Mobile robot mosaic imaging of vehicle undercarriages using catadioptric vision

Abstract

This paper describes a technique for performing imaging of car undercarriages using a mobile robot fitted with a catadioptric camera. The mobile robot follows a visual based pre-computed path which trisects the undercarriage whilst the catadioptric camera sensor records a wide angle view of the undercarriage as the robot traverses the undercarriage forwards and backwards. Odometry data from the mobile robot is used to dynamically generate a region of interest, reducing the search space for image registration where the images are stitched together to form a complete undercarriage mosaic, reducing computation time by 18%. In contrast to a regular camera which would require 7 passes, the proposed catadioptric imaging system can image the majority of the undercarriage in 2, reducing image acquisition time by 70%. The technique is verified through mosaicing algorithms applied to experimental data and shows promise for undercarriage explosive detection robots.