Research On Real-time Phase Unpacking Algorithm Based On Unscented Kalman Filter

As a key technique in interferometry,phase unwrapping algorithm is widely used in optical interferometry,synthetic aperture radar and reflection measurement.At present,there are a lot of researches on phase unwrapping algorithms.All kinds of algorithms have different performance under different noise conditions,so they have different adaptability.Generally speaking,the algorithms with better overall performance have higher implementation complexity and cost,which will become a bandwidth bottleneck in the real-time interference systems.The main work of this paper includes the implementation of a phase unwrapping algorithm based on the unscented Kalman filter,the focus of this paper is to study the high throughput architecture based on FPGA platform.Finally,the feasibility of this algorithm is verified by experimental analysis.In view of the shortcomings of several kinds of phase unwrapping algorithms,a phase unwrapping algorithm based on unscented Kalman filter is adopted in this paper.On the one hand,according to the characteristics of Kalman filter algorithm,phase unwrapping process and phase de-noising process are carried out simultaneously,eliminating the pre-processing of noise before phase unwrapping and ensuring that no other noise is introduced in the phase unwrapping process.On the other hand,the unscented Kalman filter is an improved algorithm based on Kalman filter,which adopts UT transformation as the basis and eliminates the linearization of nonlinear process,so that the result is closer to the real value.Among them,this paper mainly researches the unscented Kalman filter algorithm suitable for phase unwrapping and focuses on solving the phase recovery distortion problem under the condition of strong noise.This paper proposes to implement the unscented Kalman filter phase unwrapping algorithm on the FPGA platform to improve the speed of the phase unpacking algorithm.Firstly,the fixed-point simulation function is designed based on the software unwrapping algorithm,and the optimal word length and decimal number selected on the FPGA platform are determined according to the SNR of phase curve recovery at different fixed-points.Then,in modelsim simulation software,the implementation of unscented Kalman filter algorithm is mainly divided into the following modules: system state module,sigma point generation module,state prediction module and state update module.Finally,based on the Quartus ? platform,complete operations such as logic synthesis and pin constraints,and complete data unwrapping processing on the FPGA chip platform.Finally,the algorithm is applied to the phase unpacking unit of a phase-modulated fiberoptic sensing system,and an Michelson interferometric fiber optic sensor experimental system is built,using the rectangular pulse binary phase modulation demodulation technology to obtain wrapped phase,adopted a system of Altera corporation EP3SE110F1152I3 FPGA as the core of the real time control and processing,using 32 Q16 fixed-point format,get the signal to noise ratio of 31.7478 d B,which is about 30 to 40 times faster than the PC,realizing large-scale data real-time throughput operations.This topic improves the accuracy of the phase unpacking algorithm and uses FPGA hardware circuit board to achieve faster phase unpacking speed,which has important application value in various interferometric fields.