Title :
Stereo analysis using individual evolution strategy
Author_Institution :
Ecole Nat. Superieure de Tech. Avancees, Paris, France
Abstract :
Presents an individual evolutionary strategy devised for image analysis applications. The example problem chosen is obstacle detection using a pair of cameras. The algorithm evolves a population of three-dimensional points (`flies´) in the cameras fields of view, using a low complexity fitness function giving highest values to flies likely to be on the surfaces of 3-D obstacles. The algorithm uses classical sharing, mutation and crossover operators. The final result is a fraction of the population rather than a single individual. Some test results are presented and potential extensions to real-time image sequence processing and mobile robotics are discussed
Keywords :
genetic algorithms; image motion analysis; image sequences; object detection; stereo image processing; crossover; image analysis; individual evolution strategy; low complexity fitness function; mobile robotics; mutation; obstacle detection; real-time image sequence processing; sharing; stereo analysis; Application software; Cameras; Computer vision; Genetic mutations; Image processing; Image sequence analysis; Layout; Pixel; Stereo vision; Testing;
Conference_Titel :
Pattern Recognition, 2000. Proceedings. 15th International Conference on
Conference_Location :
Barcelona
Print_ISBN :
0-7695-0750-6
DOI :
10.1109/ICPR.2000.905580