Enhancement of TEM Data and Noise Characterization by Principal Component Analysis.

摘要

This is the final report for the SERDP project MM-1640; it covers the research results accomplished since the inception of the project in 2007. The basic premise of this project is the theoretical understanding and algorithm development of principal component analysis (PCA) as a de-noising and signal-separation tool for transient electromagnetic (TEM) data in unexploded ordnance (UXO) applications. There is an express need for techniques to reduce the presence of random noise in TEM data as well as reduce the influence of correlated noise due to a wide variety of sources on automatic anomaly-picking routines for more accurate detection with fewer false anomalies. We have developed an algorithm and workflow for the processing and inversion of TEM data that attenuates signal from undesired sources and accurately inverts TEM data for diagnostic UXO parameters. First, we developed a principal component analysis algorithm tailored to unexploded ordnance applications. Decay characteristics of TEM data preclude the standard Karhunen-Loeve transform; we have addressed these issues with algorithm modifications and incorporated these into the workflow. Secondly, we identified the optimum choice of principal components for the attenuation of both random noise and correlated noise, leaving the signal due to UXO intact. We show that the processed data is optimally prepared for automatic anomaly picking routines with a highly reduced number of false anomalies. We demonstrate this on both synthetic examples of UXO surveys, as well as on TEM data from Kaho'olawe, Hawaii. Finally, we have identified a critical issue with inversion of processed data that results in extremely inaccurate recovered models without the incorporation of the PCA process into the forward model. We developed an inversion algorithm which takes the processing steps into account during construction of the inverse kernels. This leads to more accurate recovered models of inverted anomalies.