Many companies are working to create inside-out marker-less tracking for virtual reality headsets. Inside-out marker-less tracking can be found on consumer augmented reality devices, but currently there is no system available to researchers, developers, or consumers that provides this feature without custom hardware and software. Our research provides a development environment for testing content that would take advantage of this feature before consumer level inside-out marker-less virtual reality systems hit the market. The solution utilizes current commercial off-the-shelf hardware systems, and by networking them together, allows the user to move through a captured environment without needing tracking towers. Ultimately, it provides researchers and developers the opportunity to test content that both takes advantage of the free movement through rooms provided by inside-out tracking and is yet constrained by real-world boundaries such as walls within the virtual environment.

This research concentrates on defense-type systems for operational awareness using virtual reality solutions. Many computer vision solutions in defense environments concentrate on removing user decision from systems for automation, but in doing so remove trained military professionals from decision making. Instead this research looks to reintegrate the military professional by providing a real-time point cloud visualization of a previously unmapped environment, in virtual reality, for remote operation of an unmanned vehicle. Due to an interest in a low-cost and modular system, that could be used on a variety of unmanned vehicles, including those used in underwater environments, the solution has moved away from LIDAR and Kinnect's laser-based systems and towards stereo cameras, particularly webcams. While looking to provide this approach, several other areas of interest become apparent, including the need to have a reliable stereo reconstruction under variant lighting conditions, particularly in the case of backlight in natural environments. Consequently, a variety of keypoint detection algorithms and point cloud reconstruction methods are being explored in conjunction with their usefulness within real-time constraints and with low-resolution images. Also, it will require evaluating different point cloud render techniques such as splatting, and storing the data in reliable data structures for effective LODs for real-time rendering. Then, finally, it will be important to explore research on effective user operated virtual environments and simulations to determine best practices for navigation and interface in the front-facing part of the system that allows the user operation.