Abstract:

A new 6-DoF state ego-motion estimation algorithm was presented based on the lunar rover stereovision image sequence to realize the autonomous navigation of the rover. A scale invariant feature transform SIFT feature tracking and matching method was proposed and studied. 2 new matching and restraint conditions were introduced to establish the accurate environment feature correspondence before and after the motion of the rover. A robust estimation technique was proposed to solve the motion parameters of the rover based on the obtained feature correspondence. A method combining random sampling consensus and unit quaternions was used to reject the matching error and get the preliminary estimation of the rover body motion parameters, and from this preliminary estimation, a Levenberg-Marquardt non-linear optimization algorithm was used to find the accurate solution of the motion parameters. The effectiveness of the proposed motion parameter estimation algorithm was verified by the simulated field test.

Key words: engineering of communications and transportation system, lunar rover, feature tracking and matching, ego-motion estimation