Fund Project:Project supported by Ningbo University Discipline Project, China (Grant Nos. XKL14D2058, XYL15008) and the K. C. Wong Magna Fund in Ningbo University, China.
Received Date:01 December 2018
Accepted Date:28 December 2018
Available Online:01 March 2019
Published Online:05 March 2019
Abstract:In the recent years, the theory and technologies of electromagnetic computational imaging have been well developed and several novel imaging methods have been proposed, one of which is known as the microwave imaging under random field illumination. In order to solve the matrix equation of imaging model, the key of such an imaging system is to generate the random electromagnetic radiation field distribution, implementing the independent measurements under random field illuminations. In this work, an optimal microwave imaging system for the desired imaging region and resolution is theoretically analyzed and experimentally implemented. In the randomness analysis, the correlation between different measurements is evaluated by the singular value decomposition, which is also adopted as a criterion for choosing the optimal parameters of the imaging system. Based on random field illuminations generated by the least number of antenna elements, a full-rank matrix equation can be used to reconstruct the object by direct matrix inversion, which can be completed in nearly real-time once the system calibration is implemented in advance. The numerical simulation and experimental investigation are performed, and the results prove the effectiveness of the proposed optimal imaging system. By using the traditional array theory, it is found that for an N-element phase array, N illuminations with each element excited by a single frequency, equal amplitude and randomized 0 or ${\text{π}}$ phase signal will result in N independent measurements. Theoretically, any additional measurement under random illumination will be correlated with the previous N measurements. Since the random field illumination is obtained by array antennas with 1-bit random phase modulation, the power radiated by each transmitting element is not sacrificed, resulting in an optimal power efficiency of the imaging system compared with those of earlier metasurface-based imaging systems. Besides, a single frequency signal source is used in the system, which also realizes the optimal spectrum efficiency. In conclusion, there are two major innovations of the proposed imaging system: 1) the completely random field illuminations based on 1-bit phase modulation; 2) the approach to optimizing the system on desired demand. The compact and low-cost imaging system promises to have various imaging applications, such as public security and indoor localization. Keywords:microwave imaging/ array antenna/ random field/ phase modulation
全文HTML
--> --> -->
2.理论分析在基于随机场照射的成像体制中, 利用随机电磁特征的场分布对目标进行非相关测量, 从而实现对目标散射信息的有效拾取. 该成像系统的核心在于随机场的产生, 随机相位调制的阵列天线是其中的一种实现形式. 如图1所示, x-y平面内的阵列天线单元数量为N, 并标记第i个单元的空间位置为${\bar r_i}$, $i=1,2, \cdots, N $. 假设每个阵列单元的激励电流为${I_0}{{\rm{e}}^{ - {\rm{j}}{\varphi _i}}}$, 其中I0是一个常数. 为了实现随机照射, 阵列天线中的每个单元相位${\varphi _i}$在0或${\text{π}}$之间做1 bit相位切换. 成像计算过程中, 成像区域被离散成N' 个网格, 其中第$ i'$个网格位于$\bar r_i'$, $i'= 1, 2,\cdots,N'$. 根据电磁传播理论, 由第i个天线单元辐射到成像区域$\bar r_i'$处的电场表达式为 图 1 基于随机场照射的成像系统示意图 Figure1. Schematic diagram of the imaging system based on randomized field illuminations.
上述的仿真结果验证了: 当$M=N=N' $时, 完全随机照射得到的H矩阵是一个满秩方阵, 从而实现对目标散射信息的有效测量. 在成像系统中, 作为指标的成像分辨率$\Delta r' $和成像区域通常是给定的, 而待确定的系统参数包括测量次数M、成像距离R、天线单元数量N及其组阵间距$\Delta r$. 因此, 一旦给定了成像分辨率$ \Delta r' $和成像区域, 就得到了离散化网格数量$ N'$, 而天线单元数量N和测量次数M也随之确定. 余下的未知参数可以通过奇异值来进一步优化确定. 为了探究不同系统参数之间的相互关系, 仿真计算中设置天线单元数量N = 5 × 5, 组阵间距$ \Delta r$ = 1$\lambda $, 成像分辨率$\Delta r' $ = 2$\lambda $, 得到如图4(a)所示的随测量次数和成像距离变化的H矩阵奇异值分布曲线图. 图4(b)为图4(a)在测量次数为25处的二维剖面图. 根据奇异值曲线, 对于不同的测量次数, 存在最优的成像距离R. 具体地, 对于组阵间距$\lambda $, 最优成像距离为10$\lambda $. 因此, 可以依据不同的组阵间距$\Delta r$得到相应的最优成像距离R和分辨率$\Delta r' $的关系曲线, 见图5(a). 对于给定的成像分辨率, 最优成像距离随着组阵间距增大而增大. 同样, 保持组阵间距$\Delta r$为典型的半波长, 还可以得到如图5(b)所示的不同天线单元个数N和最优成像距离R、分辨率$\Delta r' $之间关系曲线. 对于给定的成像分辨率, 最优成像距离随着阵列天线单元数量增大而增大. 依据阵列天线口径D和组阵间距$ \Delta r$、天线单元数量N之间关系, 结合图5还能够得出分辨率$\Delta r' $随阵列天线口径D增大而减小的变化规律. 以上性质与传统雷达的角分辨率一致. 图 5 成像系统参数之间关系 (a)最优成像距离R与组阵间距Δ r、分辨率Δ r′之间的变化关系; (b)最优成像距离R与天线单元数量N、分辨率Δ r′之间的变化关系 Figure5. Dependence analysis: (a) Dependence of the optimal imaging distance with respect to Δ r and Δ r′; (b) dependence of the optimal imaging distance with respect to Δ r′ and N when Δ r = 0.5$\lambda $.
图 4 随测量次数和成像距离变化的H矩阵奇异值分布曲线(a)以及25测量处的二维剖面(b) Figure4. (a) Dependence of the normalized singular value with respect to the measurement times M and the imaging distance R; (b) profiles of Fig. 4(a) for M = 25.
在成像之前, 首先需要得到包含随机场特性的H矩阵[18]. 目前构建该矩阵主要有两种方式: 1)近场扫描法, 对发射/接收天线进行近场扫描, 通过傅里叶变换将近场信息转换到位于远场的成像区域, 此处的电磁场信息即为H矩阵[20]; 2)逐点校准法, 将分辨率大小的校准反射体逐点放置于成像区域中所有网格, 综合每个网格处的传输信息就构成了H矩阵[24]. 从实际操作角度, 本文采用易于实现的第二种方法. 具体地, 选择2$\lambda $ × 2$\lambda $的金属片作为校准反射体. 在第一次的校准测量中, 校准反射体放置在成像区域中的第一个网格处. 随后, 25次随机照射得到25个S21测量值, 从而构成H矩阵的第一列. 在剩余的24个网格处分别重复上述校准过程, 最终得到完整的H矩阵. 图11为实际测量H矩阵的归一化奇异值曲线, 结果表明, 尽管实际实验中测量次数和发射天线单元个数并不相等, 但是由于测量误差和系统噪声的影响, 使得第25次测量与先前测量值并不完全相关, 因此仍可以通过伪逆矩阵来求解方程(9). 图 11 实测H矩阵的归一化奇异值 Figure11. Normalized singular values of the H matrix using the measured data.
24.3.实验结果 -->
4.3.实验结果
图12为利用VNA、放大器等仪器以及随机照射天线阵列搭建的成像实验系统, 其中步进转台用于精确地得到不同成像距离. 测量中, 首先在成像区域放置不同形状的金属片作为待成像目标体(如图12中所示的倒“L”形图案); 同时移动步进转台使目标体位于成像距离R处; 随后计算机向阵列天线依次发送25次随机照射指令并读取VNA的S21测量数据. 最终, 利用测量得到的散射测量值g和校准得到的H矩阵, 通过矩阵求逆的方法得到反演结果. 图 12 基于随机多波束照射的微波成像实验系统 Figure12. Experimental setup of the imaging system based on random field illuminations.
图13(a)所示分别为成像距离7$\lambda $, 10$\lambda $, 13$\lambda $处的离散点目标和倒“L”形状目标重建图像, 其中位于10$\lambda $处的成像效果最好. 由于倒“L”形的目标更加复杂, 使得该目标在7$\lambda $和13$\lambda $处的成像结果难以分辨. 以上结论仅仅通过直观观察得到. 图13(b)为上述两种原始目标在不同距离下的NRMSE结果, 曲线清楚地显示, 成像距离为10$\lambda $处的重建图像的误差最小, 验证了仿真和理论中的最优成像距离. 为了探究该成像系统的更多性能, 在最优成像距离处对多个离散点、“T”形、“十”字目标等较复杂目标进行成像实验. 图14(a)和图14(b)分别为原始目标以及对应的图像重建结果, 原始目标的位置和形状都能够准确反演. 图 14 在最优成像距离处的成像实验结果 (a)原始目标; (b)重建图像 Figure14. Imaging results at the optimal distance using experimental data: (a) The original objects; (b) reconstructed images.
图 13 不同成像距离处的成像实验结果 (a)离散点目标和倒“L”形状目标在不同成像距离R的重建图像; (b)重建图像误差随成像距离R的变化曲线 Figure13. Experimental results with different R: (a) Reconstructed imageof two discrete objectsand inverted L-shape objectat 7$\lambda $, 10$\lambda $, and 13$\lambda $ distances, respectively; (b) NRMSE analysis.