Fast High Resolution Stereo Matching for Tunnel Survey
Gerhard Paar (Joanneum Research , Austria)
Three-dimensional reconstruction of highly textured surfaces on unvegetated
(rock-like) terrain is of major interest for stereo vision based mapping
applications. We describe a system for automatic high-resolution measurement
of such scenes,
the application to tunnel surfaces is emphasized. The system is based on two
frame CCD cameras, which are tightly attached to each other ensuring
constant relative orientation. The image acquisition hardware
is portable to keep data acquisition as short as possible.
Exterior orientation is provided by automatic theodolite measurements of
prisms tightly attached to the camera system.
The robust matching method HFVM (Hierarchical Feature Vector Matching) is
using local texture features that,
for every pixel, build up a feature vector. Similarity is based on
the comparison of these feature vectors in both stereo images. It allows a
flexible optimization of the feature sets, genetic algorithms proved to be a
valuable tool in this respect.
Dense disparities together with the Locus Method as reconstruction
algorithm render an accurate grid of 3D points on arbitrarily shaped
surfaces. The results of several stereo reconstructions are merged.
Projection onto the global cavity shape allows quick visualization and easy
evaluation of volumes, as well as thematic mapping with respect to the
desired surface geometry in construction processes. We report on accuracy
and computational complexity, and emphasize on some detail aspects in the
image processing chain.
The integrated system has been developed by JOANNEUM RESEARCH, Graz, in the
frame of an R&D contract for DIBIT Messtechnik GmbH, Mils, Austria.