Microwave Imaging of Non-Rigid Moving Target Using 2D Sparse MIMO Array
Microwave Imaging of Non-Rigid Moving Target Using 2D Sparse MIMO Array
复制标题
使用 2D 稀疏 MIMO 阵列对非刚性移动目标进行微波成像
DOI：
10.1109/access.2019.2945968
复制
发表时间：
2019-01-01
期刊：
影响因子：
3.9
通讯作者：
Xu, Feng
中科院分区：
文献类型：
--
作者：
Zhu, Zhanyu;Kuang, Lei;Xu, Feng
提出了一种适用于大交通场景的非刚体运动目标的微波/毫米波成像方法。该方法是针对由两个正交线性阵列组成的二维稀疏MIMO(多输入多输出)阵列提出的。在使用时分复用(TDM)信号传输技术时，与接收天线相比，使用较少的发送天线来减少系统的数据采集时间。在二维稀疏MIMO阵列上传输宽带信号，利用合成孔径成像技术重建三维空间的后向散射系数。然后在目标运动过程中生成一系列合成的3D快照图像。本文将非刚体人体目标分解为25个关节，利用Kinect光学设备进行跟踪。分解后，每个关节的运动可以描述为Kinect跟踪的时空轨迹，提供了每个关节的3D位置的初步指示。在此基础上，将每幅微波/毫米波图像分成25幅对应于关节位置的子图，形成每个关节的图像序列。随后，提出了一种分段联合估计方法，从图像序列中准确地估计出每个关节的运动参数。通过分段估计可以显着降低大角度观测中图像的去相关性，从而实现运动参数的全局优化。因此，将估计的参数纳入每个部件的有效运动补偿中。同时，提出了图像重聚焦的方法来生成各分量的聚焦良好的图像。最后，将所有部件的图像融合成一幅完整目标的图像，从而重建出高分辨率的3D图像。成像仿真和测量数据实验表明，该方法是有效的。
This paper develops a microwave/mmw (millimeter-wave) imaging method for moving objects of non-rigid body which can be used in the heavy traffic scenarios. This method is proposed with a 2D sparse MIMO (multiple-input multiple-output) array composed of two orthogonal linear arrays. In comparison to the RX antennas, a smaller number of TX antennas is used to decrease the data acquisition time of this system when using TDM (time division multiplexing) signal transmitting technique. The 2D sparse MIMO array is deployed with wide-band signals transmitted to reconstruct the backscattering coefficient in 3D space using synthetic aperture imaging technique. Then a series of synthesized 3D snapshot images are generated during the motion of the target. The non-rigid human body target is decomposed into 25 joints tracked by an optical device which is Kinect in this paper. After the decomposition, the movement of each joint can be described as a space-time trajectory tracked by Kinect, providing a preliminary indication of the 3D position of each joint. With the indication, we divide every microwave/mmw image into 25 sub-images corresponding to the joints’ position to form an image sequence of each joint. Subsequently, a segmental joint estimation method is proposed to estimate the accurate motion parameters of each joint from the image sequence. The decorrelation of the images in large angle observation can be significantly reduced by the segmental estimation which leads global optimized motion parameters. Consequently, the estimated parameters are taken into the effective movement compensation of each component. Meanwhile, the image refocusing method is presented to generate the well-focused image of each component. Finally, all the images of components are fused into one image of the whole target to reconstruct the 3D high-resolution image. Imaging simulations and measuring data experiments show that this method is efficient.