BP Neural Networks Based Sensor Registration Algorithm

In multi-sensor multi-target fusion systems the sensor registration is the major problem which is generally inadequate. The typical registration algorithms are based on statistics, such as LSE, ML etc. and on filtering theory, such as KF, EKF etc. They all are parametric methods namely. We propose a new algorithm that is based on ANN, nonparametric and many different kinds of sensor error variables to be solved. The algorithm is modified by adjusting the learning rate in which fuzzy logic is applied. That is to say, the learning rate of the NN is adaptive.