Research On InSAR Deformation Monitoring Technology Based On GPU

Ground-based Synthetic Aperture Radar Interferometry has been an important monitoring method of deformation monitoring with the superiority of all-weather,convenient installation and high precision.This method is applicable for real-time monitoring of deformations in mountains,bridges,and buildings,warning potential disaster and protecting people’s lives and property security.Accuracy and real-time are both important features and How to balance them is fully challenging in current radar deformation monitoring and processing.In this paper,Ground-based Synthetic Aperture Radar is employed as the platform of deformation monitoring.The main work is as follows:First of all,establish a signal model of Ground-based SAR deformation monitoring system.Determine the signal of the monitoring system as Stepped Frequency Continuous Wave(SFCW)by comparing and analyzing the signal characteristics of pulse and continuous wave,Linear Frequency Modulation Continuous Wave and Stepped Frequency Continuous Wave.Then,work methods of the Ground-based SAR system and the relationship of three-dimensional geometry is designed.Finally,analyze the principle of Ground-based SAR deformation monitoring and summarize the basic process of deformation monitoring.What’s more,analyze the specific process of Ground-based SAR deformation monitoring.Firstly,study the echo focusing imaging of the Ground-based SAR deformation monitoring system that based on SFCW.By using the high-accuracy time-domain BP imaging algorithm,the echo imaging simulation of the point target is performed.Then design the pixel-level and sub-pixel-level image registration algorithms and generate interferograms by the registered complex images.Weighted circular periodic median filter algorithm is used for noise filtering of interferograms.Focus on the analysis of the phase unwrapping algorithm,perform respectively the phase unwrapping processing of Branch-cut algorithm and Quality-guide algorithm by the simulated interferogram,and compare the time-effect and mean-square error of the two algorithms.Finally,according to the Quality-guide algorithm,a fast implementation method based on heap data structure that takes into account both timeliness and accuracy is proposed as the unwrapping algorithm in this paper.Lastly,analyze the algorithm parallelism of ground-based SAR deformation monitoring and processing flow,and propose to use CPU+GPU heterogeneous platform for accelerated implementation to improve the timeliness of deformation monitoring.The parallel implementation of BP imaging algorithm,interferogram generation,interferogram filtering,and phase unwrapping algorithm is designed respectively.The GPU accelerated processing is implemented by optimization strategies such as converting multiple loops to appropriate parallel granularity,adaptively dividing the optimal thread block size,and using shared memory to reduce the reading time of duplicate data.Compared with the realization of the CPU+GPU heterogeneous platforms and the CPU platform,the accelerating effect ranging from a dozen to a hundred times was accelerated,and the feasibility of the GPU in improving the timeliness of deformation monitoring was verified.