Zu diesem lizenzpflichtigen Artikel gibt es eine Open Access Version, die kostenlos und ohne Lizenzbeschränkung gelesen werden kann. Die Open Access Version kann inhaltlich von der lizenzpflichtigen Version abweichen.
Preisinformation
Bitte wählen Sie ihr Lieferland und ihre Kundengruppe
Imaging sensors are powerful tools enabling remote control, by tele-operation, of numerous tasks where the operator requires an appreciation of the three-dimensional structure of the viewed scene. Passive video sensors also lend themselves to tasks where covert operation or electromagnetic compatibility is required. A commonly mooted tele-operational task is that of driving a known vehicle through an unknown terrain - or keeping station on a known object moving through an unknown terrain. The computer vision aspects of automating this task are divided into two separate vision functions, which are the subjects of this paper: (1) analysis of image sequences of a general scene to extract its three dimensional (3D) structure without any prior information; and (2) analysis of images of a well defined object, to extract its 3D position and orientation relative to the sensor. For both these functions, the paper provides a brief introduction to possible techniques followed by further descriptions of particular systems, DROID and RAPiD, developed by Roke Manor Research Limited. DROID is a general, feature-based 3D vision system using the structure-from-motion principle. That is, it uses the apparent image-plane movement of localized features viewed by a moving sensor to extract the three-dimensional structure of the scene. RAPiD is a model-based real-time tracker which extracts the position (X,Y,Z) and orientation (roll, pitch, yaw) of a known object from image data. The system operates iteratively, using prediction of object pose (position and orientation) to cue the search for selected edge features in subsequent imagery. This approach results in minimal processing of image pixels, so that the system can be implemented at full video rate using modest hardware.