GPU-accelerated circular SAR echo data simulation of large scenes

In circular synthetic aperture radar (CSAR), radar collects data over a circular trajectory. This wide-angle non-linear collection geometry makes CSAR capable of obtaining both two-dimensional (2D) high resolution image and three-dimensional (3D) image. In this paper, compressive sensing technique is used in 3D CSAR imaging. Not only echo samples used in imaging are greatly reduced, but also the imaging results are with good performance. However, the consumed time of reconstruction is very long. In order to overcome this problem, graphic process unit (GPU) is furtherly used in 3D compressive sensing CSAR imaging in this paper. Simulation results show that GPU can improve the efficiency of the 3D reconstruction.

Circular synthetic aperture radar (CSAR) has a 360° observation capability on the central observation scenario. A typical way to process CSAR imaging is to cut data into small sub-apertures because most targets are only coherent at a very small observation angle. There are many sub-aperture imaging methods after development in recent years. The back-projection algorithm is widely used because it is simple and can be applied to an arbitrary trajectory. Because of the limitation of the Nyquist sampling frequency and influence of the antenna sidelobe, azimuth ambiguity is a phenomenon that may occur in the radar imaging process. The existing researches typically choose the back-projection (BP) imaging area according to the SAR platform flight path and the antenna beam width. The limitation of the CSAR imaging area and its azimuth ambiguity region are rarely analyzed theoretically. This paper focus on the sub-aperture imaging of CSAR, based on the BP algorithm, which derives the relationship of azimuth ambiguity with CSAR parameters such as the pause repeat frequency (PRF), slant range angle, velocity of radar platform, etc. This paper proposes an equation for the calculation of the azimuth ambiguity region and analyzes the limitations, which provides theoretical support for CSAR parameter design, imaging area selection, and azimuth ambiguity analysis.

Circular synthetic aperture radar (CSAR) is a high resolution 2D and 3D imaging mode, to obtain the demand of all observation information. Circular motion trajectory make motion compensation problem is very difficult. However, existing method, such as autofocus back projection (ABP) algorithm, cannot effectively solve the problem of motion compensation in CSAR. In this paper, an autofocus algorithm based on minimum entropy is used to solve this problem. It can accurately estimate phase errors in CSAR. The simulated CSAR data are used to validate performance of minimum entropy autofocus (MEA) algorithm.

Circular synthetic aperture radar (CSAR) has attracted much attention in the field of high-resolution SAR imaging. However, the CSAR image focusing is affected by the motion deviations of platform. In the processing of experimental CSAR data to deal with motion errors, the main way is using setup calibrators, which restricts its widespread applications. In this paper, based on the estimation of motion errors, an autofocus CSAR imaging strategy is proposed without using any setup calibrator. The first step is to split the entire aperture into several subapertures, the second step is to process the data in subapertures with an autofocus backprojection method, and the last step is to obtain the final CSAR image by merging the subimages obtained from the subaperture processing. The CSAR data processing results prove that the proposed strategy can remove the motion errors accurately and acquire well-focused CSAR images.

A three dimensional (3D) imaging method which is a complex-valued generalized Radon transform (CGRT)-CLEAN algorithm for circular synthetic aperture radar (CSAR) is proposed in this chapter. Based on the analysis of CSAR echo model, the echo of a scatterer in the range-azimuth domain can be modeled as a single cycle sinusoid. Inspired by the generalized Radon transform, this chapter utilizes the CGRT to estimate for scatterers’ positions. Then, combining the CGRT with CLEAN technique, the parameters of a scatterer which include position information and reflection coefficient can be obtained. Thus, a 3D image of a target can also be reconstructed. The performances of the CGRT-CLEAN method have been verified by simulations, and the results confirm the effectiveness of the proposed algorithm for processing CSAR data.

An airborne circular synthetic aperture radar system captured data for a 5 km diameter area over 31 orbits. For this challenge problem, the phase history for 56 targets was extracted from the larger data set and placed on a DVD for public release. The targets include 33 civilian vehicles of which many are repeated models, facilitating training and classification experiments. The remaining targets include an open area and 22 reflectors for scattering and calibration research. The circular synthetic aperture radar provides 360 degrees of azimuth around each target. For increased elevation content, the collection contains two nine-orbit volumetric series, where the sensor reduces altitude between each orbit. Researchers are challenged to further the art of focusing, 3D imaging, and target discrimination for circular synthetic aperture radar.

This document discusses the evaluation of a consolidated linear performance model of Texas State University's HPC LEAP Cluster, and the development of a non-linear model to represent the performance of the NVIDIA V100 Graphics Processing Unit (GPU). The consolidated model constructed for the LEAP Cluster was evaluated by using it to compute execution times for different sizes of processes and comparing those times to the actual processes run on LEAP. The model itself is found to effectively represent the performance trends of LEAP but fail to match the actual value. The GPU's performance analysis was done by first collecting performance data then using non-linear methods to model it. The number of floating-point operations required to perform a square matrix multiplication is readily known, so the performance data was gathered by running a simple square matrix multiplication program and measuring its execution time. The size of the matrices was then used to determine the number of floating-point operations and divided by the execution time to calculate FLOPS. The final model for the GPU was generated by using a set of logarithmic functions that mapped the GPU performance to a semi-logarithmic scale of the number of floating-point operations.

This paper deals with the issue of ground moving target imaging and motion parameter estimation with an airborne dual-channel circular stripmap synthetic aperture radar (CSSAR) system. Although several methods of ground moving target motion parameter estimation have been proposed for the conventional airborne linear stripmap SAR, they cannot be applied to airborne CSSAR because the range history of a ground moving target for airborne CSSAR is different than that for airborne linear stripmap SAR. In this paper, the moving target’s range history for airborne dual-channel CSSAR and the target signal model after the displaced phase center antenna processing are derived, and a new ground moving target imaging and motion parameter estimation algorithm is developed. In this algorithm, the estimation of baseband Doppler centroid and its compensation are first performed. Then focusing is implemented in the 2-D frequency domain via phase multiplication, and the target is focused in the SAR image without azimuth displacement due to the compensation of the Doppler shift caused by its motion. Finally, the target’s motion parameters are estimated with its Doppler parameters and its position in the SAR image. Numerical simulations are conducted to validate the derived range history and the performance of the proposed algorithm.

The graphics processing unit (GPU) is extensively used in diverse domains, such as finance, machine learning, and image processing. The GPU can be underutilized as multiple applications may not share the same GPU concurrently owing to a memory oversubscription issue. For example, when applications that require fewer computational resources but a larger GPU memory are running instantaneously, the GPU memory may be insufficient; consequently, the number of GPU applications running simultaneously is restricted, decreasing GPU utilization. Further, it can even stop the execution of applications that are running on the GPU. To this end, we propose FlexGPU, which schedules the kernels of the GPU applications that run on the same GPU according to their features. This framework 1) schedules the kernel at the launching time according to its features to improve GPU utilization and 2) temporarily checkpoints and restores non-dependent content in the GPU memory to/from the host memory, which avoids oversubscription of the GPU when out-of-memory failure occurs and allows more kernels to run concurrently on the GPU. The experimental results show that compared to existing methods, our approach demonstrates a 7 times improvement in performance in terms of execution time and enables a 2.5 times increase in the concurrent execution of applications.

The calculation of pairwise correlation coefficient on a dataset, known as the correlation matrix, is often used in data analysis, signal processing, pattern recognition, image processing, and bioinformatics. With the state-of-the-art Graphic Processing Units (GPUs) that consist of massive cores capable to do processing up to several Gflops, the calculation of correlation matrix can be accelerated several times over traditional CPUs. However, due to the rapid growth of the data in the digital era, the correlation matrix calculation becomes computing intensive which needs to be executed on multiple GPUs. As of now, GPUs are common components in data center at many institutions. Their GPU deployment tends towards a GPU cluster which each node is equipped with GPUs. In this paper, we propose a parallel computing based on the hybrid MPI/CUDA programming for fast and efficient Pearson correlation matrix calculation on GPU clusters. At coarse grain parallelism, the correlation matrix is partitioned into tiles which are distributed to execute concurrently on many GPUs using MPI. At fine grain level, the CUDA kernel function on each node performs massively parallel computing on a GPU. To balance load across all GPUs, we adopt the work pool model which there is a master node that manages tasks in the work pool and dynamically assign tasks to worker nodes. The result of the evaluation shows that the proposed work can ensure the load balance across different GPUs and thus gives better execution time than using a simple static data partitioning.

Circular synthetic aperture radar (CSAR) can obtain more complete scattering characteristics by observing the target with different azimuth angles. Therefore, extracting the complete structure of the target from CSAR images is of great significance for accurate interpretation. At present, the artificial target extraction based on CSAR images mostly uses anisotropic scattering features. For special targets such as buildings, as the walls and the ground form dihedral corner structures, there are also obvious strong scattering features such as double-scattering lines in SAR images. Therefore, combining the strong scattering features of buildings at specific aspects with anisotropic scattering characteristics at different aspects can obtain better extraction results, and how to extract these features accurately and efficiently is the key point. Based on this, this paper proposes a novel method for building contour extraction based on CSAR images. For strong scattering features, a fast fuzzy C-means (FCM) clustering algorithm was used to extract them. For anisotropic scattering features, aspect entropy was used to characterize the anisotropy degree, and K-means clustering was combined to extract. Finally, a more accurate result is obtained by merging the two feature extraction results. In order to verify the effectiveness and practicability of the proposed method, a lot of measured data acquired by the self-developed airborne L-band and Ku-band CSAR systems were processed. The experiments show that, compared with state-of-the-art algorithms, the proposed method can obtain more accurate results in less time.

Circular synthetic aperture radar (CSAR) enables 360-degree observation of the scene. Compared with coherent imaging, noncoherent imaging can suppress speckle and obtain a large number of texture features. However, due to the existence of a man-made anisotropic targets and the unique imaging geometry of CSAR, the different subaperture images are quite different, which makes the registration very difficult. Therefore, this manuscript proposed a factorized registration strategy. First, the full aperture is divided into multiple subapertures; then the adjacent subaperture images are registered and the noncoherent superimposition of the subaperture images yields the next stage image; finally, the previous step is performed iteratively until the final imaging result is obtained. Experimental results based on Ka-band CSAR data verify the effectiveness of the strategy.

In this paper, two 3-D (three-dimensional) imaging techniques, MB-TomoSAR (multi-baseline SAR tomography) and CSAR (circular synthetic aperture radar), are introduced firstly. Secondly, the indoor experimental systems for acquiring SAR data in the microwave anechoic chamber are described, then the experiments of the two 3-D imaging mode for the metallic spheres and human body are performed, respectively. Finally, the 3-D imaging capability of the MB-TomoSAR and CSAR is demonstrated with the experimental data.

Moving target brings out the different position shifts and defocusing across the image sequences acquired by circular synthetic aperture radar (CSAR) due to the Doppler shift and range smear effects. In this paper, a novel moving target detection approach for single-channel CSAR is proposed based on deep neural network (DNN). A dual-channel densely connected convolutional network (DenseNet) in consideration of complex-valued information is exploited for distinguishing the ground clutter and moving target. In terms of limited CSAR measure data set available for training the DNN network, simulated moving target samples are generated and fused into the measured ones under the various motion parameters. Finally, experiments have demonstrated that the proposed DenseNet for single-channel CSAR system processes an accepted detection performance and effectively overcomes the insufficiency of the limited dataset applications.

3D scene reconstruction from single‐pass circular synthetic aperture radar (CSAR) data has been a research hot spot in various remote sensing applications. Among existing literature, radargrammetry techniques and model assistant methods are two dominant approaches. However, the performance of these methods is limited in practical applications: radargrammetry techniques obtain merely one height value along vertical; model assistant approaches only work when valid models are given. A single‐pass CSAR 3D scene point cloud reconstruction method is proposed. First, the connection between an object and its projections in multi‐aspect SAR images is established with no auxiliary model. On the basis of the connection, voxels are classified into inner voxels (voxels beneath the target profile) and outside voxels (voxels outside of the target profile). Discarded all the outside voxels, the remained inner voxels are combined to form the 3D point cloud of the target. It is similar to the sculpturing process from a raw material to a piece of craft. Finally, a 3D scene is retrieved through assembling all reconstructed targets into a unified coordinate system. The feasibility of the proposed method is validated by the Gotcha Volumetric SAR Data Set.