Next: An Adaptive Computer Up: Sensor and Processor Previous: Range Measurement Systems

Stereo Vision for the Robot Eye

Authors: [tex2html_wrap4216]M. Ezzati, M.D. Levine

Investigator username: levine

Category: perception

Subcategory: sensor and processor design

Binocular stereo vision is a biologically motivated approach that uses two slightly different views of a scene to extract information about its three-dimensional properties. The first stage in stereopsis is obtaining the vector displacement of the corresponding ``sections'' of the stereo image pair. We approach the problem using cepstral filtering, a disparity calculation algorithm that offers significant computational efficiency. The images from the two eyes are represented in the form of windows motivated by the ocular dominance column representation in the primary visual cortex. The parameters of the algorithm are chosen based on those of human stereo vision. The algorithm will be implemented in real-time and used for depth perception in the Robot Eye. The Robot Eye is a miniature active vision system being developed at CIM for use with autonomous mobile robots.