Powder Bed Fusion (PBF) techniques are additive manufacturing (AM) technologies with paradigm-shifting potential for the production industry. However, before they can become viable production solutions for quality critical industries, issues with process consistency and repeatability need to be addressed. There is a need for in-situ sensing systems that characterize process variation and analytical methods that relate sensor data back to input parameters and final part quality. This dissertation describes an in-situ stereovision-based metrology technique for characterizing 3D build surface variation in metal PBF AM. The natural optical characteristics of metal powders and fused metal regions are leveraged to extract high-precision 3D surface measurements from stereoscopic images of the powder bed. Measurement results are shown to detect spreader blade wear, broad powder surface height variations, powder spread interactions with part geometry, and powder bed surface irregularities linked to a focal part defect. Build surface measurements are also used to calculate layer-wise measures of powder layer thickness, material densification, and incremental build height which provide the information necessary to perform localized in-process parameter optimization with closed-loop control. Finally, layer-wise stereoscopic surface measurement data are rendered volumetrically to produce quasi-tomographic representations of build process variation that may assist with final part qualification. This dissertation provides a foundation of knowledge for applying in-situ stereo vision metrology to metal PBF, but work remains to fully realize its potential.