Applications for realtime stereo exist anywhere timely range data is useful.  After several years of making SVS available to the research community, there are many applications at various stages of progress, from completed systems such as MIT's robotic wheelchair, to advanced research on indoor environment modeling at SRI International.  This page has some sample applications and links to further information.  If you have an application you would like to share, please let us know!

Robotics

The STH-V2 is an ideal sensor for mobile robots, because of its small size, low power, and ability to operate with just a single framegrabber.  The new MEGA-D sensor is also becoming popular on robotics projects.

ActivMedia Robotics
The STH-V3 and MEGA-D are available on mobile robots from ActivMedia, fully integrated on their P2 and PeopleBot robots.


SRI International AI Center's SenseNet Project

The SenseNet Project is part of DARPA's Tactical Mobile Robotics initiative. SRI International's Artificial Intelligence Center is developing a team of mobile robots to cooperatively map indoor areas. The STH-V1 is mounted on the top front of the robot. It's being used to avoid obstacles, to map and model the environment, and to detect and track people.

There are a lot of other sensors on the robot, too -- check out the SenseNet web pages.

 

MIT AI Center: the Wheelesley Project

Holly Yanco's robotics thesis is about assistive wheelchairs with semi-autonomy.     

The picture shows Wheelesley, a robot wheelchair project she started at Wellesley College, and then carried over as her thesis work at MIT's Artificial Intelligence Center.  Wheelesley uses an STH-V1 stereo head, coupled to an onboard computer running SVS, to detect obstacles in realtime.

 

CMU Robotics Institute: SunSync Project and the Hyperion Robot

Under a NASA grant, CMU is prototyping a sun-synchronous vehicle for planetary exploration.  Name Hyperion, this robot will continuously travel at an appropriate latitude to stay in the sunlight, supplying continous power.

The SVS was chose to operate a stereo system on the robot, because of its outstanding processing efficiency and stereo performance.  A tech report describing the use of SVS for terrain mapping can be found here.

 

 

People Tracking

One of the most natural areas to use stereo is in detecting and tracking people and their parts. Typical applications include head tracking and pose estimation for video conferencing, limb tracking for gesture recognition, and body tracking to analyze gate and actions.

 

SRI International AI Center's Extra Sets of Eyes Project


The Extra Sets of Eyes Project was part of DARPA's VSAM effort. Under this project, SRI developed people detection and tracking algorithms that work robustly in crowded environments.

University of Maryland Computational Vision Center
This is another people-tracking project that was part of DARPA's VSAM effort. The W4S system reliably tracks people and identifies the head, torso, and limbs. W4S uses stereo from the STH-V2 head to augment its basic motion detection algorithm.

University of Illinois Beckman Institute and ECE Department / Microsoft Research
This joint research project tracks articulated 3D objects (e.g., humans) using the STH-V2 and SRI's Small Vision System to acquire realtime range data. Check out the MPEG movies of articulated arm/head/torso tracking!

 

Space Applications

As spacecraft become more automated, they need expanded sensor capabilities to perform their tasks. Stereo analysis is a natural sensor, because potentially it can provide detailed range data, at high data rates, with very little power.


NASA AMES: Personal Satellite Assistant


The PSA is envisioned as a device about the size of a softball, with advanced sensors and actuators that enable it to maneuver in the zero gravity of a space station or other space structure. Its ultimate task is to detect and fix problems with space structures.

The STH-V2 and SRI's Small Vision System were used on the first working prototype, to perform station-keeping tasks.  The next version is being designed with the MEGA-D sensor for increased performance.