Fusion of GPS and InSAR data to derive robust 3D deformation maps based on MRF L1-regularization

ABSTRACT

To achieve the three-dimensional surface deformation maps, interferometric synthetic aperture radar (InSAR) and global position system (GPS) data are integrated. In this letter, Markov random field-based L1 regularization is proposed to reconstruct the Gibbs energy equation for improving the previous three-dimensional optimized model. Furthermore, differential evolution algorithm is used to minimize the energy function for deriving the three optimal components of the ground motion. The new method proposed in this paper is then applied to derive the three-dimensional deformation maps of the 2016 Kumamoto earthquake. The Experimental results show that our new method achieves considerably higher performance than the more standard L2-based method. The root mean square errors for the three components of east, north and up are 1.45, 1.87 and 2.14 cm, respectively, which are smaller than the L2-based method, especially in the vertical direction.

Acknowledgments

The authors would like to thank the European Space Agency for the Sentinel-1A data and the Nevada Geodetic Laboratory for the GPS time series.

Additional information

Funding

This work is supported by the National Key R&D Program of China under Grant [2018YFC1505101], the National Natural Science Foundation of China under Grant [41801356], and the China Academy of Railway Sciences Fund [2017YJ040].