Stereo vision and 3D reconstruction on a processor network

Surface measurements during outdooor construction processes ar very costly whenever the measurement process interferes with the construction activities, since machine and man power resources are idle during the data acquisition procedure. Using frame cameras as sensors to provide a rneasurement data base can avoid, this problem. In this paper a system is introduced which uses stereo vision for three dimensional (3D) surlace reconstruction. The hardware and software issues are focussed, which are crucial in terms of the time performance limits given by the construction process. Stereo matching as major part is solved in a feature vector based opproach implemented on dedicated hardware. A multi-processor solution is described, which satisfies all requirements in terms of price, time behavior and accuracy as well as physical robustness and indipendence in the industrial environment. We conclude with examples and realistic figures as exerimental results ol a prototype solution