【作者】 王本庆；

【导师】 李兴国；

【作者基本信息】 南京理工大学 ， 信息与通信工程， 2010， 博士

【摘要】 为了防范恐怖袭击,许多国家都加强了机场和车站等公共场所的安检措施,其中近程毫米波成像技术就是最简捷有效的安检方式之一。毫米波兼具有微波与红外的优点,有一定的穿透能力,能够根据散射能量的大小区分不同物理属性的物体。近年来毫米波器件的不断发展和人们需求的不断提高,使得近程毫米波成像技术能够在医疗、导航和交管等领域得到越来越广泛的应用。根据应用环境和要求的不同,以成像距离为依据研究了被动焦平面全功率交流辐射计机械扫描成像、被动合成孔径成像和主动阵列全息成像等三种不同的近程毫米波成像体制,主要对它们的成像系统、成像特性及其信号处理技术进行了深入分析和讨论。全文围绕提高成像质量展开,主要包括以下几个方面的内容。1)近程毫米波成像系统的研究。为了提高成像速度,基于94GHz被动焦平面全功率交流辐射计机械扫描成像系统,设计了旋转圆楔无惯性机械扫描装置,大大提高了成像速度和质量。然后还设计了基于机械扫描的双通道被动合成孔径成像系统,系统由在水平和垂直方向上扫描的一对辐射计以及后端的信号处理系统组成,给出了系统设计参数和结构以及优化设计方案。2)近程毫米波成像特性的研究。近程成像是本次论文的特点,也带来了不同于远场成像的独有性质。通过研究可知,在三种近程成像体制中以不同形式都存在非线性相位项,它使得系统脉冲响应变宽,所成图像幅值变低,产生模糊,对于不同的成像体制要分别采用不同方法去除它的影响。3)近程毫米波图像恢复的研究。图像恢复和成像算法是本次论文研究的重点,主要是以成像特性为依据,研究提高成像质量的相关信号处理技术,针对三种不同的成像体制具有不同的侧重点。对于被动焦平面全功率辐射计机械扫描成像,重点研究了点扩散函数估计和图像恢复算法,如基于盲极大似然估计算法和偏微分方程方法的图像恢复。盲极大似然估计把图像看作一个随机场,各像素之间独立同分布,可以在去除噪声的同时保持图像细节。而偏微分方程方法利用原始图像和降晰图像之间差分,采用迭代算法,有效克服图像中由于离焦产生的模糊等现象。4)近程毫米波成像算法的研究。在能够同时得到电磁波幅值和相位信息的情况下,可以通过多种方式来进行信号处理,大大提高了系统的灵活性。对于被动合成孔径成像,采用直角坐标系推导了成像公式,简化了表达式和相位补偿运算；在计算相关函数时提出采用多相滤波算法,有效地克制了相关函数虚部的波动；采用多种成像算法提高成像质量,如分数阶Fourier变换和Radon-Wigner变换,同时提出了基于互强度传播方程的近程成像算法,以及基于四阶累积量的孔径扩展；提出用二阶的高阶双正交Fourier变换算法提取距离信息,可以实现被动测距等。对于主动阵列全息成像,研究基于阵列天线接收的窄带及宽带全息成像算法。全息成像能够在波数域完全补偿近程电磁波的波前球面弯曲,特别适用于超近程成像。窄带全息成像算法以Fourier变换算法为主,同时以边界元法、分布源边界点法和Helmholtz方程最小二乘法等为有益补充。还研究了宽带毫米波三维成像算法,使得目标像具有立体感,也没有焦距和焦深限制,这样可以克服离焦模糊等现象,较为真实地反映目标的物理特性。5)高阶双正交Fourier变换(High-order Biorthogonal Fourier Transform, HBFT)的研究。HBFT是本次论文首次提出的一种全新算法,其特征是把信号在高阶调频信号组成的双正交基下展开,在这组双正交基下,对应不同参数的高阶调频信号彼此正交,因此特别适合对多分量高阶调频信号的调频参数进行估计,得到调频参数密度谱。这个算法没有交叉项,不用搜索,具有较好的性能。通过以上研究,对近程毫米波成像中存在的主要问题进行了详细分析,同时提出了切实有效的解决方案,进而提高了成像质量,推动了成像技术的发展。更多还原

【Abstract】 In order to prevent terrorist attacks, many countries have reinforced the security check in the public places such as airports and stations, and the Near Range Millimeter wave Imaging (NRMI) technology is one of the most simple and effective ways. Millimeter wave (MMW) has the advantages of both the infrared and Microwave, and can distinguish objects with different physical properties according to their scattering energy. In recent years, with the improvement of MMW devices and people’s needs, NRMI is more and more widely used such as in the medical, navigation, traffic control and other areas.Three different NRMI systems of the passive focal plane full power AC radiometer mechanical scanning imaging, passive synthetic aperture imaging and active array holographic imaging are researched for different application environments and requirements in the dissertation, and their imaging characteristics and signal processing are detailed discussed. The work of the whole dissertation mainly includes the following items.1) Research on the NRMI systems. A non-inertial mechanical scanning device is designed based on rotating circular wedge in order to improve imaging system speed, which is a 94GHz passive full power AC radiometer imaging system based on the mechanical scanning. The dissertation also designs a dual-channel passive synthetic aperture imaging system based on mechanical scanning, which consists of a pair of horizontal and vertical scanning radiometers, as well as signal processing system, and gives the system hardware parameters and structure.2) Research on the NRMI characteristics. There exists nonlinear phase in the three NRMI system, which makes the system impulse response broaden and amplitude become lower. It needs different methods to remove its effects for different systems. For the passive full power imaging, it uses image restoration to improve image quality, and for passive synthetic aperture imaging and active array holographic imaging, it reduces its impacts by the phase compensation respectively in the relevant domain and wave number domain.3) Research on the NRMI image restoration. Image restoration and imaging algorithm are the keys of this dissertation, and three different systems emphasize different aspects.For passive focal plane full power radiometer mechanical scanning imaging, it focuses on the point spread function estimation and image restoration algorithms such as maximum likelihood (ML) estimation algorithm and partial differential equation (PDE) method. The ML estimation takes the image as a random field, and can remove the noise while retaining the edges. While the PDE method uses the difference between the original and blurred images; it can effectively overcome the image blur due to defocus iteratively. 4) Research on the NRMI imaging algorithms. If both the amplitude and phase information of the electromagnetic waves can be obtained, it could bring a variety of ways to carry out signal processing, which greatly increase system flexibility.For the passive synthetic aperture imaging, it deduces the imaging formula in the Cartesian coordinate system, which greatly simplifies the expression and phase compensation. Otherwise, it introduces polyphase filtering to calculate the correlation function for restraint fluctuations of imaginary part to improve image quality and system efficiency. It also introduces a variety of imaging algorithms to improve image quality, such as the fractional Fourier transform and Radon-Wigner transform, and put forward a new algorithm based on the mutual intensity propagation equation, moreover introduce the fourth order cumulant for aperture expansion. Finally it proposes the principle to extract the range information using the second order HBFT, which can achieve passive ranging.For the active array holographic imaging, it researches both the narrow-band and wideband imaging algorithms. Holographic imaging can fully compensate the spherical wavefront bending in the wave number domain, especially suitable for ultra-near range imaging. Narrow-band holographic imaging algorithm is based on the Fourier transform, while the boundary element method for arbitrary shape, distributed source boundary point method and the Helmholtz equation least squares method are the useful complement. The wideband MMW three-dimensional imaging algorithm also is discussed, while there is no restrictions on the focal length and focus depth, so that defocus blur can be overcome, and it can more truly reflect the physical properties of objects.5) Research on the High-order Biorthogonal Fourier Transform (HBFT). HBFT is a new algorithm proposed by this dissertation for the first time, which expands signal with biorthogonal basis functions composed with the high order FM signal, the corresponding high order FM signals with different parameters are orthogonal each other under this biorthogonal basis. Therefore, HBFT is particularly suitable for parameters analysis of multi-component high order FM signals, and obtain the density spectrum. It has a good performance of no cross-term and no need search.Through the above research, it detailed analyses the main problems in NRMI, and puts forward some practical and effective solutions, which improves the imaging quality and promotes the development of NRMI technology.更多还原