Phase noise filtering and phase unwrapping method based on unscented Kalman filter

This paper presents a new phase unwrapping algorithm based on the unscented Kalman filter（UKF） for synthetic aperture radar（SAR） interferometry.This method is the result of combining an UKF with path-following strategy and an omni-directional local phase slope estimator.This technique performs simultaneously noise filtering and phase unwrapping along the high-quality region to the low-quality region,which is also able to avoid going directly through the noisy regions.In addition,phase slope is estimated directly from the sample frequency spectrum of the complex interferogram,by which the underestimation of phase slope is overcome.Simulation and real data processing results validate the effectiveness of the proposed method,and show a significant improvement with respect to the extended Kalman filtering（EKF） algorithm and some conventional phase unwrapping algorithms in some situations.

THE COHERENCE COEFFICIENT MAP AND RESIDUE GUIDED LEAST SQUARE PHASE UNWRAPPING ALGORITHM

The generation of high-resolution DEM from interferometric SAR has resulted in the need for accurate and efficient methods of 2-dimensional phase unwrapping. In this paper, we give a brief description of the mathematical base of phase unwrapping, and a detailed description of the unweighted and weighted least square phase unwrapping algorithm.Then our algorithm combining with the weighted least square phase unwrapping guided by the branch-cuts derived from Goldstein’ s algorithm and coherence coefficient map derived from the INSAR data is provided. In our experiment we write subroutines of the Goldstein’s branch-cut algorithm,unweighted and weighted least square phase unwrapping algorithm as well as our algorithm,and construct a small experiment system to resolve the phase unwrapping problem. Finally we test our algorithm on some INSAR data. The result shows that our approach can obtain unwrapped phase correctly and efficiently.

Enhanced multi-baseline unscented Kalman filtering phase unwrapping algorithm

This paper presents an enhanced multi-baseline phase unwrapping algorithm by combining an unscented Kalman filter with an enhanced joint phase gradient estimator based on the amended matrix pencil model, and an optimal path-following strategy based on phase quality estimate function. The enhanced joint phase gradient estimator can accurately and effectively extract the phase gradient information of wrapped pixels from noisy interferograms, which greatly increases the performances of the proposed method. The optimal path-following strategy ensures that the proposed algorithm simultaneously performs noise suppression and phase unwrapping along the pixels with high-reliance to the pixels with low-reliance. Accordingly, the proposed algorithm can be predicted to obtain better results, with respect to some other algorithms, as will be demonstrated by the results obtained from synthetic data.

A new method of weight choice in InSAR least squares unwrapping

The de-coherence phenomena such as Low-SNR radar signal, shadows and layover caused by topography, etc. , causing phase data discontinuity, makes the result of unwrapping phase inaccuracy or even completely wrong. Based on the analysis of influencing factors to weight choice, this thesis develops a new method to choose the weights based on the measure of the confidence in the frequency domain. Experiments show that it could overcome the defect of sub-estimate to the slope of least squares method very well, which has a better rationale, stability and performance.

A new method of weighted choice in InSAR Least Squares unwrapping

The decorelation phenomena such as Low-SNR radar signal, shadows and layover caused by topography etc, causes phase data discontinuous and makes the result of unwrapping phase inaccurate or completely wrong. Based on the analysis of influencing factors to the weight selection, this paper develops a new method to choose the weights based on the measurement of confidence in frequency domain. Results show that it is more precise and robust than other methods, and can make up for the defect of sub-estimate to the slope of least squares method.

Comparison of phase unwrapping algorithms for living cell's interference pattern

A multiphase microscopic interference system is designed to measure the height of cell which is important to the research of collective cell migration in physiology and medicine. This sys- tem can quantitatively measure cell height across a living monolayer without knowing the refractive index of cells. For the interference pattern, because the phases are all wrapped between - π to π, it is necessary to get the real phase through phase unwrapping,a method to restore the wrapped phase data of the object by using numerical calculations. Three representative algorithms are selected to unwrap the interference pattern of ceils： branch-cut method, quality-guided method and network method. Although each of them can restore the phase, their performances are obviously different. We compare these methods and find that branch-cut method needs the smallest execution time and can obtain good unwrapped patterns when noises are not serious.

Multi-baseline extended particle filtering phase unwrapping algorithm based on amended matrix pencil model and quantized path-following strategy

This paper proposes a new multi-baseline extended particle filtering phase unwrapping algorithm which combines an extended particle filter with an amended matrix pencil model and a quantized path-following strategy. The contributions to multibaseline synthetic aperture radar(SAR) interferometry are as follows: a new recursive multi-baseline phase unwrapping model based on an extended particle filter is built, and the amended matrix pencil model is used to acquire phase gradient information with a higher precision and lower computational cost, and the quantized path-following strategy is introduced to guide the proposed phase unwrapping procedure to efficiently unwrap wrapped phase image along the paths routed by a phase derivative variance map.

Quality Guided Phase Unwrapping for Fringe Projection Profilometry

Fringe projection technique is a non-contact, full field 3-D shape measurement method. The object depth information is recorded in one or several deformed fringe patterns. The phase-shifting algorithm or the Fourier transform method can be used to extract the wrapped phase data. A phase unwrapping process is then applied to retrieve a continuous phase distribution, which represents the surface profile of the test object. In this paper, a quality-guided phase unwrapping approach is incorporated and two novel phase quality evaluation methods are proposed to facilitate the phase unwrapping process.

Investigation on Three Types of Residues in Phase-unwrapping

In phase unwrapping, how to control the residues and cut lines is critical to unwrap ill-state wrapped phase map successfully. Some new concepts in phase unwrapping field are proposed. Firstly, the residues in three types, i.e., singular residues, normal residues and virtual residues, are classified, which leads to some rational unwrapping paths. Secondly, some differences between cut-line and curved line are analyzed. And finally, experiment is carried out to compare our new enhanced Goldstein algorithm with the original Goldstein algorithm.

Phase Unwrapping in Magnetic Resonance Elastography

Phase Unwrapping (PU) is an ill-posed problem in Magnetic Resonance Elastography (MRE). The phase information is not usable until the phases are retrieved by using PU algorithms. In this present study, we attempt to determine the ideal PU method for MRE using both phantom and volunteer psoas major (PM) muscle images. All the MRE experiments were carried out in Philips MRI (Achieva 3.0 T, Best, The Netherlands). A multi-echo gradient-echo MRE pulse sequence was employed and the four PU methods were considered based on their easy user platform. They are namely, Minimum Discontinuity (MD), Laplacian-Based Estimate (LBE), Region Growing (RG) and Dilate-Erode (DE) Propagate. Phantom images were successfully unwrapped by all four methods, whereas MD and LBE could only unwrap PM muscle images properly. RG and DE failed to unwrap the PM muscle images.

InSAR Kalman Filter Phase Unwrapping Algorithm Based on SRTM DEM

PU （phase unwrapping） is the key step and important problem in DEM （digital elevation model） extraction and the measurement of surface deformation of InSAR （Interferometric synthetic aperture radar）. The CKFPUA （conventional Kalman filter phase unwrapping algorithm） can obtain reliable results in the flat terrain areas, but it caused error transmission not making the accurate inversion of surface deformation information in the steep terrain. Considering this situation, so it needs to introduce topographic information for guiding phase unwrapping. Here the 90 m resolution DEM data have been used and it is obtained by SRTM （shuttle radar topography mission） measured jointly by NASA （National Aeronautics and Space Administration） and NIMA （National Imaging Mapping Agency） of U.S. Department of Defense. This paper presents a SD-KFPUA （Kalman filter phase unwrapping algorithm） based on SRTM DEM. With SRTM DEM directing InSAR image to implement phase unwrapping, the speed and accuracy are improved. By analyzing with the conventional Kalman filter phase unwrapping algorithms, it is shown that the proposed method can achieve good results in particular to improve unwrapping accuracy in the low coherence region.

MCJ-UNet:一种双/多通道联合InSAR相位解缠网络

干涉合成孔径雷达(InSAR)可实现地表高程的高效获取,在地形测绘中应用广泛。双/多通道InSAR技术可借助不同通道(基线、频点)的高程模糊度差异,解决相位欠采样问题,完成高程陡变区域的干涉相位解缠,实现InSAR技术在测绘困难区域的有效应用。该文即面向高效高精度相位解缠需求,利用深度学习这一有力工具,结合不同通道的相位特征及相互约束关系,提出了一种双/多通道联合干涉相位解缠网络:Multi-Channel-Joint-UNet(MCJ-UNet)。该网络的构建以双通道(双频、双基线)InSAR为基本观测构型,并可实现向多通道构型的扩展,其构建的核心思路主要包括3点:首先,将干涉相位解缠中的模糊数估计问题转化为语义分割问题,并采用UNet网络完成分割处理;其次,引入挤压激励模块(SE)动态调整信息权重,以增强网络不同通道对其所需信息的感知能力;最后,利用多通道联合约束下的相位残差优化损失函数,实现网络调谐。此外,为避免语义分割结果的边缘细节误差对解缠效果的影响,该文还提出了一种基于多通道联合约束的解缠误差自修正方法,以保证解缠质量。模拟地形仿真数据、真实地形仿真数据以及TerraSAR-X实测数据验证了所提方法的有效性。

融合长短基线的干涉图解缠错误自动改正

合成孔径雷达干涉测量(interferometric synthetic aperture radar,InSAR)技术因具有全天时、全天候、大范围的独特优势,已被广泛应用于地球科学研究、自然灾害监测和环境管理等领域。然而,受地表失相干、干涉条纹密集等因素影响,InSAR数据处理关键的步骤——干涉图相位解缠常出现错误,进而影响高程反演和地表形变监测精度。为此,文章提出了一种融合长短基线干涉相位闭合环的解缠错误自动探测与改正方法,并利用湖南省-江西省高植被覆盖区的Sentinel-1雷达影像开展了算法验证。实验结果表明,该方法可有效地改正干涉相位解缠错误,这对InSAR技术提升高程反演和形变监测精度具有重要意义。

改进的三频四步相移面结构光三维重建算法

针对面结构光中传统三频四步相移算法投影图片过多、效率较慢这一问题,提出了改进的三频四步相移面结构光重建算法。改进的算法需要3个高频率的正弦条纹图,但所需投影的条纹数量从传统算法的12张减少到了7张,最高频率的条纹图4张,次高频和最低条纹图分别为2张和1张,投影效率比传统算法提高了41.7%。介绍了改进算法的原理,将最高频率4张图所计算的包裹相位用来三维重建,其余两个频率条纹图计算的包裹相位用来进行相位展开,理论上该方法与传统算法精度一致。最后,通过实验验证了该算法的重建精度以及对具有复杂表面物体的三维重建能力。实验结果表明改进的面结构光算法在提高投影效率的同时与传统算法保持着一致的精度。

基于改进U-Net网络的相位解包裹技术研究

提出了一种结合深度学习的空间相位解包裹方法,采用基于改进U-Net网络的编码器-解码器架构,同时加入包含双向长短期记忆网络(BILSTM)的CBiLSTM模块,并且结合注意力机制,避免了典型卷积神经网络学习全局空间依赖关系的固有缺陷的同时增强了深度学习模型对相位解包裹任务中的关键信息的关注能力。通过大量的模拟数据,验证了文中方法在严重噪声(SNR=0)、不连续条件和混叠条件下的鲁棒性,在以上三种情况下,同其他深度学习网络模型进行对比,文中所提出的网络模型的归一化均方根误差(NRMSE)分别为0.75%、1.81%和1.68%;结构相似性指数(SSIM)分别为0.98、0.92和0.94;峰值信噪比(PSNR)分别为40.87、32.56、37.38;同时计算时间显著减少,适合应用到需要快速准确的空间相位解包裹任务中去。通过实际测量数据,验证了文中提出网络模型的可行性。该研究将双向长短期记忆网络(BILSTM)和注意力机制同时引入光学相位解包裹问题中,为解决复杂相位场的解包裹提供了新的思路和方案。

InSAR相位解缠最大流/最小割权值改进算法

InSAR已经在高精度DEM反演中得到了广泛应用,相位解缠技术是影响DEM获取精度的关键步骤之一,然而大梯度变化区域一直是影响解缠结果精度的核心问题。针对该问题,本文提出一种基于势函数改进权值的最大流/最小割相位解缠算法。首先,针对PUMA模型权值设置不合理的问题进行研究,利用外部已有DEM获取相位梯度变化的先验信息,将窗口最大绝对相位梯度值代入相应的势函数公式中获取权值;然后,通过对势函数权值的设置进行阈值调节,解决由于势函数权值设置不合理导致PUMA势函数无法发挥作用引起的解缠错误问题,进而提高大梯度变化区域相位解缠精度;最后,通过仿真数据和真实TanDEM-X InSAR数据对本文算法进行验证,并与已有方法进行对比。结果表明,仿真数据中本文算法可以提高至少44.93%的解缠精度,真实数据中在大梯度变化区域,本文算法较已有算法可以获得更大范围的有效解缠结果。

基于空洞空间卷积网络的点衍射干涉图像相位解包技术

为满足点衍射干涉测量对解包算法高精度、高效以及抗干扰的检测需求。提出一种基于空洞空间卷积的相移点衍射干涉图像的相位解包方法,通过将自编码器结构和空洞空间卷积结合获得更高的相位解包精度,实现对包裹相位图像可控的多尺度特征提取。根据点衍射图像特点制作的大量多样化数据集对其进行训练和优化,从而实现准确识别包裹相位所在阶次,最终可以快速处理包裹图像得到高精度的解包结果。利用所提方法对实际点衍射干涉图像进行处理,并与ESDI专业干涉图像处理软件以及其他解包算法处理结果进行比对,结果表明:本文解包结果与软件枝切法解包处理结果均方根误差值为0.0222 rad,面形拟合结果与软件面形拟合结果峰谷差值仅为0.0121λ、均方根差值仅为0.0042λ;时间效率上,完成一幅图像的处理平均仅需0.035 s,而传统方法均大于1 s。与其他方法相比,所提方法在处理包裹相位方面具有快速、高精度的特性,为点衍射干涉图像处理的高精度相位解包提供了新的可行方案。

合成孔径声呐时延估计误差校正改进算法

由相位解缠错误造成的时延估计误差会严重降低合成孔径声呐时延估计的精度。现有的时延估计误差校正算法采用二次函数作为时延的拟合模型,该模型在距离向近端和远端处不符合时延空变规律,拟合误差较大且无法估计载体的横荡和升沉运动。针对该问题,文章提出一种改进的时延估计误差校正算法,利用时延的距离空变函数代替二次函数作为拟合模型。数值仿真和实验结果表明,相较于参考算法,改进算法校正时延估计误差的效果更好、速度更快,同时还能准确地估计载体的横荡和升沉运动。运动补偿结果显示了改进算法能较好地提升成像质量。

Segmentation-aided phase unwrapping for 3D reconstruction with airborne array-InSAR images

Unmanned aerial vehicle(UAV)array InSAR is a new type of single-flight 3D SAR imaging system with the advantages of high coherence and resolution.However,due to the low altitude of the platform,the elevation ambiguity of the system is smaller than the maximal terrain elevation.Since the spatial phase unwrapping is conducted based on the assumption of phase continuity,the inappropriate ambiguity will cause significant unwrapping errors.In this work,a 3D phase unwrapping algorithm assisted by image segmentation is proposed to address this issue.The Markov random field(MRF)is utilized for image segmentation.The optimal spatial phase unwrapping network is achieved based on the segmentation results.The asymptotic 3D phase unwrapping is adopted to get the refined 3D reconstruction.Results based on the real airborne array-InSAR data show that the proposed method effectively improves the elevation ambiguity.