SBIR Project Overviews
Current Phase II SBIR (DARPA): micro UAV Collision Avoidance
A small real time processor is being developed that will convert the images from a single
conventional camera moving relative to a scene into range images of the scene (Patent Pending). The intensity of each
point in the range image is proportional to the range from the camera to the corresponding point in
the scene. Range images will warn of looming collisions with fixed or moving objects anywhere within
the FOV and provide the 3-D scene structure as needed for autonomous operation.
||For collision avoidance and autonomous operation
||For safe operation in the National Air Space
||To warn of looming collisions on the highway and of low speed looming impacts when parking
||To enable autonomous robots to move safely and effectively in complex environments in which people may also be moving. (E.g. A robot vehicle moving though the corridors of a hospital to deliver and collect supplies,)
||In this case the camera is fixed and the processor will locate and report the projected time to impact of any objects moving towards the camera. This information could substantially improve the efficiency (probability of detection)/(probability of false alarm) of automatic pre screening of surveillance video.
Single Camera Stereo Vision
The processor can also be applied to generate range images from
the output of a single camera at the end of a snake-like (serpentine) robot (used to inspect
confined spaces), or at the end of an endoscope (used by surgeons to see and operate within the
body) when these inspection devices are moving. In this case the range image will be used to convert
the images from the single-camera in these systems into stereo pairs, and the operator will view
these through a binocular eyepiece. The user's visual capability is thereby enhanced with a 3-D
stereo view of the scene. (Patent Pending). The range images generated when the camera in these
inspection devices is moving will be used to create stereo pairs. These range images will also be
stored and used to create stereo pairs when the camera stops moving, and range images are
consequently not then generated.