Crevasse detection is crucial for glacier and climate research, and provides essential guidance for activities in glacier regions. Ground-penetrating radar(GPR) and machine learning are used to automatically detect crevasse. In this study, a Siam-Gabor-ResNet deep learning framework is proposed to detect crevasse automatically using GPR data. A contrast learning with Siamese network framework is proposed to improve the accuracy of crevasse detection, which aims to increase the feature similarity between crevasses while simultaneously enhancing the feature distinctiveness between crevasse and continuous snow layers. Additionally, a trainable Gabor-ResNet feature extraction module is built by integrating the Gabor filter bank into the ResNet network and used to further reduce the complexity of model training while extracting multi-scale features in a real-time manner. Experiments are performed on the Greenland dataset and the 2015 McMurdo dataset, which illustrate the effectiveness of the proposed method. The average accuracy rate of crevasse detection reaches 94.38%, which can detect the narrowest crevasse (0.6 meters) in two datasets, with an average detection time of only 6.8 milliseconds. Experimental results show that the proposed method can detect crevasse in real-time, automatically, and accurately.

To achieve efficient ice sheet detection, this study develops a frequency-modulated continuous wave (FMCW) ice sounding radar imaging algorithm. The proposed algorithm represents an improved range-Doppler algorithm (RDA) that can effectively handle multi-layer medium imaging models and address the range migration variations caused by multi-layer media. The theoretical analysis is conducted to explain the imaging principle of the proposed algorithm, and a detailed description of its processing flow is provided. Moreover, point target simulation experiments and actual data processing tests are performed to demonstrate the effectiveness of the proposed algorithm in ice sheet imaging. The proposed algorithm is also compared with the frequency scaling algorithm (FSA), which is a commonly used FMCW ice sounding radar imaging algorithm, in terms of signal-to-noise ratio (SNR) and imaging time. The comparison results show that the proposed algorithm can improve imaging efficiency while maintaining accuracy.

This article proposes a three-dimensional (3D) human pose estimation method using through-wall radar (TWR) systems, which extends and supplements new applications in the era of the Internet of Things (IoT). TWR system can penetrate non-metallic obstacles and perceive wall-occlusive human targets, but the physical characteristics of radio frequency (RF) signals, such as poor imaging resolution and specularity effect, make the pose estimation process highly ill-posed. In this work, we propose a mutual information maximizing deeply-supervised network (MIMDSN), which aims to extract accurate and robust 3D human skeletons from TWR images. Inspired by past works, an optical system is attached to the TWR system to provide cross-modal pseudo labels. Based on a depth design philosophy of convolutional neural networks that meets radar resolution constraints, we design a resolution-guided pose estimation network for keypoint coordinate regression. To alleviate the ill-posed problem, supervising solely the network output is insufficient. The cross-modal supervision is not only built on predictions, but also on features of the network’s hidden layer. With the help of information theory, the mutual information between features and pseudo labels is maximized for feature alignment and discriminability enhancement. Experiments show competitive performance against state-of-the-art RF-based human pose estimation methods and can reconstruct accurate 3D skeletons in multi-target, low-visibility, and wall-occlusive scenes.

The rational Krylov subspace methods can improve the computational speed compared to conventional time-stepping approaches for calculating 3-D transient electromagnetic (TEM) method forward modeling. However, the rational Krylov subspace method simulates only the step-off response. Because primary source waveforms have nonnegligible effects on the induced responses, it is crucial to model the response induced by any given source waveform. The electric field (e) and the time derivative of the magnetic induction ($\mathrm {d} {\mathbf { b}} / \mathrm {d}t$) are commonly measured TEM responses. Case studies also show the magnetic induction ($\bf b$) response measured by magnetometers has a good resolution for exploring conductive mineral deposits. Therefore, modern TEM forward modeling algorithms should be able to simulate different types of responses. We present a new algorithm for TEM modeling using the rational Krylov subspace method. The following improvements are implemented in our approach: 1) the algorithm can efficiently compute the e and$\mathrm {d} {\mathbf { b}} / \mathrm {d}t$responses, and especially the$\bf b$response, which was less considered in other 3-D TEM studies; 2) a convolution approach is employed that allows the Krylov subspace method to simulate the source waveform effects on all three types of responses; and 3) we present the approach for computing the initial condition of b in cases of using galvanic sources. This work extends the flexibility of existing 3-D TEM modeling algorithms. Numerical examples demonstrate that the new algorithm is accurate and computationally efficient.

Among the inversion methods for airborne transient electromagnetic (ATEM) data, the hybrid inversion method integrates the iterative optimization framework with artificial neural networks (ANNs), ensuring inversion accuracy while enhancing the generalization capability of neural networks. However, this method faces challenges in terms of slow computation speeds due to its lower updated step length and the lack of consideration for electromagnetic response laws. Our method adopts a supervised descent method (SDM) framework to supervise the ANNs, obtaining a longer updated step length. On the basis of the SDM framework, we have considered the electromagnetic response laws and designed RNN-ResNet and 1-D-UNet networks to update the conductivity model, improving the computing speed. Through numerical ablation experiments, we validated the effectiveness of our proposed method and compared the inversion results with those obtained using the traditional hybrid method. Additionally, we conducted tests on bundle fringe distribution, inversion fitting loss, noise sensitivity, and inversion speed for both methods using measured data to evaluate their performance in practical applications. The experimental findings demonstrate that our method achieves the same level of inversion accuracy and generalization ability as the traditional hybrid method while enhancing inversion speed by up to 68.5%.

For fine-grained human perception tasks such as pose estimation and activity recognition, radar-based sensors show advantages over optical cameras in low-visibility, privacy-aware, and wall-occlusive environments. Radar transmits radio frequency signals to irradiate the target of interest and store the target information in the echo signals. One common approach is to transform the echoes into radar images and extract the features with convolutional neural networks. This article introduces RadarFormer, the first method that introduces the self-attention (SA) mechanism to perform human perception tasks directly from radar echoes. It bypasses the imaging algorithm and realizes end-to-end signal processing. Specifically, we give constructive proof that processing radar echoes using the SA mechanism is at least as expressive as processing radar images using the convolutional layer. On this foundation, we design RadarFormer, which is a Transformer-like model to process radar signals. It benefits from the fast-/slow-time SA mechanism considering the physical characteristics of radar signals. RadarFormer extracts human representations from radar echoes and handles various downstream human perception tasks. The experimental results demonstrate that our method outperforms the state-of-the-art radar-based methods both in performance and computational cost and obtains accurate human perception results even in dark and occlusive environments.

Through-wall radar (TWR) can image the target of interest and capture the human sensing information. However, the poor human interpretability of TWR images and the lack of effective supervision make the extraction of complete body contour intractable. This letter proposes dual UNet, an unsupervised human contour extraction method for TWR images. Specifically, the method adopts two UNets with the same structure. One serves as the encoder to convert the TWR images into the latent representation. Another serves as the decoder to reconstruct the latent representation into the original images. Reconstruction loss and smooth normalized cut loss are optimized together to offset the dependence on labels and supplement global segment constraints. After training and post-processing, the latent representation can be used as the result of contour extraction. Experimental results show that dual UNet stands out among unsupervised human contour extraction methods in both free space and wall-occlusive scenarios, opening the possibility of learning useful human sensing information from raw TWR images without manual annotations.

The transient electromagnetic (TEM) method utilizes artificial transmitters and measures electromagnetic (EM) responses to reveal the resistivity information of the subsurface. The current waveform of transmitters has nonnegligible effects on induced fields. Therefore, 3-D TEM forward modeling algorithms need the capability of simulating arbitrary waveforms to obtain accurate responses. In time-stepping-based 3-D TEM forward modeling, the source term (ST) approach is frequently used, which employs the source current density to model the waveform variation during time-stepping. The ST approach, however, requires fine-time discretization to describe complex waveforms, which could significantly raise the computational cost. We present a robust convolution (Conv) approach that computes the convolution between the time derivative of the waveform and the step-off response to incorporate the waveform effects in 3-D TEM modeling. The Conv approach does not discretize the waveform using time steps. Hence, it is advantageous when modeling full-waveform cases. The developed algorithm is based on the finite-element (FE) method using unstructured grids and the implicit backward Euler approach. Both galvanic and inductive transmitters are incorporated. Ground and airborne TEM surveys are tested using an actual airborne TEM waveform, a full waveform of the$2^{(n)}$-sequence pseudorandom signal, and various synthetic waveforms. Accuracy is validated against the 1-D and 3-D solutions of published studies. The ST and Conv approaches are compared. Synthetic examples show that the latter approach simplifies the waveform incorporation in TEM modeling and substantially improves time-stepping efficiency without sacrificing accuracy.

The 3D object detector usually uses a framework similar to 2D detection and benefits from the advancements of 2D detection tasks. In these frameworks, it is necessary to make the unstructured, sparse point cloud features into dense grids to be compatible with popular 2D operators such as convolution and transformers, which also causes extra computational costs. In this paper, we propose a simple and efficient Fully Sparse TRansformer (FSTR) for LiDAR-based 3D object detection, which is able to combine with state-of-the-art sparse backbones to form a fully sparse, end-to-end, simple, and efficient detection framework. FSTR uses the sparse voxel feature from the sparse backbone as the input token without any custom operators. Further, we introduce the dynamic queries to provide a priori location and context of the foreground for the decoder and drop the high-confidence background tokens to further reduce redundant computations. We propose Gaussian denoising queries to speed up the decoder training and make it more adaptable to the distribution of sparse voxel features. Extensive experiments on the nuScenes benchmark and the Argoverse2 benchmark validate the effectiveness of the proposed method. FSTR outperforms all LiDAR real-time methods by 69.5 mAP and 72.9 NDS on the official benchmark of nuScenes dataset. On the long-range detection benchmark Argoverse2, the proposed method achieves a new state-of-art performance of 39.9 mAP which outperforms the existing LiDAR detectors, even the LiDAR-Camera detectors by a large margin (+9.4 mAP and +7.5mAP), showing the great advantage of the proposed method for long-range detection.

Recent through-wall radar (TWR) systems can reconstruct the pose of human targets blocked by occlusion. They rely on the fusion of optical and radar data to avoid the painful annotation burden. However, the fusion process is not always reliable, especially for human joint coordinates that carry 3-D spatial information. Inspired by cross-modal learning and self-supervised learning, this letter proposes a two-stage 3-D human pose reconstruction method for TWR systems. In the cross-modal supervision stage, the pretrained optical model provides initial noisy labels extracted from optical images. In the self-supervision stage, supervised labels and the model weight are corrected circularly with radar images. The self-supervision enhances the robustness of the model and the reliability of labels. It can be directly extended to existing radar-based pose reconstruction methods, and hardly requires extra training time. Experiments show the model beats state of the art (SOTA) for reconstructing 3-D poses from TWR images and contains robust generalization in unseen wall-occlusive scenes.

In the past decades, radio-echo sounding (RES) data have been used to predict basal dry-wet distributions in glaciated regions through manual inspection of the records. Extending such work, we propose a semi-automatic method for predicting such distributions. The method improves previous work in two ways: (1) subglacial water bodies are taken as reference to correct the thresholds of dry and wet beds identification at a regional scale; and (2) five distinct features are defined and used to automatically identify the dry-wet transition, allowing a classification model based on a support vector machine. To demonstrate its effectiveness, the method is applied to airborne RES data collected in recent years over Princess Elizabeth Land in East Antarctica. A comparative analysis of the new vs previous method was carried out in the Ridge B region of the East Antarctica and at the Thwaites Glacier region of West Antarctica. The results show the method can obtain more accurate subglacial dry-wet distribution results with larger coverage and has the potential to determine dry-wet transitions at a continental scale if applied to the full set of known Antarctic RES data.

The spatial correlation of ice-sounding data can be used to trace internal isochronic layers and synchronise the age-depth relationship between different ice core sites, which is difficult using existing imaging methods. In this study, we propose a new algorithm to address the opportunity that applying spatial correlation to ice-sheet imaging offers. The algorithm is a spatial-correlation–based ice-sounding imaging method using fast back-projection that successfully improves the spatial correlation of imaging results with high efficiency. We give the specific steps to implement the algorithm and apply it to simulate both point targets and ice-sounding radar data to demonstrate its validity in imaging ice sheets. Furthermore, compared with two previous methods the proposed algorithm improves the spatial correlation without degrading the ability of signal-to-noise ratio improvement and processing efficiency.

Aeromagnetic noise compensation is a vital part of aerial survey measurement, and its compensation effect directly determines the quality of aeromagnetic survey data. At present, the commonly used compensation model is the T-L model, and the least squares method is used to solve for the coefficients. However, the noise source modeled in the T-L model is incomplete. Since the tail boom cannot be completely rigid, tail-boom swing is an unavoidable problem in aeromagnetic measurement. This kind of swing is the most obvious when the aircraft is maneuvering, and it will significantly interfere with the measurement data of the sensor. In this article, two causes of the swing noise are analyzed, and the nonlinear relationship between the swing displacement and the noise is derived. Since it is difficult to express the nonlinear relationship with mathematical forms to compensate for the aeromagnetic data, we propose a new compensation method that uses a 1-D convolutional neural network to perform secondary compensation on the data already compensated by the T-L model in order to remove the effect of tail-boom swing. The flight experiment data show that the proposed method can significantly improve the quality of aeromagnetic data. Compared with the T-L method, the improve ratio is increased by 60%–100%. It shows that the proposed method has a remarkable compensation effect for aeromagnetic noise.

Existing unsupervised domain adaptive (UDA) 3D detection methods only address the domain gap caused by the prior size of 3D bounding boxes between different datasets, which ignore the difference in the distribution of point clouds. To address this challenge, we propose an unsupervised domain adaptive 3D detection by data adaption, which trains the model by transferring the source domain instances into the target domain scenes by adaptive point distribution. First, an instance transferring method is proposed for selecting and transferring suitable instances from the source domain into the target domain scene; Second, we propose an adaptive downsampling method to adjust the point cloud distribution of the transferred instances to approximate the points distribution of the target domain. Finally, our method trains the randomly initialized detector with the pseudo-instances in the target domain. To the best of our knowledge, we first address the UDA problem of the 3D detectors from the perspective of data. Extensive experiments on several popular datasets show that the proposed method outperforms the existing state-of-the-art methods by a large margin. Further experiments also show our approach is detector-agnostic and achieves consistent and significant gains on all types of 3D detectors.

In this paper, we propose a novel processing algorithm for FMCW ice-sounding radar which addresses the along-track focusing of internal reflecting horizons. It is a modified frequency scaling algorithm that can handle the two-layer medium imaging model as well as the Doppler impact of the motion within the sweep in a nadir-looking radar. The proposed algorithm’s comprehensive derivation and implementation processes are described. In the range-Doppler domain, a really effective formulation for radar signals is also presented, which takes into account refraction effects and electromagnetic wave propagation velocity changes at the interface of two different mediums. Point target simulations are carried out to demonstrate the algorithm’s performance. The proposed method is also tested on the real data acquired by the Shallow-Layers-Detection Ice Sounding Radar (SLDISR). The result shows the applicability of the proposed method for imaging the ice sheet.

The accumulation rate is a key parameter in computing the mass balance of glaciers and ice sheets to estimate sea level rise. A shallow-layers-detection ice sounding radar (SLDISR) is developed to measure the accumulation rate and shape of near-surface internal layers with high resolution. With a transmitting frequency from 500 to 2000 MHz, this frequency-modulated continuous wave (FMCW) radar provides a range resolution of about 16 cm in free space by using a Hanning window and a penetrating depth about 150 m under polar ice. The spectral analysis and coherent integration techniques are used to obtain a high processing gain and to improve the signal-to-noise ratio of the system. A phase-locked loop with wideband yttrium iron garnet (YIG) oscillator is applied to generate a sweeping chirp signal as an input source for the transmitter. A stable, low-frequency reference chirp signal is generated with a direct digital synthesizer (DDS) integrated in field-programmable gate array (FPGA). To reduce the high-speed requirement to the analog-to-digital converter (ADC), dechirp technology is adopted at the RF section of the receiver. The implementation of the digital unit is based on an FPGA chip. The designed radar has been successfully deployed in Antarctica during the 31st Chinese Antarctic Research Expedition (CHINARE 31) and CHINARE 33, mainly over the East Antarctic Ice Sheet (EAIS). The echograms indicate the effectiveness of the radar system on detecting clear internal reflecting horizons (IRHs) over ice sheets.

Although the through-the-wall radar imaging (TWRI) system working in the appropriate frequency band can penetrate the nonmetallic obstacles and sense the targets behind, its low imaging spatial resolution hinders the acquisition of more detailed information, such as human pose and shape. This article mainly discusses a deep learning-based human pose and shape recovery method from TWRI images. Inspired by cross-modal learning, the method follows a teacher–student learning pipeline that avoids the heavy cost of manual labeling. Specifically, a camera is attached to the self-develop radar system to simultaneously capture paired red-green-blue (RGB) images and TWRI images in a scenario without wall occlusion. A pose estimation framework (Hourglass) and a semantic segmentation framework (UNet) serve as the teacher network to convert the RGB images into the pose keypoints and the shape masks. By taking inspiration from the topological architecture of these frameworks, a student network radar pose shape network (RPSNet) is designed to extract the information from the corresponding radar images and predict the keypoints and masks that are close to the results above. Instead of learning two single-task objectives independently, multitasking learning is introduced to adaptatively learn common features. When applied to wall-occlusive scenarios, only the radar images are collected and fed into the student network for pose and shape recovery. The advantages of this method over computer vision-based methods for human recovery are demonstrated in scenarios both without and with wall occlusion.

The characteristics of the range cell migration (RCM) under single-input-multiple-output (SIMO) bi-static geometry are studied and an omega-K algorithm for near-field 3-D imaging based on planar SIMO or multiple-input-multiple-output (MIMO) array is proposed. The RCM and the linear part of the range offset (RO) within SIMO data are corrected by employing a 3-D Stolt transformation. The residual range error is further compensated by phase multiplication and an image-domain interpolation. Reconstruction for a MIMO aperture can be achieved by adding together all the focusing results from its SIMO subarrays. The implementation details of the algorithm are described. The imaging resolution and the sampling requirements for a MIMO aperture are discussed. Since the RO correction is achieved by an approximate way, the spatial limitation for accurate reconstruction is also given. The imaging accuracy and the high efficiency of the algorithm are demonstrated both by simulations and experiments with various distributed targets based on planar MIMO arrays. Real-time 3-D imaging for planar SIMO/MIMO aperture is expected to be achieved by using the algorithm.

A modified omega-K algorithm for planar multiple-input-multiple-output (MIMO) array-based 3-D image reconstruction is proposed. By applying several proper approximations in wavenumber domain, the complex and time-consuming bistatic Stolt transformation within the traditional MIMO-$\omega \text{K}$is changed into a relatively simple monostatic one, and therefore the amount of the interpolation given for the data is greatly reduced. Combining with multidimensional fast Fourier transformation, extremely high computation efficiency can be achieved. Implementation details are well described. The imaging accuracy and the high efficiency of the algorithm are demonstrated by a near-field imaging experiment with a flat mannequin target based on a planar MIMO array radar.

This paper presents the ice thickness, fine resolution internal reflecting horizons (IRHs), and distinct bottom topography measurements of Chinese Kunlun Station and Grove Mountains, Antarctica, derived from sounding these glaciers with a high-resolution radar. To enable the development of next-generation ice-sheet models, we need information on IRHs, bottom topography, and basal conditions. To this end, we performed measurements with the progressively improved ice-sounding radar system, currently known as the high-resolution ice-sounding radar developed by the Key Laboratory of Electromagnetic Radiation and Sensing Technology of Institute of Electronics, Chinese Academy of Sciences, Beijing, China. We processed the collected data using focused synthetic aperture radar (SAR) algorithm named the modified range migration algorithm using curvelets and the modified nonlinear chirp scaling algorithm to improve radar sensitivity and reduce along-track surface clutter. Representative results from selected transects indicate that we successfully sounded 3-km-thick ice with a fine resolution of 0.75 m. In this paper, we provide a brief description of the radar system, discuss the focused SAR processing algorithms, and provide sample results to demonstrate the successful sounding of the ice sheet in Antarctica.

The polar layered deposits (PLDs) provide a wealth of information about the past climate evolution of Mars. Surface mass fluxes and ice flow mainly governed topography and layering of the PLD. China's Mars probe including an orbiter and a landing rover will be launched by 2020. A new type satellite-borne Mars penetrating radar instrument has been selected to be a part of the payloads on the orbiter. Its main scientific objectives are to map the distribution of water, water-ice and to detect the soil characteristics at global scale on the Martian crust. Compared with Mars Advanced Radar for Subsurface and Ionospheric Sounding and Shallow Radar, the biggest difference is that the antenna system of this Mars penetrating radar consists of two dipole antennas mutually perpendicular. This special configuration enables the investigation of the ice flow of PLD by detecting and analyzing the features of anisotropic crystal-orientation fabric (COF). Thus, relying on the fact that the radio waves are depolarized while passing through an anisotropic COF layer, in this paper, a method for anisotropic COF detection based on this radar system is proposed. The radar echo formulation of anisotropic COF is derived and the ratio of the signals measured by the two perpendicular antennas is used to analyze the anisotropy of COF. We demonstrate that the ratio is an ideal criterion for the detection and analysis of COF, since it contains all parameters about the anisotropy feature of COF and it is independent of the attenuation in the propagation path and the reflection coefficient. In order to verify the validity of the derived analytical expression of the ratio for the detection of COF, finite-different time-domain simulations are carried out based on a simple model of the subsurface of PLD which contains an anisotropic COF layer. The advantages of this method, the potential application scenarios, and the effects of the Martian environment are also discussed.

Surface mass fluxes and ice flow mainly governed topography and layering of the PLD. A new type satellite-borne Mars penetrating radar is designed and will be launched in 2020. The radar system transmits linear polarization wave and receives by two dipole antennas mutually perpendicular. In this paper, the model of PLD with an anisotropic COF is build. Then, a simulation experiment is done to verify the feasibility of this method. The result shows that the Mars penetrating radar can be used to detect the depolarization effect caused by the anisotropic COF.

We develop and then demonstrate a modified planar subarray processing algorithm based on the inverse scaled Fourier transform applied to very high frequency ice-sounding data that produces swath measurements of ice sheet surface topography, ice thickness, and radar reflectivity of both internal reflecting horizons and bedrock of the ice sheet. It is a real-time ice-sounding imaging method. First, theory analysis has been carried on the proposed algorithm. Then, we give the particular realizing steps to implement this algorithm. Finally, we apply this algorithm to the simulation point targets and real data collected during the 29th Chinese Antarctic Research Expedition to prove its validity of imaging of ice sheets. Furthermore, compared with two previous algorithms in two major aspects—the power of azimuth clutter reduction and calculating time—the proposed algorithm could considerably reduce the imaging time to meet the requirement of real-time imaging of ice-sounding data without degrading the ability of azimuth clutter reduction via C+MPI language on a parallel computer system.

In this paper, we propose an approach to integrate the radio-frequency interference (RFI) mitigation technique into a real-time synthetic aperture radar (SAR) imaging algorithm applied to very high frequency (VHF) ice-sounding data that produces swath measurements of ice sheet surface topography, ice thickness, and radar reflectivity of both Internal Reflecting Horizons (IRHs) and bedrock of the ice sheet. According to the proposal, the RFI suppression method-an adaptive line enhancer (ALE) controlled by the normalized least mean square (NLMS) , is incorporated into the real-time imaging algorithm named modified Planar Subarray Processing (PSAP) algorithm based on inverse scaled Fourier transform (ISFT). The approach is tested successfully on real data recorded over Chinese Kunlun Station during the 29thChinese Antarctic Research Expedition (CHINARE 29). The experimental result indicates that the narrowband RFI could be subtracted from the desired chirp signal return during the imaging processing, and the computational efficiency of the imaging processing is preserved.

In this paper, we propose a new algorithm to address the speckle noise problem in imaging of ice sheets. It is a wave-equation-based ice-sounding imaging method using curvelets as building blocks of ice-sounding data, which successfully images the topography of ice sheets. First, theory analysis has been carried on to the proposed algorithm. Then, we give the specific steps to implement this algorithm. Finally, we apply this algorithm to the simulation point targets and High-Resolution Ice-Sounding Radar data to prove its validity of imaging of ice sheets. Furthermore, compared with five previous methods in two major aspects—the power of clutter reduction and the equivalent number of looks—the proposed algorithm reduces the speckle noise during the imaging processing without degrading the ability in clutter reduction.