Advancing acoustography by multidimensional signal processing techniques

Bach, Michael (2006) Advancing acoustography by multidimensional signal processing techniques. Doctoral thesis, University of Central Lancashire.

[thumbnail of Thesis document] PDF (Thesis document) - Submitted Version
Restricted to Repository staff only
Available under License Creative Commons Attribution Non-commercial Share Alike.

7MB

Abstract

The research presented in this thesis is the investigation into multidimensional signal processing for acoustography. Acoustography is a novel inspection technique similar to x-ray. However, instead of using hazardous ionising radiation, acoustography is based on sound. Inspection data are intensity images of the interior of components under inspection.
Acoustography is a novel screening technique to inspect components without physically damaging them. Multidimensional signal processing refers to the processing of inspection data by signal and image processing techniques. The acoustographic imaging system is characterised with focus on signal and image processing. This system characterisation
investigates into various degradations and image influencing properties. Signal and image processing techniques are then formally defined for the context of processing acoustographic data.
The applicability of denoising and segmentation techniques is demonstrated. Filter primitives are demonstrated to be able to remove certain noise features. Particular focus rests on denoising and segmentation techniques based on physical analogy with diffusion.
The diffusivity is locally controlled by data gradient measures. The diffusion technique using certain parameter settings performs the denoising of the inspection data without manipulating true image features. The same algorithm with a different parameter settings is employed to perform segmentation and thus separating fault features from background.
Based on response characteristics of the acoustographic system, a data fusion algorithm has been developed to merge multiple observations into one datum, thereby increasing the dynamic range. The two stage algorithm consists of an iterative curve fitting algorithm followed by a reverse calculation using the curve parameter to yield a single observation.
The algorithm has been improved further towards robustness to noise. Further, the fusion of denoised data is demonstrated.
As a direct result of the work presented, future work is suggested to improve inferring from observation data to the state of the component under investigation. Further work is suggested to improve the understanding of the imaging system and inverse methods are proposed which take into account various particularities of the acoustographic system.


Repository Staff Only: item control page