Monostatic/Bistatic 미세도플러 영상을 이용한 효과적인 탄도미사일 식별 연구

Abstract

A ballistic missile (BM) is a missile that follows a sub-orbital ballistic flightpath. Due to the high speed and low radar cross section (RCS), it is very difficult to intercept the ballistic warhead so the defensive interception of ballistic missiles has been a major topic of research in the radar society. The flight trajectory of an intercontinental ballistic missile is generally divided into three phases: boost phase, mid-course phase, and terminal phase. If the warhead can be detected, confirmed and tracked in the boost phase, the danger can be minimized. However, because detection in boost phase is a very difficult task, interception is performed in the mid-course phase when the warhead is outside the atmosphere and the missile is divided into the warhead and the decoy. These two targets have the characteristics of small size, high height, fast moving speed, and similar RCS values which cause difficulty in interception. In order to defend against such high-threat units, a significant identification is through different micro-motion of warhead and decoy by high frequency resolution. The ballistic warhead has three kinds of micro-motion during the flight, which are spinning, conning and nutation whereas the decoy is generally engaged in tumbling and wobbling motion. Because these two motion components are time-varying, the micro-Doppler (MD) in time-frequency domain can be a very good signature to discriminate the warhead and the decoy. This thesis proposes an efficient method to simulate the micro-Doppler (MD) frequency of the ballistic warhead based on the real flight scenario at monostatic and bistatic observations. Because it is very difficult to obtain the radar signal by changing the observation angle as the conventional electromagnetic software does to obtain the reflected signal for a fixed target, we transformed the pose of the model engaged in micro-motion in a local coordinate into that on the trajectory by constructing the transformation matrix. Then, we obtained the radar signal by using the point scatterer (PS) model and the high frequency estimation method, physical optics (PO), and the MD results were compared by using the short-time Fourier transform. In addition, this thesis proposes an efficient classification method to identify the warhead and the decoy for various observation scenarios by using the convolution neural network (CNN) classifier and the MD image. In simulations for various observation scenarios, MD signatures were successfully obtained, scattering characteristics were accurately analyzed, and high classification results close to 100 % were obtained.