|
|
|
“A Two-Step Approach to Passive Navigation Using a Monocular Image Sequence”
by S. Chandrashekhar and R. Chellappa
February 1991
This paper discusses the design and application of estimation techniques to the problem of obtaining the kinematics of a moving vehicle and the structure of the external environment, based on a sequence of images taken by a camera attached rigidly to the vehicle. ``Structure'' refers to the 3-D locations of selected points of interest (landmarks) in the environment, expressed in an inertial, or ``world'' coordinate system. The (noisy) image plane trajectories of these feature points are assumed to be available, and the 3-D positions of a small number (typically 3 or 4) of them are assumed to be known. Using simple models for the motion of the vehicle it is possible to utilize effectively the information contained in image sequences of arbitrary length. Batch and recursive estimation techniques are developed to estimate the motion and structure parameters. In the first step, the batch estimation method is applied to obtain an initial estimate and its approximate error covariance, which is then used in the second step to initialize the recursive filter for tracking the parameters of interest. Experimental results on synthetic data and on one of the UMASS ALV image sequences are given.