New least squares registration algorithm for data fusion

Abstract

A new least-squares registration (NLSR) algorithm is developed to accurately estimate and correct the systematic errors of a data fusion system. First, the two-sensor registration problem is expressed by an averaged least-squares (LS) criterion function of the sensor measurements. The criterion function is optimized by a Newton algorithm. Then, the algorithm is extended to multiple-sensor case. The accuracy of the proposed estimation scheme achieves the Cramer-Rao bound (CRB). Theoretical analysis and simulations are employed to assess the performance of the proposed algorithm.