Most of the problems of industrial inspection, reverse engineering, virtual reality require data about the geometrical shape of objects in 3-D space. Such 3-D data offer advantages over 2-D data: shape data are invariant against alteration of the illumination, soiling, and object rotation... Unfortunately, those data are much more difficult to acquire than video data about the 2-D local reflectivity of objects. We will discuss the physics of 3 -D sensing and will address the following subjects: coherent and incoherent illumination interaction of light with matter, at rough surfaces and at smooth surfaces the consequences of Heisenberg's uncertainty relation.From the knowledge of the underlying physical principles that define the limitations of measuring uncertainty, one can design optimal sensors that work just at those limits, as well as judge available sensors. We will show that the vast number of known 3-D sensors are based on only three different principles. The three principles are different in terms of how the measuring uncertainty scales with the object distance. We will further learn that with only two or three different sensors a great majority of problems from automatic inspection to virtual reality can be solved.We will specifically discuss laser triangulation, phase measuring triangulation (pmt) and white light interferometry at rough surfaces, from the viewpoint of the physicist, as well as for the benefit of the user of 3-D sensors, with emphasis on the potentials and limitations of the major sensor principles. It turns out that with those sensors 3-D data of a wide range of objects and materials can be acquired. The measuring uncertainty ranges from about one nanometer to a few mill i-meters, depending on the principle and the measuring range.We will briefly discuss some advantagous technical implementations of the principles mentioned above: a pmt-sensor that can deliver a 3-D image with 250.000 pixels within 32 ms, a pmt-sensor that can measure teeth, intraorally. Fast sensors are very helpful to ac-quire full 360? 3-D views by registering many single 3-D views. Since all triangulation sensors are severely limited by ubiquituous coherence, even if we use white, extended light sources, we have to switch to other physical principles if we need a measuring uncertainty of about 1 micrometer, on rough surfaces: we will discuss the coherence radar that essentially evaluates the time-of-flight by white light interferometry on rough surfaces. We will show examples of measured objects, with difficult features: very smooth metallic surfaces that are a big challenge for metrology, translucent objects such as ceramics, teeth or skin, layered volume scatterers, measurements in deep boreholes, living persons.