A Sensor Registration Method Using Improved Bayesian Regularization Algorithm

We consider the multi-sensor tracking systems. In order to solve the sensor registration in multi-sensor tracking system, we propose a new solution based on improved Bayesian regularization algorithm using neural networks in this paper. The nonparametric nature of this approach guarantees that many different kinds of sensor biases can be registered adequately; Levenberg-Marquardt optimum algorithm integrated with Bayesian regularization is applied to solve the registration problem with quick convergence rate and high resolution. Simulation results show the advantage of convergence and generalization as compared to the parametric algorithms and LM optimum algorithm.