An Improvement of the Performance of SAR Offset Tracking Approach to Measure Optimal Surface Displacements

Sung-Ho Chae,Won-Kyung Baek,Won-Jin Lee,Hyung-Sup Jung

doi:10.1109/access.2019.2938831

Sung-Ho Chae, Won-Kyung Baek + Show 2 more

Open Access

https://doi.org/10.1109/access.2019.2938831

Copy DOI

Abstract

For the measurement of abrupt and large surface movements caused by earthquakes, volcanic eruption and melting glacier, Synthetic aperture radar (SAR) offset tracking method would be a feasible solution because it can provide unambiguous ground displacements in both the ground range and azimuth directions when the interferometric phase is not coherent. However, the measurement performance of the method largely depends on the kernel size, which denotes the size of search window to estimate the azimuth and range offsets between reference and target SAR images. Thus, there is a trade-off between sensitivity and measurement density depending on the search kernel size. In this study, an enhanced SAR offset tracking method based on multi-kernel processing has been developed to find an optimized measurement from the trade-off between resolution and measurement accuracy. It can obtain optimal surface displacement measurements by calculating multiple offset measurements and determining a final measurement from the statistical properties of the multiple measurements. The measurement performance of the proposed method was evaluated by using European Remote Sensing 2 (ERS-2) satellite SAR data sets of the Hector Mine earthquake event in 1999 and Advanced Land Observing Satellite-2 Phased Array type L-band Synthetic Aperture Radar-2 (ALOS-2 PALSAR-2) data sets of the 2016 Kumamoto earthquake event. Our results showed that an optimized measurement from the trade-off between the observation accuracy and resolution can be effectively determined by our proposed processing strategy. The results are improved results for measurement density and accuracy over previously published results. It further confirmed that our new method is allowed for the optimal measuring the large-scale fast surface displacements that cannot be sufficiently observed with the phase-based SAR method.

Highlights

Surface deformation mapping is very important because it is a fundamental information to acquire detailed knowledge about the mechanism of geological activities
APPLICATION TO THE 1999 HECTOR MINE EARTHQUAKE The 16 range and 16 azimuth displacement maps were obtained from the Synthetic aperture radar (SAR) offset tracking method using multiple search kernel sizes varying from 30 × 60 to 180 × 360 pixels increasing the search kernel sizes of 10 pixels and 20 pixels in range and azimuth directions, respectively
The smaller the kernel sizes, the lower the signal-to-noise ratio (SNR) value that indicates the quality of the offset estimation in the region where the large surface displacement

Summary

INTRODUCTION

Surface deformation mapping is very important because it is a fundamental information to acquire detailed knowledge about the mechanism of geological activities. SAR offset tracking method, especially intensity tracking, can overcome the weakness of the phase-based InSAR method It can measure fast surface displacements in both the range and azimuth directions by using the cross-correlation optimization procedure, similar to optical feature tracking method, even if the interferometric phase is decorrelated [4]–[6]. It is important to determine the proper kernel sizes in order to effectively observe the two-dimensional (2D) displacements using the SAR offset tracking method because there are many unreliable parts using only a single search kernel. We focused on determining proper range of kernel sizes and generating optimal offset measurements through statistical processing based on the multiple pieces of offset measurements from the multiple search kernels to improve the trade-off between cross-correlation accuracy and resolution. Application and performance validation of the proposed approach were carried out by using ERS raw data spanning the 1999 Hector Mine earthquake and ALOS-2 PALSAR2 raw data spanning the 2016 Kumamoto earthquake through a comparison with GPS data

DATASET AND METHODS

RESULTS AND DISCUSSION

CONCLUSION