Catadioptric vehicle undercarriage imaging with visual path planning

This paper describes a mobile robot system used to image vehicle undercarriages. The approach uses a visual image processing technique to identify and paramaterise tyres as fiduciary markers from which pose estimation can be performed. After pose estimation is performed a path planning algorithm is used to navigate the mobile robot underneath the vehicle whilst imaging the vehicle undercarriage using a wide angle catadioptric imaging system. The images of the undercarriage are then re-projected to a flat plane and template matching assisted by odometry is used to generate a mosaic view of the vehicle undercarriage. The mobile robot imaging approach is evaluated visually, with images including imitation improvised explosive devices (IED´s) and is shown to be significantly more versatile and safer for operators compared to alternate methods of undercarriage inspection.