The problem of rapidly generating accurate object material descriptor databases of the earth's surface has been approached using a quantized rendering transform to code photogrammetric image measurements into physical surface parameters. The approach eliminates the lighting effect inherent in aerial photos which view earth surface elements from different perspectives. It reduces the multi-aspect photographic spectral measurements to objective surface properties which are then used for automated object and surface material classification. This paper presents the algorithms and design for a terrain database creation workstation used to generate 1 meter resolution data. The system was used to digitize approximately 200 aerial photos covering a 400 sq km area of Ft. Hunter Liggett, Calif. and translated into a 1.2 gigabyte surface descriptor database. Included in the workstation is a parallel processing, transputer- based, perspective view generator which uses the rendering transform to calculate side views at real-time rates. The use of this subsystem as a real-time feedback and quality control mechanism during database creation is described and the technique extended to real-time terrain database update systems.