Title :
A generic bridge finder [object recognition]
Author :
Vergnet, R.L. ; Saint-Marc, P. ; Jezouin, J.L.
Author_Institution :
Signal & Image Process. Lab., MATRA-LTIS, St.-Quentin-en-Yvelines, France
Abstract :
The authors show that for some objects whose geometry is not precisely known, it is possible to design a robust recognition scheme based on discriminating features through an exhaustive tree-search. They deal with the problem of finding one instantiation of an object in a single image that is the best one according to a predefined quality-score. It is also assumed that the object surely lies within the field of the sensor, and that the view point is known, up to uncertainties involving distortions in the image that cannot be neglected. The example of bridges is put forward and a system is developed for detecting bridges of a priori unknown shape with an airborne camera. The results shown are representative of various sensing conditions
Keywords :
computational geometry; computer vision; computerised pattern recognition; computerised picture processing; search problems; solid modelling; trees (mathematics); bridges; discriminating features; exhaustive tree-search; generic bridge finder; object recognition; Bridges; Feature extraction; Geometry; Image recognition; Image sensors; Laboratories; Object recognition; Robustness; Signal design; Solid modeling;
Conference_Titel :
Automated CAD-Based Vision, 1991., Workshop on Directions in
Conference_Location :
Maui, HI
Print_ISBN :
0-8186-2147-8
DOI :
10.1109/CADVIS.1991.148774