A LOW COST 3D-MEASUREMENT TOOL FOR ARCHITECTURAL AND ARCHAEOLOGICAL APPLICATIONS Klaus Hanke, Associate Professor Institute of Geodesy, University of Innsbruck, Austria Mostafa Abdel-Bary Ebrahim, M. Sc. Assiut University, Egypt, Doctoral Student at Innsbruck University, Austria KEY WORDS: Accuracy, Architectural photogrammetry, Archaeological surveying, PhotoModeler. ABSTRACT: An important question to be answered for any measurement tool is ” How accurate is it ?”. Because of this question, an accuracy investigation has been done to obtain the accuracy of a low cost tool of measurements for the architectural and archaeological applications. The aim of this investigation is to study the accuracy of a low cost photogrammetric softcopy close-range program ”PhotoModeler” developed and distributed by Eos Systems, Vancouver, Canada. The software contains a bundle adjustment program for digital images and a camera calibration part. It is easy to use and requires only short time to familiarize with. Different types of cameras with different types of calibration of these cameras and different numbers and arrangements of photographs have been tested. A brick wall with a good contrast of its brick corners has been chosen to be a test field because it is of the same type as in many architectural and archaeological applications. To properly document the various levels of accuracy one can achieve with this program, the results have been compared with high accuracy measurements from other standard surveying methods, the accuracy of the control points lies between 1.0 and 1.5 mm absolute. Two types of cameras have been used, a common small format ”amateur camera” Ashai Pentax and a metric camera WILD P32. The non-metric camera has been calibrated in three different ways to show the effect of the lens distortion on the accuracy of the measurements. Two different camera configurations have been used for both camera types, once for top and lower camera positions and another for just lower camera positions only. The results are very promising, the achieved accuracy in the distances between measured points is in the range of 1:1700 to 1:6500 of the object’s size. Regarding to the points’ co-ordinates, the average accuracy reached up to 1:8000. INTRODUCTION In theory of errors of measurements the term accuracy is traditionally used for a concept that measures the closeness of derived estimated or predicted data to reality. Photogrammetrists very often estimate the accuracy of a method by controlled experiments, where the photogrammetrically determined co-ordinates are compared with so called given co-ordinates that have an accuracy that is considerably higher than that of the method to be checked (Ebrahim M., 1992). An accuracy investigation has been carried out to obtain the accuracy of the 3D-models that can be expected from the PhotoModeler software for the architectural and archaeological applications. Both, a metric camera and a non-metric camera have been investigated. To obtain the accuracy of the non-metric camera, a calibration of this camera has been done using PhotoModeler's own calibrator software. Different calibration projects for the non-metric camera have been used to obtain the accuracy with different camera parameters. The first project has been done with full camera lens distortion, the second with radial lens distortion only and the third without any lens distortion parameters. Also two different configurations have been used. One with upper and lower camera positions and the other with lower positions only. THE NON-METRIC CAMERA CALIBRATION An Ashai Pentax non-metric camera has been used to achieve the accuracy of the non-metric camera with the used bundle adjustment software. PhotoModeler's calibrator software was the tool to calibrate the non-metric camera i.e. to obtain the parameters of the interior orientation in the three projects (with full lens distortion, with radial lens distortion only and without lens distortion). A special calibration pattern that is available with the software has been used. This grid is available as a slide that one can project on a screen for photographing, but the screen must be flat. It is available also as DXF file that one can plot for photographing. The latter has been used in this investigation. At least six photographs must be exposed, one from each corner and two from each of the two sides. After scanning the photos or digitizing them via Photo-CD, the software marks the points automatically using pattern recognition algorithms and stores the calibrated camera parameters in a "camera file" which will be used by the bundle adjustment software afterwards. Figure (1) illustrates the calibration pattern. Ten photographs have been used, four from corners, four from sides and additional two from sides as a vertical photos. The result of this calibration projects was a set of