Classification of radar clutter using neural networks

Abstract

A classifier that incorporates both preprocessing and postprocessing procedures as well as a multilayer feedforward network (based on the back-propagation algorithm) in its design to distinguish between several major classes of radar returns including weather, birds, and aircraft is described. The classifier achieves an average classification accuracy of 89% on generalization for data collected during a single scan of the radar antenna. The procedures of feature selection for neural network training, the classifier design considerations, the learning algorithm development, the implementation, and the experimental results of the neural clutter classifier, which is simulated on a Warp systolic computer, are discussed. A comparative evaluation of the multilayer neural network with a traditional Bayes classifier is presented.

An Improved Neural Image Compression Approach With Cluster Based Pre-Processing

“…Based on the calculated error correction term the weights are updated by back propagation [24]. With updated weights again the error is calculated and back propagated.…”

Neural Model for Far-Field 1D Localization of Mobile Stochastic EM Sources with Partially Correlated Radiation

“…In far-field, the electric field strength at the selected sampling point can be calculated as: (2) where M represents the mapping by Green function (3) θ and j represent the spatial angles of stochastic source location with respect to the selected sampling point in far-field, F(θ,j) is the radiation pattern of the short dipole, r S is the distance between s-th stochastic EM source and selected sampling point, z 0 is the impedance of free space and k is the phase constant (k=2π/λ).…”

“…The neural model is based on MLP ANN [3][4][5], and for the case which considers the two stochastic sources, its architecture is shown in Fig.2. The main purpose of the model is to perform the mapping from the space of S signals described by correlation matrix C E to the 1D DoA space (9) where…”

“…In [9][10][11] the neural models for 1D DoA [9,10] and 2D DoA estimation [11] and spatial position estimation of stochastic radiation sources are presented, developed for sources whose radiation is mutually uncorrelated. This paper continues research work of [9,10] with the aim to investigate the possibility of multilayer perceptron neural network (MLP) application (MLP) [3][4][5] in DoA estimation for stochastic radiation sources for the case when there is some level of partial correlation in their radiations.…”

“…These methods provide a high accuracy in space location estimation of the radiation sources, but because of complex matrix estimations they demand powerful hardware resources and are not convenient for real time applications. Artificial neural networks [3][4][5] may be applied in the process of DoA estimation of radiation EM sources, as neural models that avoid intensive matrix estimations may be a good alternative to superresolution algorithms. These neural solutions may achieve same accuracy while gaining higher speeds in response.…”