A deep-learning (DL)-based data calibration technique applied to quantitative microwave inverse-scattering analysis is presented. This technique aims at subsurface inspection for the buried object under a concrete road or soil. The inverse-scattering analysis provides a complex permittivity profile, which is useful for object identification such as air gap or water. Contrast source inversion (CSI) is one of the most promising inverse-scattering methods. This method is capable of avoiding the iterative use of highly computational forward solvers. However, when applied to the measured data, an appropriate calibration capable of converting measured data to simulation data is required. In this work, a DL-based calibration suitable for nonlinear inverse problems is proposed. Its efficiency is experimentally demonstrated using a concrete cylinder containing water with different salinities.

Super-resolution subsurface imaging based on sparse regularization is presented in assuming the terahertz (THz) band multilayer structure analysis. The THz wave subsurface imaging with$\mu $-scale spatial resolution and penetration depth are promising for several applications, such as nondestructive testing and chemical/biomedical compound analyses. The sparse regularization-based compressed sensing (CS) approach has considerable potential to provide super-resolution subsurface imaging in a time-of-flight estimation. However, using optical lens-based measurements, e.g., THz time-domain spectroscopic (THz-TDS) systems, a depth resolution is highly dependent on the depth of each layer, which becomes more critical in the out-of-focus case. This study demonstrated that the above depth-dependence could be solved by using an appropriate depth-dependent reference signal, by using the THz-TDS measured data.

A complex permittivity profile reconstruction for multilayered objects was presented in this study by incorporating compressed sensing (CS)-based thickness estimation with contrast source inversion (CSI) non-linear inverse scattering (IS) method using a terahertz (THz) frequency band. Several studies investigated permittivity estimation for multiple layers. However, they require a prior knowledge of the thickness of each layer. Moreover, a critical problem in this field is the simultaneous estimation of both the dielectric constant and the thickness of each layer. To address this, a super-resolution thickness estimator using a CS filter and the CSI-based dielectric profile reconstruction scheme was used. This problem was effectively solved by introducing the cost function estimated using the CSI scheme, where the number of layers is given. The finite-difference time-domain (FDTD) numerical test indicated that the proposed method provides an accurate estimation of the thickness and dielectric profile in double-layered objects.

This letter presents an auto-focusing imaging algorithm for terahertz (THz) band imaging by extending the range point migration (RPM) method using a dielectric lens model. The traditional THz time-domain spectroscopy (THz-TDS) imaging system has an essential problem for depth dependence of the azimuth resolution, which becomes more crucial in subsurface imaging, because the depth of target is unknown in most cases. This letter thus introduces an RPM-based auto-focusing imaging method using an equivalent virtual source model to accurately represent both near- and far-range areas from the focal point. Experimental validation using THz-TDS demonstrated that the proposed algorithm accurately compensates image distortions by out-of-focus effect with much lower complexity than traditional radar-imaging approaches.

In this article, we present a novel method that incorporates the range points migration (RPM) method,$k$-space decomposition-based accurate, and noise-robust range extraction filter for microwave or millimeter-wave (MMW) short-range radar using a considerably lower fractional bandwidth signal. The advantage for higher angular resolution in higher frequency systems, such as MMW radar, has been implemented to the incoherent-based RPM method, using the simple 1-D or 2-D Fourier transform-based processing to maintain the imaging accuracy in RPM processing for both the range and the angular directions. As an additional advantage of our method, it also offers data clustering in$k$-space, which can enhance the imaging accuracy of the RPM method. The numerical and experimental tests demonstrated that the proposed method offers numerous advantages over the Capon-based super-resolution algorithm or coherent-based imaging approaches.

Super-resolution and highly accurate Doppler and imaging algorithm for microwave or millimeter wave (MMW) short range radar is presented here. As human body detection or recognition in self-driving, through-the-wall imaging, or security system, the micro-Doppler analysis is one of the promising approaches. In the previous study, we have proposed an innovative Doppler velocity estimation, named as weighted kernel density (WKD) estimator, which simultaneously achieves higher temporal and Doppler velocity resolutions. This paper focuses on bi-directional processing between the WKD based Doppler velocity estimator and the range points migration (RPM) based radar imaging, so that both estimation accuracies could be enhanced. 3-D Numerical simulation demonstrate the effectiveness of our method.

Millimeter wave (MMW) radar is highly expected to environmentally robust automotive sensor, especially in optically blurred vision. Non-line-of-sight (NLOS) sensing for human detection in automotive radar is another distinct feature of millimeter wave, where a diffraction effect would be exploited to detect an unique signal of human body characterized by respiration or attitude control. In this paper, the machine learning based recognition algorithm is introduced to deal with a diffraction signal of human body in NLOS situation. Several feature extraction schemes are implemented in support vector machine (SVM) recognition to address with lower signal-to-noise ratio (SNR) problem. The experimental data, using MMW radar in NLOS case, show the effectiveness for the use of diffraction signal to discriminate human body from other objects.

Lower band ultrawideband (UWB) Doppler radar is promising for through-wall imaging, e.g., human body detection in rescue scenarios. The inherent problem with pulse-Doppler radar is the tradeoff between the Doppler velocity resolution and the resulting temporal resolution that makes it difficult to conduct real-time target tracking, because the separation of micro-Doppler velocities of the human body requires a higher Doppler velocity resolution. This problem is particularly severe for lower band UWB radar systems, which are required to attain a sufficient penetration depth in concrete material in the through-the-wall imaging scenario. Because UWB signals generally have large fractional bandwidths, the reflected pulse is located over a range gate along the slow-time direction; this is well known as the range walk problem. As a promising solution to this problem, this article newly introduces a technique for a super-resolution Doppler velocity estimation algorithm based on Gaussian kernel density estimation, which converts observed range-τ points to Doppler-associated ranges. In addition, this approach makes an important contribution for super-resolution range extraction with a compressed sensing (CS) filter, which is combined with the range-point migration (RPM) method for human body imaging associated with micro-Doppler components. 2-D or 3-D numerical simulations, including human body imaging scenario, demonstrate that the proposed method allows both accurate Doppler velocity estimation and human body imaging, which can be updated at the pulse-repetition interval.

Millimeter wave short range radar is a great potential for monitoring sensor being applicable to optically challenging environments (e.g., smog, fog or darkness). While a number of imaging algorithms have been intensively developed, the range points migration(RPM) method is regarded as one of the most promising options to achieve accurate and fast three-dimensional imaging. However, since the RPM is based on incoherent processing, it has disadvantage for angular resolution in considering millimeter wave radar with lower fractional bandwidth. To address with the above problem, this paper proposes a k-space decomposition based RPM method, where the received data is converted and decomposed into the k-space to retain the desired angular resolution. As a notable feature of the proposed method, it offers a clustered range points decomposed by k-space, which contributes to enhance an imaging accuracy by the RPM. The two-dimensional and three-dimensional numerical tests demonstrate the effectiveness of our proposed method, compared with other super resolution algorithms.

This paper focuses on super-temporal resolution velocity vector estimation in the through-the-wall imaging (TWI) scenario, for human body detection in rescue or crime scenes. In the pulse-Doppler radar system, there is trade-off between Doppler velocity and temporal resolution, and it leads difficulty for real-time target tracking for motion vector, especially for in the lower frequency band ultra-wideband (UWB) radar system, which is required in TWI radar to maintain an enough penetration depth in concrete material. Since general UWB signal has larger fractional bandwidth, then reflection pulse is located over a range gate along slow-time direction, that is known as the range walk (RW) effect. As a substantial solution for this problem, this paper introduces the Gaussian-kernel based Doppler velocity estimation for motion estimation by exploiting the range points migration (RPM) method, where instantaneous motion vector of target can be calculated by least square approach. The finite-difference time-domain (FDTD) based numerical simulation demonstrates that our proposed method achieves accurate velocity vector estimation updated with pulse repetition interval.

High-resolution, short-range sensors that can be applied in optically challenging environments (e.g., in the presence of clouds, fog, and/or dark smog) are in high demand for various applications. Ultrawideband radar is a promising sensor that is suitable for short-range surveillance or watching sensors. Range-point migration (RPM) has been recently established as a promising imaging approach to achieve accurate and real-time 3-D imaging. However, when objects with many scattering points are dealt with, such as a human body, RPM suffers from high computational costs. In this letter, we propose an algorithm with a lower complexity for an RPM-based 3-D imaging method by introducing a sampling-based scattering center extraction with a simplified evaluation function, in which an efficient sample pattern is provided by a golden ratio. The results from a finite-difference time-domain-based numerical test, which introduces a realistic human body object, demonstrate that our proposed method remarkably reduces the computational cost without sacrificing the reconstruction accuracy.

Ultrawideband radar with high-range resolution is a promising technology for use in short-range 3-D imaging applications, in which optical cameras are not applicable. One of the most efficient 3-D imaging methods is the range-point migration (RPM) method, which has a definite advantage for the synthetic aperture radar approach in terms of computational burden, high accuracy, and high spatial resolution. However, if an insufficient aperture size or angle is provided, these kinds of methods cannot reconstruct the whole target structure due to the absence of reflection signals from large part of target surface. To expand the 3-D image obtained by RPM, this paper proposes an image expansion method by incorporating the RPM feature and fully polarimetric data-based machine learning approach. Following ellipsoid-based scattering analysis and learning with a neural network, this method expresses the target image as an aggregation of parts of ellipsoids, which significantly expands the original image by the RPM method without sacrificing the reconstruction accuracy. The results of numerical simulation based on 3-D finite-difference time-domain analysis verify the effectiveness of our proposed method, in terms of image-expansion criteria.

High-resolution, short-range sensors that can be applied in optically challenging environments (e.g., in the presence of clouds, fog, and/or dark smog) are in high demand. Ultrawideband (UWB) millimeter-wave radars are one of the most promising devices for the above-mentioned applications. For target recognition using sensors, it is necessary to convert observational data into full 3-D images with both time efficiency and high accuracy. For such conversion algorithm, we have already proposed the range points migration (RPM) method. However, in the existence of multiple separated objects, this method suffers from inaccuracy and high computational cost due to dealing with many observed RPs. To address this issue, this letter introduces Doppler-based RPs clustering into the RPM method. The results from numerical simulations, assuming 140-GHz band millimeter radars, show that the addition of Doppler velocity into the RPM method results in more accurate 3-D images with reducing computational costs.

Through-the-wall radar (TWR) technique with UWB (ultra wide-band) signals are promising candidates for nondestructive testing or reliable human detection buried under collapsed walls in disaster scenes. As an efficient 3-dimensional imaging approach, the range points migration (RPM) method has been established, which can transcend the performance limitation of the conventional delay-and-sum approaches, in terms of computational burden, accuracy and spatial resolution. This paper extends the RPM method to the TWR imaging model, called as TW-RPM, where the principle of the original RPM is appropriately extended to this model by considering the propagation path and delay in wall. In addition, this paper introduces the image expansion scheme, based on ellipse fitting approach. As a notable point, this fitting process is not carried out in real space but in data space, which is spanned by range points (a set of antenna location and range) to avoid the extrapolating error due to TW-RPM imaging process. The FDTD (finite time domain difference) based numerical simulation shows that our proposed method accurately expands the TW-RPM image even in narrower aperture case.

Satellite-borne or aircraft-borne synthetic aperture radar (SAR) technique is useful for high resolution imaging analysis for terrain surface monitoring or surveillance, even in optically harsh environment. For surveillance application, there are various approaches for automatic target recognition (ATR) of SAR images aiming at monitoring unidentified ships or aircrafts. In addition, various types of analyses using full polarimetric data have been developed recently because it can provide significant information to identify structure of targets, such as vegetation, urban, sea surface areas. In this paper, the circular polarization basis conversion is adopted to improve the robustness especially to variation of target rotation angles. The experimental data, assuming the 1/100 scale model of X-band radar, demonstrate that our proposed method significantly improves an accuracy of target area extraction and classification, even in noisy or angular fluctuated situations.

Synthetic aperture radar (SAR) is one of the most powerful tools for microwave imaging issue, being applicable to terrain surface measurement regardless of the weather conditions. Recently, the coherent change detection (CCD) method has been widely developed aiming at surface change detection by comparing the plural complex SAR images with the same scanning orbit. However, in the case of a general damage assessment by an earthquake or a mudslide, it requires not only a change detection but also a height change quantity of changed surface. To address with this issue, this paper proposes a novel height change estimation method with a CCD model based on the Pauli decomposition of fully polarimetric band-divided SAR images. The experimental results, assuming the 1/100 scale down model of the X-band SAR system in anechoic chamber, show that the proposed method achieves more accurate height change estimation, compared with those obtained by the method using single polarimetric data.

The microwave imaging technique, especially for synthetic aperture radar (SAR), has significant advantages in providing high-resolution complex target images, even in darkness or adverse weather conditions. Nevertheless, it is still difficult for human operators to identify targets on SAR images because they are generated using radio signals with wavelengths at the order of cm. To deal with this, various approaches for efficient automatic target recognition (ATR), based on neural networks or support vector machines (SVM), have been developed. Previously we proposed a promising ATR method using a supervised self-organizing map (SOM), where a binarized SAR image is accurately classified by exploiting the unified distance matrix (U-matrix) metric. Although this method enhances ATR performance considerably, even with SAR images heavily contaminated by random noise, the calculation burden is enormous under expansions of scale and then cannot maintain the ATR performance, especially in cases with azimuth angle variations. In this letter, we propose a constrained learning scheme for generating the SOM and introduce the A-star algorithm to handle SOM scale expansion. Experimental investigations demonstrate the effectiveness of our proposed method.

Synthetic aperture radar (SAR) is an indispensable tool for low visibility ground surface measurement such as in the case of adverse weather or darkness. In recent years, the coherent change detection (CCD) technique has been recently established, which detect a temporal change in the same region according to the phase interferometry of two complex SAR images. However, in the case of general damage assessment following an earthquake or mudslide, the technique requires a height change quantity of the changed surface, such as for a building collapse or road subsidence. To address this issue, this paper proposes a novel height estimation method by exploiting the frequency characteristic of coherence phases obtained using each multiple divided SAR image. The results obtained from experimental data demonstrate that our proposed method offers accurate height change estimation while avoiding degradation of spatial resolution.

Ultrawideband pulse radar has a definite advantage over optical ranging techniques in harsh optical environments, such as a dark smog or strong backlight. In security or rescue situations with blurry visibility, it is particularly promising for identifying human bodies. One of the most promising approaches for this type of application is the recently proposed range point migration (RPM) method, which is beneficial for nonparametric imaging and is robust in noisy or heavy interference situations. However, the original RPM requires a discretization of the direction-of-arrival variables in its search operation. The resulting coarse discretization seriously degrades the imaging accuracy, particularly for 3-D problems and far-field observations. Consequently, in this approach, there is a major tradeoff between the amount of computation and accuracy. To overcome this difficulty, this letter proposes a more efficient RPM method, where the extraction of the point of intersection of spheres is adopted. A distinct advantage of this method is that the accuracy is basically invariant to the observation range when avoiding the aforementioned discretization. Numerical simulations including noisy cases prove that our proposed RPM significantly reduces the computation complexity while retaining imaging accuracy.

The 3-D reconstruction radar techniques are required for coast-guard patrols. While many 3-D reconstruction methods exploiting ISAR images have been proposed, they assume only pointwise targets. For objects with continuous boundaries, such as a wire, their accuracy is, however, severely degraded owing to scattering centers shifting on target surface. Solving this difficulty, this paper extends RPM (Range Points Migration) method to ISAR model. Numerical and experimental validations demonstrate that our method accomplishes accurate 3-D imaging even for non-point targets.

Ultrawideband pulse radars are a promising technology for high-quality imaging sensors for rescue robots in the near field because they have the advantage of high range resolution. We have already proposed the accurate and fast imaging algorithm as range point migration (RPM), which employs the direction of arrival (DOA) with the global characteristic of the multiple observed ranges. However, this algorithm assumes the line scanning of an omnidirectional antenna and limits the imaging range. To overcome this limitation, this letter derives an extended RPM method, which accomplishes omnidirectional imaging and accurate target positioning. As a false-image-reduction scheme in this method, this letter introduces a postprocessing algorithm, which constrains the searching range for DOA estimation by using the initial RPM image. The results of numerical simulations, including noisy situations, show that the proposed method accomplishes accurate and omnidirectional imaging on the order of 1/100 wavelength and enhances the imaging range while avoiding false images.

UWB (Ultra Wideband) radars have great promise for near field sensing systems, holding its high range resolution. It is particularly suitable for robotic or security sensors that must identify a target in optically blurry visions. Some recently developed radar imaging algorithms proactively employ multiple scattered components, which can enhance an imaging range compared to synthesizing a single scattered component. We have already proposed the SAR (Synthetic Aperture Radar) method considering a double scattered, which successfully expanded a reconstructible range of radar imagery with no preliminary knowledge of target or surroundings. However, this method requires an intensive computation and its spatial resolution is insufficient for clear boundary ex traction such as edges or specular surfaces. As a substantial solution, this paper proposes a novel shadow region imaging algorithm based on a range derivative of double scattered signals. This new method accomplishes high-speed imaging, including a shadow region without any integration process, and enhances the accuracy with respect to clear boundary extraction. Some results from numerical simulations verify that the proposed method remarkably decreases the computation amount compared to that for the conventional method, enhancing the visible range of radar imagery.

Ultrawideband (UWB) pulse radar with high range resolution is suitable for near-field sensing. Applications of UWB pulse radar include human body identification in blurry vision for security or rescue purposes and accurate spatial measurements for industrial products such as a reflector antenna. The synthetic aperture radar is still promising for these applications because it creates an accurate image even for near-field targets in free space. However, for complex-shaped or multiple objects, this algorithm suffers from increased shadow region because it employs only a single-scattered signal for imaging. To resolve this difficulty, this paper proposes a novel imaging algorithm based on aperture synthesis for double-scattered signals. In general, double-scattered waves include independent information on target points, which are not obtained by a single-scattered wave. Based on this principle, the proposed method effectively synthesizes the double-scattered signals and enhances the reconstructible range of a target shape, part of which becomes a shadow in the former approach. In order to enhance accuracy, a false image suppression approach based on the Fresnel zone theory is also incorporated in the proposed method. The results from numerical simulations and an experiment verify that our method significantly enhances the visible range of target surfaces without either a priori knowledge of target shapes or preliminary observation of their surroundings.

Ultra-wide band (UWB) pulse radar has a definite advantage over optical ranging techniques, as to applicability to the harsh optical environment, such as the dark smog, or strong back-light. We have already proposed the extended Synthetic Aperture Radar (SAR) algorithm employing the multiple scattered waves, which aims at enhancing the reconstructible region of the target boundary including the shadow. However, it still suffers from the shadow region in the case of the target with a sharp inclination or deep concave boundary, because it assumes the antenna scanning whose real aperture size is too small. To resolve this difficulty, this paper proposes an extension algorithm using the array antenna model. While this extension is quite simple, the effectiveness of the proposed method is nontrivial regarding to the expansion of the imaging range. The results from numerical simulations verify that our method remarkably enhances the visible range of target surfaces without a priori knowledge of target shapes or a preliminary observation of its surroundings.

Ultrawide-band pulse radars have immeasurable potential for a high-range-resolution imaging in the near field and can be used for noncontact measurement of industrial products with specular or precision surfaces, such as reflector antenna or aircraft fuselage, or identifying and locating the human body in security systems. In our previous work, we developed a stable and high-speed 3-D imaging algorithm, Envelope, which is based on the principle that a target boundary can be expressed as inner or outer envelopes of spheres, which are determined using antenna location and observed ranges. Although Envelope produces a high-resolution image for a simple shape target that may include edges, it requires an exact connection for observed ranges to maintain the imaging quality. For complex shapes or multiple targets, this connection becomes a difficult task because each antenna receives multiple echoes from many scattering points on the target surface. This paper proposes a novel imaging algorithm without range point connection to accomplish high-quality and flexible 3-D imaging for various target shapes. The algorithm uses an accurate estimation for the direction of arrival using signal amplitudes and realizes direct mapping from observed ranges to target points. Several comparative studies of conventional algorithms clarify that our proposed method accomplishes accurate and reliable 3-D imaging even for complex or multiple boundaries.

Abstract—Ultrawideband pulse radars have a great potential for high-range resolution in near field imaging and can be used for noncontact measuring in precision or specular products such as reflector antennas and aircraft fuselages. We have already proposed a high-speed 3-D imaging algorithm, SEABED, which is based on a reversible transform, which is the boundary scattering transform, between the received signals and the target shape. However, the estimated image with SEABED is unstable with random noise because it utilizes a derivative of the received data. In this paper, we propose a robust 3-D imaging algorithm with an envelope of spheres that completely resolves the instability due to derivative operations. Moreover, to enhance the resolution of estimated images, this method is combined with a direct waveform compensation method that does not sacrifice high-speed calcula-tion. Numerical simulations and an experiment confirm that the proposed method can realize fast, robust, and high-resolution 3-D imaging for arbitrary targets. Index Terms—Direct waveform compensation, envelope of spheres, high-resolution and fast 3-D imaging, scattered waveform deformation, ultrawideband (UWB) pulse radars. I.

An imaging algorithm with a transform between the real and data spaces was proposed by Greenhalgh and Marescot for georadar data in 2006. This technique utilizes a reversible transform between the real space (X, Z = F(X)) and the data space (x, z = f(x)). The inverse transform is equivalent to the imaging method that is proposed by Li et al. (2005) if an antenna interval is short. This method was applied to preprocessing for breast cancer detection in 2005. These transforms were originally proposed by Sakamoto and Sato (2004) for an imaging with ultrawideband radar systems. By utilizing these transforms, a high-speed imaging algorithm, which is called as SEABED algorithm, was developed, which was extended to compensate for the phase rotation at caustic points, which was also described by Greenhalgh and Marescot. In addition, the SEABED algorithm was extended to apply to noisy data, 3D systems, experimental data, and bistatic radars. Note that the transform is fundamentally sensitive to noise because it includes derivative operations. A new algorithm was developed by extending the SEABED algorithm to avoid the derivative operations, which is stable and has a high resolution even for noisy data. The reversible transform in is now simultaneously and independently studied by some research groups because it is the sole solution for the imaging with wave fields. In addition, the transform has a variety of applications because it can be applied to electromagnetic waves, sonic and ultrasonic waves, seismic waves, and other waves.

Target estimation methods with UWB pulse signals are promising as imaging techniques for household or rescue robots. We have already proposed an efficient al-gorithm for shape estimation, SEABED (Shape Estimation Algorithm based on BST and Extraction of Directly scattered waves), which is based on a reversible transform BST (Boundary Scattering Transform) between the time delay and the target shape. In this method, we determine quasi wavefronts from received signals with the matched filter of the transmitted waveform. However, the scattered wave-form is in general different from the transmitted one depending on the shape of targets. These differences cause estimation errors in SEABED method. In this paper, we propose a high-resolution algorithm for general convex targets based on the scattered waveform estimation, and evaluate the method by numerical simulations.