This thesis presents the construction of active stereo vision system and its evaluation. Active stereo vision is a method of obtaining three-dimensional information about observed textureless objects. System consists of cameras, usually a pair, and a light source that can be manipulated. Therefore the aim was to test, if we could use two monochrome cameras and a projector, which were available, to successfully detect and reconstruct the surface of objects. Overview of the field was done and we decided to test three techniques; Gray coding, random pattern and line shift of light beams. Their performance was analyzed by means of quantitative and qualitative experiments. Based on analysis we can conclude that the choice of object material has little affect on the operation of technique with Gray coding. Surfaces with variable depths are problematic for random pattern technique, by contrast technique with line shift can find corresponding areas on those objects very well. Because line shift technique is time-consuming and inconsistent and random pattern technique has its limitations, we can conclude that the performance of Gray coding is the most versatile.
|
