Gathering Intel

A few months ago I posted about the Kinect device from Microsoft. Kinect, since its launch in November, has sold over 2.5 million units in less than 30 days. This makes it the fastest adopted consumer electronic device ever, surpassing the recent title holder the iPad. Now what is really impressing me is how the hacking community has gone wild with the device and I recommend visiting KinectHacks.net to see what I mean.

The technology behind the Kinect intrigues me. I know very little about 3D depth sensors and RGB cameras, so to learn more I decided to visit the Intel Lab Seattle that has been doing research in this area long before Kinect was released.

Located by the University of Washington, Intel Lab Seattle is a Center of Excellence in Context-Aware Sensor-Driven Systems. Interdisciplinary projects and an emphasis on prototyping novel systems and user evaluation drive its agenda. Led by Director Dieter Fox, the research staff are recognized experts across a range of topics from sensing and wireless systems, through robotics and computer vision, to human-computer interaction. I got the opportunity to meet with Dieter and learn more first hand about the lab and the research they are conducting there.

As stated in the video, PrimeSense is the maker of the depth cameras that the Intel team currently works with and the hardware that is in Kinect.

A nice writeup on how the PrimeSense sensor works can be found at the MirrorImage blog. Depth cameras provide a stream of color images and depth per pixel. The sensors can be used to generate 3D maps of indoor environments. The Intel Lab team has used this to create some 3D maps built in their lab.

What you see is not the raw data collected by the camera, but a walk through the model generated by their mapping technique. As can be seen in the video, the maps are not complete. They were generated by simply carrying a depth camera through the lab in one pass and aligning the data into a globally consistent model using statistical estimation techniques.

3D indoor models could be used to automatically generate architectural drawings, allow virtual fly-throughs for real estate, or remodeling and furniture shopping by inserting 3D furniture models into the map. Being someone that works with the digital built environment, this could come in very handy.

Many thanks to Dieter and the staff at the Intel Lab Seattle for the lesson in depth sensors. With the work this group is doing in this domain and other, I am sure I will find myself back in Seattle to gather more intel.