Visual terrain matching for a Mars rover

A method of matching unequally spaced height maps is described. This method would be useful in a Mars rover which tries to refine the estimate of its position by matching the data from its stereo vision system or laser rangefinder to data obtained from a camera orbiting Mars. The refined position can be used to merge the two sets of data with proper registration. The method is designed to make full use of the information contained in the data, including accuracy estimates in the form of covariance matrices and reliability estimates in the form of probabilities. Means of extracting this information from stereo vision are discussed. The terrain matching process uses a coarse-to-fine strategy, and it includes automatic editing to remove points with excessive disagreement. An example using real data from an experimental vehicle is presented.<
>