Abstract
An inverse synthetic aperture radar (ISAR) image shows a two-dimensional (2D) distribution of the reflectivity of a target and is formed by projecting the reflectivity onto a 2D image plane. It is widely used to identify unknown targets. However, it is very difficult to construct the training database of the ISAR image by estimating the 2D image plane. This paper proposes an efficient method of constructing an ISAR training database based on the flight scenario, which can be applied to recognition of real targets. Simulation results for three flight scenarios using real-sized five targets prove that the proposed method is very efficient and insensitive to noise and training data reduction especially when the flight is at an angle to the viewing vector.