It is recording a real image, in real time, and then using technical wizardry to add virtual objects to the scene. A  real-world positioning sensor allows the virtual objects to be manipulated via computer to create an image that unites the real video with the virtual — fixing the virtual objects in the real landscape.

Video to explain

It’s a small but revolutionary advance for architecture, industry, defense, cinema, leisure…

Here is one of many examples:
YouTube Preview Image