Quality-preserved Ambient Noise Imaging in Distributed Sensor Networks
with Limited Bandwidth
Abstract
Distributed sensor networks empower real-time in-situ computing and
seismic imaging without transmitting the raw data to the remote data
center. This attribute is valuable for planet exploration that has
stringent bandwidth. The previously distributed ambient noise imaging
algorithm computes results using data from near neighbors, while the
information from distant neighbors is not utilized, hence the image
quality is compromised. To overcome this problem, this work proposes an
innovative common-receiver-based decentralized ambient noise imaging
algorithm. In this algorithm, the basic imaging algorithm is still
Eikonal tomography, but the distant neighbors are also used to computing
the seismic image so that the quality of the output image can be
preserved. An in-situ computing and clustering algorithm is created to
optimize the data transition and computation while meeting the bandwidth
constrains. The experiments were performed on both synthetic data from
Enceladus and real data from the USArray archives. The new algorithm
generates higher-resolution images under the same bandwidth constraints,
comparing to previous algorithms, and the quality of the output image is
satisfactorily preserved. The communication cost reduction over the raw
data collection is in several orders of magnitude (e.g., 1: 1600). It
meets the desired bandwidth constraint in planetary exploration
applications.