Journal, 45(No. 4), 
pp. 275-286. 
Ishikawa, S., Kuwamoto, H., et al., 1988. Visual navigation of 
an autonomous vehicle using white line recognition. IEEE 
Transactions on Pattern Analysis and Machine Intelligence, 
10(5), pp. 743-749. 
Kenue, S. K., 1991, LANELOK. An algorithm for extending 
the lane sensing operating range to 100 feet. Proceedings of 
SPIE - The International Society for Optical Engineering, 
Boston, MA, USA, 1388, pp. 222-233. 
Jochem, T. M., Pomerleau, D. A., et ah, 1993, "MANIAC": A 
Next Generation Neurally Based Autonomous Road Follower. 
Image Understanding Workshop, pp. 473—479. 
Figure 13: ARVEE result overlapped on map 
In order to evaluate the misdetection rate and false detection 
rate of ARVEE, over 25 kilometres (more than 100,000 meters 
lane lines) survey data are first automatically processed by 
ARVEE, and then corrected manually. The manually corrected 
result (MCR) is viewed as a reference, and compared with the 
automatic result (AR). All the lane lines appear in MCR but not 
in AR are count as misdetection (MD), and all the lane lines 
appear in AR but not in MCR are viewed as false detection. The 
misdetection and false detection rate is calculate based on the 
length of the lane lines. The statistics of the test are shown in 
Table 1. 
Length (meter) 
Percentage 
Total Lane Line 
102,732 
100% 
False detection 
9,889 
9.62% 
Miss-detection 
2,510 
2.4% 
Table 1: Performance statistics of ARVEE 
The major causes of the misdetection are worn-out lane lines, 
and heavy occlusion. The major causes of false detection are 
lane-line-similar structures near the road, such as the edge of 
side walks, or the line shapes in the nearby vehicles. Figure 14 
and Figure 15 show examples of misdetection and false 
detection. In Figure 14, there is a worn-out dashed white lane 
line in the right side of the road, and is misdetected. 
Chen, K. H. and Tsai, W. H., 1997. Vision-based autonomous 
land vehicle guidance in outdoor road environments using 
combined line and road following techniques. Journal of 
Robotic Systems, 14(10), pp. 711-728. 
Beauvais, M. and Lakshmanan, S., 2000. CLARK: A 
heterogeneous sensor fusion method for finding lanes and 
obstacles. Image and Vision Computing, 18(5), pp. 397-413. 
Paetzold, F. and Franke, U., 2000. Road recognition in urban 
environment. Image and Vision Computing, 18(5), pp. 377. 
Yim, Y. U. and Oh, S.-Y., 2003. Three-feature based automatic 
lane detection algorithm (TFALDA) for autonomous driving. 
IEEE Transactions on Intelligent Transportation Systems, 4(4), 
pp. 219-225.