地面运动目标检测、成像与识别中的关键技术

Battlefield awareness is very important for the success of military operations. This project is focused on some enabling techniques for the detection, imaging, and classification of ground moving and stationary targets. A ground moving target detection scheme is presented, which combines keystone formatting with space-time adaptive processing and can compensate for the range walk of moving targets and clutter in high resolution radar. A computationally efficient method, referred to as DS-WRELAX, is proposed for the delay and Doppler scale estimation of multiple closely spaced moving targets. Later, it has been extended to the continuous high resolution pavement profiling with ground penetrating radar, which can compensate for the vertical vibration of testing vehicles. A novel approach based on signal separation estimation theory with special relaxation structure is proposed for the three-dimensional super-resolution imaging and autofocusing in curvilinear SAR (synthetic aperture radar), which can also be used for the autofocus of conventional SAR, ISAR(inverse synthetic aperture radar), and interferometric SAR (IFSAR) with two antennas. Automatic target recognition (ATR) techniques are studied with the MSTAR (Moving and Stationary Target Acquisition and Recognition) data set collected by US DARPA (Defense Advanced Research Project Agency) and AFRL (Air Force Research Laboratory). A novel scheme is proposed for the ATR based on high range resolution (HRR) radar profiles. Various kinds of kernel-based classification techniques (especially the SVM (support vector machine)) are studied with their potential applications to SAR ATR. KFD (Kernel Fisher Discriminant) is first applied to SAR ATR. The effect of power transform on the performance of ATR is studied systematically and an adaptive preprocessing scheme is proposed with variable power transform coefficients, which can significantly improve the ATR performance. The above achievements are not only of great academic significance, but also of great military application potential

本项目拟用系统统一设计与优化的方法，研究可用于现代战场感知的地面运动目标检测、成像与识别中的若干关键技术，包括统一距离走动补偿、能控制或补偿信号失真的空时自适应处理、稳健的动目标运动补偿与超分辨率成像、基于二维动目标像与新信息论测度的自动目标识别和利用一维高距离分辨率像、基于模型的自动目标识别。