Ruge’s summary of “KinectFusion and “Going Out”
The KinectFusion paper gave a detailed description of a 3d mapping technology based on the Kinect hardware. It provided use cases, explanations of existing hardware, and how the product would be used in an operational sense. Beyond that it captured the the mathematical and computer programming fundamentals that are pivotal to both its operation and usability. The system used the Kinect hardware in combination with realtime video and image based 3d mapping technologies. By leveraging the strengths and problems with each individual solution the KinectFusion architecture can achieve unprecedented levels of detail. The only systems that were capable before demanded smaller objects and/or large amounts of infrastructure to develop the 3d environments. The paper discuses two large features that could be useful in many practical applications. The first is the handheld 3d mapping usability. The system could be used to rapidly create detailed models of rooms or objects without calibration of electronics or even prior information about the environment. When combined with Augmented Reality techniques this could be both incredibly useful and interesting. Possibilities such as the ability to “mount” information and virtual 3d objects onto existing furniture or desktops. In games virtual characters could interact with the room just as the player would. By creating maps of existing environments and being able to input them into virtual computer generated systems, the immersion features of VR could be combined with the feasibility and usefulness of actual environments. Users of full virtual reality systems could safely interact with the world around them.
“Going Out” discussed a system developed at Cambridge, that utilized a tablet and a camera to provide an image based orientation and position experience. By using known models, and video edge detection techniques it is possible to achieve a detailed summary of the users position and orientation in the environment. The system was built with an older Sony tablet with graphics and processing power far less than is available in similar platforms today. I feel there are 2 large flaws with the system. The system utilizes known models and placement information about the environment in use. The example in the paper utilized a modeled courtyard. To make the system operate, the needed 3d models of the structures, as well as knowledge about preexisting surface features, and surveys to determine the orientation with respect to the rest of the world. Given the software used, and the information provided, this technology could be used to develop easily real-time image tracking, however the practical uses of such technology are limited. The system would not satisfy the environment agnostic nature that most use cases desire. However many of the edge detection techniques in combination with modern hardware and more available simple models (Google earth, Bing, apple maps, etc.) could be used to augment other AR applications and tools to gain both speed, reliability, accuracy, and resilience. As this paper and the KinectFusion Paper, discussed above, mentioned, the best systems involve a combination and synergy among many different technologies.
Comments are closed.