English/Japanese

Reliable Stereo Vision

Stereo vision is a fundamental method for measuring objects' 3D positions (range from cameras) by observing them from two or more cameras placed on different positions. This method enables robots to acquire 3D depth information, which is a fundamental information for recoginzing their environment. The most difficult problem in stereo vision is to find the corresponding areas between two (or more) images.

We have been proposed several reliable matching methods and its applications as follows,

Self-calibration of stereo cameras using correspondences of edges
An efficient matching method for detecting dangerous objects with trinocular vision
Stereo matching with the geometry constraint of the occlusion
Detecting moving object and measuring the range from camera by time series stereo images
Uncertainty/Ambiguity Modeling of Stereo Vision

researchers: To Jun Miura's home page; To Osafumi Nakayama's home page; To Yukihiro Ayaki's home page; To Makoto Miyata's home page

Go back to Shirai lab. homepage