This paper is about real-time refinement of the 3D positions of a large number of stationary point-targets from a sequence of 2D images which are taken by a hand-held, calibrated camera group. To cope with the large data quantity arriving rapidly, an efficient iterative algorithm was developed. The problem and solution are expressed entirely within the computational framework of conformal geometric algebra. The iterative solution requires a pose estimation step of which two strategies are investigated. Experiments are performed to evaluate the algorithm based on synthetic and real data.