Utilization of a Stripe Pattern for Dynamic Scene Analysis

This paper describes a new idea to project a stripe pattern onto a time-varying scene to find moving objects and acquire scene features in the consecutive frames for estimating 3-D motion parameters. At first, a simple temporal difference method detects objects moving against a complex background. A 2(1/2)D representation of moving objects at each frame is then obtained by estimating surface normals from the slopes and intervals of stripes in the image. The 2(1/2)D image is further divided into planar or singly curved surfaces by examining the distribution of the surface normals in the gradient space. Then, the rotational motion parameters of the objects are estimated from changes in the geometry of these surfaces between frames. Determining translational ones is also discussed.