Frame Registration of In-Vehicle Normal Camera with Omni-Directional Camera for Self-Position Estimation

We propose a method for frame registration between in-vehicle omni-directional and normal cameras aimed at self-position estimation of a vehicle. We assume that the position of a vehicle is estimated by frame registration between the inputted normal camera images and the omni-directional video database with accurate position information. A DTW- based algorithm is used for nonlinear time-series matching because we consider that time-series information contributes to robustness. According to an evaluation experiment, the proposed method showed precise self-position estimation ability.