Theoretical accuracy analysis of N-ocular vision systems for scene reconstruction, motion estimation, and positioning

Theoretical models are derived to analyze the accuracy of N-Ocular vision systems for scene reconstruction, motion estimation and self positioning. Covariance matrices are given to estimate the uncertainty bounds for the reconstructed points in 3-D space, motion parameters, and 3-D position of the vision system. Simulation results of various experiments, based on synthetic and real data acquired with a 12-camera stereo panoramic imaging system, are given to demonstrate the application of these models, as well as to evaluate the performance of the panoramic system for high-precision 3-D mapping and positioning.