The RealSense™ D400 series depth cameras, as has been previously explained on this blog, are stereo depth cameras. Stereo depth cameras solve for depth on a pixel by pixel basis by using the disparities between two different images of the same scene. A stereo depth camera, by definition, has two imaging sensors that are a known distance apart. By using that known information, the cameras can output depth information by triangulating specific points of interest that are shared between the two imagers. Just as the human brain computes depth by taking information from our eyes and adjusting our impression of how far away objects are based on the different images from each eye, stereo depth cameras operate in the same way.
One of the issues that both human eyes and depth cameras share however is that in order to perform this kind of triangulation, we need enough visual detail visible within our field of view. In darker environments, or places where visual details are minimal, we can have a harder time distinguishing distance or depth. The D400 series cameras solve this problem by adding visual noise to a scene, projected from the camera itself. This noise is invisible to the human eye, consisting of near infrared light projecting onto the scene. In this whitepaper, the parameters and ideal projector are considered, as well as methods to improve performance by tuning and refining these patterns.
Options for additional external projectors are also explored, including when and where using an external projector or additional projector is advisable or desirable to increase range or accuracy, as well as the right configuration of these external projectors in terms of noise pattern, temporal adjustments or variable power levels.
