Final Report Summary - ADAPTIVES (Algorithmic Development and Analysis of Pioneer Techniques for Imaging with waVES)
Waves have been used in detection and imaging for many years. We all know the radar, which uses electromagnetic waves, the sonar that is based on acoustic waves and medical ultrasound where usually acoustic and elastic waves coexist. Other applications concern seismic imaging, i.e. imaging of the earth's subsurface, and non-destructive evaluation of materials. In most of these applications the materials encountered are often complex and their properties are not known – and cannot be estimated – in every detail.
We model, therefore, the propagation medium as a random process for which we may know some statistical properties, such as the mean speed of propagation. Our goal is to solve the imaging problem in a regime where multiple scattering due to the medium heterogeneities is significant. Imaging in such regimes is quite challenging and requires very different methods from the usual ones in homogeneous or known deterministic environments. The challenge is to produce reliable, i.e. statistically stable results, especially when there is no a priori knowledge about the propagation medium.
We have introduced and developed extensively, a Coherent INTerferometric imaging methodology, CINT, that is based on back propagating cross-correlations of array data rather than the data itself as traditionally done in imaging. CINT is designed for imaging with partially coherent array data recorded in richly scattering media. It uses statistical smoothing techniques to obtain results that are independent of the random medium realization.
CINT works well when the coherent signal due to the reflector that we wish to image is strong enough. In strongly scattering media, however, this is no longer true as the back-scattered field from the medium heterogeneities overwhelms the signal received on the array. In the project ADAPTIVES, to extend the applicability range of CINT, and of other coherent imaging methods to such media, we have developed data filtering techniques that enhance the coherent signal which is useful for imaging. To this end we have developed two approaches: the layer anihilators, a filtering technique that suppresses scattered echoes from layered structures and an adaptive time-frequency filtering technique that can be used in any type of clutter.
Another way to overcome heavy clutter effects is by using auxiliary arrays that are near the reflector that we wish to image. The first step towards imaging under such circumstances is the computation of cross-correlations of the noisy recorded signals, which have attracted a lot of attention recently because of their numerous applications in seismic imaging, volcano monitoring, and petroleum prospecting. We use these cross-correlations for imaging reflectors hidden below very complex structures.
Cross-correlations of passive recordings due to ambient noise sources can be used as well in structural health monitoring for detecting and imaging defects. The quality of the obtained image, measured by the signal to noise ratio (SNR), inherits the SNR of the cross-correlations. It is therefore proportional to the square root of the bandwidth of the noise times the recording time. Moreover, we showed that it is also proportional to the array size. This is important in applications such as structural health monitoring as it implies that for short-time recordings the quality of the image can be improved by increasing the array size.
As the medium backscatter increases, coherence is lost in the array data and this prohibits the use of any coherent imaging method. In that case incoherent imaging should be used. Incoherent methods are based in general on a model that describes how energy propagates through the scattering medium. In the project ADAPTIVES we developed such methods for imaging in random acoustic waveguides.
To resume, in this project we developed and analyzed novel coherent and incoherent imaging techniques for realistic problems. The application focus is predominantly on non-destructive testing, underwater acoustics and geophysics. These are areas of great interest as they constitute European, as well as global, priorities due to their potential economic and societal impact.
We model, therefore, the propagation medium as a random process for which we may know some statistical properties, such as the mean speed of propagation. Our goal is to solve the imaging problem in a regime where multiple scattering due to the medium heterogeneities is significant. Imaging in such regimes is quite challenging and requires very different methods from the usual ones in homogeneous or known deterministic environments. The challenge is to produce reliable, i.e. statistically stable results, especially when there is no a priori knowledge about the propagation medium.
We have introduced and developed extensively, a Coherent INTerferometric imaging methodology, CINT, that is based on back propagating cross-correlations of array data rather than the data itself as traditionally done in imaging. CINT is designed for imaging with partially coherent array data recorded in richly scattering media. It uses statistical smoothing techniques to obtain results that are independent of the random medium realization.
CINT works well when the coherent signal due to the reflector that we wish to image is strong enough. In strongly scattering media, however, this is no longer true as the back-scattered field from the medium heterogeneities overwhelms the signal received on the array. In the project ADAPTIVES, to extend the applicability range of CINT, and of other coherent imaging methods to such media, we have developed data filtering techniques that enhance the coherent signal which is useful for imaging. To this end we have developed two approaches: the layer anihilators, a filtering technique that suppresses scattered echoes from layered structures and an adaptive time-frequency filtering technique that can be used in any type of clutter.
Another way to overcome heavy clutter effects is by using auxiliary arrays that are near the reflector that we wish to image. The first step towards imaging under such circumstances is the computation of cross-correlations of the noisy recorded signals, which have attracted a lot of attention recently because of their numerous applications in seismic imaging, volcano monitoring, and petroleum prospecting. We use these cross-correlations for imaging reflectors hidden below very complex structures.
Cross-correlations of passive recordings due to ambient noise sources can be used as well in structural health monitoring for detecting and imaging defects. The quality of the obtained image, measured by the signal to noise ratio (SNR), inherits the SNR of the cross-correlations. It is therefore proportional to the square root of the bandwidth of the noise times the recording time. Moreover, we showed that it is also proportional to the array size. This is important in applications such as structural health monitoring as it implies that for short-time recordings the quality of the image can be improved by increasing the array size.
As the medium backscatter increases, coherence is lost in the array data and this prohibits the use of any coherent imaging method. In that case incoherent imaging should be used. Incoherent methods are based in general on a model that describes how energy propagates through the scattering medium. In the project ADAPTIVES we developed such methods for imaging in random acoustic waveguides.
To resume, in this project we developed and analyzed novel coherent and incoherent imaging techniques for realistic problems. The application focus is predominantly on non-destructive testing, underwater acoustics and geophysics. These are areas of great interest as they constitute European, as well as global, priorities due to their potential economic and societal impact.