Search

Objectives: Synthetic aperture sonar (SAS) is one kind of high resolution underwater imaging sonar. Its principle is to use small array to simulate large aperture array by uniforming linear motion along the direction of flight track, and realize coherent processing of echo signal to obtain high resolution two-dimensional sonar image, which is independent of imaging distance and acoustic wavelength. The back projection algorithm is an accurate point-by-point imaging algorithm, which is characterized by accurate focusing ability, suitable for wide beam and large scene imaging, and easy to realize motion compensation, but it has the disadvantages of low computational efficiency. In recent years, with the development of computer technology, especially the appearance of graphics processor unit, it has powerful computing power and provides a new way to accelerate the back projection algorithm. Methods: We propose a fast accurate back projection algorithm for multi-receiver SAS in heterogeneous environment. On the basis of analyzing the principle of accurate back-projection imaging algorithm, the pulse compression and time-consuming azi-muth accumulation are transformed into a single instruction multi-threaded mode, and the imaging process is accelerated by the powerful multi-core computing ability of graphics processor unit. Results: The validity and efficiency of the proposed fast imaging algorithm are verified by the imaging experiments performed on simulated and real SAS data. Compared with the serial imaging algorithm, its acceleration ratio is as high as 326.3 and 333.6. For large-scale data imaging processing, it shows excellent acceleration performance. Conclusions: The proposed back projection imaging algorithm greatly improves the imaging efficiency and meets the needs of real-time SAS imaging, which lays the foundation for subsequent motion compensation. [ABSTRACT FROM AUTHOR]

Objectives: Complex image registration is crucial in interferometric synthetic aperture sonar (InSAS) signal processing. The registration quality directly influences the generation of interferogram and the subsequent reconstruction of digital elevation model. The traditional global polynomial surface fitting method has the disadvantage of low accuracy for large scene InSAS images. Therefore, a registration method based on piecewise surface fitting is proposed to overcome the problem. Methods: Firstly, the whole image is partitioned into several pieces along range direction with partly overlapped. Then, polynomial surface fitting is performed using control points in each piece. Finally, the global offset surface is obtained by merging those local surface fitting results. Results: Through the simulation and real data experiments, the accuracy of complex image registration using the proposed method is higher than the result of global polynomial surface fitting method. Conclusions: The evaluation has been carried out with the root mean squared error, correlation coefficient, residue numbers, which indicates the validity of the proposed approach. [ABSTRACT FROM AUTHOR]

We propose a combined minimum discontinuity phase unwrapping algorithm based on clusters. The wrapped phase image is tessellated into small regular blocks by the main node according to the com - puting resources and then sent to the idle computing nodes. During the unwrapping process of single block, the wrapped phase quality map is firstly computed and the wrapped phase image is partitioned into high and low quality areas; then the quality-guided and minimum discontinuity combined phase unwrapping algorithm is performed to get the initial unwrapped result. After the local unwrapping process is finished, the initial unwrapped result and the region partition result are sent to the main node. When all the blocked phases are unwrapped, the main node performs the minimum discontinuity optimization process on the bor- der of different blocks and the adjacent low quality areas to get the final unwrapped result. Tests performed on the simulated and the real InSAS (interferometric synthetic aperture sonar) interferograms based on clus- ters confirm the validity and efficiency of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Complex image registration is one of the most important steps during the process of interferometric signal processing, and its efficiency directly affects the performance of interferometric signal processing system. In order to improve the efficiency of complex image registration algorithm, we propose an opti⁃ mized complex image registration algorithm for interferometric synthetic aperture sonar (InSAS) in shared memory environment. Firstly, after analyzing the characteristic of calculation and parallelism of each pro⁃ cessing step for complex image registration, an optimized computation method for coarse registration and refine registration is proposed for the massive sliding operation based on the coherence of data in adjacent windows. Then, in order to make full use of multi-core computing resource to accelerate the process of complex image registration, OpenMP instructions are used to parallel design and task allocation for coarse and refine registration, complex image interpolation and wrapped phase extraction. Finally, the correctness and efficiency of the proposed method are verified by the parallel complex image registration for InSAS. [ABSTRACT FROM AUTHOR]

Complex image registration is a crucial step in interferometric synthetic aperture sonar (InSAS) signal processing. The registration quality directly influences the generation of interferogram and subsequent reconstruction of digital elevation model (DEM). Utilizing the geometric model of imaging, the rational function relationship between offset and slant range is deduced, thus this paper propose a new InSAS complex image registration method based on second order rational function surface fitting. Compared with the traditional polynomial surface fitting method, the proposed method in this paper has the advantages of higher precision and smaller calculation cost. The evaluation has been carried out using the root mean squared error(RMSE), correlation coefficient,residue numbers and computing time as the criteria. Simulation and real data experiments results indicate the validity of the proposed method. [ABSTRACT FROM AUTHOR]

Aiming to solve the low efficiency imaging of multi-array synthetic aperture sonar, this paper proposed a fast imaging method for multi-array synthetic aperture sonar in heterogeneous environment. It considered the range Doppler algorithm characteristic for multi-array synthetic aperture sonar and different calculation features between CPU and GPU, the intensive calculation such as range compression, phase compensation, range cell migration compensation and azimuth compression was implemented on GPU, which greatly enhanced the imaging efficiency for multi-array synthetic aperture sonar. In the end, it verifies the validity and efficiency of the proposed method through the imaging experiment performed on real data and the speedup is up to 14.45 times compared to the serial algorithm. [ABSTRACT FROM AUTHOR]

Phase unwrapping is one of the key processing steps during reconstruction the digital elevation model from the signal of interferometric synthetic aperture radar(InSAR) or interferometric synthetic aperture sonar(InSAS). In order to solve the problem of low unwrapping efficiency with large interferogram, we propose a blocked minimum discontinuity phase unwrapping algorithm in shared memory environment. The whole phase quality map is computed firstly through multi-cores of CPU, then wrapped phase image and phase quality map are tessellated into small regular blocks, which are unwrapped by the quality guided and minimum discontinuity optimization method simultaneously. In the end, the minimum discontinuity optimization process is performed again on the border of different blocks and the low quality areas located on the border to get the final merged unwrapped result in the main thread. Tests performed on the simulated and real InSAS interferogram show that the speed up reaches 3.98 and 2.26 respectively. [ABSTRACT FROM AUTHOR]

This paper proposed an efficient range Doppler imaging algorithm for multi-array synthetic aperture sonar (SAS) in shared memory environment based on OpenMP, which solved the imaging problem of low efficiency by multi-core CPU resources. Considered the parallel characteristic of range Doppler algorithm for multi-array SAS, it parallelized the signal processing process such as preprocessing, range compression, phase compensation, range cell migration compensation and azimuth compression based on OpenMP, which greatly enhanced the reconstruction process of SAS image by multi-core CPU resources. The imaging test performed on real data shows that the speedup of the parallel imaging algorithm is up to 19. 86 times, which meets the demand for real-time processing of SAS imaging. [ABSTRACT FROM AUTHOR]

We propose a combined phase unwrapping algorithm using region partitioning. The wrapped phase image is partitioned into high and low quality areas according to its quality map- and a quality guided algorithm is performed to gel the initial unwrapped result. A quality guided strategy is adopted to identify low quality areas- and each low quality area is stored separately. The low quality areas are optimized by a minimum discontinuity phase unwrapping algorithm one by one. and then the final unwrapped result is obtained. An unwrapping test performed on an InSAR interferogram shows that the proposed algorithm is more efficient than the minimum discontinuity phase unwrapping algorithm, and the unwrapped result was more accurate than results obatained from a quality guided algorithm. [ABSTRACT FROM AUTHOR]

We proposed a fast back-projection algorithm for multi-receiver synthetic aperture sonar (SAS) in a shared memory environment. A two-dimensional dynamic data structure was used to keep the delay time and the compensation phase information was designed based on the characteristics of the back-projection algorithm, which helps to eliminate repeated delay time computation for the same range. In the shared memory environment, we combine the focusing operation for the near and far distance of the image scene to achieve a computation load balance. An imaging test performed on real SAS data confirms the proposed algorithm is accurate and efficient. [ABSTRACT FROM AUTHOR]

In order to solve the unwrapping problem of low efficiency for large interferogram, we propose a new phase unwrapping algorithm in shared memory environment, which is designed based on quality-guided and minimum discontinuity optimization strategies. The three processes, phase quality map computing, high and low quality area’s partition and low quality area’s optimization, are parallelized in shared memory environment. In order to fully use the multi-core computing ability, a dynamic allocation method for computing task is designed for the optimization process of low quality are- as. Unwrapping tests performed on simulated and InSAR interferogram show that the proposed algo rithm improves the high quality of unwrapped result and greatly optimizes the efficiency. [ABSTRACT FROM AUTHOR]

We propose a non-recursive minimum discontinuity phase unwrapping algorithm for im­proved efficiency. After analyzing the principles of minimum discontinuity phase unwrapping algo­rithm* a stack is used to store the intermediate results during edge spreading and circle canceling processes and thena non-recursive algorithm for phase unwrapping is implemented. The new algorithm was combined with the quantized quality-guided phase unwrapping algorithm, which enhances the un­wrapping efficiency by restricting the optimization area. Unwrapping tests performed on InSAR and InSAS interferograms show that the proposed method maintains high quality unwrapping results and greatly improves efficiency. [ABSTRACT FROM AUTHOR]

We proposed a parallel quality-guided phase unwrapping algorithm in shared memory environment. The intrinsic relationship between the neighboring phase points about quality value computing is analyzed firstly, and then the row and column arrays are used to store the intermediate results in order to eliminate the repeated computation of gradient. The computing task is allocated by row, which can simplify the computing of row and column gradient mean values. Finally, the allocation of computing task is realized using OpenMP instructions, and the quality map is computed in parallel. The inherent parallelism of quality guided process is also analyzed in-depth, but subject to the repetitive thread startup and exit overhead, it is difficult to reflect the advantage of speed. Unwrapping tests performed on InSAR and InSAS interferogram show that the proposed method greatly improves the efficiency of phase unwrapping, and provides a foundation for improving the solution precision under real-time conditions. [ABSTRACT FROM AUTHOR]