Currently radar target identification is effected by finding the downrange distribution of scatterers on a platform using high definition radar and referencing this distribution to a look-up table of measured radar returns. This method is difficult due to the rapid scintillation of the radar signal of a flying target at short wavelengths. We investigate a method of target identification in the low definition, relatively long wavelength, limit using techniques derived from the field of non-linear dynamics. In this limit the radar reflection is much less sensitive to the orientation of the target to the radar and scintillation is greatly reduced. Modern detection advances allow the digitization and processing of raw real time non-pulse compressed waveforms. We simulate the raw radar returns from several very similar small (4 to 5 wavelength) objects (i.e.. cylinder, cone-cylinder, cone-ogive and double ogive) illuminated with chaotic or random modulated waveforms or a linear chirp using a commercial finite difference time domain code. In this aggressive limit we embed the target return in two dimensions using the method of delays to form an attractor. The geometric properties of the attractors are analyzed to distinguish between these similar target objects in the low definition limit. Our method selects random points on a reference attractor (or strands of points) with many Euclidean nearest neighbors and compares these with the nearest neighbor density of points (or strands) with the same time index on an attractor from an unknown target. Comparing the density of nearest neighbors at these points on two attractors allows us to calculate probabilities of correct identification. (Is the unknown signal from the reference target or from one of the other targets?) Simulation data and analyses derived from three different transmit waveforms are compared. The robustness of our target ID method is probed by the addition of white noise to the signals.
展开▼