Robot Motion Vision by Fixation
Abstract
In numerous current and future applications ranging from autonomous navigation of mobile robots to collision avoidance systems for cars, an imaging system (installed on a moving vehicle) takes 2D images of an environment with the aim of finding the motion of the vehicle (translational and rotational velocities) as well as the structure of the environment (shape). In machine vision, this problem is referred to as the general motion vision problem. This thesis introduces a direct method called fixation for solving this general motion vision problem, arbitrary motion relative to an arbitrary environment. Avoiding feature correspondence and optical flow has been the motivation behind this direct method which uses the spatio-temporal brightness gradients of the images directly. The fixation method results in a linear constraint equation (Fixation Constraint Equation) which explicitly expresses the rotational velocity in terms of the translational velocity. The combination of this constraint equation with the Brightness-Change Constraint Equation (a fundamental equation which relates the motion to the brightness gradients at any image point) solves the general motion vision problem.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1992
- Accession Number
- ADA259456
Entities
People
- M. A. Taalebinezhaad
Organizations
- Massachusetts Institute of Technology