Detection of outliers from spacecraft tracking data using GP-RBF network

The data, accounting for a small proportion of the measured data, which deviate from the tendency showed by the measured data as a whole are called outliers. Outlier detection is significant to improving accuracy of target-tracking data. A new method based on a radial basis function neural network with general parameter (GP-RBF) is proposed for detecting outliers. The GP-RBF neural network can do on-line outlier detection by adjusting the general parameter adaptively. Simulation results show that this new method can resolve the problem of outlier detection efficiently and quickly.