Target detection improvement by using a class decision algorithm for synthetic aperture radar

This paper proposes a classification method to improve the target detection accuracy in a synthetic aperture radar processing algorithm. In the imaging algorithm, Fourier based processing algorithm is used to obtain the processed image from the measured 2D Cartesian backscattered frequency domain data. In case of measured data with limited frequency band and aspect angle interval, radar target detection accuracy will be reduced. Besides, the noise in measured data affects the result and thus it is hard to identify the target. In this paper, a decision rule, which is based on a classification algorithm by using neural networks, is proposed to improve the target identification accuracy by comparing the processed image with the images in the data bank. In this work, the effect of noise, frequency and aspect angle limitation in the decision accuracy are investigated and the results are presented. Some other decision methods are also given to compare the results.