Abstract:

Radar highresolution range profile (HRRP) provide potentially discriminative information on the geometry of target, which has been shown to be promising signatures for radar automatic target recognition (ATR) application．As traditional algorithms are not robust to noise, an autocorrelation wavelet support vector machine is proposed, the kernel of which is constructed with a compactly supported wavelet satisfies the translation invariant property. Adaptive power transformation is adopted to enhance the classifier noiserobustness with the estimation method for realtime SNR and an empirical formula for selecting the power exponent. The simulation results show that the average recognition rate with the proposed classifier is higher than SVM with Gaussian RBF kernel under Gaussian white noise background.