雷达接收系统目标信号实时校正仿真研究
|
摘要
合成孔径雷达(SAR)接收系统由于电路结构和器件的非理想性无法避免地存在误差,这些误差会在一定程度上影响目标信号的幅度和相位,从而影响雷达成像性能。由于系统传递函数的随机性和非共轭对称性,传统设计方法以及实系数FIR滤波器都不能满足要求。BP神经网络具备逼近任意非线性函数的能力,可利用其根据系统传递函数设计出相应的复系数滤波器。基于此,提出了将滤波器设计问题转化为实权值神经网络的训练问题,在实数域对误差函数的实部和虚部分别进行最小化的设计方法。结果显示,经过校正后的目标信号的评价指标和脉压波形均接近理想信号。最后结合实测数据验证了该方法能完成任意雷达接收系统的误差校正,是一种有效的设计方法。
Because the circuits of the system and the electrical elements are not ideal, it is unavoidable that the errors exist in the synthetic aperture radar(SAR) receiver, which may affect the amplitude and phase of the target signals to a certain extent, thus affecting the radar imaging performance. Due to the randomness and non-conjugate symmetry of the system transfer function, the traditional design methods and the real coefficient FIR filters cannot meet the requirements. BP neural network has the ability of approximating any non-linear function, and can be used to design the corresponding complex coefficient filters according to the system transfer function. Based on this, the method was proposed, which transforms the filter design problem into the training problem of real weights neural network and minimizes the real part and imaginary part of the error function in the real domain respectively. The results show that the evaluation indexes and the pulse pressure waveforms of the target signals are close to the ideal signals after the correction. Finally, it is verified that this method can complete error correction of any radar receiving system with the measured data, which is of great effectiveness.
Because the circuits of the system and the electrical elements are not ideal, it is unavoidable that the errors exist in the synthetic aperture radar(SAR) receiver, which may affect the amplitude and phase of the target signals to a certain extent, thus affecting the radar imaging performance. Due to the randomness and non-conjugate symmetry of the system transfer function, the traditional design methods and the real coefficient FIR filters cannot meet the requirements. BP neural network has the ability of approximating any non-linear function, and can be used to design the corresponding complex coefficient filters according to the system transfer function. Based on this, the method was proposed, which transforms the filter design problem into the training problem of real weights neural network and minimizes the real part and imaginary part of the error function in the real domain respectively. The results show that the evaluation indexes and the pulse pressure waveforms of the target signals are close to the ideal signals after the correction. Finally, it is verified that this method can complete error correction of any radar receiving system with the measured data, which is of great effectiveness.
引文
[1] John C.Curlander,Robert N.Mcdonough.合成孔径雷达——系统与信号处理[M].北京:电子工业出版社,2006-12.
[2] 张月,鲍庆龙,杨剑,陈曾平.宽带数字整列雷达通道均衡方法的设计与实现[J].信号处理,2010,26(3):453-457.
[3] 矫伟,等.基于内定标信号的合成孔径雷达系统幅相误差的提取和校正[J].电子与信息学报,2005,27(15):1883-1886.
[4] 常文革,祝明波,梁甸农.宽带线性调频信号产生技术研究[J].信号处理,2002,18(2):113-117.
[5] 何志华,等.SAR回波信号模拟器幅相误差实时校正方法[J].雷达科学与技术,2011,9(2):125-129.
[6] 何志华,等.干涉SAR模拟器通道幅相特性实时校正方法[J].电子学报,2013,41(9):1710-1715.
[7] L J Karam,J H Mc Clellan.Chebyshev digital FIR filter design[J].Signal processing,1999,76(1):17-36.
[8] 陆必应,等.用线性规划算法设计复系数FIR滤波器[J].国防科技大学学报,2001,23(4):93-97.
[9] 王炎,等.FIR数字滤波器设计的复数神经网络方法[J].信号处理,1999,15(3):193-98.
[10] 李季檩,吕宝粮.复系数FIR数字滤波器的神经网络设计方法[J].计算机仿真,2008,25(2):185-189.
[2] 张月,鲍庆龙,杨剑,陈曾平.宽带数字整列雷达通道均衡方法的设计与实现[J].信号处理,2010,26(3):453-457.
[3] 矫伟,等.基于内定标信号的合成孔径雷达系统幅相误差的提取和校正[J].电子与信息学报,2005,27(15):1883-1886.
[4] 常文革,祝明波,梁甸农.宽带线性调频信号产生技术研究[J].信号处理,2002,18(2):113-117.
[5] 何志华,等.SAR回波信号模拟器幅相误差实时校正方法[J].雷达科学与技术,2011,9(2):125-129.
[6] 何志华,等.干涉SAR模拟器通道幅相特性实时校正方法[J].电子学报,2013,41(9):1710-1715.
[7] L J Karam,J H Mc Clellan.Chebyshev digital FIR filter design[J].Signal processing,1999,76(1):17-36.
[8] 陆必应,等.用线性规划算法设计复系数FIR滤波器[J].国防科技大学学报,2001,23(4):93-97.
[9] 王炎,等.FIR数字滤波器设计的复数神经网络方法[J].信号处理,1999,15(3):193-98.
[10] 李季檩,吕宝粮.复系数FIR数字滤波器的神经网络设计方法[J].计算机仿真,2008,25(2):185-189.