Estimating the kinematics and structure of a rigid object from a sequence of monocular images
IEEE Transactions on Pattern Analysis and Machine Intelligence
The problem considered here involves the use of a sequence of noisy monocular images of a three-dimensional (3-D) moving object to estimate both its structure and kinematics. The object is assumed to be rigid, and its motion is assumed to be "smooth." A set of object match points is assumed to be available, consisting of fixed features on the object, the image plane coordinates of which have been extracted from successive images in the sequence. Structure is defined as the 3-D positions of
... D positions of these object feature points, relative to each other. Rotational motion occurs about the origin of an object-centered coordinate system, while translational motion is that of the origin of this coordinate system. In previous work [SI-  we have developed a model based approach for motion/structure estimation using a long sequence of monocular images. This approach provides a great deal of flexibility, by allowing the use of arbitrarily many image frames and feature points, and each model can easily be modified or extended for different problems. Our earlier work involved assumptions about object structure and/or motion, were primarily tested on simulated imagery, and did not address the issue of uniqueness of the model parameters. In this paper, which is a continuation of the research started in [q, results of an experiment with real imagery are presented, involving estimation of 28 unknown translational, rotational, and structural parameters, based on 12 images with 7 feature points. Uniqueness results are summarized for the case of purely translational motion. A test based on a singular value decomposition is described that determines whether or not noise-free data from an image sequence uniquely determines the elements of any given parameter vector, and empirical support of this test is given.