Evaluation of vision systems for teleoperated land vehicles

A pilot study that compares vision systems that might be used to control a teleoperated land vehicle is described. The study compares three forward-looking vision systems, both for actual remote driving and for noninteractive video simulation. Remote driving has the advantage of realism but is subject to variability in driving strategies and can be hazardous to equipment. Video simulation provides a more controlled environment in which to compare vision-system parameters, but at the expense of some realism. Results demonstrate that relative differences in performance among the visual systems are generally consistent in the two test modes. A detection-range metric was found to be sensitive enough to demonstrate performance differences viewing large obstacles using black-and-white and color vision systems. Consequently, the author plans experimentation, aimed at optimizing vision-system parameters, that will rely to a greater extent on the more cost-effective, video-simulation approach.<>